feat(ext/crypto): support importing ECSDA and ECDH

[cryptographix]

importKey implemented for ECDSA and ECDH, extracted from #13013 with modifications

• pkcs8 : curve P-256 only
• spki : curves P-256 and P-384
• jwk :
  public key - curves P-256 and P-384
  private key - curve P-256 only

Restrictions on to P-384 curve due to lack of support in p384 crate.

[lucacasonato]

Good start. Thanks @seanwykes.

[lucacasonato]

Are the version bumps necessary? I would rather wait for them until rsa has gotten a new release, so we don't duplicate so many dependencies.

[cryptographix]

Are the version bumps necessary? I would rather wait for them until rsa has gotten a new release, so we don't duplicate so many dependencies.

Not as such .. bumped during attempts to improve p384 coverage (pkcs8 etc). Will back-out the changes ..

[lucacasonato]

@seanwykes I have unbumped the crates locally - will push that shortly.

[lucacasonato]

There are some WPT regressions it seems that I need to be dealt with before merge. (I can do this)

I will try to get this landed for 1.17.0 tomorrow. If it doesn't make it, I'll try to land it for 1.17.1 in a week from now. Maybe we can have web crypto "done" by the end of the year.

[tarcieri]

I can take a look at getting the rsa crate bumped to the latest versions of pkcs1/pkcs8

[cryptographix]

There are some WPT regressions it seems that I need to be dealt with before merge. (I can do this)

I will try to get this landed for 1.17.0 tomorrow. If it doesn't make it, I'll try to land it for 1.17.1 in a week from now. Maybe we can have web crypto "done" by the end of the year.

@lucacasonato WPT - import_export/ec will only pass the supported cases when #13104 is in, since the tests need exportKey to function. PTAL @ #13104 it should hopefully be in reasonable shape. ;)

[lucacasonato]

import_export/ec will only pass the supported cases when #13104 is in

I am aware. I meant the cases in WebCryptoAPI/sign_verify (I have no idea yet what is going on there).

[cryptographix]

import_export/ec will only pass the supported cases when #13104 is in

I am aware. I meant the cases in WebCryptoAPI/sign_verify (I have no idea yet what is going on there).
@lucacasonato, I think there maybe something strange going on with key-hash combinations. In op_crypto_sign_key ECDSA, the comment states underlying ring::signature crate only supports two algo pairs
// We only support P256-SHA256 & P384-SHA384. These are recommended signature pairs.
// https://briansmith.org/rustdoc/ring/signature/index.html#statics
However, the hash param passed from js is checked (SHA-256 or SHA-384), independent of the curve used.

[lucacasonato]

Yeah, just found that too. This code needs a general rewrite in the future.

I'll clean up this PR now, and then land it.

[lucacasonato]

LGTM, thanks @seanwykes.

[cryptographix]

// id-secp521r1

[cryptographix]

LGTM, thanks @seanwykes.

@lucacasonato - thanks for your help, cleanups and fixes. Learnt a lot about paradigmatic Rust from this exercise.