nhyatt/putty-source
1
0
Fork 0
mirror of https://git.tartarus.org/simon/putty.git synced 2025-01-09 17:38:00 +00:00
Code Issues Projects Releases Wiki Activity
74f792e00b
putty-source/sshecc.c

2773 lines
77 KiB
C
Raw Normal View History

Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/*
* Elliptic-curve crypto module for PuTTY
* Implements the three required curves, no optional curves
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
*
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
* NOTE: Only curves on prime field are handled by the maths functions
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
* in Weierstrass form using Jacobian co-ordinates.
*
* Montgomery form curves are supported for DH. (Curve25519)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
*
* Edwards form curves are supported for DSA. (Ed25519)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
*/
/*
* References:
*
* Elliptic curves in SSH are specified in RFC 5656:
* http://tools.ietf.org/html/rfc5656
*
* That specification delegates details of public key formatting and a
* lot of underlying mechanism to SEC 1:
* http://www.secg.org/sec1-v2.pdf
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
*
* Montgomery maths from:
* Handbook of elliptic and hyperelliptic curve cryptography, Chapter 13
* http://cs.ucsb.edu/~koc/ccs130h/2013/EllipticHyperelliptic-CohenFrey.pdf
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
*
Add a reference to a spec for Curve25519. It doesn't seem to be all that good a spec, in that it seems to be specified in terms of functions in libssh and hence based on the assumption that you already know exactly what those functions do. But it's something, at least.
2015-05-19 08:56:56 +00:00
* Curve25519 spec from libssh (with reference to other things in the
* libssh code):
* https://git.libssh.org/users/aris/libssh.git/tree/doc/curve25519-sha256@libssh.org.txt
*
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
* Edwards DSA:
* http://ed25519.cr.yp.to/ed25519-20110926.pdf
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
*/
#include <stdlib.h>
#include <assert.h>
#include "ssh.h"
/* ----------------------------------------------------------------------
* Elliptic curve definitions
*/
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
static void initialise_wcurve(struct ec_curve *curve, int bits,
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
const unsigned char *p,
const unsigned char *a, const unsigned char *b,
const unsigned char *n, const unsigned char *Gx,
const unsigned char *Gy)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
int length = bits / 8;
if (bits % 8) ++length;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
curve->type = EC_WEIERSTRASS;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
curve->fieldBits = bits;
curve->p = bignum_from_bytes(p, length);
/* Curve co-efficients */
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
curve->w.a = bignum_from_bytes(a, length);
curve->w.b = bignum_from_bytes(b, length);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Group order and generator */
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
curve->w.n = bignum_from_bytes(n, length);
curve->w.G.x = bignum_from_bytes(Gx, length);
curve->w.G.y = bignum_from_bytes(Gy, length);
curve->w.G.curve = curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
curve->w.G.infinity = false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
static void initialise_mcurve(struct ec_curve *curve, int bits,
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
const unsigned char *p,
const unsigned char *a, const unsigned char *b,
const unsigned char *Gx)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
int length = bits / 8;
if (bits % 8) ++length;
curve->type = EC_MONTGOMERY;
curve->fieldBits = bits;
curve->p = bignum_from_bytes(p, length);
/* Curve co-efficients */
curve->m.a = bignum_from_bytes(a, length);
curve->m.b = bignum_from_bytes(b, length);
/* Generator */
curve->m.G.x = bignum_from_bytes(Gx, length);
curve->m.G.y = NULL;
curve->m.G.z = NULL;
curve->m.G.curve = curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
curve->m.G.infinity = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
static void initialise_ecurve(struct ec_curve *curve, int bits,
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
const unsigned char *p,
const unsigned char *l, const unsigned char *d,
const unsigned char *Bx, const unsigned char *By)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
int length = bits / 8;
if (bits % 8) ++length;
curve->type = EC_EDWARDS;
curve->fieldBits = bits;
curve->p = bignum_from_bytes(p, length);
/* Curve co-efficients */
curve->e.l = bignum_from_bytes(l, length);
curve->e.d = bignum_from_bytes(d, length);
/* Group order and generator */
curve->e.B.x = bignum_from_bytes(Bx, length);
curve->e.B.y = bignum_from_bytes(By, length);
curve->e.B.curve = curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
curve->e.B.infinity = false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
static struct ec_curve *ec_p256(void)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
static struct ec_curve curve = { 0 };
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool initialised = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!initialised)
{
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char p[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x01,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char a[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x01,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfc
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char b[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x5a, 0xc6, 0x35, 0xd8, 0xaa, 0x3a, 0x93, 0xe7,
0xb3, 0xeb, 0xbd, 0x55, 0x76, 0x98, 0x86, 0xbc,
0x65, 0x1d, 0x06, 0xb0, 0xcc, 0x53, 0xb0, 0xf6,
0x3b, 0xce, 0x3c, 0x3e, 0x27, 0xd2, 0x60, 0x4b
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char n[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xbc, 0xe6, 0xfa, 0xad, 0xa7, 0x17, 0x9e, 0x84,
0xf3, 0xb9, 0xca, 0xc2, 0xfc, 0x63, 0x25, 0x51
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Gx[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x6b, 0x17, 0xd1, 0xf2, 0xe1, 0x2c, 0x42, 0x47,
0xf8, 0xbc, 0xe6, 0xe5, 0x63, 0xa4, 0x40, 0xf2,
0x77, 0x03, 0x7d, 0x81, 0x2d, 0xeb, 0x33, 0xa0,
0xf4, 0xa1, 0x39, 0x45, 0xd8, 0x98, 0xc2, 0x96
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Gy[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x4f, 0xe3, 0x42, 0xe2, 0xfe, 0x1a, 0x7f, 0x9b,
0x8e, 0xe7, 0xeb, 0x4a, 0x7c, 0x0f, 0x9e, 0x16,
0x2b, 0xce, 0x33, 0x57, 0x6b, 0x31, 0x5e, 0xce,
0xcb, 0xb6, 0x40, 0x68, 0x37, 0xbf, 0x51, 0xf5
};
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
initialise_wcurve(&curve, 256, p, a, b, n, Gx, Gy);
Log which elliptic curve we're using for ECDH kex. It seems like quite an important thing to mention in the event log! Suppose there's a bug affecting only one curve, for example? Fixed- group Diffie-Hellman has always logged the group, but the ECDH log message just told you the hash and not also the curve. To implement this, I've added a 'textname' field to all elliptic curves, whether they're used for kex or signing or both, suitable for use in this log message and any others we might find a need for in future.
2015-05-19 07:42:23 +00:00
curve.textname = curve.name = "nistp256";
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Now initialised, no need to do it again */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
initialised = true;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
return &curve;
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
static struct ec_curve *ec_p384(void)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
static struct ec_curve curve = { 0 };
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool initialised = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!initialised)
{
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char p[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe,
0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char a[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe,
0xff, 0xff, 0xff, 0xff, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0xff, 0xff, 0xff, 0xfc
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char b[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xb3, 0x31, 0x2f, 0xa7, 0xe2, 0x3e, 0xe7, 0xe4,
0x98, 0x8e, 0x05, 0x6b, 0xe3, 0xf8, 0x2d, 0x19,
0x18, 0x1d, 0x9c, 0x6e, 0xfe, 0x81, 0x41, 0x12,
0x03, 0x14, 0x08, 0x8f, 0x50, 0x13, 0x87, 0x5a,
0xc6, 0x56, 0x39, 0x8d, 0x8a, 0x2e, 0xd1, 0x9d,
0x2a, 0x85, 0xc8, 0xed, 0xd3, 0xec, 0x2a, 0xef
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char n[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xc7, 0x63, 0x4d, 0x81, 0xf4, 0x37, 0x2d, 0xdf,
0x58, 0x1a, 0x0d, 0xb2, 0x48, 0xb0, 0xa7, 0x7a,
0xec, 0xec, 0x19, 0x6a, 0xcc, 0xc5, 0x29, 0x73
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Gx[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0xaa, 0x87, 0xca, 0x22, 0xbe, 0x8b, 0x05, 0x37,
0x8e, 0xb1, 0xc7, 0x1e, 0xf3, 0x20, 0xad, 0x74,
0x6e, 0x1d, 0x3b, 0x62, 0x8b, 0xa7, 0x9b, 0x98,
0x59, 0xf7, 0x41, 0xe0, 0x82, 0x54, 0x2a, 0x38,
0x55, 0x02, 0xf2, 0x5d, 0xbf, 0x55, 0x29, 0x6c,
0x3a, 0x54, 0x5e, 0x38, 0x72, 0x76, 0x0a, 0xb7
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Gy[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x36, 0x17, 0xde, 0x4a, 0x96, 0x26, 0x2c, 0x6f,
0x5d, 0x9e, 0x98, 0xbf, 0x92, 0x92, 0xdc, 0x29,
0xf8, 0xf4, 0x1d, 0xbd, 0x28, 0x9a, 0x14, 0x7c,
0xe9, 0xda, 0x31, 0x13, 0xb5, 0xf0, 0xb8, 0xc0,
0x0a, 0x60, 0xb1, 0xce, 0x1d, 0x7e, 0x81, 0x9d,
0x7a, 0x43, 0x1d, 0x7c, 0x90, 0xea, 0x0e, 0x5f
};
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
initialise_wcurve(&curve, 384, p, a, b, n, Gx, Gy);
Log which elliptic curve we're using for ECDH kex. It seems like quite an important thing to mention in the event log! Suppose there's a bug affecting only one curve, for example? Fixed- group Diffie-Hellman has always logged the group, but the ECDH log message just told you the hash and not also the curve. To implement this, I've added a 'textname' field to all elliptic curves, whether they're used for kex or signing or both, suitable for use in this log message and any others we might find a need for in future.
2015-05-19 07:42:23 +00:00
curve.textname = curve.name = "nistp384";
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Now initialised, no need to do it again */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
initialised = true;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
return &curve;
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
static struct ec_curve *ec_p521(void)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
static struct ec_curve curve = { 0 };
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool initialised = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!initialised)
{
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char p[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x01, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char a[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x01, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xfc
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char b[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x00, 0x51, 0x95, 0x3e, 0xb9, 0x61, 0x8e, 0x1c,
0x9a, 0x1f, 0x92, 0x9a, 0x21, 0xa0, 0xb6, 0x85,
0x40, 0xee, 0xa2, 0xda, 0x72, 0x5b, 0x99, 0xb3,
0x15, 0xf3, 0xb8, 0xb4, 0x89, 0x91, 0x8e, 0xf1,
0x09, 0xe1, 0x56, 0x19, 0x39, 0x51, 0xec, 0x7e,
0x93, 0x7b, 0x16, 0x52, 0xc0, 0xbd, 0x3b, 0xb1,
0xbf, 0x07, 0x35, 0x73, 0xdf, 0x88, 0x3d, 0x2c,
0x34, 0xf1, 0xef, 0x45, 0x1f, 0xd4, 0x6b, 0x50,
0x3f, 0x00
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char n[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x01, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xfa, 0x51, 0x86, 0x87, 0x83, 0xbf, 0x2f,
0x96, 0x6b, 0x7f, 0xcc, 0x01, 0x48, 0xf7, 0x09,
0xa5, 0xd0, 0x3b, 0xb5, 0xc9, 0xb8, 0x89, 0x9c,
0x47, 0xae, 0xbb, 0x6f, 0xb7, 0x1e, 0x91, 0x38,
0x64, 0x09
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Gx[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x00, 0xc6, 0x85, 0x8e, 0x06, 0xb7, 0x04, 0x04,
0xe9, 0xcd, 0x9e, 0x3e, 0xcb, 0x66, 0x23, 0x95,
0xb4, 0x42, 0x9c, 0x64, 0x81, 0x39, 0x05, 0x3f,
0xb5, 0x21, 0xf8, 0x28, 0xaf, 0x60, 0x6b, 0x4d,
0x3d, 0xba, 0xa1, 0x4b, 0x5e, 0x77, 0xef, 0xe7,
0x59, 0x28, 0xfe, 0x1d, 0xc1, 0x27, 0xa2, 0xff,
0xa8, 0xde, 0x33, 0x48, 0xb3, 0xc1, 0x85, 0x6a,
0x42, 0x9b, 0xf9, 0x7e, 0x7e, 0x31, 0xc2, 0xe5,
0xbd, 0x66
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Gy[] = {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
0x01, 0x18, 0x39, 0x29, 0x6a, 0x78, 0x9a, 0x3b,
0xc0, 0x04, 0x5c, 0x8a, 0x5f, 0xb4, 0x2c, 0x7d,
0x1b, 0xd9, 0x98, 0xf5, 0x44, 0x49, 0x57, 0x9b,
0x44, 0x68, 0x17, 0xaf, 0xbd, 0x17, 0x27, 0x3e,
0x66, 0x2c, 0x97, 0xee, 0x72, 0x99, 0x5e, 0xf4,
0x26, 0x40, 0xc5, 0x50, 0xb9, 0x01, 0x3f, 0xad,
0x07, 0x61, 0x35, 0x3c, 0x70, 0x86, 0xa2, 0x72,
0xc2, 0x40, 0x88, 0xbe, 0x94, 0x76, 0x9f, 0xd1,
0x66, 0x50
};
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
initialise_wcurve(&curve, 521, p, a, b, n, Gx, Gy);
Log which elliptic curve we're using for ECDH kex. It seems like quite an important thing to mention in the event log! Suppose there's a bug affecting only one curve, for example? Fixed- group Diffie-Hellman has always logged the group, but the ECDH log message just told you the hash and not also the curve. To implement this, I've added a 'textname' field to all elliptic curves, whether they're used for kex or signing or both, suitable for use in this log message and any others we might find a need for in future.
