Make Linux sched_getaffinity(2) and sched_setaffinity(2) accept short masks
AbandonedPublic
Actions

Authored by dchagin on Jul 16 2020, 3:35 PM.

Details

Reviewers

kib
trasz

Group Reviewers

Linux Emulation

Summary

Make Linux sched_getaffinity(2)/sched_setaffinity() , and also our native cpuset_getaffinity(2)/cpuset_setaffinity(2), accept short masks. Fixes Linux Python regression test, and possibly some real applications.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 32367
Build 29848: arc lint + arc unit

Event Timeline

trasz created this revision.Jul 16 2020, 3:35 PM

Herald added a subscriber: imp. · View Herald TranscriptJul 16 2020, 3:35 PM

trasz requested review of this revision.Jul 16 2020, 3:35 PM

Harbormaster completed remote builds in B32360: Diff 74526.Jul 16 2020, 3:35 PM

trasz edited the summary of this revision. (Show Details)Jul 16 2020, 3:38 PM

trasz added reviewers: kib, Linux Emulation.

emaste added a subscriber: emaste.Jul 16 2020, 3:52 PM

emaste added inline comments.

sys/kern/kern_cpuset.c
1884	we still copyin cpusetsize?

kib added inline comments.Jul 16 2020, 6:30 PM

sys/compat/linux/linux_misc.c
2077	`cpuset *) args->` remove space
sys/kern/kern_cpuset.c
1877	cpusetsize cannot be < 0.
1882–1883	How can cpusetsize be >= sizeof(cpuset_t) ?
1884	Don't you need to initialize not-copied in bits ?

trasz added inline comments.Jul 16 2020, 7:12 PM

sys/kern/kern_cpuset.c
1882–1883	No idea, but the existing code (`if (cpusetsize < sizeof(cpuset_t) \|\| cpusetsize > CPU_MAXSIZE / NBBY)` suggests it can be variable in size?
1884	We copy as much data as the userspace is willing to give us. The rest is initialized to zero just above (M_ZERO).

Also handle sched_getaffinity().

Harbormaster completed remote builds in B32367: Diff 74537.Jul 16 2020, 7:42 PM

trasz retitled this revision from Make Linux kern_cpuset(4) accept short masks to Make Linux sched_getaffinity(2) and sched_setaffinity(2) accept short masks.Jul 16 2020, 7:45 PM

trasz edited the summary of this revision. (Show Details)

trasz marked an inline comment as done.

(Tinderboxed.)

kib added inline comments.Jul 24 2020, 7:42 PM

sys/kern/kern_cpuset.c
1877	This still not handled. Also, CPU_MAXSIZE in your formula should really be CPU_SETSIZE.
1882–1883	Look at the definitions of cpuset_t and CPU_MAXSIZE/CPU_SETSIZE. After the check at the start of the function, it is never can be larger then cpuset_t. In fact it is shorter.