twx-linux/include
Peter Zijlstra f97bb5272d sched: Fix data-race in wakeup
Mel reported that on some ARM64 platforms loadavg goes bananas and
Will tracked it down to the following race:

  CPU0					CPU1

  schedule()
    prev->sched_contributes_to_load = X;
    deactivate_task(prev);

					try_to_wake_up()
					  if (p->on_rq &&) // false
					  if (smp_load_acquire(&p->on_cpu) && // true
					      ttwu_queue_wakelist())
					        p->sched_remote_wakeup = Y;

    smp_store_release(prev->on_cpu, 0);

where both p->sched_contributes_to_load and p->sched_remote_wakeup are
in the same word, and thus the stores X and Y race (and can clobber
one another's data).

Whereas prior to commit c6e7bd7afaeb ("sched/core: Optimize ttwu()
spinning on p->on_cpu") the p->on_cpu handoff serialized access to
p->sched_remote_wakeup (just as it still does with
p->sched_contributes_to_load) that commit broke that by calling
ttwu_queue_wakelist() with p->on_cpu != 0.

However, due to

  p->XXX = X			ttwu()
  schedule()			  if (p->on_rq && ...) // false
    smp_mb__after_spinlock()	  if (smp_load_acquire(&p->on_cpu) &&
    deactivate_task()		      ttwu_queue_wakelist())
      p->on_rq = 0;		        p->sched_remote_wakeup = Y;

We can be sure any 'current' store is complete and 'current' is
guaranteed asleep. Therefore we can move p->sched_remote_wakeup into
the current flags word.

Note: while the observed failure was loadavg accounting gone wrong due
to ttwu() cobbering p->sched_contributes_to_load, the reverse problem
is also possible where schedule() clobbers p->sched_remote_wakeup,
this could result in enqueue_entity() wrecking ->vruntime and causing
scheduling artifacts.

Fixes: c6e7bd7afaeb ("sched/core: Optimize ttwu() spinning on p->on_cpu")
Reported-by: Mel Gorman <mgorman@techsingularity.net>
Debugged-by: Will Deacon <will@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20201117083016.GK3121392@hirez.programming.kicks-ass.net
2020-11-17 13:15:27 +01:00
..
acpi
asm-generic This tree introduces static_call(), which is the idea of static_branch() 2020-10-12 13:58:15 -07:00
clocksource clocksource/drivers/sp804: Remove unused sp804_timer_disable() and timer-sp804.h 2020-09-24 10:51:04 +02:00
crypto
drm drm-misc-fixes for v5.9: 2020-10-06 12:38:28 +10:00
dt-bindings
keys
kunit
kvm KVM: arm64: pmu: Make overflow handler NMI safe 2020-09-28 19:00:17 +01:00
linux sched: Fix data-race in wakeup 2020-11-17 13:15:27 +01:00
math-emu
media media: media/v4l2: remove V4L2_FLAG_MEMORY_NON_CONSISTENT flag 2020-09-14 15:28:06 +02:00
memory
misc
net netlink: fix policy dump leak 2020-10-02 13:00:38 -07:00
pcmcia
ras
rdma
scsi
soc net: mscc: ocelot: fix fields offset in SG_CONFIG_REG_3 2020-09-24 20:00:40 -07:00
sound ASoC: core: Do not cleanup uninitialized dais on soc_pcm_open failure 2020-09-07 12:40:24 +01:00
target
trace Clean up the paravirt code after the removal of 32-bit Xen PV support. 2020-10-12 15:15:24 -07:00
uapi Scheduler changes for v5.10: 2020-10-12 12:56:01 -07:00
vdso
video
xen arm/arm64: xen: Fix to convert percpu address to gfn correctly 2020-10-07 07:08:43 +02:00