2015-05-19 07:42:23 +00:00
curve.textname = curve.name = "nistp521";
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
/* Now initialised, no need to do it again */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
initialised = true;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
return &curve;
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
static struct ec_curve *ec_curve25519(void)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
static struct ec_curve curve = { 0 };
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool initialised = false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (!initialised)
{
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char p[] = {
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xed
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char a[] = {
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x07, 0x6d, 0x06
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char b[] = {
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char gx[32] = {
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x09
};
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
initialise_mcurve(&curve, 256, p, a, b, gx);
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* This curve doesn't need a name, because it's never used in
* any format that embeds the curve name */
curve.name = NULL;
Log which elliptic curve we're using for ECDH kex. It seems like quite an important thing to mention in the event log! Suppose there's a bug affecting only one curve, for example? Fixed- group Diffie-Hellman has always logged the group, but the ECDH log message just told you the hash and not also the curve. To implement this, I've added a 'textname' field to all elliptic curves, whether they're used for kex or signing or both, suitable for use in this log message and any others we might find a need for in future.
2015-05-19 07:42:23 +00:00
curve.textname = "Curve25519";
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Now initialised, no need to do it again */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
initialised = true;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
return &curve;
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
static struct ec_curve *ec_ed25519(void)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
static struct ec_curve curve = { 0 };
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool initialised = false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (!initialised)
{
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char q[] = {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff,
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xed
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char l[32] = {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x14, 0xde, 0xf9, 0xde, 0xa2, 0xf7, 0x9c, 0xd6,
0x58, 0x12, 0x63, 0x1a, 0x5c, 0xf5, 0xd3, 0xed
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char d[32] = {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
0x52, 0x03, 0x6c, 0xee, 0x2b, 0x6f, 0xfe, 0x73,
0x8c, 0xc7, 0x40, 0x79, 0x77, 0x79, 0xe8, 0x98,
0x00, 0x70, 0x0a, 0x4d, 0x41, 0x41, 0xd8, 0xab,
0x75, 0xeb, 0x4d, 0xca, 0x13, 0x59, 0x78, 0xa3
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char Bx[32] = {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
0x21, 0x69, 0x36, 0xd3, 0xcd, 0x6e, 0x53, 0xfe,
0xc0, 0xa4, 0xe2, 0x31, 0xfd, 0xd6, 0xdc, 0x5c,
0x69, 0x2c, 0xc7, 0x60, 0x95, 0x25, 0xa7, 0xb2,
0xc9, 0x56, 0x2d, 0x60, 0x8f, 0x25, 0xd5, 0x1a
};
Add missing consts in elliptic curve setup code. All those static arrays giving the curves' constants ought to be 'static const' and go in the data segment, of course.
2015-05-18 20:04:46 +00:00
static const unsigned char By[32] = {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66,
0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66,
0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66,
0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x58
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* This curve doesn't need a name, because it's never used in
* any format that embeds the curve name */
curve.name = NULL;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
initialise_ecurve(&curve, 256, q, l, d, Bx, By);
Log which elliptic curve we're using for ECDH kex. It seems like quite an important thing to mention in the event log! Suppose there's a bug affecting only one curve, for example? Fixed- group Diffie-Hellman has always logged the group, but the ECDH log message just told you the hash and not also the curve. To implement this, I've added a 'textname' field to all elliptic curves, whether they're used for kex or signing or both, suitable for use in this log message and any others we might find a need for in future.
2015-05-19 07:42:23 +00:00
curve.textname = "Ed25519";
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Now initialised, no need to do it again */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
initialised = true;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
return &curve;
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Return 1 if a is -3 % p, otherwise return 0
* This is used because there are some maths optimisations */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool ec_aminus3(const struct ec_curve *curve)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool ret;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Bignum _p;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (curve->type != EC_WEIERSTRASS) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
_p = bignum_add_long(curve->w.a, 3);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ret = !bignum_cmp(curve->p, _p);
freebn(_p);
return ret;
}
/* ----------------------------------------------------------------------
* Elliptic curve field maths
*/
static Bignum ecf_add(const Bignum a, const Bignum b,
const struct ec_curve *curve)
{
Bignum a1, b1, ab, ret;
a1 = bigmod(a, curve->p);
b1 = bigmod(b, curve->p);
ab = bigadd(a1, b1);
freebn(a1);
freebn(b1);
ret = bigmod(ab, curve->p);
freebn(ab);
return ret;
}
static Bignum ecf_square(const Bignum a, const struct ec_curve *curve)
{
return modmul(a, a, curve->p);
}
static Bignum ecf_treble(const Bignum a, const struct ec_curve *curve)
{
Bignum ret, tmp;
/* Double */
tmp = bignum_lshift(a, 1);
/* Add itself (i.e. treble) */
ret = bigadd(tmp, a);
freebn(tmp);
/* Normalise */
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
while (bignum_cmp(ret, curve->p) >= 0)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
tmp = bigsub(ret, curve->p);
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
assert(tmp);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(ret);
ret = tmp;
}
return ret;
}
static Bignum ecf_double(const Bignum a, const struct ec_curve *curve)
{
Bignum ret = bignum_lshift(a, 1);
if (bignum_cmp(ret, curve->p) >= 0)
{
Bignum tmp = bigsub(ret, curve->p);
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
assert(tmp);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(ret);
return tmp;
}
else
{
return ret;
}
}
/* ----------------------------------------------------------------------
* Memory functions
*/
void ec_point_free(struct ec_point *point)
{
if (point == NULL) return;
point->curve = 0;
if (point->x) freebn(point->x);
if (point->y) freebn(point->y);
if (point->z) freebn(point->z);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
point->infinity = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
sfree(point);
}
static struct ec_point *ec_point_new(const struct ec_curve *curve,
const Bignum x, const Bignum y, const Bignum z,
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool infinity)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
struct ec_point *point = snewn(1, struct ec_point);
point->curve = curve;
point->x = x;
point->y = y;
point->z = z;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
point->infinity = infinity;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return point;
}
static struct ec_point *ec_point_copy(const struct ec_point *a)
{
if (a == NULL) return NULL;
return ec_point_new(a->curve,
a->x ? copybn(a->x) : NULL,
a->y ? copybn(a->y) : NULL,
a->z ? copybn(a->z) : NULL,
a->infinity);
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool ec_point_verify(const struct ec_point *a)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (a->infinity) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
} else if (a->curve->type == EC_EDWARDS) {
/* Check y^2 - x^2 - 1 - d * x^2 * y^2 == 0 */
Bignum y2, x2, tmp, tmp2, tmp3;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool ret;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
y2 = ecf_square(a->y, a->curve);
x2 = ecf_square(a->x, a->curve);
tmp = modmul(a->curve->e.d, x2, a->curve->p);
tmp2 = modmul(tmp, y2, a->curve->p);
freebn(tmp);
tmp = modsub(y2, x2, a->curve->p);
freebn(y2);
freebn(x2);
tmp3 = modsub(tmp, tmp2, a->curve->p);
freebn(tmp);
freebn(tmp2);
ret = !bignum_cmp(tmp3, One);
freebn(tmp3);
return ret;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else if (a->curve->type == EC_WEIERSTRASS) {
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Verify y^2 = x^3 + ax + b */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool ret = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Bignum lhs = NULL, x3 = NULL, ax = NULL, x3ax = NULL, x3axm = NULL, x3axb = NULL, rhs = NULL;
Bignum Three = bignum_from_long(3);
lhs = modmul(a->y, a->y, a->curve->p);
/* This uses montgomery multiplication to optimise */
x3 = modpow(a->x, Three, a->curve->p);
freebn(Three);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
ax = modmul(a->curve->w.a, a->x, a->curve->p);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
x3ax = bigadd(x3, ax);
freebn(x3); x3 = NULL;
freebn(ax); ax = NULL;
x3axm = bigmod(x3ax, a->curve->p);
freebn(x3ax); x3ax = NULL;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
x3axb = bigadd(x3axm, a->curve->w.b);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(x3axm); x3axm = NULL;
rhs = bigmod(x3axb, a->curve->p);
freebn(x3axb);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ret = !bignum_cmp(lhs, rhs);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(lhs);
freebn(rhs);
return ret;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
}
/* ----------------------------------------------------------------------
* Elliptic curve point maths
*/
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
/* Returns true on success and false on memory error */
static bool ecp_normalise(struct ec_point *a)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
if (!a) {
Fix some rogue // comments. That's what you get for changing things at the last minute...
2014-11-03 18:41:56 +00:00
/* No point */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (a->infinity) {
Fix some rogue // comments. That's what you get for changing things at the last minute...
2014-11-03 18:41:56 +00:00
/* Point is at infinity - i.e. normalised */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (a->curve->type == EC_WEIERSTRASS) {
/* In Jacobian Coordinates the triple (X, Y, Z) represents
the affine point (X / Z^2, Y / Z^3) */
Bignum Z2, Z2inv, Z3, Z3inv, tx, ty;
if (!a->x || !a->y) {
/* No point defined */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else if (!a->z) {
/* Already normalised */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Z2 = ecf_square(a->z, a->curve);
Z2inv = modinv(Z2, a->curve->p);
if (!Z2inv) {
freebn(Z2);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
tx = modmul(a->x, Z2inv, a->curve->p);
freebn(Z2inv);
Z3 = modmul(Z2, a->z, a->curve->p);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(Z2);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Z3inv = modinv(Z3, a->curve->p);
freebn(Z3);
if (!Z3inv) {
freebn(tx);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
ty = modmul(a->y, Z3inv, a->curve->p);
freebn(Z3inv);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
freebn(a->x);
a->x = tx;
freebn(a->y);
a->y = ty;
freebn(a->z);
a->z = NULL;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else if (a->curve->type == EC_MONTGOMERY) {
/* In Montgomery (X : Z) represents the x co-ord (X / Z, ?) */
Bignum tmp, tmp2;
if (!a->x) {
/* No point defined */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else if (!a->z) {
/* Already normalised */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
tmp = modinv(a->z, a->curve->p);
if (!tmp) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
tmp2 = modmul(a->x, tmp, a->curve->p);
freebn(tmp);
freebn(a->z);
a->z = NULL;
freebn(a->x);
a->x = tmp2;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
} else if (a->curve->type == EC_EDWARDS) {
/* Always normalised */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static struct ec_point *ecp_doublew(const struct ec_point *a, bool aminus3)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Bignum S, M, outx, outy, outz;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (bignum_cmp(a->y, Zero) == 0)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
/* Identity */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, NULL, NULL, NULL, true);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
/* S = 4*X*Y^2 */
{
Bignum Y2, XY2, _2XY2;
Y2 = ecf_square(a->y, a->curve);
XY2 = modmul(a->x, Y2, a->curve->p);
freebn(Y2);
_2XY2 = ecf_double(XY2, a->curve);
freebn(XY2);
