DIT, for 'Data-Independent Timing', is a bit you can set in the
processor state on sufficiently new Arm CPUs, which promises that a
long list of instructions will deliberately avoid varying their timing
based on the input register values. Just what you want for keeping
your constant-time crypto primitives constant-time.
As far as I'm aware, no CPU has _yet_ implemented any data-dependent
optimisations, so DIT is a safety precaution against them doing so in
future. It would be embarrassing to be caught without it if a future
CPU does do that, so we now turn on DIT in the PuTTY process state.
I've put a call to the new enable_dit() function at the start of every
main() and WinMain() belonging to a program that might do
cryptography (even testcrypt, in case someone uses it for something!),
and in case I missed one there, also added a second call at the first
moment that any cryptography-using part of the code looks as if it
might become active: when an instance of the SSH protocol object is
configured, when the system PRNG is initialised, and when selecting
any cryptographic authentication protocol in an HTTP or SOCKS proxy
connection. With any luck those precautions between them should ensure
it's on whenever we need it.
Arm's own recommendation is that you should carefully choose the
granularity at which you enable and disable DIT: there's a potential
time cost to turning it on and off (I'm not sure what, but plausibly
something of the order of a pipeline flush), so it's a performance hit
to do it _inside_ each individual crypto function, but if CPUs start
supporting significant data-dependent optimisation in future, then it
will also become a noticeable performance hit to just leave it on
across the whole process. So you'd like to do it somewhere in the
middle: for example, you might turn on DIT once around the whole
process of verifying and decrypting an SSH packet, instead of once for
decryption and once for MAC.
With all respect to that recommendation as a strategy for maximum
performance, I'm not following it here. I turn on DIT at the start of
the PuTTY process, and then leave it on. Rationale:
1. PuTTY is not otherwise a performance-critical application: it's
not likely to max out your CPU for any purpose _other_ than
cryptography. The most CPU-intensive non-cryptographic thing I can
imagine a PuTTY process doing is the complicated computation of
font rendering in the terminal, and that will normally be cached
(you don't recompute each glyph from its outline and hints for
every time you display it).
2. I think a bigger risk lies in accidental side channels from having
DIT turned off when it should have been on. I can imagine lots of
causes for that. Missing a crypto operation in some unswept corner
of the code; confusing control flow (like my coroutine macros)
jumping with DIT clear into the middle of a region of code that
expected DIT to have been set at the beginning; having a reference
counter of DIT requests and getting it out of sync.
In a more sophisticated programming language, it might be possible to
avoid the risk in #2 by cleverness with the type system. For example,
in Rust, you could have a zero-sized type that acts as a proof token
for DIT being enabled (it would be constructed by a function that also
sets DIT, have a Drop implementation that clears DIT, and be !Send so
you couldn't use it in a thread other than the one where DIT was set),
and then you could require all the actual crypto functions to take a
DitToken as an extra parameter, at zero runtime cost. Then "oops I
forgot to set DIT around this piece of crypto" would become a compile
error. Even so, you'd have to take some care with coroutine-structured
code (what happens if a Rust async function yields while holding a DIT
token?) and with nesting (if you have two DIT tokens, you don't want
dropping the inner one to clear DIT while the outer one is still there
to wrongly convince callees that it's set). Maybe in Rust you could
get this all to work reliably. But not in C!
DIT is an optional feature of the Arm architecture, so we must first
test to see if it's supported. This is done the same way as we already
do for the various Arm crypto accelerators: on ELF-based systems,
check the appropriate bit in the 'hwcap' words in the ELF aux vector;
on Mac, look for an appropriate sysctl flag.
On Windows I don't know of a way to query the DIT feature, _or_ of a
way to write the necessary enabling instruction in an MSVC-compatible
way. I've _heard_ that it might not be necessary, because Windows
might just turn on DIT unconditionally and leave it on, in an even
more extreme version of my own strategy. I don't have a source for
that - I heard it by word of mouth - but I _hope_ it's true, because
that would suit me very well! Certainly I can't write code to enable
DIT without knowing (a) how to do it, (b) how to know if it's safe.
Nonetheless, I've put the enable_dit() call in all the right places in
the Windows main programs as well as the Unix and cross-platform code,
so that if I later find out that I _can_ put in an explicit enable of
DIT in some way, I'll only have to arrange to set HAVE_ARM_DIT and
compile the enable_dit() function appropriately.
Jacob reports that bullseye objected to the change from
G_APPLICATION_FLAGS_NONE to G_APPLICATION_DEFAULT_FLAGS, on the
grounds that it only has the former defined. Sigh. Added a cmake
check.
This was requested by a downstream of the code, who wanted to change
the time/space tradeoff in the terminal. I currently have no plans to
change this setting for upstream PuTTY, although there is a cmake
option for it just to make testing it easy.
To avoid sprinkling ifdefs over the whole terminal code, the strategy
is to keep the separate type 'compressed_scrollback_line', and turn it
into a typedef for a 'termline *'. So compressline() becomes almost
trivial, and decompressline() even more so.
