kern_domainset: Split domainset validation logic into a separate function

[bnovkov]

Summary

This patch splits the validation and struct domainset-filling logic from kern_cpuset_setdomain into a separate function - domainset_populate.
This function's main use is to validate user-provided domainset(9) policies and populate a struct domainset before handing it off to domainset_create. No functional change intended.

[markj]

share/man/man9/domainset.9

25

157

161

sys/kern/kern_cpuset.c

2412–2413

Can mask be a pointer to const?

2498

It feels a bit weird to include this last section here--is it appropriate for a generic helper function, or is it really a policy of the syscall layer?

[bnovkov]

Address @markj 's comments.

[bnovkov]

sys/kern/kern_cpuset.c

2412–2413

It can, done!

2498

Hm right, it's a bit unclear where the cutoff point should be.

I guess it'd be useful to give other consumers of this interface a chance to return an error when faced with an empty domainset, the only issue with this is that it would require making domainset_vm_empty (and perhaps domainset_copy) public functions. Does that sound reasonable?

[markj]

sys/kern/kern_cpuset.c

2434–2435

[annoying style nitpicking] This function consistently omitted braces for single statements, but now it has a mix of both styles. Either style is ok, but it'd be nice to be consistent.

2498

I think making those symbols public is fine, yes.

[bnovkov]

Fix style issues and make domainset_empty_vm public.

[ziaee]

share/man/man9/domainset.9

29

Right?

[markj]

sys/sys/domainset.h

116

It'd be nice to have a comment above this function, like the others do.