S = ecf_double(_2XY2, a->curve);
freebn(_2XY2);
}
/* Faster calculation if a = -3 */
if (aminus3) {
/* if a = -3, then M can also be calculated as M = 3*(X + Z^2)*(X - Z^2) */
Bignum Z2, XpZ2, XmZ2, second;
if (a->z == NULL) {
Z2 = copybn(One);
} else {
Z2 = ecf_square(a->z, a->curve);
}
XpZ2 = ecf_add(a->x, Z2, a->curve);
XmZ2 = modsub(a->x, Z2, a->curve->p);
freebn(Z2);
second = modmul(XpZ2, XmZ2, a->curve->p);
freebn(XpZ2);
freebn(XmZ2);
M = ecf_treble(second, a->curve);
freebn(second);
} else {
/* M = 3*X^2 + a*Z^4 */
Bignum _3X2, X2, aZ4;
if (a->z == NULL) {
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
aZ4 = copybn(a->curve->w.a);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
} else {
Bignum Z2, Z4;
Z2 = ecf_square(a->z, a->curve);
Z4 = ecf_square(Z2, a->curve);
freebn(Z2);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
aZ4 = modmul(a->curve->w.a, Z4, a->curve->p);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(Z4);
}
X2 = modmul(a->x, a->x, a->curve->p);
_3X2 = ecf_treble(X2, a->curve);
freebn(X2);
M = ecf_add(_3X2, aZ4, a->curve);
freebn(_3X2);
freebn(aZ4);
}
/* X' = M^2 - 2*S */
{
Bignum M2, _2S;
M2 = ecf_square(M, a->curve);
_2S = ecf_double(S, a->curve);
outx = modsub(M2, _2S, a->curve->p);
freebn(M2);
freebn(_2S);
}
/* Y' = M*(S - X') - 8*Y^4 */
{
Bignum SX, MSX, Eight, Y2, Y4, _8Y4;
SX = modsub(S, outx, a->curve->p);
freebn(S);
MSX = modmul(M, SX, a->curve->p);
freebn(SX);
freebn(M);
Y2 = ecf_square(a->y, a->curve);
Y4 = ecf_square(Y2, a->curve);
freebn(Y2);
Eight = bignum_from_long(8);
_8Y4 = modmul(Eight, Y4, a->curve->p);
freebn(Eight);
freebn(Y4);
outy = modsub(MSX, _8Y4, a->curve->p);
freebn(MSX);
freebn(_8Y4);
}
/* Z' = 2*Y*Z */
{
Bignum YZ;
if (a->z == NULL) {
YZ = copybn(a->y);
} else {
YZ = modmul(a->y, a->z, a->curve->p);
}
outz = ecf_double(YZ, a->curve);
freebn(YZ);
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, outx, outy, outz, false);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
static struct ec_point *ecp_doublem(const struct ec_point *a)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Bignum z, outx, outz, xpz, xmz;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
z = a->z;
if (!z) {
z = One;
}
/* 4xz = (x + z)^2 - (x - z)^2 */
{
Bignum tmp;
tmp = ecf_add(a->x, z, a->curve);
xpz = ecf_square(tmp, a->curve);
freebn(tmp);
tmp = modsub(a->x, z, a->curve->p);
xmz = ecf_square(tmp, a->curve);
freebn(tmp);
}
/* outx = (x + z)^2 * (x - z)^2 */
outx = modmul(xpz, xmz, a->curve->p);
/* outz = 4xz * ((x - z)^2 + ((A + 2) / 4)*4xz) */
{
Bignum _4xz, tmp, tmp2, tmp3;
tmp = bignum_from_long(2);
tmp2 = ecf_add(a->curve->m.a, tmp, a->curve);
freebn(tmp);
_4xz = modsub(xpz, xmz, a->curve->p);
freebn(xpz);
tmp = modmul(tmp2, _4xz, a->curve->p);
freebn(tmp2);
tmp2 = bignum_from_long(4);
tmp3 = modinv(tmp2, a->curve->p);
freebn(tmp2);
if (!tmp3) {
freebn(tmp);
freebn(_4xz);
freebn(outx);
freebn(xmz);
return NULL;
}
tmp2 = modmul(tmp, tmp3, a->curve->p);
freebn(tmp);
freebn(tmp3);
tmp = ecf_add(xmz, tmp2, a->curve);
freebn(xmz);
freebn(tmp2);
outz = modmul(_4xz, tmp, a->curve->p);
freebn(_4xz);
freebn(tmp);
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, outx, NULL, outz, false);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Forward declaration for Edwards curve doubling */
static struct ec_point *ecp_add(const struct ec_point *a,
const struct ec_point *b,
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool aminus3);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static struct ec_point *ecp_double(const struct ec_point *a, bool aminus3)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
if (a->infinity)
{
/* Identity */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, NULL, NULL, NULL, true);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (a->curve->type == EC_EDWARDS)
{
return ecp_add(a, a, aminus3);
}
else if (a->curve->type == EC_WEIERSTRASS)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
return ecp_doublew(a, aminus3);
}
else
{
return ecp_doublem(a);
}
}
static struct ec_point *ecp_addw(const struct ec_point *a,
const struct ec_point *b,
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool aminus3)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
Bignum U1, U2, S1, S2, outx, outy, outz;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* U1 = X1*Z2^2 */
/* S1 = Y1*Z2^3 */
if (b->z) {
Bignum Z2, Z3;
Z2 = ecf_square(b->z, a->curve);
U1 = modmul(a->x, Z2, a->curve->p);
Z3 = modmul(Z2, b->z, a->curve->p);
freebn(Z2);
S1 = modmul(a->y, Z3, a->curve->p);
freebn(Z3);
} else {
U1 = copybn(a->x);
S1 = copybn(a->y);
}
/* U2 = X2*Z1^2 */
/* S2 = Y2*Z1^3 */
if (a->z) {
Bignum Z2, Z3;
Z2 = ecf_square(a->z, b->curve);
U2 = modmul(b->x, Z2, b->curve->p);
Z3 = modmul(Z2, a->z, b->curve->p);
freebn(Z2);
S2 = modmul(b->y, Z3, b->curve->p);
freebn(Z3);
} else {
U2 = copybn(b->x);
S2 = copybn(b->y);
}
/* Check if multiplying by self */
if (bignum_cmp(U1, U2) == 0)
{
freebn(U1);
freebn(U2);
if (bignum_cmp(S1, S2) == 0)
{
freebn(S1);
freebn(S2);
return ecp_double(a, aminus3);
}
else
{
freebn(S1);
freebn(S2);
/* Infinity */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, NULL, NULL, NULL, true);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
}
{
Bignum H, R, UH2, H3;
/* H = U2 - U1 */
H = modsub(U2, U1, a->curve->p);
freebn(U2);
/* R = S2 - S1 */
R = modsub(S2, S1, a->curve->p);
freebn(S2);
/* X3 = R^2 - H^3 - 2*U1*H^2 */
{
Bignum R2, H2, _2UH2, first;
H2 = ecf_square(H, a->curve);
UH2 = modmul(U1, H2, a->curve->p);
freebn(U1);
H3 = modmul(H2, H, a->curve->p);
freebn(H2);
R2 = ecf_square(R, a->curve);
_2UH2 = ecf_double(UH2, a->curve);
first = modsub(R2, H3, a->curve->p);
freebn(R2);
outx = modsub(first, _2UH2, a->curve->p);
freebn(first);
freebn(_2UH2);
}
/* Y3 = R*(U1*H^2 - X3) - S1*H^3 */
{
Bignum RUH2mX, UH2mX, SH3;
UH2mX = modsub(UH2, outx, a->curve->p);
freebn(UH2);
RUH2mX = modmul(R, UH2mX, a->curve->p);
freebn(UH2mX);
freebn(R);
SH3 = modmul(S1, H3, a->curve->p);
freebn(S1);
freebn(H3);
outy = modsub(RUH2mX, SH3, a->curve->p);
freebn(RUH2mX);
freebn(SH3);
}
/* Z3 = H*Z1*Z2 */
if (a->z && b->z) {
Bignum ZZ;
ZZ = modmul(a->z, b->z, a->curve->p);
outz = modmul(H, ZZ, a->curve->p);
freebn(H);
freebn(ZZ);
} else if (a->z) {
outz = modmul(H, a->z, a->curve->p);
freebn(H);
} else if (b->z) {
outz = modmul(H, b->z, a->curve->p);
freebn(H);
} else {
outz = H;
}
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, outx, outy, outz, false);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
static struct ec_point *ecp_addm(const struct ec_point *a,
const struct ec_point *b,
const struct ec_point *base)
{
Bignum outx, outz, az, bz;
az = a->z;
if (!az) {
az = One;
}
bz = b->z;
if (!bz) {
bz = One;
}
/* a-b is maintained at 1 due to Montgomery ladder implementation */
/* Xa+b = Za-b * ((Xa - Za)*(Xb + Zb) + (Xa + Za)*(Xb - Zb))^2 */
/* Za+b = Xa-b * ((Xa - Za)*(Xb + Zb) - (Xa + Za)*(Xb - Zb))^2 */
{
Bignum tmp, tmp2, tmp3, tmp4;
/* (Xa + Za) * (Xb - Zb) */
tmp = ecf_add(a->x, az, a->curve);
tmp2 = modsub(b->x, bz, a->curve->p);
tmp3 = modmul(tmp, tmp2, a->curve->p);
freebn(tmp);
freebn(tmp2);
/* (Xa - Za) * (Xb + Zb) */
tmp = modsub(a->x, az, a->curve->p);
tmp2 = ecf_add(b->x, bz, a->curve);
tmp4 = modmul(tmp, tmp2, a->curve->p);
freebn(tmp);
freebn(tmp2);
tmp = ecf_add(tmp3, tmp4, a->curve);
outx = ecf_square(tmp, a->curve);
freebn(tmp);
tmp = modsub(tmp3, tmp4, a->curve->p);
freebn(tmp3);
freebn(tmp4);
tmp2 = ecf_square(tmp, a->curve);
freebn(tmp);
outz = modmul(base->x, tmp2, a->curve->p);
freebn(tmp2);
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, outx, NULL, outz, false);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
static struct ec_point *ecp_adde(const struct ec_point *a,
const struct ec_point *b)
{
Bignum outx, outy, dmul;
/* outx = (a->x * b->y + b->x * a->y) /
* (1 + a->curve->e.d * a->x * b->x * a->y * b->y) */
{
Bignum tmp, tmp2, tmp3, tmp4;
tmp = modmul(a->x, b->y, a->curve->p);
tmp2 = modmul(b->x, a->y, a->curve->p);
tmp3 = ecf_add(tmp, tmp2, a->curve);
tmp4 = modmul(tmp, tmp2, a->curve->p);
freebn(tmp);
freebn(tmp2);
dmul = modmul(a->curve->e.d, tmp4, a->curve->p);
freebn(tmp4);
tmp = ecf_add(One, dmul, a->curve);
tmp2 = modinv(tmp, a->curve->p);
freebn(tmp);
if (!tmp2)
{
freebn(tmp3);
freebn(dmul);
return NULL;
}
outx = modmul(tmp3, tmp2, a->curve->p);
freebn(tmp3);
freebn(tmp2);
}
/* outy = (a->y * b->y + a->x * b->x) /
* (1 - a->curve->e.d * a->x * b->x * a->y * b->y) */
{
Bignum tmp, tmp2, tmp3, tmp4;
tmp = modsub(One, dmul, a->curve->p);
freebn(dmul);
tmp2 = modinv(tmp, a->curve->p);
freebn(tmp);
if (!tmp2)
{
freebn(outx);
return NULL;
}
tmp = modmul(a->y, b->y, a->curve->p);
tmp3 = modmul(a->x, b->x, a->curve->p);
tmp4 = ecf_add(tmp, tmp3, a->curve);
freebn(tmp);
freebn(tmp3);
outy = modmul(tmp4, tmp2, a->curve->p);
freebn(tmp4);
freebn(tmp2);
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return ec_point_new(a->curve, outx, outy, NULL, false);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
static struct ec_point *ecp_add(const struct ec_point *a,
const struct ec_point *b,
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool aminus3)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
if (a->curve != b->curve) {
return NULL;
}
/* Check if multiplying by infinity */
if (a->infinity) return ec_point_copy(b);
if (b->infinity) return ec_point_copy(a);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (a->curve->type == EC_EDWARDS)
{
return ecp_adde(a, b);
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (a->curve->type == EC_WEIERSTRASS)
{
return ecp_addw(a, b, aminus3);
}
return NULL;
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static struct ec_point *ecp_mul_(
const struct ec_point *a, const Bignum b, bool aminus3)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
struct ec_point *A, *ret;
int bits, i;
A = ec_point_copy(a);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ret = ec_point_new(a->curve, NULL, NULL, NULL, true);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
bits = bignum_bitcount(b);
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
for (i = 0; i < bits; ++i)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
if (bignum_bit(b, i))
{
struct ec_point *tmp = ecp_add(ret, A, aminus3);
ec_point_free(ret);
ret = tmp;
}
if (i+1 != bits)
{
struct ec_point *tmp = ecp_double(A, aminus3);
ec_point_free(A);
A = tmp;
}
}
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
ec_point_free(A);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return ret;
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
static struct ec_point *ecp_mulw(const struct ec_point *a, const Bignum b)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
struct ec_point *ret = ecp_mul_(a, b, ec_aminus3(a->curve));
if (!ecp_normalise(ret)) {
ec_point_free(ret);
return NULL;
}
return ret;
}
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
static struct ec_point *ecp_mule(const struct ec_point *a, const Bignum b)
{
int i;
struct ec_point *ret;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ret = ec_point_new(a->curve, NULL, NULL, NULL, true);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
for (i = bignum_bitcount(b); i >= 0 && ret; --i)
{
{
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
struct ec_point *tmp = ecp_double(ret, false);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec_point_free(ret);
ret = tmp;
}
if (ret && bignum_bit(b, i))
{
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
struct ec_point *tmp = ecp_add(ret, a, false);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec_point_free(ret);
ret = tmp;
}
}
return ret;
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
static struct ec_point *ecp_mulm(const struct ec_point *p, const Bignum n)
{
struct ec_point *P1, *P2;
int bits, i;
/* P1 <- P and P2 <- [2]P */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
P2 = ecp_double(p, false);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
P1 = ec_point_copy(p);
/* for i = bits − 2 down to 0 */
bits = bignum_bitcount(n);
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
for (i = bits - 2; i >= 0; --i)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
if (!bignum_bit(n, i))
{
/* P2 <- P1 + P2 */
struct ec_point *tmp = ecp_addm(P1, P2, p);
ec_point_free(P2);
P2 = tmp;
/* P1 <- [2]P1 */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
tmp = ecp_double(P1, false);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
ec_point_free(P1);
P1 = tmp;
}
else
{
/* P1 <- P1 + P2 */
struct ec_point *tmp = ecp_addm(P1, P2, p);
ec_point_free(P1);
P1 = tmp;
/* P2 <- [2]P2 */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
tmp = ecp_double(P2, false);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
ec_point_free(P2);
P2 = tmp;
}
}
Remove pointless NULL checks in the ECC code. snew(), and most of the bignum functions, are deliberately written to fail an assertion and terminate the program rather than return NULL, so there's no point carefully checking their every return value for NULL. This removes a huge amount of pointless error-checking code, and makes the elliptic curve arithmetic almost legible in places :-) I've kept error checks after modinv(), because that can return NULL if asked to invert zero. bigsub() can also fail in principle, because our bignums are non-negative only, but in the couple of cases where it's used there's a preceding compare that should prevent it, so I've just added assertions.
