Handles managed by winhandl.c have a 'busy' flag, which is used to
mean two things: (a) is a subthread currently blocked on this handle
so various operations in the main thread have to be deferred until it
finishes? And (b) is this handle currently one that should be returned
to the main loop to be waited for?
For HT_INPUT and HT_OUTPUT, those things are either both true or both
false, so a single flag covering both of them is fine. But HT_FOREIGN
handles have the property that they should always be waited for in the
main loop, but no subthread is blocked on them. The latter means that
operations done on them in the main thread should not be deferred; the
only such operation is cleaning them up in handle_free().
handle_free() was failing to spot this, and was deferring freeing
HT_FOREIGN handles until their subthread terminated - which of course
never happened. As a result, when a named pipe server was closed, its
actual Windows event object got destroyed, but winhandl.c still kept
passing it back to the main thread, leading to a tight loop because
MsgWaitForMultipleObjects would return ERROR_INVALID_HANDLE and never
block.
If you use the new 'plink -shareexists' feature, then on Unix at least
it's possible for the upstream to receive EPIPE, because the
downstream makes a test connection and immediately closes it, so that
upstream fails to write its version string.
This looks a bit ugly in the upstream's Event Log, so I'm making a
special case: an error of 'broken pipe' type, which occurs on a socket
from a connection sharing downstream, before we've received a version
string from that downstream, is treated as an unusual kind of normal
connection termination and not logged as an error.
A Plink invocation of the form 'plink -shareexists <session>' tests
for a currently live connection-sharing upstream for the session in
question. <session> can be any syntax you'd use with Plink to make the
actual connection (a host/port number, a bare saved session name,
-load, whatever).
I envisage this being useful for things like adaptive proxying - e.g.
if you want to connect to host A which you can't route to directly,
and you might already have a connection to either of hosts B or C
which are viable proxies, then you could write a proxy shell script
which checks whether you already have an upstream for B or C and goes
via whichever one is currently active.
Testing for the upstream's existence has to be done by actually
connecting to its socket, because on Unix the mere existence of a
Unix-domain socket file doesn't guarantee that there's a process
listening to it. So we make a test connection, and then immediately
disconnect; hence, that shows up in the upstream's event log.
On Windows, colons are illegal in filenames, because they're part of
the path syntax. But colons can appear in automatically constructed
log file names, if an IPv6 address is expanded from the &H placeholder.
Now we coerce any such illegal characters to '.', which is a bit of a
bodge but should at least cause a log file to be generated.
I noticed that Unix PSCP was unwantedly renaming downloaded files
which had a backslash in their names, because pscp.c's stripslashes()
treated \ as a path component separator, since it hadn't been modified
since PSCP ran on Windows only.
It also turns out that pscp.c, psftp.c and winsftp.c all had a
stripslashes(), and they didn't all have quite the same prototype. So
now there's one in winsftp.c and one in uxsftp.c, with appropriate
OS-dependent behaviour, and the ones in pscp.c and psftp.c are gone.
We are passing pointers as third argument to AppendMenu. Do not
truncate them to UINT, use UINT_PTR instead which has the required
size on 64bit Windows.
We're passing a pointer as 4th argument to WinHelp. Do not cast it to
DWORD which would truncate the pointer. Instead use UINT_PTR as that
is what WinHelp expects.
We've had several reports that launching saved sessions from the
Windows 10 jump list fails; Changyu Li reports that this is because we
create those IShellLink objects with a command line string starting
with @, and in Windows 10 that causes the SetArguments method to
silently do the wrong thing.
This is generated in response to the SendInput() Windows API call, if
that in turn is passed an KEYBDINPUT structure with KEYEVENTF_UNICODE
set. That method of input generation is used by programs such as
'WinCompose' to send an arbitrary Unicode character as if it had been
typed at the keyboard, even if the keyboard doesn't actually provide a
key for it.
Like VK_PROCESSKEY, this key code is an exception to our usual policy
of manually translating keystrokes: we handle it by calling
TranslateMessage, to get back the Unicode character it contains as a
WM_CHAR message.
(If that Unicode character in turn is outside the BMP, it may come
back as a pair of WM_CHARs in succession containing UTF-16 surrogates;
if so, that's OK, because the new Unicode WM_CHAR handler can cope.)
This causes WM_CHAR messages sent to us to have a wParam containing a
16-bit value encoded in UTF-16, rather than an 8-bit value encoded in
the system code page.
As far as I can tell, there aren't many other knock-on effects - e.g.
you can still interact with the window using ordinary char-based API
functions such as SetWindowText, and the Windows API will do the
necessary conversions behind the scenes. However, even so, I'm half
expecting some sort of unforeseen bug to show up as a result of this.
