Make sys/geom/eli more easily reusable in userspace

[allanjude]

Summary

• Break the HMAC parts of geli out into a separate file, so they can be used without requiring openssl or the entire crypto framework
• some ifdef _KERNEL's to make remaining stuff work in userland
• Need a better resolution for struct g_eli_softc

[allanjude]

Add change to sys/conf/files required for when GEOM_ELI is compiled into the kernel (found with LINT kernels during universe build)

[delphij]

Thanks very much for working on this! It looks good overall except the softc definition (I think it would be better if we just ifdef the software crypto portion out, this way we don't have two places to define the same structure, making it easier for future maintenance).

sys/geom/eli/g_eli.h

211–239

Usually ! _KERNEL for consistency.

213

Can the software portion be #ifdef'ed instead?

[allanjude]

sys/geom/eli/g_eli.h

213

not sure why I didn't do it this way originally. At first I was cutting too much stuff out, like struct mtx, and RB_HEAD, to avoid those includes.

[allanjude]

do struct g_eli_softc properly

[delphij]

Looks good to me now, thanks!

[pjd]

It looks good overall, but please be sure you run GELI regression tests. Thanks!

sys/geom/eli/g_eli.h

178

Maybe we could just use 'bool' now?

[allanjude]

Add missing reference to g_eli_hmac.c to /lib/geom/geom_eli.so (sbin/geom/class/eli)

[allanjude]

With that fix, it now passed the geom_eli regression suite

Files=20 Tests=14996
Result: PASS

[allanjude]

Instead of typedef'ing boolean_t in the !_KERNEL case, just use 'int' in that case