Replace sys/crypto/sha2/sha2.c with lib/libmd/sha512c.c

[allanjude]

Summary

cperciva's libmd implementation is 5-30% faster (also seen here: https://svnweb.freebsd.org/changeset/base/r263218 )

cperciva's implementation was lacking SHA-384 which I implemented
Extend sbin/md5 to create sha384(1)

In a separate commit, I will also add SHA-512/256 (required by new ZFS features)

Chase dependancies on sys/crypto/sha2/sha2.{c,h} and replace them with sha512{c.c,.h}

Will extend reviewers to include secteam@ once the first round of issues are dealt with

Test Plan

benchmark results: http://www.allanjude.com/bsd/sha512_results.txt

The output of the new SHA-384 implementation were verified against openssl sha384 and http://csrc.nist.gov/groups/ST/toolkit/documents/Examples/SHA384.pdf

[jmg]

Just a few comments.

lib/libmd/Makefile

40

is this tab/space matching others or is this just a phab misformating issue?

sys/conf/files

559

I don't see bde using SHA256, we should probably remove it, though maybe as a different commit?

Looks like the module also compiles it. A quick compile of the module does look like it uses it.

561

I don't think you need zfs here yet. I'd add ZFS when you finally enable sha512/256 support.

[allanjude]

lib/libmd/Makefile

40

i put a tab when I shouldn't have, fixed in my local copy, will update with other changes

sys/conf/files

559

yeah, that was already there, I didn't change it. geom_eli uses both 256 and 512, should it be listed in both places?

561

the zfs SRCS= line references both, and again, I just updated sha2.c -> sha512c.c (and changed the order to make sense). But yeah, I could remove ZFS here and readd it in the next commit.

[allanjude]

Minor updates to address feedback from jmg@

[jmg]

sys/conf/files

559

bde != eli.. They are different, so commenting that eli uses 256 and 512 is unrelated to my bde comment.

561

Since you're adding the line fresh, it's best to keep it minimal.

[allanjude]

sys/conf/files

559

that should have been a separate comment.

Unrelated to geom_bde:

geom_eli uses both SHA256 and SHA512, should it be listed in this file?

[allanjude]

Update the diff to put sha384 prototypes in their own file as requested by cperciva

[cperciva]

Looks good to me.

[delphij]

Approved by so, too. Sorry for the delay and thanks for working on this.