[ML-KEM] Match mainline APIs #112
Conversation
|
Blocked on #117 getting merged, which updates the C extraction scripts. |
keks
left a comment
keks
left a comment
There was a problem hiding this comment.
I think this looks like it should work, just the copying is unfortunate. Do you think there is something we can do about with little effort or is that a bigger thing?
| macro_rules! impl_kem_trait { | ||
| ($variant:ty, $pk:ty, $sk:ty, $ct:ty) => { | ||
| impl | ||
| libcrux_traits::kem::arrayref::Kem< | ||
| CPA_PKE_PUBLIC_KEY_SIZE, | ||
| SECRET_KEY_SIZE, | ||
| CPA_PKE_CIPHERTEXT_SIZE, | ||
| SHARED_SECRET_SIZE, | ||
| KEY_GENERATION_SEED_SIZE, | ||
| SHARED_SECRET_SIZE, | ||
| > for $variant | ||
| { | ||
| fn keygen( | ||
| ek: &mut [u8; CPA_PKE_PUBLIC_KEY_SIZE], | ||
| dk: &mut [libcrux_secrets::U8; SECRET_KEY_SIZE], | ||
| rand: &[libcrux_secrets::U8; KEY_GENERATION_SEED_SIZE], | ||
| ) -> Result<(), libcrux_traits::kem::owned::KeyGenError> { | ||
| let key_pair = generate_key_pair(*rand); | ||
| ek.copy_from_slice(key_pair.pk()); | ||
| dk.copy_from_slice(key_pair.sk()); | ||
|
|
||
| Ok(()) | ||
| } | ||
|
|
||
| fn encaps( | ||
| ct: &mut [u8; CPA_PKE_CIPHERTEXT_SIZE], | ||
| ss: &mut [libcrux_secrets::U8; SHARED_SECRET_SIZE], | ||
| ek: &[u8; CPA_PKE_PUBLIC_KEY_SIZE], | ||
| rand: &[libcrux_secrets::U8; SHARED_SECRET_SIZE], | ||
| ) -> Result<(), libcrux_traits::kem::owned::EncapsError> { | ||
| let public_key: $pk = ek.into(); | ||
|
|
||
| let (ct_, ss_) = encapsulate(&public_key, *rand); | ||
| ct.copy_from_slice(ct_.as_slice()); | ||
| ss.copy_from_slice(ss_.as_slice()); | ||
|
|
||
| Ok(()) | ||
| } | ||
|
|
||
| fn decaps( | ||
| ss: &mut [libcrux_secrets::U8; SHARED_SECRET_SIZE], | ||
| ct: &[u8; CPA_PKE_CIPHERTEXT_SIZE], | ||
| dk: &[libcrux_secrets::U8; SECRET_KEY_SIZE], | ||
| ) -> Result<(), libcrux_traits::kem::owned::DecapsError> { | ||
| let secret_key: $sk = dk.into(); | ||
| let ciphertext: $ct = ct.into(); | ||
|
|
||
| let ss_ = decapsulate(&secret_key, &ciphertext); | ||
|
|
||
| ss.copy_from_slice(ss_.as_slice()); | ||
|
|
There was a problem hiding this comment.
It's a bit unfortunate that we have to copy everywhere here. I suppose we don't have &mut in this implementation yet?
There was a problem hiding this comment.
I agree, it is unfortunate. At the moment, we do not have any top-level APIs here that operate on pre-allocated inputs. This would be good to have and I'll file an issue for it, but since this PR is about matching mainline APIs I would rather not add new top-level APIs here.
There was a problem hiding this comment.
Makes sense! Let's get this in as-is
2 participants
This PR ports over some of the APIs that are provided for ML-KEM in mainline libcrux, namely:
tls_codecThe PR does not port the incremental API.
Fixes https://github.com/cryspen/home/issues/431