2015-05-15 12:27:15 +00:00
ec_point_free(P2);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (!ecp_normalise(P1)) {
ec_point_free(P1);
return NULL;
}
return P1;
}
/* Not static because it is used by sshecdsag.c to generate a new key */
struct ec_point *ecp_mul(const struct ec_point *a, const Bignum b)
{
if (a->curve->type == EC_WEIERSTRASS) {
return ecp_mulw(a, b);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
} else if (a->curve->type == EC_EDWARDS) {
return ecp_mule(a, b);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else {
return ecp_mulm(a, b);
}
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
static struct ec_point *ecp_summul(const Bignum a, const Bignum b,
const struct ec_point *point)
{
struct ec_point *aG, *bP, *ret;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool aminus3;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (point->curve->type != EC_WEIERSTRASS) {
return NULL;
}
aminus3 = ec_aminus3(point->curve);
aG = ecp_mul_(&point->curve->w.G, a, aminus3);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!aG) return NULL;
bP = ecp_mul_(point, b, aminus3);
if (!bP) {
ec_point_free(aG);
return NULL;
}
ret = ecp_add(aG, bP, aminus3);
ec_point_free(aG);
ec_point_free(bP);
if (!ecp_normalise(ret)) {
ec_point_free(ret);
return NULL;
}
return ret;
}
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
static Bignum *ecp_edx(const struct ec_curve *curve, const Bignum y)
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
{
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Get the x value on the given Edwards curve for a given y */
Bignum x, xx;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* xx = (y^2 - 1) / (d * y^2 + 1) */
{
Bignum tmp, tmp2, tmp3;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
tmp = ecf_square(y, curve);
tmp2 = modmul(curve->e.d, tmp, curve->p);
tmp3 = ecf_add(tmp2, One, curve);
freebn(tmp2);
tmp2 = modinv(tmp3, curve->p);
freebn(tmp3);
if (!tmp2) {
freebn(tmp);
return NULL;
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
tmp3 = modsub(tmp, One, curve->p);
freebn(tmp);
xx = modmul(tmp3, tmp2, curve->p);
freebn(tmp3);
freebn(tmp2);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* x = xx^((p + 3) / 8) */
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Bignum tmp, tmp2;
tmp = bignum_add_long(curve->p, 3);
tmp2 = bignum_rshift(tmp, 3);
freebn(tmp);
x = modpow(xx, tmp2, curve->p);
freebn(tmp2);
}
/* if x^2 - xx != 0 then x = x*(2^((p - 1) / 4)) */
{
Bignum tmp, tmp2;
tmp = ecf_square(x, curve);
tmp2 = modsub(tmp, xx, curve->p);
freebn(tmp);
freebn(xx);
if (bignum_cmp(tmp2, Zero)) {
Bignum tmp3;
freebn(tmp2);
tmp = modsub(curve->p, One, curve->p);
tmp2 = bignum_rshift(tmp, 2);
freebn(tmp);
tmp = bignum_from_long(2);
tmp3 = modpow(tmp, tmp2, curve->p);
freebn(tmp);
freebn(tmp2);
tmp = modmul(x, tmp3, curve->p);
freebn(x);
freebn(tmp3);
x = tmp;
} else {
freebn(tmp2);
}
}
/* if x % 2 != 0 then x = p - x */
if (bignum_bit(x, 0)) {
Bignum tmp = modsub(curve->p, x, curve->p);
freebn(x);
x = tmp;
}
return x;
}
/* ----------------------------------------------------------------------
* Public point from private
*/
struct ec_point *ec_public(const Bignum privateKey, const struct ec_curve *curve)
{
if (curve->type == EC_WEIERSTRASS) {
return ecp_mul(&curve->w.G, privateKey);
} else if (curve->type == EC_EDWARDS) {
/* hash = H(sk) (where hash creates 2 * fieldBits)
* b = fieldBits
* a = 2^(b-2) + SUM(2^i * h_i) for i = 2 -> b-2
* publicKey = aB */
struct ec_point *ret;
unsigned char hash[512/8];
Bignum a;
int i, keylen;
SHA512_State s;
SHA512_Init(&s);
keylen = curve->fieldBits / 8;
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
for (i = 0; i < keylen; ++i)
put_byte(&s, bignum_byte(privateKey, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
SHA512_Final(&s, hash);
Clear an extra low bit in EdDSA exponent calculation. The source paper, and OpenSSH, agree that the lowest bit index used from the hash of the private key is bit 3, i.e. bits 0,1,2 at the bottom are all zero. We were only clearing bits 0 and 1, which would have worked for about half of keys. I must have got lucky during testing!
2015-05-10 13:04:16 +00:00
/* The second part is simply turning the hash into a Bignum,
* however the 2^(b-2) bit *must* be set, and the bottom 3
* bits *must* not be */
hash[0] &= 0xf8; /* Unset bottom 3 bits (if set) */
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
hash[31] &= 0x7f; /* Unset above (b-2) */
hash[31] |= 0x40; /* Set 2^(b-2) */
/* Chop off the top part and convert to int */
a = bignum_from_bytes_le(hash, 32);
ret = ecp_mul(&curve->e.B, a);
freebn(a);
return ret;
} else {
return NULL;
}
}
/* ----------------------------------------------------------------------
* Basic sign and verify routines
*/
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool _ecdsa_verify(const struct ec_point *publicKey,
const unsigned char *data, const int dataLen,
const Bignum r, const Bignum s)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
int z_bits, n_bits;
Bignum z;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool valid = false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (publicKey->curve->type != EC_WEIERSTRASS) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
/* Sanity checks */
if (bignum_cmp(r, Zero) == 0 || bignum_cmp(r, publicKey->curve->w.n) >= 0
|| bignum_cmp(s, Zero) == 0 || bignum_cmp(s, publicKey->curve->w.n) >= 0)
{
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* z = left most bitlen(curve->n) of data */
z = bignum_from_bytes(data, dataLen);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
n_bits = bignum_bitcount(publicKey->curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
z_bits = bignum_bitcount(z);
if (z_bits > n_bits)
{
Bignum tmp = bignum_rshift(z, z_bits - n_bits);
freebn(z);
z = tmp;
}
/* Ensure z in range of n */
{
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Bignum tmp = bigmod(z, publicKey->curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(z);
z = tmp;
}
/* Calculate signature */
{
Bignum w, x, u1, u2;
struct ec_point *tmp;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
w = modinv(s, publicKey->curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!w) {
freebn(z);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
u1 = modmul(z, w, publicKey->curve->w.n);
u2 = modmul(r, w, publicKey->curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(w);
tmp = ecp_summul(u1, u2, publicKey);
freebn(u1);
freebn(u2);
if (!tmp) {
freebn(z);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
x = bigmod(tmp->x, publicKey->curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec_point_free(tmp);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
valid = (bignum_cmp(r, x) == 0);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(x);
}
freebn(z);
return valid;
}
static void _ecdsa_sign(const Bignum privateKey, const struct ec_curve *curve,
const unsigned char *data, const int dataLen,
Bignum *r, Bignum *s)
{
unsigned char digest[20];
int z_bits, n_bits;
Bignum z, k;
struct ec_point *kG;
*r = NULL;
*s = NULL;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (curve->type != EC_WEIERSTRASS) {
return;
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* z = left most bitlen(curve->n) of data */
z = bignum_from_bytes(data, dataLen);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
n_bits = bignum_bitcount(curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
z_bits = bignum_bitcount(z);
if (z_bits > n_bits)
{
Bignum tmp;
tmp = bignum_rshift(z, z_bits - n_bits);
freebn(z);
z = tmp;
}
/* Generate k between 1 and curve->n, using the same deterministic
* k generation system we use for conventional DSA. */
SHA_Simple(data, dataLen, digest);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
k = dss_gen_k("ECDSA deterministic k generator", curve->w.n, privateKey,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
digest, sizeof(digest));
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
kG = ecp_mul(&curve->w.G, k);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!kG) {
freebn(z);
freebn(k);
return;
}
/* r = kG.x mod n */
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
*r = bigmod(kG->x, curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec_point_free(kG);
/* s = (z + r * priv)/k mod n */
{
Bignum rPriv, zMod, first, firstMod, kInv;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
rPriv = modmul(*r, privateKey, curve->w.n);
zMod = bigmod(z, curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(z);
first = bigadd(rPriv, zMod);
freebn(rPriv);
freebn(zMod);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
firstMod = bigmod(first, curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(first);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
kInv = modinv(k, curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(k);
if (!kInv) {
freebn(firstMod);
freebn(*r);
return;
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
*s = modmul(firstMod, kInv, curve->w.n);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
freebn(firstMod);
freebn(kInv);
}
}
/* ----------------------------------------------------------------------
* Misc functions
*/
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
static Bignum BinarySource_get_mp_le(BinarySource *src)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ptrlen mp_str = get_string(src);
return bignum_from_bytes_le(mp_str.ptr, mp_str.len);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
#define get_mp_le(src) BinarySource_get_mp_le(BinarySource_UPCAST(src))
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool decodepoint_ed(const char *p, int length, struct ec_point *point)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
/* Got some conversion to do, first read in the y co-ord */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool negative;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
point->y = bignum_from_bytes_le((const unsigned char*)p, length);
if ((unsigned)bignum_bitcount(point->y) > point->curve->fieldBits) {
freebn(point->y);
point->y = NULL;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
/* Read x bit and then reset it */
negative = bignum_bit(point->y, point->curve->fieldBits - 1);
bignum_set_bit(point->y, point->curve->fieldBits - 1, 0);
Fix an assertion failure when loading Ed25519 keys. "amax == 0 || a[amax] != 0" Essentially, when decodepoint_ed() clears the top bit of the key, it needs to call bn_restore_invariant() in case that left the high-order word zero. Bug found with the help of afl-fuzz.
2015-10-12 22:43:49 +00:00
bn_restore_invariant(point->y);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Get the x from the y */
point->x = ecp_edx(point->curve, point->y);
if (!point->x) {
freebn(point->y);
point->y = NULL;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
if (negative) {
Bignum tmp = modsub(point->curve->p, point->x, point->curve->p);
freebn(point->x);
point->x = tmp;
}
/* Verify the point is on the curve */
if (!ec_point_verify(point)) {
freebn(point->x);
point->x = NULL;
freebn(point->y);
point->y = NULL;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool decodepoint(const char *p, int length, struct ec_point *point)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (point->curve->type == EC_EDWARDS) {
return decodepoint_ed(p, length, point);
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (length < 1 || p[0] != 0x04) /* Only support uncompressed point */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Skip compression flag */
++p;
--length;
/* The two values must be equal length */
if (length % 2 != 0) {
point->x = NULL;
point->y = NULL;
point->z = NULL;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
length = length / 2;
Make lots of generic data parameters into 'void *'. This is a cleanup I started to notice a need for during the BinarySink work. It removes a lot of faffing about casting things to char * or unsigned char * so that some API will accept them, even though lots of such APIs really take a plain 'block of raw binary data' argument and don't care what C thinks the signedness of that data might be - they may well reinterpret it back and forth internally. So I've tried to arrange for all the function call APIs that ought to have a void * (or const void *) to have one, and those that need to do pointer arithmetic on the parameter internally can cast it back at the top of the function. That saves endless ad-hoc casts at the call sites.
2018-05-26 07:31:34 +00:00
point->x = bignum_from_bytes(p, length);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
p += length;
Make lots of generic data parameters into 'void *'. This is a cleanup I started to notice a need for during the BinarySink work. It removes a lot of faffing about casting things to char * or unsigned char * so that some API will accept them, even though lots of such APIs really take a plain 'block of raw binary data' argument and don't care what C thinks the signedness of that data might be - they may well reinterpret it back and forth internally. So I've tried to arrange for all the function call APIs that ought to have a void * (or const void *) to have one, and those that need to do pointer arithmetic on the parameter internally can cast it back at the top of the function. That saves endless ad-hoc casts at the call sites.
2018-05-26 07:31:34 +00:00
point->y = bignum_from_bytes(p, length);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
point->z = NULL;
/* Verify the point is on the curve */
if (!ec_point_verify(point)) {
Fixes to memory management in the elliptic curve code. There was an error-handling path testing the wrong variable; an inappropriate call to ec_point_free in decodepoint() (in fact, that function always gets passed a pointer to an ec_point structure that's not a dynamically allocated block at all or not in its own right, so we should have just cleared its contents without freeing the structure itself); a missing return on an error path which would have caused the same structure to be freed a second time; and two missing freebn in ecdsa_sign. Patch due to Tim Kosse.
2014-12-20 17:07:17 +00:00
freebn(point->x);
point->x = NULL;
freebn(point->y);
point->y = NULL;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return true;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool BinarySource_get_point(BinarySource *src, struct ec_point *point)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ptrlen str = get_string(src);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
if (get_err(src)) return false;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
return decodepoint(str.ptr, str.len, point);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
#define get_point(src, pt) BinarySource_get_point(BinarySource_UPCAST(src), pt)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* ----------------------------------------------------------------------
* Exposed ECDSA interface
*/
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
struct ecsign_extra {
struct ec_curve *(*curve)(void);
Turn SSH hashes into a classoid. The new version of ssh_hash has the same nice property as ssh2_mac, that I can make the generic interface object type function directly as a BinarySink so that clients don't have to call h->sink() and worry about the separate sink object they get back from that.
