Support ECDiffieHellman on X509Certificate2

[vcsjones]

This adds ECDiffieHellman functionality to X509Certificate2. In addition, this also adds arbitrary SubjectPublicKeyInfo support for PublicKey.

This also makes some under-the-hood changes. Previously, we were always able to know what type of algorithm was used with a given certificate based on the certificate SPKI's OID. In the case of id-ecPublicKey, this is not true anymore. Other plumbing changes include changes to support ImportFromPem on X509Certificate2.

Work left to be done:

• XML documentation
• Windows PFX tests

Closes #413

[Dotnet-GitSync-Bot]

Note regarding the new-api-needs-documentation label:

This serves as a reminder for when your PR is modifying a ref *.cs file and adding/modifying public APIs, to please make sure the API implementation in the src *.cs file is documented with triple slash comments, so the PR reviewers can sign off that change.

[ghost]

Tagging subscribers to this area: @bartonjs, @vcsjones, @krwq, @jeffhandley
See info in area-owners.md if you want to be subscribed.

[vcsjones]

Hm. CNG is behaving differently between Win10 and Win8.1/7. I suppose this is related to CNG's strong opinion that EC keys are either ECDH or ECDSA.

[vcsjones]

@bartonjs I think I fixed the Windows 7 / 8.1 key import problem. Does the change in 2d4fda5 make sense?

[bartonjs]

Does the change in 2d4fda5 make sense?

Depends on your definition of "make sense" 😄. Apparently the private keys default to ECDH but the public keys were defaulting to ECDSA, and whatever lookup it did in the CryptFindOID universe returned multiple hits, ECDSA first... and that flag says "if something can't encrypt, I don't want it".. and while ECDH can't encrypt, it's purpose is eventually encryption, so it's marked as "encryption". Does any of that make sense? Not especially 😄.

Does the code read cleanly? Sure. Does it make tests pass? Apparently 😄.

[vcsjones]

@bartonjs thanks as always for the thoughtful feedback. I think I have addressed all of the points you have made so far.

[vcsjones]

Guessing macOS is going to blow up with a bad DSA at usage time, not import time. No idea what what it is actually doing with an 8-bit DSA key.

[bartonjs]

Guessing macOS is going to blow up with a bad DSA at usage time, not import time. No idea what what it is actually doing with an 8-bit DSA key.

Well, p is 0, so that's a 0-bit key, no? 😄.

I'd probably delete one of the integers out of the DSA-Params struct so it doesn't decode correctly, making the whole "bad DSA" SPKI

30 1B 30 13 06 07 2A 86 48 CE 38 04 01 30 08 02 
01 00 02 03 00 FF FF 03 04 00 02 01 03 

[bartonjs]

Once again: Thanks, @vcsjones!