Revision 6e2df0581f569038719cf2bc2b3baa3fcc83cab4 authored by Peter Zijlstra on 08 November 2019, 10:11:52 UTC, committed by Peter Zijlstra on 08 November 2019, 21:34:14 UTC
Commit 67692435c411 ("sched: Rework pick_next_task() slow-path")
inadvertly introduced a race because it changed a previously
unexplored dependency between dropping the rq->lock and
sched_class::put_prev_task().

The comments about dropping rq->lock, in for example
newidle_balance(), only mentions the task being current and ->on_cpu
being set. But when we look at the 'change' pattern (in for example
sched_setnuma()):

	queued = task_on_rq_queued(p); /* p->on_rq == TASK_ON_RQ_QUEUED */
	running = task_current(rq, p); /* rq->curr == p */

	if (queued)
		dequeue_task(...);
	if (running)
		put_prev_task(...);

	/* change task properties */

	if (queued)
		enqueue_task(...);
	if (running)
		set_next_task(...);

It becomes obvious that if we do this after put_prev_task() has
already been called on @p, things go sideways. This is exactly what
the commit in question allows to happen when it does:

	prev->sched_class->put_prev_task(rq, prev, rf);
	if (!rq->nr_running)
		newidle_balance(rq, rf);

The newidle_balance() call will drop rq->lock after we've called
put_prev_task() and that allows the above 'change' pattern to
interleave and mess up the state.

Furthermore, it turns out we lost the RT-pull when we put the last DL
task.

Fix both problems by extracting the balancing from put_prev_task() and
doing a multi-class balance() pass before put_prev_task().