2018-09-13 15:41:46 +00:00
const struct ssh_hashalg *hash;
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* These fields are used by the OpenSSH PEM format importer/exporter */
const unsigned char *oid;
int oidlen;
};
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
static void ecdsa_freekey(ssh_key *key)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
struct ec_key *ec;
if (!key) return;
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
ec = container_of(key, struct ec_key, sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (ec->publicKey.x)
freebn(ec->publicKey.x);
if (ec->publicKey.y)
freebn(ec->publicKey.y);
if (ec->publicKey.z)
freebn(ec->publicKey.z);
if (ec->privateKey)
freebn(ec->privateKey);
sfree(ec);
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static ssh_key *ecdsa_new_pub(const ssh_keyalg *self, ptrlen data)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct ecsign_extra *extra =
(const struct ecsign_extra *)self->extra;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource src[1];
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
struct ec_key *ec;
struct ec_curve *curve;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource_BARE_INIT(src, data.ptr, data.len);
get_string(src);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
curve = extra->curve();
assert(curve->type == EC_WEIERSTRASS || curve->type == EC_EDWARDS);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
/* Curve name is duplicated for Weierstrass form */
if (curve->type == EC_WEIERSTRASS) {
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (!ptrlen_eq_string(get_string(src), curve->name))
return NULL;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec = snew(struct ec_key);
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ec->sshk = self;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec->publicKey.curve = curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ec->publicKey.infinity = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec->publicKey.x = NULL;
ec->publicKey.y = NULL;
ec->publicKey.z = NULL;
Another ecdsa_newkey crash: initialise ec->privateKey earlier. This one might be exploitable, since without the fix, ecdsa_freekey() tries to wipe the bignum pointed to by an uninitialised pointer. Bug found with the help of afl-fuzz.
2015-10-09 23:20:51 +00:00
ec->privateKey = NULL;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (!get_point(src, &ec->publicKey)) {
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return NULL;
}
if (!ec->publicKey.x || !ec->publicKey.y ||
bignum_cmp(ec->publicKey.x, curve->p) >= 0 ||
bignum_cmp(ec->publicKey.y, curve->p) >= 0)
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec = NULL;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
return &ec->sshk;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static char *ecdsa_cache_str(ssh_key *key)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
char *p;
int len, i, pos, nibbles;
static const char hex[] = "0123456789abcdef";
if (!ec->publicKey.x || !ec->publicKey.y || !ec->publicKey.curve)
return NULL;
len = 4 + 2 + 1; /* 2 x "0x", punctuation, \0 */
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
if (ec->publicKey.curve->name)
len += strlen(ec->publicKey.curve->name); /* Curve name */
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
len += 4 * (bignum_bitcount(ec->publicKey.x) + 15) / 16;
len += 4 * (bignum_bitcount(ec->publicKey.y) + 15) / 16;
p = snewn(len, char);
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
pos = 0;
if (ec->publicKey.curve->name)
pos += sprintf(p + pos, "%s,", ec->publicKey.curve->name);
pos += sprintf(p + pos, "0x");
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
nibbles = (3 + bignum_bitcount(ec->publicKey.x)) / 4;
if (nibbles < 1)
nibbles = 1;
for (i = nibbles; i--;) {
p[pos++] =
hex[(bignum_byte(ec->publicKey.x, i / 2) >> (4 * (i % 2))) & 0xF];
}
pos += sprintf(p + pos, ",0x");
nibbles = (3 + bignum_bitcount(ec->publicKey.y)) / 4;
if (nibbles < 1)
nibbles = 1;
for (i = nibbles; i--;) {
p[pos++] =
hex[(bignum_byte(ec->publicKey.y, i / 2) >> (4 * (i % 2))) & 0xF];
}
p[pos] = '\0';
return p;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
static void ecdsa_public_blob(ssh_key *key, BinarySink *bs)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
int pointlen;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
int i;
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (ec->publicKey.curve->type == EC_EDWARDS) {
/* Edwards compressed form "ssh-ed25519" point y[:-1] + x[0:1] */
pointlen = ec->publicKey.curve->fieldBits / 8;
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(pointlen >= 2);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
put_stringz(bs, ec->sshk->ssh_id);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_uint32(bs, pointlen);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
for (i = 0; i < pointlen - 1; ++i)
put_byte(bs, bignum_byte(ec->publicKey.y, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, ((bignum_byte(ec->publicKey.y, i) & 0x7f) |
(bignum_bit(ec->publicKey.x, 0) << 7)));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
} else if (ec->publicKey.curve->type == EC_WEIERSTRASS) {
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
assert(ec->publicKey.curve->name);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
pointlen = (bignum_bitcount(ec->publicKey.curve->p) + 7) / 8;
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
put_stringz(bs, ec->sshk->ssh_id);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_stringz(bs, ec->publicKey.curve->name);
put_uint32(bs, (2 * pointlen) + 1);
put_byte(bs, 0x04);
for (i = pointlen; i--;)
put_byte(bs, bignum_byte(ec->publicKey.x, i));
for (i = pointlen; i--;)
put_byte(bs, bignum_byte(ec->publicKey.y, i));
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else {
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(0 && "Bad key type in ecdsa_public_blob");
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
static void ecdsa_private_blob(ssh_key *key, BinarySink *bs)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
int keylen;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
int i;
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(ec->privateKey);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (ec->publicKey.curve->type == EC_EDWARDS) {
/* Unsigned */
keylen = (bignum_bitcount(ec->privateKey) + 7) / 8;
} else {
/* Signed */
keylen = (bignum_bitcount(ec->privateKey) + 8) / 8;
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_uint32(bs, keylen);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (ec->publicKey.curve->type == EC_EDWARDS) {
/* Little endian */
for (i = 0; i < keylen; ++i)
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, bignum_byte(ec->privateKey, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
} else {
for (i = keylen; i--;)
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, bignum_byte(ec->privateKey, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static ssh_key *ecdsa_new_priv(const ssh_keyalg *self, ptrlen pub, ptrlen priv)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource src[1];
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
ssh_key *sshk;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
struct ec_key *ec;
struct ec_point *publicKey;
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
sshk = ecdsa_new_pub(self, pub);
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
if (!sshk)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return NULL;
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
ec = container_of(sshk, struct ec_key, sshk);
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource_BARE_INIT(src, priv.ptr, priv.len);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (ec->publicKey.curve->type != EC_WEIERSTRASS
&& ec->publicKey.curve->type != EC_EDWARDS) {
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
return NULL;
}
if (ec->publicKey.curve->type == EC_EDWARDS) {
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ec->privateKey = get_mp_le(src);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
} else {
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ec->privateKey = get_mp_ssh2(src);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
if (!ec->privateKey) {
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
return NULL;
}
/* Check that private key generates public key */
publicKey = ec_public(ec->privateKey, ec->publicKey.curve);
if (!publicKey ||
bignum_cmp(publicKey->x, ec->publicKey.x) ||
bignum_cmp(publicKey->y, ec->publicKey.y))
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec = NULL;
}
ec_point_free(publicKey);
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
return &ec->sshk;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static ssh_key *ed25519_new_priv_openssh(const ssh_keyalg *self,
BinarySource *src)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
struct ec_key *ec;
struct ec_point *publicKey;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ptrlen p, q;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
p = get_string(src);
q = get_string(src);
if (get_err(src) || p.len != 32 || q.len != 64)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
return NULL;
ec = snew(struct ec_key);
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ec->sshk = self;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec->publicKey.curve = ec_ed25519();
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ec->publicKey.infinity = false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec->privateKey = NULL;
ec->publicKey.x = NULL;
ec->publicKey.z = NULL;
ec->publicKey.y = NULL;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (!decodepoint_ed(p.ptr, p.len, &ec->publicKey))
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
return NULL;
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ec->privateKey = bignum_from_bytes_le(q.ptr, 32);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
/* Check that private key generates public key */
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
publicKey = ec_public(ec->privateKey, ec->publicKey.curve);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!publicKey ||
bignum_cmp(publicKey->x, ec->publicKey.x) ||
bignum_cmp(publicKey->y, ec->publicKey.y))
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec = NULL;
}
ec_point_free(publicKey);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* The OpenSSH format for ed25519 private keys also for some
* reason encodes an extra copy of the public key in the second
* half of the secret-key string. Check that that's present and
* correct as well, otherwise the key we think we've imported
* won't behave identically to the way OpenSSH would have treated
* it. */
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (0 != memcmp((const char *)q.ptr + 32, p.ptr, 32)) {
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
return NULL;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
return &ec->sshk;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static void ed25519_openssh_blob(ssh_key *key, BinarySink *bs)
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
strbuf *pub;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
int pointlen;
int keylen;
int i;
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(ec->publicKey.curve->type == EC_EDWARDS);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
pointlen = (bignum_bitcount(ec->publicKey.y) + 7) / 8;
keylen = (bignum_bitcount(ec->privateKey) + 7) / 8;
/* Encode the public point */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
pub = strbuf_new();
put_uint32(pub, pointlen);
for (i = 0; i < pointlen - 1; ++i)
put_byte(pub, bignum_byte(ec->publicKey.y, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(pub, ((bignum_byte(ec->publicKey.y, i) & 0x7f) |
(bignum_bit(ec->publicKey.x, 0) << 7)));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_data(bs, pub->s, pub->len);
put_uint32(bs, keylen + pointlen);
for (i = 0; i < keylen; ++i)
put_byte(bs, bignum_byte(ec->privateKey, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Now encode an extra copy of the public point as the second half
* of the private key string, as the OpenSSH format for some
* reason requires */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_data(bs, pub->s + 4, pub->len - 4);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
strbuf_free(pub);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static ssh_key *ecdsa_new_priv_openssh(const ssh_keyalg *self,
BinarySource *src)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct ecsign_extra *extra =
(const struct ecsign_extra *)self->extra;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
struct ec_key *ec;
struct ec_curve *curve;
struct ec_point *publicKey;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
get_string(src);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
curve = extra->curve();
assert(curve->type == EC_WEIERSTRASS);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec = snew(struct ec_key);
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ec->sshk = self;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec->publicKey.curve = curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ec->publicKey.infinity = false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec->publicKey.x = NULL;
ec->publicKey.y = NULL;
ec->publicKey.z = NULL;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (!get_point(src, &ec->publicKey)) {
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return NULL;
}
ec->privateKey = NULL;
if (!ec->publicKey.x || !ec->publicKey.y ||
bignum_cmp(ec->publicKey.x, curve->p) >= 0 ||
bignum_cmp(ec->publicKey.y, curve->p) >= 0)
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return NULL;
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
ec->privateKey = get_mp_ssh2(src);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (ec->privateKey == NULL)
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return NULL;
}
/* Now check that the private key makes the public key */
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
publicKey = ec_public(ec->privateKey, ec->publicKey.curve);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!publicKey)
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return NULL;
}
if (bignum_cmp(ec->publicKey.x, publicKey->x) ||
bignum_cmp(ec->publicKey.y, publicKey->y))
{
/* Private key doesn't make the public key on the given curve */
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec_point_free(publicKey);
Fixes to memory management in the elliptic curve code. There was an error-handling path testing the wrong variable; an inappropriate call to ec_point_free in decodepoint() (in fact, that function always gets passed a pointer to an ec_point structure that's not a dynamically allocated block at all or not in its own right, so we should have just cleared its contents without freeing the structure itself); a missing return on an error path which would have caused the same structure to be freed a second time; and two missing freebn in ecdsa_sign. Patch due to Tim Kosse.
2014-12-20 17:07:17 +00:00
return NULL;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
ec_point_free(publicKey);
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
return &ec->sshk;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
static void ecdsa_openssh_blob(ssh_key *key, BinarySink *bs)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
int pointlen;
int i;
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(ec->publicKey.curve->type == EC_WEIERSTRASS);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
pointlen = (bignum_bitcount(ec->publicKey.curve->p) + 7) / 8;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_stringz(bs, ec->publicKey.curve->name);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_uint32(bs, 1 + (pointlen * 2));
put_byte(bs, 0x04);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
for (i = pointlen; i--; )
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, bignum_byte(ec->publicKey.x, i));
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
for (i = pointlen; i--; )
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, bignum_byte(ec->publicKey.y, i));
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_mp_ssh2(bs, ec->privateKey);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
static int ecdsa_pubkey_bits(const ssh_keyalg *self, ptrlen blob)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
ssh_key *sshk;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
struct ec_key *ec;
int ret;
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
sshk = ecdsa_new_pub(self, blob);
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
if (!sshk)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return -1;
Move null pointer checks to before FROMFIELD. This fixes an oversight in commit 0fc2d3b45: if a key creation function returns a null 'ssh_key *', then adjusting the pointer's address using FROMFIELD is a mistake, both in technical C terms (undefined behaviour) and practically speaking because it will foil the subsequent check against NULL. Instead, if we're going to check a pointer against NULL, we must do it _before_ applying this kind of address-adjusting type conversion.
