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.
This provides a convenient hook to be called between SSH messages, for
the crypto components to do any per-message processing like
incrementing a sequence number.
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.
