Add pre-TLS13 libcrypto PRF implementation #4020
Conversation
Squashed commit of the following: commit 1678bff Author: Sam Clark <[email protected]> Date: Wed May 24 13:21:23 2023 -0400 cleanup commit ebea361 Author: Sam Clark <[email protected]> Date: Wed May 24 13:12:19 2023 -0400 clang-format commit e12b4ac Author: Sam Clark <[email protected]> Date: Wed May 24 13:09:12 2023 -0400 move s2n_libcrypto_supports_tls_prf to s2n_prf commit e9ae9ae Author: Sam Clark <[email protected]> Date: Wed May 24 13:01:43 2023 -0400 cleanup dynamic seed_b allocation commit 2e8ecc0 Author: Sam Clark <[email protected]> Date: Tue May 23 16:32:12 2023 -0400 remove openssl implementation commit 21b77d2 Author: Sam Clark <[email protected]> Date: Tue May 23 11:45:20 2023 -0400 gate libcrypto prf with > OpenSSL 1.0.2 commit ab20222 Author: Sam Clark <[email protected]> Date: Tue May 23 11:04:10 2023 -0400 discard const commit 287ecc1 Author: Sam Clark <[email protected]> Date: Wed May 17 19:01:56 2023 -0400 Add pre-TLS13 libcrypto PRF implementation commit 9711902 Author: Sam Clark <[email protected]> Date: Wed May 17 16:35:41 2023 -0400 wip
tls/s2n_prf.c
Outdated
| /* BoringSSL and AWSLC define the CRYPTO_tls1_prf API in a private header. This function is | ||
| * forward-declared to make it accessible. | ||
| */ |
There was a problem hiding this comment.
Do you have any concerns about visibility / this method not being stable? For example, this would be a very bad idea to do with any s2n-tls method not officially in the api.
If this is safe, we need to explain why in a comment.
There was a problem hiding this comment.
I added a comment to explain this. I also removed the feature probe. After testing the probe by rebuilding AWSLC with modified API definitions, the try_compile was not sufficient to catch the changes and disable the API. If the modification just hid the symbol, the try_compile worked, but if an argument was added or something the try_compile failed to catch it. Since BoringSSL could theoretically change this API definition, I gated it behind just AWSLC rather than a feature probe.
There was a problem hiding this comment.
And just to confirm, since we're not using a try_compile, the method is available and exported in all versions of awslc? Even the oldest?
There was a problem hiding this comment.
Correct - CRYPTO_tls1_prf was made visible by aws/aws-lc@37c6eb4 in 2018. The first AWSLC release was v0.1-alpha, released in 2020. The API is exported in this version here: https://github.com/aws/aws-lc/blob/v0.1-alpha/third_party/boringssl/crypto/fipsmodule/tls/internal.h#L27-L33
goatgoose
force-pushed
the
fips-prf
branch
3 times, most recently
from
a136896 to
46237a4
Compare
Description of changes:
s2n-tls currently uses a custom PRF implementation to generate secret data for the TLS connection. This PR adds the AWSLC/BoringSSL libcrypto implementation, and uses this when s2n-tls is operating in FIPS mode.
Call-outs:
EVP_PKEY_CTXAPIs, separate from the AWSLC/BoringSSL API. However, the TLS PRF was added in OpenSSL 1.1.0. Since the only FIPS-compatible OpenSSL version is 1.0.2, I didn't implement the OpenSSL version.Testing:
s2n_tls_prf_testensure the libcrypto implementation is correct. Thes2n_prffunction is provided 3 seeds in the case of calculating a hybrid PQ master secret, and this is tested ins2n_tls_hybrid_prf_test.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.