2018-05-31 17:32:09 +00:00
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
ec = container_of(sshk, struct ec_key, sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ret = ec->publicKey.curve->fieldBits;
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&ec->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return ret;
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
static bool ecdsa_verify(ssh_key *key, ptrlen sig, ptrlen data)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
const struct ecsign_extra *extra =
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
(const struct ecsign_extra *)ec->sshk->extra;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource src[1];
ptrlen sigstr;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool ret;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
if (!ec->publicKey.x || !ec->publicKey.y || !ec->publicKey.curve)
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource_BARE_INIT(src, sig.ptr, sig.len);
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* Check the signature starts with the algorithm name */
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
if (!ptrlen_eq_string(get_string(src), ec->sshk->ssh_id))
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
sigstr = get_string(src);
if (get_err(src))
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (ec->publicKey.curve->type == EC_EDWARDS) {
struct ec_point *r;
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
int pointlen = ec->publicKey.curve->fieldBits / 8;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Bignum s, h;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Check that the signature is two times the length of a point */
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (sigstr.len != pointlen * 2) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Check it's the 256 bit field so that SHA512 is the correct hash */
if (ec->publicKey.curve->fieldBits != 256) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
/* Get the signature */
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
r = ec_point_new(ec->publicKey.curve, NULL, NULL, NULL, false);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (!r) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
if (!decodepoint(sigstr.ptr, pointlen, r)) {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec_point_free(r);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
s = bignum_from_bytes_le(
(const char *)sigstr.ptr + pointlen, pointlen);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Get the hash of the encoded value of R + encoded value of pk + message */
{
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
int i;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
unsigned char digest[512 / 8];
SHA512_State hs;
SHA512_Init(&hs);
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
/* Add encoded r (no need to encode it again, it was in
* the signature) */
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
put_data(&hs, sigstr.ptr, pointlen);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Encode pk and add it */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
for (i = 0; i < pointlen - 1; ++i)
put_byte(&hs, bignum_byte(ec->publicKey.y, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
put_byte(&hs, ((bignum_byte(ec->publicKey.y, i) & 0x7f) |
(bignum_bit(ec->publicKey.x, 0) << 7)));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Add the message itself */
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
put_data(&hs, data.ptr, data.len);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Get the hash */
SHA512_Final(&hs, digest);
/* Convert to Bignum */
h = bignum_from_bytes_le(digest, sizeof(digest));
}
/* Verify sB == r + h*publicKey */
{
struct ec_point *lhs, *rhs, *tmp;
/* lhs = sB */
lhs = ecp_mul(&ec->publicKey.curve->e.B, s);
freebn(s);
if (!lhs) {
ec_point_free(r);
freebn(h);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* rhs = r + h*publicKey */
tmp = ecp_mul(&ec->publicKey, h);
freebn(h);
if (!tmp) {
ec_point_free(lhs);
ec_point_free(r);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
rhs = ecp_add(r, tmp, false);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec_point_free(r);
ec_point_free(tmp);
if (!rhs) {
ec_point_free(lhs);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
/* Check the point is the same */
ret = !bignum_cmp(lhs->x, rhs->x);
if (ret) {
ret = !bignum_cmp(lhs->y, rhs->y);
if (ret) {
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
ret = true;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
}
ec_point_free(lhs);
ec_point_free(rhs);
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
} else {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Bignum r, s;
unsigned char digest[512 / 8];
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
int digestLen;
Turn SSH hashes into a classoid. The new version of ssh_hash has the same nice property as ssh2_mac, that I can make the generic interface object type function directly as a BinarySink so that clients don't have to call h->sink() and worry about the separate sink object they get back from that.
2018-09-13 15:41:46 +00:00
ssh_hash *hashctx;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
BinarySource_BARE_INIT(src, sigstr.ptr, sigstr.len);
r = get_mp_ssh2(src);
s = get_mp_ssh2(src);
if (get_err(src)) {
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
freebn(r);
Clean up ssh_keyalg APIs and implementations. Quite a few of the function pointers in the ssh_keyalg vtable now take ptrlen arguments in place of separate pointer and length pairs. Meanwhile, the various key types' implementations of those functions now work by initialising a BinarySource with the input ptrlen and using the new decode functions to walk along it. One exception is the openssh_createkey method which reads a private key in the wire format used by OpenSSH's SSH-2 agent protocol, which has to consume a prefix of a larger data stream, and tell the caller how much of that data was the private key. That function now takes an actual BinarySource, and passes that directly to the decode functions, so that on return the caller finds that the BinarySource's read pointer has been advanced exactly past the private key. This let me throw away _several_ reimplementations of mpint-reading functions, one in each of sshrsa, sshdss.c and sshecc.c. Worse still, they didn't all have exactly the SSH-2 semantics, because the thing in sshrsa.c whose name suggested it was an mpint-reading function actually tolerated the wrong number of leading zero bytes, which it had to be able to do to cope with the "ssh-rsa" signature format which contains a thing that isn't quite an SSH-2 mpint. Now that deviation is clearly commented!
2018-05-31 17:40:51 +00:00
freebn(s);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
return false;
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
digestLen = extra->hash->hlen;
assert(digestLen <= sizeof(digest));
Turn SSH hashes into a classoid. The new version of ssh_hash has the same nice property as ssh2_mac, that I can make the generic interface object type function directly as a BinarySink so that clients don't have to call h->sink() and worry about the separate sink object they get back from that.
2018-09-13 15:41:46 +00:00
hashctx = ssh_hash_new(extra->hash);
put_data(hashctx, data.ptr, data.len);
ssh_hash_final(hashctx, digest);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Verify the signature */
ret = _ecdsa_verify(&ec->publicKey, digest, digestLen, r, s);
freebn(r);
freebn(s);
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
return ret;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
static void ecdsa_sign(ssh_key *key, const void *data, int datalen,
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
BinarySink *bs)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Rename FROMFIELD to 'container_of'. Ian Jackson points out that the Linux kernel has a macro of this name with the same purpose, and suggests that it's a good idea to use the same name as they do, so that at least some people reading one code base might recognise it from the other. I never really thought very hard about what order FROMFIELD's parameters should go in, and therefore I'm pleasantly surprised to find that my order agrees with the kernel's, so I don't have to permute every call site as part of making this change :-)
2018-10-05 22:49:08 +00:00
struct ec_key *ec = container_of(key, struct ec_key, sshk);
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
const struct ecsign_extra *extra =
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
(const struct ecsign_extra *)ec->sshk->extra;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
unsigned char digest[512 / 8];
int digestLen;
Bignum r = NULL, s = NULL;
int i;
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(ec->privateKey);
assert(ec->publicKey.curve);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
if (ec->publicKey.curve->type == EC_EDWARDS) {
struct ec_point *rp;
int pointlen = ec->publicKey.curve->fieldBits / 8;
/* hash = H(sk) (where hash creates 2 * fieldBits)
* b = fieldBits
* a = 2^(b-2) + SUM(2^i * h_i) for i = 2 -> b-2
* r = H(h[b/8:b/4] + m)
* R = rB
* S = (r + H(encodepoint(R) + encodepoint(pk) + m) * a) % l */
{
unsigned char hash[512/8];
Bignum a;
SHA512_State hs;
SHA512_Init(&hs);
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
for (i = 0; i < pointlen; ++i)
put_byte(&hs, bignum_byte(ec->privateKey, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
SHA512_Final(&hs, hash);
Clear an extra low bit in EdDSA exponent calculation. The source paper, and OpenSSH, agree that the lowest bit index used from the hash of the private key is bit 3, i.e. bits 0,1,2 at the bottom are all zero. We were only clearing bits 0 and 1, which would have worked for about half of keys. I must have got lucky during testing!
2015-05-10 13:04:16 +00:00
/* The second part is simply turning the hash into a
* Bignum, however the 2^(b-2) bit *must* be set, and the
* bottom 3 bits *must* not be */
hash[0] &= 0xf8; /* Unset bottom 3 bits (if set) */
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
hash[31] &= 0x7f; /* Unset above (b-2) */
hash[31] |= 0x40; /* Set 2^(b-2) */
/* Chop off the top part and convert to int */
a = bignum_from_bytes_le(hash, 32);
SHA512_Init(&hs);
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
put_data(&hs, hash+(ec->publicKey.curve->fieldBits / 8),
((ec->publicKey.curve->fieldBits / 4) -
(ec->publicKey.curve->fieldBits / 8)));
put_data(&hs, data, datalen);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
SHA512_Final(&hs, hash);
r = bignum_from_bytes_le(hash, 512/8);
rp = ecp_mul(&ec->publicKey.curve->e.B, r);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(rp);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Now calculate s */
SHA512_Init(&hs);
/* Encode the point R */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
for (i = 0; i < pointlen - 1; ++i)
put_byte(&hs, bignum_byte(rp->y, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
put_byte(&hs, ((bignum_byte(rp->y, i) & 0x7f) |
(bignum_bit(rp->x, 0) << 7)));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Encode the point pk */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
for (i = 0; i < pointlen - 1; ++i)
put_byte(&hs, bignum_byte(ec->publicKey.y, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
put_byte(&hs, ((bignum_byte(ec->publicKey.y, i) & 0x7f) |
(bignum_bit(ec->publicKey.x, 0) << 7)));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Add the message */
Replace all uses of SHA*_Bytes / MD5Update. In fact, those functions don't even exist any more. The only way to get data into a primitive hash state is via the new put_* system. Of course, that means put_data() is a viable replacement for every previous call to one of the per-hash update functions - but just mechanically doing that would have missed the opportunity to simplify a lot of the call sites.
2018-05-24 09:03:36 +00:00
put_data(&hs, data, datalen);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
SHA512_Final(&hs, hash);
{
Bignum tmp, tmp2;
tmp = bignum_from_bytes_le(hash, 512/8);
tmp2 = modmul(tmp, a, ec->publicKey.curve->e.l);
freebn(a);
freebn(tmp);
tmp = bigadd(r, tmp2);
freebn(r);
freebn(tmp2);
s = bigmod(tmp, ec->publicKey.curve->e.l);
freebn(tmp);
}
}
/* Format the output */
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
put_stringz(bs, ec->sshk->ssh_id);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
pointlen = ec->publicKey.curve->fieldBits / 8;
put_uint32(bs, pointlen * 2);
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Encode the point */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
for (i = 0; i < pointlen - 1; ++i)
put_byte(bs, bignum_byte(rp->y, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Unset last bit of y and set first bit of x in its place */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, ((bignum_byte(rp->y, i) & 0x7f) |
(bignum_bit(rp->x, 0) << 7)));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ec_point_free(rp);
/* Encode the int */
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
for (i = 0; i < pointlen; ++i)
put_byte(bs, bignum_byte(s, i));
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
freebn(s);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
} else {
Turn SSH hashes into a classoid. The new version of ssh_hash has the same nice property as ssh2_mac, that I can make the generic interface object type function directly as a BinarySink so that clients don't have to call h->sink() and worry about the separate sink object they get back from that.
2018-09-13 15:41:46 +00:00
ssh_hash *hashctx;
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
strbuf *substr;
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
digestLen = extra->hash->hlen;
assert(digestLen <= sizeof(digest));
Turn SSH hashes into a classoid. The new version of ssh_hash has the same nice property as ssh2_mac, that I can make the generic interface object type function directly as a BinarySink so that clients don't have to call h->sink() and worry about the separate sink object they get back from that.
2018-09-13 15:41:46 +00:00
hashctx = ssh_hash_new(extra->hash);
put_data(hashctx, data, datalen);
ssh_hash_final(hashctx, digest);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Do the signature */
_ecdsa_sign(ec->privateKey, ec->publicKey.curve, digest, digestLen, &r, &s);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
assert(r);
assert(s);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
/* Format the output */
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
put_stringz(bs, ec->sshk->ssh_id);
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
substr = strbuf_new();
put_mp_ssh2(substr, r);
put_mp_ssh2(substr, s);
put_stringsb(bs, substr);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
freebn(r);
freebn(s);
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct ecsign_extra sign_extra_ed25519 = {
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
ec_ed25519, NULL,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
NULL, 0,
};
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const ssh_keyalg ssh_ecdsa_ed25519 = {
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_new_pub,
ecdsa_new_priv,
ed25519_new_priv_openssh,
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ecdsa_freekey,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_sign,
ecdsa_verify,
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ecdsa_public_blob,
ecdsa_private_blob,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ed25519_openssh_blob,
ecdsa_cache_str,
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
ecdsa_pubkey_bits,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
"ssh-ed25519",
"ssh-ed25519",
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
&sign_extra_ed25519,
Support public keys using the "ssh-ed25519" method. This introduces a third system of elliptic curve representation and arithmetic, namely Edwards form.
2015-05-09 14:02:54 +00:00
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* OID: 1.2.840.10045.3.1.7 (ansiX9p256r1) */
static const unsigned char nistp256_oid[] = {
0x2a, 0x86, 0x48, 0xce, 0x3d, 0x03, 0x01, 0x07
};
const struct ecsign_extra sign_extra_nistp256 = {
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
ec_p256, &ssh_sha256,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
nistp256_oid, lenof(nistp256_oid),
};
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const ssh_keyalg ssh_ecdsa_nistp256 = {
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_new_pub,
ecdsa_new_priv,
ecdsa_new_priv_openssh,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_freekey,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_sign,
ecdsa_verify,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_public_blob,
ecdsa_private_blob,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_openssh_blob,
ecdsa_cache_str,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_pubkey_bits,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
"ecdsa-sha2-nistp256",
"ecdsa-sha2-nistp256",
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
&sign_extra_nistp256,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* OID: 1.3.132.0.34 (secp384r1) */
static const unsigned char nistp384_oid[] = {
0x2b, 0x81, 0x04, 0x00, 0x22
};
const struct ecsign_extra sign_extra_nistp384 = {
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
ec_p384, &ssh_sha384,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
nistp384_oid, lenof(nistp384_oid),
};
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const ssh_keyalg ssh_ecdsa_nistp384 = {
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_new_pub,
ecdsa_new_priv,
ecdsa_new_priv_openssh,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_freekey,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_sign,
ecdsa_verify,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_public_blob,
ecdsa_private_blob,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_openssh_blob,
ecdsa_cache_str,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_pubkey_bits,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
"ecdsa-sha2-nistp384",
"ecdsa-sha2-nistp384",
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
&sign_extra_nistp384,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* OID: 1.3.132.0.35 (secp521r1) */
static const unsigned char nistp521_oid[] = {
0x2b, 0x81, 0x04, 0x00, 0x23
};
const struct ecsign_extra sign_extra_nistp521 = {
Clean up hash selection in ECDSA. Removed another set of ad-hoc tests of the key size to decide which hash to use for the signature system, and replaced them with a straightforward pointer to an ssh_hash structure in the 'extra' area.