Coverity complained that some paths through the loop in the
WM_INITDIALOG handler might leave firstpath==NULL. In fact this can't
happen because the input data to that loop is largely static and we
know what it looks like, but it doesn't seem unreasonable to add an
assertion anyway, to keep static checkers happy and as an explanatory
quasi-comment for humans.
Our config boxes are constructed using the CreateDialog() API
function, rather than the modal DialogBox(). CreateDialog() is not
that different from CreateWindow(), so windows created with it don't
appear on the screen automatically; MSDN says that they must be shown
via ShowWindow(), just like non-dialog windows have to be. But we
weren't doing that at any point!
So how was our config box ever getting displayed at all? Apparently by
sheer chance, it turns out. The handler for a selection change in the
tree view, which has to delete a whole panel of controls and creates a
different set, surrounds that procedure with some WM_SETREDRAW calls
and an InvalidateRect(), to prevent flicker while lots of changes were
being made. And the creation of the _first_ panelful of controls, at
dialog box setup, was done by simply selecting an item in the treeview
and expecting that handler to be recursively called. And it appears
that calling WM_SETREDRAW(TRUE) and then InvalidateRect was
undocumentedly having an effect equivalent to the ShowWindow() we
should have called, so that we never noticed the latter was missing.
But a recent Vista update (all reports implicate KB3057839) has caused
that not to work any more: on an updated Vista machine, in some
desktop configurations, it seems that any attempt to fiddle with
WM_SETREDRAW during dialog setup can leave the dialog box in a really
unhelpful invisible state - the window is _physically there_ (you can
see its taskbar entry, and the mouse pointer changes as you move over
where its edit boxes are), but 100% transparent.
So now we're doing something a bit more sensible. The first panelful
of controls is created directly by the WM_INITDIALOG handler, rather
than recursing into code that wasn't really designed to run at setup
time. To be on the safe side, that handler for treeview selection
change is also disabled until the WM_INITDIALOG handler has finished
(like we already did with the WM_COMMAND handler), so that we can be
sure of not accidentally messing about with WM_SETREDRAW at all during
setup. And at the end of setup, we show the window in the sensible
way, by a docs-approved call to ShowWindow().
This appears (on the one machine I've so far tested it on) to fix the
Vista invisible-window issue, and also it should be more API-compliant
and hence safer in future.
(cherry picked from commit 6163710f04)
When anyone connects to a PuTTY tool's listening socket - whether it's
a user of a local->remote port forwarding, a connection-sharing
downstream or a client of Pageant - we'd like to log as much
information as we can find out about where the connection came from.
To that end, I've implemented a function sk_peer_info() in the socket
abstraction, which returns a freeform text string as best it can (or
NULL, if it can't get anything at all) describing the thing at the
other end of the connection. For TCP connections, this is done using
getpeername() to get an IP address and port in the obvious way; for
Unix-domain sockets, we attempt SO_PEERCRED (conditionalised on some
moderately hairy autoconfery) to get the pid and owner of the peer. I
haven't implemented anything for Windows named pipes, but I will if I
hear of anything useful.
(cherry picked from commit c8f83979a3)
Conflicts:
pageant.c
Cherry-picker's notes: the conflict was because the original commit
also added a use of the same feature in the centralised Pageant code,
which doesn't exist on this branch. Also I had to remove 'const' from
the type of the second parameter to wrap_send_port_open(), since this
branch hasn't had the same extensive const-fixing as master.
The last use of it, to store the contents of the saved session name
edit box, was removed nearly two years ago in svn r9923 and replaced
by ctrl_alloc_with_free. The mechanism has been unused ever since
then, and I suspect any further uses of it would be a bad idea for the
same reasons, so let's get rid of it.
(cherry picked from commit 42c592c4ef)
PuTTY now uses the updated version of Diffie-Hellman group exchange,
except for a few old OpenSSH versions which Darren Tucker reports only
support the old version.
FIXME: this needs further work because the Bugs config panel has now
overflowed.
(cherry picked from commit 62a1bce7cb)
When a winhandl.c input thread returns EOF to the main thread, the
latter might immediately delete the input thread's context. I
carefully wrote in a comment that in that case we had to not touch ctx
ever again after signalling to the main thread - but the test for
whether that was true, which also touched ctx, itself came _after_ the
SetEvent which sent that signal. Ahem.
Spotted by Minefield, which it looks as if I haven't run for a while.
