sched: spit sched_ap_entry() out of sched_throw()
ClosedPublic
Actions

Authored by kevans on Nov 4 2021, 12:38 AM.

Details

Reviewers

markj
kib
manu

Commits

rG6a8ea6d17498: sched: split sched_ap_entry() out of sched_throw()

Summary

sched_throw() can no longer take a NULL thread, APs enter through
sched_ap_entry() instead. This completely removes branching in the
common case. No functional change intended.

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

kevans created this revision.Nov 4 2021, 12:38 AM

Herald added a reviewer: manu. · View Herald TranscriptNov 4 2021, 12:38 AM

Herald added subscribers: jhibbits, emaste, andrew, imp. · View Herald Transcript

kevans requested review of this revision.Nov 4 2021, 12:38 AM

Harbormaster completed remote builds in B42573: Diff 97994.Nov 4 2021, 12:39 AM

kib accepted this revision.Nov 4 2021, 12:44 AM

kib added inline comments.

sys/kern/sched_4bsd.c
1708	I would group this MPASS together with td != NULL

This revision is now accepted and ready to land.Nov 4 2021, 12:44 AM

markj accepted this revision.Nov 4 2021, 12:49 PM

markj added inline comments.

sys/kern/sched_ule.c
2987	Maybe explain why: it's important to not enable interrupts until the CPU is running in a threaded context.
sys/sys/sched.h
105	Keep them sorted?

kevans added inline comments.Nov 4 2021, 4:22 PM

sys/kern/sched_ule.c

2987

Sure, how about (with wrapping fixed):

/*
 * sched_throw_grab() chooses a thread from the queue to switch to
 * next.  It returns with the tdq lock dropped in a spinlock section to
 * keep interrupts disabled until the CPU is running in a proper threaded
 * context.
 */

markj added inline comments.Nov 5 2021, 1:53 PM