2015-05-15 09:13:06 +00:00
ec_p521, &ssh_sha512,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
nistp521_oid, lenof(nistp521_oid),
};
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const ssh_keyalg ssh_ecdsa_nistp521 = {
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_new_pub,
ecdsa_new_priv,
ecdsa_new_priv_openssh,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_freekey,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_sign,
ecdsa_verify,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_public_blob,
ecdsa_private_blob,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
ecdsa_openssh_blob,
ecdsa_cache_str,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ecdsa_pubkey_bits,
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
"ecdsa-sha2-nistp521",
"ecdsa-sha2-nistp521",
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
&sign_extra_nistp521,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
/* ----------------------------------------------------------------------
* Exposed ECDH interface
*/
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
struct eckex_extra {
struct ec_curve *(*curve)(void);
};
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
static Bignum ecdh_calculate(const Bignum private,
const struct ec_point *public)
{
struct ec_point *p;
Bignum ret;
p = ecp_mul(public, private);
if (!p) return NULL;
ret = p->x;
p->x = NULL;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (p->curve->type == EC_MONTGOMERY) {
Fix construction of the output bignum in Curve25519 kex. We were doing an endianness flip on the output elliptic-curve point. Endianness flips of bignums, of course, have to specify how many bytes they're imagining the value to have (that's how you decide whether to convert 0xA1A2 into 0xA2A1 or 0xA2A10000 or 0xA2A1000000000000 etc), and we had chosen our byte count based on the highest set bit in the _output value_ - but in fact we should have chosen it based on the size of the curve's modulus, leading to a failure about 1/256 of the time when the MSB happened to come out zero so the two byte counts differed. (Also added a missing smemclr, while I was there.)
2015-05-19 08:54:17 +00:00
/*
* Endianness-swap. The Curve25519 algorithm definition
* assumes you were doing your computation in arrays of 32
* little-endian bytes, and now specifies that you take your
* final one of those and convert it into a bignum in
* _network_ byte order, i.e. big-endian.
*
* In particular, the spec says, you convert the _whole_ 32
* bytes into a bignum. That is, on the rare occasions that
* p->x has come out with the most significant 8 bits zero, we
* have to imagine that being represented by a 32-byte string
* with the last byte being zero, so that has to be converted
* into an SSH-2 bignum with the _low_ byte zero, i.e. a
* multiple of 256.
*/
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
int i;
Fix construction of the output bignum in Curve25519 kex. We were doing an endianness flip on the output elliptic-curve point. Endianness flips of bignums, of course, have to specify how many bytes they're imagining the value to have (that's how you decide whether to convert 0xA1A2 into 0xA2A1 or 0xA2A10000 or 0xA2A1000000000000 etc), and we had chosen our byte count based on the highest set bit in the _output value_ - but in fact we should have chosen it based on the size of the curve's modulus, leading to a failure about 1/256 of the time when the MSB happened to come out zero so the two byte counts differed. (Also added a missing smemclr, while I was there.)
2015-05-19 08:54:17 +00:00
int bytes = (p->curve->fieldBits+7) / 8;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
unsigned char *byteorder = snewn(bytes, unsigned char);
for (i = 0; i < bytes; ++i) {
byteorder[i] = bignum_byte(ret, i);
}
freebn(ret);
ret = bignum_from_bytes(byteorder, bytes);
Fix construction of the output bignum in Curve25519 kex. We were doing an endianness flip on the output elliptic-curve point. Endianness flips of bignums, of course, have to specify how many bytes they're imagining the value to have (that's how you decide whether to convert 0xA1A2 into 0xA2A1 or 0xA2A10000 or 0xA2A1000000000000 etc), and we had chosen our byte count based on the highest set bit in the _output value_ - but in fact we should have chosen it based on the size of the curve's modulus, leading to a failure about 1/256 of the time when the MSB happened to come out zero so the two byte counts differed. (Also added a missing smemclr, while I was there.)
2015-05-19 08:54:17 +00:00
smemclr(byteorder, bytes);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
sfree(byteorder);
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
ec_point_free(p);
return ret;
}
Log which elliptic curve we're using for ECDH kex. It seems like quite an important thing to mention in the event log! Suppose there's a bug affecting only one curve, for example? Fixed- group Diffie-Hellman has always logged the group, but the ECDH log message just told you the hash and not also the curve. To implement this, I've added a 'textname' field to all elliptic curves, whether they're used for kex or signing or both, suitable for use in this log message and any others we might find a need for in future.
2015-05-19 07:42:23 +00:00
const char *ssh_ecdhkex_curve_textname(const struct ssh_kex *kex)
{
const struct eckex_extra *extra = (const struct eckex_extra *)kex->extra;
struct ec_curve *curve = extra->curve();
return curve->textname;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
struct ec_key *ssh_ecdhkex_newkey(const struct ssh_kex *kex)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct eckex_extra *extra = (const struct eckex_extra *)kex->extra;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
struct ec_curve *curve;
struct ec_key *key;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
struct ec_point *publicKey;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
curve = extra->curve();
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
key = snew(struct ec_key);
Reorganise ssh_keyalg and use it as a vtable. After Pavel Kryukov pointed out that I have to put _something_ in the 'ssh_key' structure, I thought of an actually useful thing to put there: why not make it store a pointer to the ssh_keyalg structure? Then ssh_key becomes a classoid - or perhaps 'traitoid' is a closer analogy - in the same style as Socket and Plug. And just like Socket and Plug, I've also arranged a system of wrapper macros that avoid the need to mention the 'object' whose method you're invoking twice at each call site. The new vtable pointer directly replaces an existing field of struct ec_key (which was usable by several different ssh_keyalgs, so it already had to store a pointer to the currently active one), and also replaces the 'alg' field of the ssh2_userkey structure that wraps up a cryptographic key with its comment field. I've also taken the opportunity to clean things up a bit in general: most of the methods now have new and clearer names (e.g. you'd never know that 'newkey' made a public-only key while 'createkey' made a public+private key pair unless you went and looked it up, but now they're called 'new_pub' and 'new_priv' you might be in with a chance), and I've completely removed the openssh_private_npieces field after realising that it was duplicating information that is actually _more_ conveniently obtained by calling the new_priv_openssh method (formerly openssh_createkey) and throwing away the result.
2018-06-03 11:58:05 +00:00
key->sshk = NULL;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
key->publicKey.curve = curve;
if (curve->type == EC_MONTGOMERY) {
unsigned char bytes[32] = {0};
int i;
for (i = 0; i < sizeof(bytes); ++i)
{
bytes[i] = (unsigned char)random_byte();
}
bytes[0] &= 248;
bytes[31] &= 127;
bytes[31] |= 64;
Swap endianness of the Curve25519 ECDH private key. DJB's spec at http://cr.yp.to/ecdh/curve25519-20060209.pdf is clear that we should be clearing the low 3 bits of the _LSB_ of the private key bit string, and setting bit 6 and clearing bit 7 of the _MSB_. We were doing the opposite, due to feeding the resulting bit string to bignum_from_bytes() rather than bignum_from_bytes_le(). This didn't cause an interoperability issue, because the two DH exponentiations still commute, but it goes against the Curve25519 spec, in particular the care taken to fix the position of the leading exponent bit. The code is now consistent with the test vectors in RFC 7748 section 6.1: if you modify the EC_MONTGOMERY branch of ssh_ecdhkex_newkey() to replace the loop on random_byte() with a memcpy that fills bytes[] with 77076d0a7318a57d3c16c17251b26645df4c2f87ebc0992ab177fba51db92c2a and then print out the resulting publicKey->x, you find that it's (byte-reversed) the expected output value given in that RFC section, 8520f0098930a754748b7ddcb43ef75a0dbf3a0d26381af4eba4a98eaa9b4e6a.
2016-05-03 13:44:00 +00:00
key->privateKey = bignum_from_bytes_le(bytes, sizeof(bytes));
Replace an ad-hoc buffer-clearing loop with smemclr. Thanks to @ch3root on Twitter for spotting it, and thanks to Chris Emerson for bothering to let me know. I must have missed this when I code-reviewed the ECC contribution.
2016-01-25 19:24:41 +00:00
smemclr(bytes, sizeof(bytes));
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (!key->privateKey) {
sfree(key);
return NULL;
}
publicKey = ecp_mul(&key->publicKey.curve->m.G, key->privateKey);
if (!publicKey) {
freebn(key->privateKey);
sfree(key);
return NULL;
}
key->publicKey.x = publicKey->x;
key->publicKey.y = publicKey->y;
key->publicKey.z = NULL;
sfree(publicKey);
} else {
key->privateKey = bignum_random_in_range(One, key->publicKey.curve->w.n);
if (!key->privateKey) {
sfree(key);
return NULL;
}
publicKey = ecp_mul(&key->publicKey.curve->w.G, key->privateKey);
if (!publicKey) {
freebn(key->privateKey);
sfree(key);
return NULL;
}
key->publicKey.x = publicKey->x;
key->publicKey.y = publicKey->y;
key->publicKey.z = NULL;
sfree(publicKey);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
return key;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
void ssh_ecdhkex_getpublic(struct ec_key *ec, BinarySink *bs)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
int i;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
int pointlen;
pointlen = (bignum_bitcount(ec->publicKey.curve->p) + 7) / 8;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (ec->publicKey.curve->type == EC_WEIERSTRASS) {
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
put_byte(bs, 0x04);
for (i = pointlen; i--;)
put_byte(bs, bignum_byte(ec->publicKey.x, i));
for (i = pointlen; i--;)
put_byte(bs, bignum_byte(ec->publicKey.y, i));
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
} else {
Change ssh.h crypto APIs to output to BinarySink. This affects all the functions that generate public and private key and signature blobs of all kinds, plus ssh_ecdhkex_getpublic. Instead of returning a bare block of memory and taking an extra 'int *length' parameter, all these functions now write to a BinarySink, and it's the caller's job to have prepared an appropriate one where they want the output to go (usually a strbuf). The main value of this change is that those blob-generation functions were chock full of ad-hoc length-counting and data marshalling. You have only to look at rsa2_{public,private}_blob, for example, to see the kind of thing I was keen to get rid of!
2018-05-24 09:59:39 +00:00
for (i = 0; i < pointlen; ++i)
put_byte(bs, bignum_byte(ec->publicKey.x, i));
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
}
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Fix some missing void * and const in existing APIs. Several changes here that should have been in commit 7babe66a8 but I missed them.
2018-06-02 06:52:26 +00:00
Bignum ssh_ecdhkex_getkey(struct ec_key *ec,
const void *remoteKey, int remoteKeyLen)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
struct ec_point remote;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
Bignum ret;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (ec->publicKey.curve->type == EC_WEIERSTRASS) {
remote.curve = ec->publicKey.curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
remote.infinity = false;
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
if (!decodepoint(remoteKey, remoteKeyLen, &remote)) {
return NULL;
}
} else {
/* Point length has to be the same length */
if (remoteKeyLen != (bignum_bitcount(ec->publicKey.curve->p) + 7) / 8) {
return NULL;
}
remote.curve = ec->publicKey.curve;
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
remote.infinity = false;
Fix some missing void * and const in existing APIs. Several changes here that should have been in commit 7babe66a8 but I missed them.
2018-06-02 06:52:26 +00:00
remote.x = bignum_from_bytes_le((const unsigned char *)remoteKey,
remoteKeyLen);
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
remote.y = NULL;
remote.z = NULL;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
ret = ecdh_calculate(ec->privateKey, &remote);
if (remote.x) freebn(remote.x);
if (remote.y) freebn(remote.y);
return ret;
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
void ssh_ecdhkex_freekey(struct ec_key *key)
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
ecdsa_freekey(&key->sshk);
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
}
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
static const struct eckex_extra kex_extra_curve25519 = { ec_curve25519 };
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
static const struct ssh_kex ssh_ec_kex_curve25519 = {
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
"curve25519-sha256@libssh.org", NULL, KEXTYPE_ECDH,
&ssh_sha256, &kex_extra_curve25519,
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct eckex_extra kex_extra_nistp256 = { ec_p256 };
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
static const struct ssh_kex ssh_ec_kex_nistp256 = {
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
"ecdh-sha2-nistp256", NULL, KEXTYPE_ECDH,
&ssh_sha256, &kex_extra_nistp256,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct eckex_extra kex_extra_nistp384 = { ec_p384 };
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
static const struct ssh_kex ssh_ec_kex_nistp384 = {
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
"ecdh-sha2-nistp384", NULL, KEXTYPE_ECDH,
&ssh_sha384, &kex_extra_nistp384,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct eckex_extra kex_extra_nistp521 = { ec_p521 };
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
static const struct ssh_kex ssh_ec_kex_nistp521 = {
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
"ecdh-sha2-nistp521", NULL, KEXTYPE_ECDH,
&ssh_sha512, &kex_extra_nistp521,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
static const struct ssh_kex *const ec_kex_list[] = {
Support ECDH key exchange using the 'curve25519' curve. This is the kex protocol id "curve25519-sha256@libssh.org", so called because it's over the prime field of order 2^255 - 19. Arithmetic in this curve is done using the Montgomery representation, rather than the Weierstrass representation. So 'struct ec_curve' has grown a discriminant field and a union of subtypes.