Memory management is the fiddly part. To make this work sensibly on
both sides, I've broken up each of compressline() and decompressline()
into two versions, one of which takes ownership of (and logically
speaking frees) its input, and the other doesn't. So at call sites
where a function was followed by a free, it's now calling the
'and_free' version of the function, and where the input object was
reused afterwards, it's calling the 'no_free' version. This means that
in different branches of the #if, I can make one function call the
other or vice versa, and no call site is stuck with having to do
things in a more roundabout way than necessary.
The freeing of the _return_ value from decompressline() is handled for
us, because termlines already have a 'temporary' flag which is set
when they're returned from the decompressor, and anyone receiving a
termline from lineptr() calls unlineptr() when they're finished with
it, which will _conditionally_ free it, depending on that 'temporary'
flag. So in the new mode, 'temporary' is never set at all, and all
those unlineptr() calls do nothing.
However, we also still need to free compressed lines properly when
they're actually being thrown away (scrolled off the top of the
scrollback, or cleaned up in term_free), and for that, I've made a new
special-purpose free_compressed_line() function.
In some compilers (I'm told clang 10, in particular), the NEON
intrinsic vaddq_p128 is missing, even though its input type poly128_t
is provided.
vaddq_p128 is just an XOR of two vector registers, so that's easy to
work around by casting to a more mundane type and back. Added a
configure-time test for that intrinsic, and a workaround to be used in
its absence.
FreeBSD declares setpgrp() as taking two arguments, like Linux's
setpgid(). Detect that at configure time and adjust the call in
Pageant appropriately.
I only recently found out that OpenSSH defined their own protocol IDs
for AES-GCM, defined to work the same as the standard ones except that
they fixed the semantics for how you select the linked cipher+MAC pair
during key exchange.
(RFC 5647 defines protocol ids for AES-GCM in both the cipher and MAC
namespaces, and requires that you MUST select both or neither - but
this contradicts the selection policy set out in the base SSH RFCs,
and there's no discussion of how you resolve a conflict between them!
OpenSSH's answer is to do it the same way ChaCha20-Poly1305 works,
because that will ensure the two suites don't fight.)
People do occasionally ask us for this linked cipher/MAC pair, and now
I know it's actually feasible, I've implemented it, including a pair
of vector implementations for x86 and Arm using their respective
architecture extensions for multiplying polynomials over GF(2).
Unlike ChaCha20-Poly1305, I've kept the cipher and MAC implementations
in separate objects, with an arm's-length link between them that the
MAC uses when it needs to encrypt single cipher blocks to use as the
inputs to the MAC algorithm. That enables the cipher and the MAC to be
independently selected from their hardware-accelerated versions, just
in case someone runs on a system that has polynomial multiplication
instructions but not AES acceleration, or vice versa.
There's a fourth implementation of the GCM MAC, which is a pure
software implementation of the same algorithm used in the vectorised
versions. It's too slow to use live, but I've kept it in the code for
future testing needs, and because it's a convenient place to dump my
design comments.
The vectorised implementations are fairly crude as far as optimisation
goes. I'm sure serious x86 _or_ Arm optimisation engineers would look
at them and laugh. But GCM is a fast MAC compared to HMAC-SHA-256
(indeed compared to HMAC-anything-at-all), so it should at least be
good enough to use. And we've got a working version with some tests
now, so if someone else wants to improve them, they can.
Without this, the build of e.g. psusan would fail on systems without
that header (such as Termux on Android).
This is similar to how things were pre-cmake, but not identical. We used
to treat lack of updwtmpx() as a reason to OMIT_UTMP (as of f0dfa73982),
but usage of that function got conditionalised in c19e7215dd, so I
haven't restored that exclusion.
Using a new screenshot-taking module I just added in windows/utils,
these new options allow me to start up one of the tools with
demonstration window contents and automatically save a .BMP screenshot
to disk. This will allow me to keep essentially the same set of demo
images and update them easily to keep pace with the current appearance
of the real tools as PuTTY - and Windows itself - both evolve.
This applies to all of AES, SHA-1, SHA-256 and SHA-512. All those
source files previously contained multiple implementations of the
algorithm, enabled or disabled by ifdefs detecting whether they would
work on a given compiler. And in order to get advanced machine
instructions like AES-NI or NEON crypto into the output file when the
compile flags hadn't enabled them, we had to do nasty stuff with
compiler-specific pragmas or attributes.
Now we can do the detection at cmake time, and enable advanced
instructions in the more sensible way, by compile-time flags. So I've
broken up each of these modules into lots of sub-pieces: a file called
(e.g.) 'foo-common.c' containing common definitions across all
implementations (such as round constants), one called 'foo-select.c'
containing the top-level vtable(s), and a separate file for each
implementation exporting just the vtable(s) for that implementation.
One advantage of this is that it depends a lot less on compiler-
specific bodgery. My particular least favourite part of the previous
setup was the part where I had to _manually_ define some Arm ACLE
feature macros before including <arm_neon.h>, so that it would define
the intrinsics I wanted. Now I'm enabling interesting architecture
features in the normal way, on the compiler command line, there's no
need for that kind of trick: the right feature macros are already
defined and <arm_neon.h> does the right thing.
