Move implementation of ioctls into kern_*() functions.

[brooks]

Summary

Move locks from outside ioctl to the individual implementations.

This is the first step of changing the implementations to act on a
kernel-internal request struct rather than on struct md_ioctl and to
removing the use of kern_ioctl in mountroot.

[cem]

sys/dev/md/md.c

1776–1777

else break? Doesn't seem to be any point in continuing the loop once the output buffer is full. That can wait for a separate commit though.

[kib]

With this structure, I would expect the lock to be pushed into individual ioctl routines, in fact.

sys/dev/md/md.c

1747

Perhaps use the opportunity and return EFAULT if this copyout failed ?

[brooks]

I'll update following @kib's suggestions shortly.

sys/dev/md/md.c

1776–1777

The continued looping is to finish counting which is the ABI (both a count and a terminator, historically the terminator only existed if the set didn't fit). The ABI is broken by design and prior to r330880 the loop was a kernel buffer overflow.

[cem]

sys/dev/md/md.c

1776–1777

If that's what it's for, it's broken now. We won't ever increment i again. Probably need else i++; or similar.

[brooks]

sys/dev/md/md.c

1776–1777

Oops, fixed in HEAD, will fix here.

[brooks]

• Rebase
• Return immediatly if copyout of md_label fails.
• Move locking into command implementations.
• Unconditionally increment count to preserve ABI.