2015-05-09 14:02:52 +00:00
&ssh_ec_kex_curve25519,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
&ssh_ec_kex_nistp256,
&ssh_ec_kex_nistp384,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
&ssh_ec_kex_nistp521,
Elliptic-curve cryptography support. This provides support for ECDSA public keys, for both hosts and users, and also ECDH key exchange. Supported curves are currently just the three NIST curves required by RFC 5656.
2014-11-01 09:45:20 +00:00
};
const struct ssh_kexes ssh_ecdh_kex = {
sizeof(ec_kex_list) / sizeof(*ec_kex_list),
ec_kex_list
};
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
/* ----------------------------------------------------------------------
* Helper functions for finding key algorithms and returning auxiliary
* data.
*/
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const ssh_keyalg *ec_alg_by_oid(int len, const void *oid,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct ec_curve **curve)
{
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
static const ssh_keyalg *algs_with_oid[] = {
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
&ssh_ecdsa_nistp256,
&ssh_ecdsa_nistp384,
&ssh_ecdsa_nistp521,
};
int i;
for (i = 0; i < lenof(algs_with_oid); i++) {
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const ssh_keyalg *alg = algs_with_oid[i];
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
const struct ecsign_extra *extra =
(const struct ecsign_extra *)alg->extra;
if (len == extra->oidlen && !memcmp(oid, extra->oid, len)) {
*curve = extra->curve();
return alg;
}
}
return NULL;
}
Invent a struct type for polymorphic SSH key data. During last week's work, I made a mistake in which I got the arguments backwards in one of the key-blob-generating functions - mistakenly swapped the 'void *' key instance with the 'BinarySink *' output destination - and I didn't spot the mistake until run time, because in C you can implicitly convert both to and from void * and so there was no compile-time failure of type checking. Now that I've introduced the FROMFIELD macro that downcasts a pointer to one field of a structure to retrieve a pointer to the whole structure, I think I might start using that more widely to indicate this kind of polymorphic subtyping. So now all the public-key functions in the struct ssh_signkey vtable handle their data instance in the form of a pointer to a subfield of a new zero-sized structure type 'ssh_key', which outside the key implementations indicates 'this is some kind of key instance but it could be of any type'; they downcast that pointer internally using FROMFIELD in place of the previous ordinary C cast, and return one by returning &foo->sshk for whatever foo they've just made up. The sshk member is not at the beginning of the structure, which means all those FROMFIELDs and &key->sshk are actually adding and subtracting an offset. Of course I could have put the member at the start anyway, but I had the idea that it's actually a feature _not_ to have the two types start at the same address, because it means you should notice earlier rather than later if you absentmindedly cast from one to the other directly rather than by the approved method (in particular, if you accidentally assign one through a void * and back without even _noticing_ you perpetrated a cast). In particular, this enforces that you can't sfree() the thing even once without realising you should instead of called the right freekey function. (I found several bugs by this method during initial testing, so I think it's already proved its worth!) While I'm here, I've also renamed the vtable structure ssh_signkey to ssh_keyalg, because it was a confusing name anyway - it describes the _algorithm_ for handling all keys of that type, not a specific key. So ssh_keyalg is the collection of code, and ssh_key is one instance of the data it handles.
2018-05-27 07:32:21 +00:00
const unsigned char *ec_alg_oid(const ssh_keyalg *alg,
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
int *oidlen)
{
const struct ecsign_extra *extra = (const struct ecsign_extra *)alg->extra;
*oidlen = extra->oidlen;
return extra->oid;
}
Polish up the PuTTYgen user interface for ECC key types. Jacob pointed out that a free-text field for entering a key size in bits is all very well for key types where we actually _can_ generate a key to a size of your choice, but less useful for key types where there are only three (or one) legal values for the field, especially if we don't _say_ what they are. So I've revamped the UI a bit: now, in ECDSA mode, you get a dropdown list selector showing the available elliptic curves (and they're even named, rather than just given by bit count), and in ED25519 mode even that disappears. The curve selector for ECDSA and the bits selector for RSA/DSA are independent controls, so each one remembers its last known value even while temporarily hidden in favour of the other. The actual generation function still expects a bit count rather than an actual curve or algorithm ID, so the easiest way to actually arrange to populate the drop-down list was to have an array of bit counts exposed by sshecc.c. That's a bit ugly, but there we go. One small functional change: if you enter an absurdly low value into the RSA/DSA bit count box (under 256), PuTTYgen used to give a warning and reset it to 256. Now it resets it to the default key length of 2048, basically because I was touching that code anyway to change a variable name and just couldn't bring myself to leave it in a state where it intentionally chose such an utterly useless key size. Of course this doesn't prevent generation of 256-bit keys if someone still really wants one - it just means they don't get one selected as the result of a typo.
2016-03-25 07:53:06 +00:00
const int ec_nist_curve_lengths[] = { 256, 384, 521 };
const int n_ec_nist_curve_lengths = lenof(ec_nist_curve_lengths);
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool ec_nist_alg_and_curve_by_bits(
int bits, const struct ec_curve **curve, const ssh_keyalg **alg)
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
{
switch (bits) {
case 256: *alg = &ssh_ecdsa_nistp256; break;
case 384: *alg = &ssh_ecdsa_nistp384; break;
case 521: *alg = &ssh_ecdsa_nistp521; break;
Adopt C99 <stdbool.h>'s true/false. This commit includes <stdbool.h> from defs.h and deletes my traditional definitions of TRUE and FALSE, but other than that, it's a 100% mechanical search-and-replace transforming all uses of TRUE and FALSE into the C99-standardised lowercase spellings. No actual types are changed in this commit; that will come next. This is just getting the noise out of the way, so that subsequent commits can have a higher proportion of signal.
2018-10-29 19:50:29 +00:00
default: return false;
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
}
*curve = ((struct ecsign_extra *)(*alg)->extra)->curve();
Adopt C99 <stdbool.h>'s true/false. This commit includes <stdbool.h> from defs.h and deletes my traditional definitions of TRUE and FALSE, but other than that, it's a 100% mechanical search-and-replace transforming all uses of TRUE and FALSE into the C99-standardised lowercase spellings. No actual types are changed in this commit; that will come next. This is just getting the noise out of the way, so that subsequent commits can have a higher proportion of signal.
2018-10-29 19:50:29 +00:00
return true;
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
}
Convert a lot of 'int' variables to 'bool'. My normal habit these days, in new code, is to treat int and bool as _almost_ completely separate types. I'm still willing to use C's implicit test for zero on an integer (e.g. 'if (!blob.len)' is fine, no need to spell it out as blob.len != 0), but generally, if a variable is going to be conceptually a boolean, I like to declare it bool and assign to it using 'true' or 'false' rather than 0 or 1. PuTTY is an exception, because it predates the C99 bool, and I've stuck to its existing coding style even when adding new code to it. But it's been annoying me more and more, so now that I've decided C99 bool is an acceptable thing to require from our toolchain in the first place, here's a quite thorough trawl through the source doing 'boolification'. Many variables and function parameters are now typed as bool rather than int; many assignments of 0 or 1 to those variables are now spelled 'true' or 'false'. I managed this thorough conversion with the help of a custom clang plugin that I wrote to trawl the AST and apply heuristics to point out where things might want changing. So I've even managed to do a decent job on parts of the code I haven't looked at in years! To make the plugin's work easier, I pushed platform front ends generally in the direction of using standard 'bool' in preference to platform-specific boolean types like Windows BOOL or GTK's gboolean; I've left the platform booleans in places they _have_ to be for the platform APIs to work right, but variables only used by my own code have been converted wherever I found them. In a few places there are int values that look very like booleans in _most_ of the places they're used, but have a rarely-used third value, or a distinction between different nonzero values that most users don't care about. In these cases, I've _removed_ uses of 'true' and 'false' for the return values, to emphasise that there's something more subtle going on than a simple boolean answer: - the 'multisel' field in dialog.h's list box structure, for which the GTK front end in particular recognises a difference between 1 and 2 but nearly everything else treats as boolean - the 'urgent' parameter to plug_receive, where 1 vs 2 tells you something about the specific location of the urgent pointer, but most clients only care about 0 vs 'something nonzero' - the return value of wc_match, where -1 indicates a syntax error in the wildcard. - the return values from SSH-1 RSA-key loading functions, which use -1 for 'wrong passphrase' and 0 for all other failures (so any caller which already knows it's not loading an _encrypted private_ key can treat them as boolean) - term->esc_query, and the 'query' parameter in toggle_mode in terminal.c, which _usually_ hold 0 for ESC[123h or 1 for ESC[?123h, but can also hold -1 for some other intervening character that we don't support. In a few places there's an integer that I haven't turned into a bool even though it really _can_ only take values 0 or 1 (and, as above, tried to make the call sites consistent in not calling those values true and false), on the grounds that I thought it would make it more confusing to imply that the 0 value was in some sense 'negative' or bad and the 1 positive or good: - the return value of plug_accepting uses the POSIXish convention of 0=success and nonzero=error; I think if I made it bool then I'd also want to reverse its sense, and that's a job for a separate piece of work. - the 'screen' parameter to lineptr() in terminal.c, where 0 and 1 represent the default and alternate screens. There's no obvious reason why one of those should be considered 'true' or 'positive' or 'success' - they're just indices - so I've left it as int. ssh_scp_recv had particularly confusing semantics for its previous int return value: its call sites used '<= 0' to check for error, but it never actually returned a negative number, just 0 or 1. Now the function and its call sites agree that it's a bool. In a couple of places I've renamed variables called 'ret', because I don't like that name any more - it's unclear whether it means the return value (in preparation) for the _containing_ function or the return value received from a subroutine call, and occasionally I've accidentally used the same variable for both and introduced a bug. So where one of those got in my way, I've renamed it to 'toret' or 'retd' (the latter short for 'returned') in line with my usual modern practice, but I haven't done a thorough job of finding all of them. Finally, one amusing side effect of doing this is that I've had to separate quite a few chained assignments. It used to be perfectly fine to write 'a = b = c = TRUE' when a,b,c were int and TRUE was just a the 'true' defined by stdbool.h, that idiom provokes a warning from gcc: 'suggest parentheses around assignment used as truth value'!
2018-11-02 19:23:19 +00:00
bool ec_ed_alg_and_curve_by_bits(
int bits, const struct ec_curve **curve, const ssh_keyalg **alg)
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
{
switch (bits) {
case 256: *alg = &ssh_ecdsa_ed25519; break;
Adopt C99 <stdbool.h>'s true/false. This commit includes <stdbool.h> from defs.h and deletes my traditional definitions of TRUE and FALSE, but other than that, it's a 100% mechanical search-and-replace transforming all uses of TRUE and FALSE into the C99-standardised lowercase spellings. No actual types are changed in this commit; that will come next. This is just getting the noise out of the way, so that subsequent commits can have a higher proportion of signal.
2018-10-29 19:50:29 +00:00
default: return false;
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
}
*curve = ((struct ecsign_extra *)(*alg)->extra)->curve();
Adopt C99 <stdbool.h>'s true/false. This commit includes <stdbool.h> from defs.h and deletes my traditional definitions of TRUE and FALSE, but other than that, it's a 100% mechanical search-and-replace transforming all uses of TRUE and FALSE into the C99-standardised lowercase spellings. No actual types are changed in this commit; that will come next. This is just getting the noise out of the way, so that subsequent commits can have a higher proportion of signal.
2018-10-29 19:50:29 +00:00
return true;
Clean up elliptic curve selection and naming. The ec_name_to_curve and ec_curve_to_name functions shouldn't really have had to exist at all: whenever any part of the PuTTY codebase starts using sshecc.c, it's starting from an ssh_signkey or ssh_kex pointer already found by some other means. So if we make sure not to lose that pointer, we should never need to do any string-based lookups to find the curve we want, and conversely, when we need to know the name of our curve or our algorithm, we should be able to look it up as a straightforward const char * starting from the algorithm pointer. This commit cleans things up so that that is indeed what happens. The ssh_signkey and ssh_kex structures defined in sshecc.c now have 'extra' fields containing pointers to all the necessary stuff; ec_name_to_curve and ec_curve_to_name have been completely removed; struct ec_curve has a string field giving the curve's name (but only for those curves which _have_ a name exposed in the wire protocol, i.e. the three NIST ones); struct ec_key keeps a pointer to the ssh_signkey it started from, and uses that to remember the algorithm name rather than reconstructing it from the curve. And I think I've got rid of all the ad-hockery scattered around the code that switches on curve->fieldBits or manually constructs curve names using stuff like sprintf("nistp%d"); the only remaining switch on fieldBits (necessary because that's the UI for choosing a curve in PuTTYgen) is at least centralised into one place in sshecc.c. One user-visible result is that the format of ed25519 host keys in the registry has changed: there's now no curve name prefix on them, because I think it's not really right to make up a name to use. So any early adopters who've been using snapshot PuTTY in the last week will be inconvenienced; sorry about that.
2015-05-15 09:13:05 +00:00
}
Reference in New Issue Copy Permalink