[cryptotest] Fix some failing RSA verify cryptotests #28655
Conversation
There was a problem hiding this comment.
Thanks for looking into this, @nasahlpa.
When I apply the changes from my comments, the failing test vectors from before are now passing, but I'm seeing a new failure in (at least) the first test vector of sw/host/cryptotest/testvectors/data/wycheproof_rsa_pss_3072_shake128.json. I see that this is documented in #28656 - is this intended to be fixed in a separate PR?
The wycheproof testvector seem to have a leading 0. This means that a 512-byte modulo for RSA-4096 is actually 513-byte. Currently, the test framework fails because it expects a 512-byte value. Remove this leading 0. Signed-off-by: Pascal Nasahl <[email protected]>
nasahlpa
force-pushed
the
fix_rsa_cryptotest
branch
from
06e3ab9 to
d0ddb49
Compare
There was a problem hiding this comment.
LGTM, probably needs to be backported to earlgrey_1.0.0 as well?
Is the intention to fix the remaining RSA test failures in #28656 (and thus the failures in //sw/device/tests/crypto/cryptotest:rsa_kat) in separate PRs?
|
Thanks for the review, Alex! Yes, the intention is to fix this in upcoming PRs and in the meanwhile track the failing tests in #28656. |
nasahlpa
added
the
CherryPick:earlgrey_1.0.0
Some tests in the wycheproof test vector set are error tests. In one of these error tests, a larger signature (514 instead of 512-bytes) is sent to the DUT. However, we currently limit the signature payload to 512- bytes, letting the test fail. The solution to this is simple, increase the max. number of signature bytes to 512-bytes. The CL now gets this 514-byte signature and internally fails, as expected by the test. Signed-off-by: Pascal Nasahl <[email protected]>
nasahlpa
force-pushed
the
fix_rsa_cryptotest
branch
from
d0ddb49 to
9ab2711
Compare
|
Which tests does this fix? |
For the first commit: However, for RSA-4096, n is actually 513-bytes because of the leading 0.
For the second commit: |
|
Successfully created backport PR for |
3 participants
This PR fixes two different test failures:
Those failures are not related to bugs in the cryptolib but in the cryptotest framework.