Fixes: 67692435c411 ("sched: Rework pick_next_task() slow-path")
Reported-by: Quentin Perret <qperret@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Quentin Perret <qperret@google.com>
Tested-by: Valentin Schneider <valentin.schneider@arm.com>
1 parent e3b8b6a
History
File Mode Size
bpf
cgroup
configs
debug
dma
events
gcov
irq
livepatch
locking
power
printk
rcu
sched
time
trace
.gitignore -rw-r--r-- 53 bytes
Kconfig.freezer -rw-r--r-- 92 bytes
Kconfig.hz -rw-r--r-- 1.7 KB
Kconfig.locks -rw-r--r-- 4.9 KB
Kconfig.preempt -rw-r--r-- 3.0 KB
Makefile -rw-r--r-- 4.3 KB
acct.c -rw-r--r-- 15.4 KB
async.c -rw-r--r-- 10.2 KB
audit.c -rw-r--r-- 61.4 KB
audit.h -rw-r--r-- 10.4 KB
audit_fsnotify.c -rw-r--r-- 5.6 KB
audit_tree.c -rw-r--r-- 25.7 KB
audit_watch.c -rw-r--r-- 14.1 KB
auditfilter.c -rw-r--r-- 34.0 KB
auditsc.c -rw-r--r-- 69.7 KB
backtracetest.c -rw-r--r-- 1.9 KB
bounds.c -rw-r--r-- 751 bytes
capability.c -rw-r--r-- 14.6 KB
compat.c -rw-r--r-- 10.0 KB
configs.c -rw-r--r-- 2.0 KB
context_tracking.c -rw-r--r-- 6.3 KB
cpu.c -rw-r--r-- 57.4 KB
cpu_pm.c -rw-r--r-- 5.6 KB
crash_core.c -rw-r--r-- 11.4 KB
crash_dump.c -rw-r--r-- 1.3 KB
cred.c -rw-r--r-- 23.3 KB
delayacct.c -rw-r--r-- 4.7 KB
dma.c -rw-r--r-- 3.3 KB
elfcore.c -rw-r--r-- 459 bytes
exec_domain.c -rw-r--r-- 1.1 KB
exit.c -rw-r--r-- 43.0 KB
extable.c -rw-r--r-- 4.3 KB
fail_function.c -rw-r--r-- 6.9 KB
fork.c -rw-r--r-- 69.7 KB
freezer.c -rw-r--r-- 4.4 KB
futex.c -rw-r--r-- 104.3 KB
gen_kheaders.sh -rwxr-xr-x 3.1 KB
groups.c -rw-r--r-- 4.9 KB
hung_task.c -rw-r--r-- 7.3 KB
iomem.c -rw-r--r-- 4.7 KB
irq_work.c -rw-r--r-- 4.7 KB
jump_label.c -rw-r--r-- 20.8 KB
kallsyms.c -rw-r--r-- 17.6 KB
kcmp.c -rw-r--r-- 5.7 KB
kcov.c -rw-r--r-- 11.0 KB
kexec.c -rw-r--r-- 7.9 KB
kexec_core.c -rw-r--r-- 31.1 KB
kexec_elf.c -rw-r--r-- 11.4 KB
kexec_file.c -rw-r--r-- 33.0 KB
kexec_internal.h -rw-r--r-- 924 bytes
kheaders.c -rw-r--r-- 1.6 KB
kmod.c -rw-r--r-- 5.0 KB
kprobes.c -rw-r--r-- 63.3 KB
ksysfs.c -rw-r--r-- 6.3 KB
kthread.c -rw-r--r-- 34.1 KB
latencytop.c -rw-r--r-- 7.2 KB
module-internal.h -rw-r--r-- 782 bytes
module.c -rw-r--r-- 115.5 KB
module_signature.c -rw-r--r-- 1.1 KB
module_signing.c -rw-r--r-- 1.1 KB
notifier.c -rw-r--r-- 16.4 KB
nsproxy.c -rw-r--r-- 6.3 KB
padata.c -rw-r--r-- 26.7 KB
panic.c -rw-r--r-- 17.5 KB
params.c -rw-r--r-- 22.7 KB
pid.c -rw-r--r-- 12.9 KB
pid_namespace.c -rw-r--r-- 11.5 KB
profile.c -rw-r--r-- 14.8 KB
ptrace.c -rw-r--r-- 36.4 KB
range.c -rw-r--r-- 3.0 KB
reboot.c -rw-r--r-- 14.1 KB
relay.c -rw-r--r-- 32.1 KB
resource.c -rw-r--r-- 42.8 KB
rseq.c -rw-r--r-- 9.8 KB
seccomp.c -rw-r--r-- 45.6 KB
signal.c -rw-r--r-- 117.9 KB
smp.c -rw-r--r-- 22.1 KB
smpboot.c -rw-r--r-- 11.7 KB
smpboot.h -rw-r--r-- 640 bytes
softirq.c -rw-r--r-- 17.1 KB
stackleak.c -rw-r--r-- 3.9 KB
stacktrace.c -rw-r--r-- 9.5 KB
stop_machine.c -rw-r--r-- 18.0 KB
sys.c -rw-r--r-- 61.9 KB
sys_ni.c -rw-r--r-- 9.3 KB
sysctl.c -rw-r--r-- 80.5 KB
sysctl_binary.c -rw-r--r-- 50.2 KB
task_work.c -rw-r--r-- 3.1 KB
taskstats.c -rw-r--r-- 15.6 KB
test_kprobes.c -rw-r--r-- 5.9 KB
torture.c -rw-r--r-- 21.0 KB
tracepoint.c -rw-r--r-- 14.9 KB
tsacct.c -rw-r--r-- 4.6 KB
ucount.c -rw-r--r-- 5.5 KB
uid16.c -rw-r--r-- 5.1 KB
uid16.h -rw-r--r-- 442 bytes
umh.c -rw-r--r-- 19.0 KB
up.c -rw-r--r-- 2.3 KB
user-return-notifier.c -rw-r--r-- 1.3 KB
user.c -rw-r--r-- 5.5 KB
user_namespace.c -rw-r--r-- 33.2 KB
utsname.c -rw-r--r-- 3.8 KB
utsname_sysctl.c -rw-r--r-- 3.2 KB
watchdog.c -rw-r--r-- 21.4 KB
watchdog_hld.c -rw-r--r-- 7.7 KB
workqueue.c -rw-r--r-- 163.9 KB
workqueue_internal.h -rw-r--r-- 2.4 KB

back to top