Implement SHA-512 truncated (224 and 256 bits)

[allanjude]

Summary

This implements SHA-512/256, which generates a 256 bit hash, but using SHA-512 so is faster than SHA-256 on 64bit platforms.

The main goal of this implementation is to enable support for this faster hashing algorithm in ZFS

This is the follow on to r292782

Test Plan

Verified against: http://csrc.nist.gov/groups/ST/toolkit/documents/Examples/SHA512_256.pdf

[bcr]

Found a typo.

sbin/md5/md5.1

87 ↗

(On Diff #15494)

I this think should be "a version".

[allanjude]

Fix typo spotted by bcr@

[cem]

sys/crypto/sha2/sha512c.c

331 ↗

(On Diff #15780)

what's "so are skipped" mean?

352–356 ↗

(On Diff #15780)

It seems like these aren't really necessary. I'd suggest removing them and adding a comment where they were in the header to use the SHA512_Update() routine.

359 ↗

(On Diff #15780)

Use digest[static ...] here :).

369 ↗

(On Diff #15780)

don't need to cast here.

391–395 ↗

(On Diff #15780)

Same quibble here.

398 ↗

(On Diff #15780)

Use digest[static DIGEST_LEN] :)

408 ↗

(On Diff #15780)

same as above

sys/crypto/sha2/sha512t.h

106 ↗

(On Diff #15780)

[static DIGEST_LEN]

115 ↗

(On Diff #15780)

[static DIGEST_LEN]

[allanjude]

sys/crypto/sha2/sha512c.c

331 ↗

(On Diff #15780)

I didn't define the _transform function because it just uses the regular SHA512 one. I'll reword this

352–356 ↗

(On Diff #15780)

Userland applications that implement this algo expect this symbol to exist in libmd.

359 ↗

(On Diff #15780)

ok

369 ↗

(On Diff #15780)

right

[allanjude]

Update with feedback from cem@

[cem]

sys/crypto/sha2/sha512c.c

331 ↗

(On Diff #16958)

still needs rewording

369 ↗

(On Diff #16958)

cast still present

408 ↗

(On Diff #16958)

This extra cast got missed :)

352–356 ↗

(On Diff #15780)

It seems like userland-only routines and declarations should live in userland, like the _File/_Data routines. But whatever.

[allanjude]

Parts of the previous update did not get submitted