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
LSM
acpi
aoe
auxdisplay
blockdev
cgroup-v1
cifs
device-mapper
gpio
hw-vuln
kdump
laptops
mm
namespaces
perf
pm
sysctl
wimax
README.rst -rw-r--r-- 18.0 KB
bcache.rst -rw-r--r-- 23.7 KB
binderfs.rst -rw-r--r-- 2.9 KB
binfmt-misc.rst -rw-r--r-- 7.2 KB
braille-console.rst -rw-r--r-- 1.7 KB
btmrvl.rst -rw-r--r-- 2.7 KB
bug-bisect.rst -rw-r--r-- 2.1 KB
bug-hunting.rst -rw-r--r-- 15.0 KB
cgroup-v2.rst -rw-r--r-- 94.0 KB
clearing-warn-once.rst -rw-r--r-- 275 bytes
cpu-load.rst -rw-r--r-- 3.0 KB
cputopology.rst -rw-r--r-- 5.7 KB
devices.rst -rw-r--r-- 11.9 KB
devices.txt -rw-r--r-- 104.6 KB
dynamic-debug-howto.rst -rw-r--r-- 13.0 KB
efi-stub.rst -rw-r--r-- 3.8 KB
ext4.rst -rw-r--r-- 26.8 KB
highuid.rst -rw-r--r-- 2.6 KB
hw_random.rst -rw-r--r-- 3.8 KB
index.rst -rw-r--r-- 2.2 KB
init.rst -rw-r--r-- 2.6 KB
initrd.rst -rw-r--r-- 14.3 KB
iostats.rst -rw-r--r-- 8.9 KB
java.rst -rw-r--r-- 10.9 KB
jfs.rst -rw-r--r-- 2.6 KB
kernel-parameters.rst -rw-r--r-- 8.0 KB
kernel-parameters.txt -rw-r--r-- 188.1 KB
kernel-per-CPU-kthreads.rst -rw-r--r-- 13.3 KB
lcd-panel-cgram.rst -rw-r--r-- 1.2 KB
ldm.rst -rw-r--r-- 4.6 KB
lockup-watchdogs.rst -rw-r--r-- 4.1 KB
md.rst -rw-r--r-- 27.0 KB
module-signing.rst -rw-r--r-- 11.1 KB
mono.rst -rw-r--r-- 2.5 KB
numastat.rst -rw-r--r-- 1.0 KB
parport.rst -rw-r--r-- 9.3 KB
perf-security.rst -rw-r--r-- 10.3 KB
pnp.rst -rw-r--r-- 7.0 KB
ramoops.rst -rw-r--r-- 5.7 KB
rapidio.rst -rw-r--r-- 2.2 KB
ras.rst -rw-r--r-- 41.7 KB
reporting-bugs.rst -rw-r--r-- 7.5 KB
rtc.rst -rw-r--r-- 7.0 KB
security-bugs.rst -rw-r--r-- 3.9 KB
serial-console.rst -rw-r--r-- 4.3 KB
svga.rst -rw-r--r-- 11.8 KB
sysfs-rules.rst -rw-r--r-- 9.4 KB
sysrq.rst -rw-r--r-- 12.1 KB
tainted-kernels.rst -rw-r--r-- 7.6 KB
thunderbolt.rst -rw-r--r-- 11.3 KB
ufs.rst -rw-r--r-- 1.3 KB
unicode.rst -rw-r--r-- 7.1 KB
vga-softcursor.rst -rw-r--r-- 1.9 KB
video-output.rst -rw-r--r-- 1.1 KB
xfs.rst -rw-r--r-- 17.8 KB

README.rst

back to top