(cherry picked from commit 9fec2e7738)
I had set up an event object for signalling incoming connections to
the named pipe, and then called handle_add_foreign_event to get that
event object watched for connections - but when I closed down the
listening pipe, I deleted the event object without also cancelling
that foreign-event handle, so that winhandl.c would potentially call
the callback for a destroyed object.
(cherry picked from commit 6f241cef2c)
This was an old bug, fixed around 0.59, which apparently regressed
when I rewrote the main event loop using the toplevel_callback
mechanism.
Investigation just now suggests that it has to do with my faulty
assumption that Windows PeekMessage would deliver messages in its
message queue in FIFO order (i.e. that the thing calling itself a
message queue is actually a _queue_). In fact my WM_NETEVENT seems to
like to jump the queue, so that once a steady stream of them starts
arriving, we never do anything else in the main event loop (except
deal with handles).
Worked around in a simple and slightly bodgy way, namely, we don't
stop looping on PeekMessage and run our toplevel callbacks until we've
either run out of messages completely or else seen at least one that
_isn't_ a WM_NETEVENT. That way we should reliably interleave NETEVENT
processing with processing of other stuff.
(cherry picked from commit 7d97c2a8fd)
To understand the handle leak bug that I fixed in git commit
7549f2da40, I had to think fairly hard
to remind myself what all this code was doing, which means the
comments weren't good enough. Expanded and rewritten some of them in
the hope that things will be clearer next time.
(cherry picked from commit a87a14ae0f)
Cherry-picker's notes: this apparently pointless commit is required on
this branch because it's a dependency of the rather less pointless
9fec2e7738.
Our config boxes are constructed using the CreateDialog() API
function, rather than the modal DialogBox(). CreateDialog() is not
that different from CreateWindow(), so windows created with it don't
appear on the screen automatically; MSDN says that they must be shown
via ShowWindow(), just like non-dialog windows have to be. But we
weren't doing that at any point!
So how was our config box ever getting displayed at all? Apparently by
sheer chance, it turns out. The handler for a selection change in the
tree view, which has to delete a whole panel of controls and creates a
different set, surrounds that procedure with some WM_SETREDRAW calls
and an InvalidateRect(), to prevent flicker while lots of changes were
being made. And the creation of the _first_ panelful of controls, at
dialog box setup, was done by simply selecting an item in the treeview
and expecting that handler to be recursively called. And it appears
that calling WM_SETREDRAW(TRUE) and then InvalidateRect was
undocumentedly having an effect equivalent to the ShowWindow() we
should have called, so that we never noticed the latter was missing.
But a recent Vista update (all reports implicate KB3057839) has caused
that not to work any more: on an updated Vista machine, in some
desktop configurations, it seems that any attempt to fiddle with
WM_SETREDRAW during dialog setup can leave the dialog box in a really
unhelpful invisible state - the window is _physically there_ (you can
see its taskbar entry, and the mouse pointer changes as you move over
where its edit boxes are), but 100% transparent.
So now we're doing something a bit more sensible. The first panelful
of controls is created directly by the WM_INITDIALOG handler, rather
than recursing into code that wasn't really designed to run at setup
time. To be on the safe side, that handler for treeview selection
change is also disabled until the WM_INITDIALOG handler has finished
(like we already did with the WM_COMMAND handler), so that we can be
sure of not accidentally messing about with WM_SETREDRAW at all during
setup. And at the end of setup, we show the window in the sensible
way, by a docs-approved call to ShowWindow().
This appears (on the one machine I've so far tested it on) to fix the
Vista invisible-window issue, and also it should be more API-compliant
and hence safer in future.
The general plan is that if PuTTY knows a host key for a server, it
should preferentially ask for the same type of key so that there's some
chance of actually getting the same key again. This should mean that
when a server (or PuTTY) adds a new host key type, PuTTY doesn't
gratuitously switch to that key type and then warn the user about an
unrecognised key.
When anyone connects to a PuTTY tool's listening socket - whether it's
a user of a local->remote port forwarding, a connection-sharing
downstream or a client of Pageant - we'd like to log as much
information as we can find out about where the connection came from.
To that end, I've implemented a function sk_peer_info() in the socket
abstraction, which returns a freeform text string as best it can (or
NULL, if it can't get anything at all) describing the thing at the
other end of the connection. For TCP connections, this is done using
getpeername() to get an IP address and port in the obvious way; for
Unix-domain sockets, we attempt SO_PEERCRED (conditionalised on some
moderately hairy autoconfery) to get the pid and owner of the peer. I
haven't implemented anything for Windows named pipes, but I will if I
hear of anything useful.