Another change in this reorganisation is that I've stopped assuming
there's just one hardware implementation per platform. Previously, the
accelerated vtables were called things like sha256_hw, and varied
between FOO-NI and NEON depending on platform; and the selection code
would simply ask 'is hw available? if so, use hw, else sw'. Now, each
HW acceleration strategy names its vtable its own way, and the
selection vtable has a whole list of possibilities to iterate over
looking for a supported one. So if someone feels like writing a second
accelerated implementation of something for a given platform - for
example, I've heard you can use plain NEON to speed up AES somewhat
even without the crypto extension - then it will now have somewhere to
drop in alongside the existing ones.
A couple of actual checks were missing (elf_aux_info, sysctlbyname).
Several more were accidentally left out of cmake.h.in, meaning they
wouldn't be propagated from cmake's variable space into the actual
compilation. And a handful of checks in the C source were still using
the autotools-style 'if defined' in place of the cmake-style "it's
always 0 or 1" plain #if.
This brings various concrete advantages over the previous system:
- consistent support for out-of-tree builds on all platforms
- more thorough support for Visual Studio IDE project files
- support for Ninja-based builds, which is particularly useful on
Windows where the alternative nmake has no parallel option
- a really simple set of build instructions that work the same way on
all the major platforms (look how much shorter README is!)
- better decoupling of the project configuration from the toolchain
configuration, so that my Windows cross-building doesn't need
(much) special treatment in CMakeLists.txt
- configure-time tests on Windows as well as Linux, so that a lot of
ad-hoc #ifdefs second-guessing a particular feature's presence from
the compiler version can now be replaced by tests of the feature
itself
Also some longer-term software-engineering advantages:
- other people have actually heard of CMake, so they'll be able to
produce patches to the new build setup more easily
- unlike the old mkfiles.pl, CMake is not my personal problem to
maintain
- most importantly, mkfiles.pl was just a horrible pile of
unmaintainable cruft, which even I found it painful to make changes
to or to use, and desperately needed throwing in the bin. I've
already thrown away all the variants of it I had in other projects
of mine, and was only delaying this one so we could make the 0.75
release branch first.
This change comes with a noticeable build-level restructuring. The
previous Recipe worked by compiling every object file exactly once,
and then making each executable by linking a precisely specified
subset of the same object files. But in CMake, that's not the natural
way to work - if you write the obvious command that puts the same
source file into two executable targets, CMake generates a makefile
that compiles it once per target. That can be an advantage, because it
gives you the freedom to compile it differently in each case (e.g.
with a #define telling it which program it's part of). But in a
project that has many executable targets and had carefully contrived
to _never_ need to build any module more than once, all it does is
bloat the build time pointlessly!
To avoid slowing down the build by a large factor, I've put most of
the modules of the code base into a collection of static libraries
organised vaguely thematically (SSH, other backends, crypto, network,
...). That means all those modules can still be compiled just once
each, because once each library is built it's reused unchanged for all
the executable targets.
One upside of this library-based structure is that now I don't have to
manually specify exactly which objects go into which programs any more
- it's enough to specify which libraries are needed, and the linker
will figure out the fine detail automatically. So there's less
maintenance to do in CMakeLists.txt when the source code changes.
But that reorganisation also adds fragility, because of the trad Unix
linker semantics of walking along the library list once each, so that
cyclic references between your libraries will provoke link errors. The
current setup builds successfully, but I suspect it only just manages
it.
(In particular, I've found that MinGW is the most finicky on this
score of the Windows compilers I've tried building with. So I've
included a MinGW test build in the new-look Buildscr, because
otherwise I think there'd be a significant risk of introducing
MinGW-only build failures due to library search order, which wasn't a
risk in the previous library-free build organisation.)
In the longer term I hope to be able to reduce the risk of that, via
gradual reorganisation (in particular, breaking up too-monolithic
modules, to reduce the risk of knock-on references when you included a
module for function A and it also contains function B with an
unsatisfied dependency you didn't really need). Ideally I want to
reach a state in which the libraries all have sensibly described
purposes, a clearly documented (partial) order in which they're
permitted to depend on each other, and a specification of what stubs
you have to put where if you're leaving one of them out (e.g.
nocrypto) and what callbacks you have to define in your non-library
objects to satisfy dependencies from things low in the stack (e.g.
out_of_memory()).
One thing that's gone completely missing in this migration,
unfortunately, is the unfinished MacOS port linked against Quartz GTK.
That's because it turned out that I can't currently build it myself,
on my own Mac: my previous installation of GTK had bit-rotted as a
side effect of an Xcode upgrade, and I haven't yet been able to
persuade jhbuild to make me a new one. So I can't even build the MacOS
port with the _old_ makefiles, and hence, I have no way of checking
that the new ones also work. I hope to bring that port back to life at
some point, but I don't want it to block the rest of this change.
