Add support for parsing ASN.1 ECDSA key pairs without the PKCS#8 metadata.

[farcaller]

This adds support for EC PRIVATE KEY, i.e. non pkcs8-wrapped keys.

[farcaller]

Tentative ping.

[cpu]

@briansmith Would you be open to this feature if it were rebased on main? We have a bit of a hack upstream in rustls patching over this and it might be nice to drop it one day.

[briansmith]

@cpu Yes, I would. Would you do RSA and EdDSA too?

[farcaller]

I don’t mind adding the others too. Let me see if I remember how it works, was a while since I touched this code.

[cpu]

Great thank you both. Happy to support as appropriate.

[farcaller]

@briansmith do you know any good producers of RSA keys? I thought ssh ones would fit but they are actually in a different format.

[briansmith]

@briansmith do you know any good producers of RSA keys? I thought ssh ones would fit but they are actually in a different format.

Does this work? https://gist.github.com/briansmith/2ee42439923d8e65a266994d0f70180b#extracting-an-rsa-public-key-from-the-private-key-without-the-subjectpublickeyinfo-metadata

[farcaller]

Sorry, I was out of free cycles the previous week to follow up on this.

@briansmith do you mind those two being decoupled, given how rebase was pretty smooth? I think the ec keys can go forward straight on and unblock rustls, and I'll work on the RSA keys in a separate PR then.

[briansmith]

@farcaller I looked more into how Rustls is using the existing APIs. I think we should "just" refactor the ring PKCS#8 API so that it works more like what Rustls needs, to avoid Rustls needing to jump through hoops. Please see #1889.