The menu options and radio buttons for key type were not consistently
setting each other when selected: in particular, selecting from the
menu did not cause the ED25519 radio button to be either set or unset
when that would have been appropriate.
Looks as if I failed to catch in code review the fact that we should
have _one_ call to each of CheckRadioButton and CheckMenuRadioItem,
and they should both have the right 'first' and 'last' parameters
Having found a lot of unfixed constness issues in recent development,
I thought perhaps it was time to get proactive, so I compiled the
whole codebase with -Wwrite-strings. That turned up a huge load of
const problems, which I've fixed in this commit: the Unix build now
goes cleanly through with -Wwrite-strings, and the Windows build is as
close as I could get it (there are some lingering issues due to
occasional Windows API functions like AcquireCredentialsHandle not
having the right constness).
Notable fallout beyond the purely mechanical changing of types:
- the stuff saved by cmdline_save_param() is now explicitly
dupstr()ed, and freed in cmdline_run_saved.
- I couldn't make both string arguments to cmdline_process_param()
const, because it intentionally writes to one of them in the case
where it's the argument to -pw (in the vain hope of being at least
slightly friendly to 'ps'), so elsewhere I had to temporarily
dupstr() something for the sake of passing it to that function
- I had to invent a silly parallel version of const_cmp() so I could
pass const string literals in to lookup functions.
- stripslashes() in pscp.c and psftp.c has the annoying strchr nature
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.
All the name strings in ssh_cipher, ssh_mac, ssh_hash, ssh_signkey
point to compile-time string literals, hence should obviously be const
char *.
Most of these const-correctness patches are just a mechanical job of
adding a 'const' in the one place you need it right now, and then
chasing the implications through the code adding further consts until
it compiles. But this one has actually shown up a bug: the 'algorithm'
output parameter in ssh2_userkey_loadpub was sometimes returning a
pointer to a string literal, and sometimes a pointer to dynamically
allocated memory, so callers were forced to either sometimes leak
memory or sometimes free a bad thing. Now it's consistently
dynamically allocated, and should be freed everywhere too.
Adding an extra radio button to the key-type selector caused it to
wrap on to another line and push the bottom of the containing control
box down off the bottom of the window.
In the long term, should more public key formats continue to appear,
we'll probably have to replace the radio buttons with something more
extensible like a drop-down list. For the moment, though, I've fixed
it by just reducing the space per radio button to bring all five
controls back on to the same line.
To fit the text into the smaller space, I also removed the 'SSH-2'
prefix on each key type, which ought to be unnecessary these days
since SSH-2 is a well established default. Only the SSH-1 RSA key type
is still labelled with an SSH version. (And I've moved it to the far
end rather than the start of the line, while I'm here.)
There were ad-hoc functions for fingerprinting a bare key blob in both
cmdgen.c and pageant.c, not quite doing the same thing. Also, every
SSH-2 public key algorithm in the code base included a dedicated
fingerprint() method, which is completely pointless since SSH-2 key
fingerprints are computed in an algorithm-independent way (just hash
the standard-format public key blob), so each of those methods was
just duplicating the work of the public_blob() method with a less
general output mechanism.
Now sshpubk.c centrally provides an ssh2_fingerprint_blob() function
that does all the real work, plus an ssh2_fingerprint() function that
wraps it and deals with calling public_blob() to get something to
fingerprint. And the fingerprint() method has been completely removed
from ssh_signkey and all its implementations, and good riddance.
There was a fair amount of duplication between Windows and Unix
PuTTYgen, and some confusion over writing things to FILE * and
formatting them internally into strings. I think all the public-key
output code now lives in sshpubk.c, and there's only one copy of the
code to generate each format.
I've now centralised into pageant.c all the logic about trying to load
keys of any type, with no passphrase or with the passphrases used in
previous key-loading actions or with a new user-supplied passphrase,
whether we're the main Pageant process ourself or are talking to
another one as a client. The only part of that code remaining in
winpgnt.c is the user interaction via dialog boxes, which of course is
the part that will need to be done differently on other platforms.
When I implemented reading and writing of the new format a couple of
weeks ago, I kept them strictly separate in the UI, so you have to ask
for the format you want when exporting. But in fact this is silly,
because not every key type can be saved in both formats, and OpenSSH
itself has the policy of using the old format for key types it can
handle, unless specifically asked to use the new one.
So I've now arranged that the key file format enum has three values
for OpenSSH: PEM, NEW and AUTO. Files being loaded are identified as
either PEM or NEW, which describe the two physical file formats. But
exporting UIs present either AUTO or NEW, where AUTO is the virtual
format meaning 'save in the old format if possible, otherwise the new
one'.
