kern: physmem: return errors in region inclusion/exclusion

[kevans]

Summary

These have some failure modes, acknowledge that. If we run past the
end of hwregions, the caller should be able to detect that before it
potentially becomes a panic() later when attempting to populate
phys_avail.

[kib]

sys/kern/subr_physmem.c

441

I recommend to not use ENOMEM for any allocation API return code, except for true 'out of memory' conditions. There we are out of items, and users often cannot tell the difference.

463

Was the intent to change this panic to some error return?

[kevans]

sys/kern/subr_physmem.c

441

Hmm... E2BIG, maybe? ERANGE could make sense, but I already used that above to indicate out-of-bounds range.

463

Whoops, yeah, that one should also toss back an error and let the caller panic if they want to.

[kevans]

ENOMEM -> ENOBIG, don't panic in exclusion case

[kib]

sys/kern/subr_physmem.c

404

What would caller do in case of this error? Is it supposed to somehow ignore it?

405

There seems to be two spaces before '='

[kevans]

sys/kern/subr_physmem.c

404

The caller can't really do anything about it and they typically won't do anything to sanity check what's been added to phys_avail, so I suppose this should just be a silent return 0.

[kevans]

Don't return error for short page @ 0x0, fix spacing