mm, slab: reschedule cache_reap() on the same CPU
authorVlastimil Babka <vbabka@suse.cz>
Fri, 13 Apr 2018 22:35:38 +0000 (15:35 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 14 Apr 2018 00:10:27 +0000 (17:10 -0700)
cache_reap() is initially scheduled in start_cpu_timer() via
schedule_delayed_work_on(). But then the next iterations are scheduled
via schedule_delayed_work(), i.e. using WORK_CPU_UNBOUND.

Thus since commit ef557180447f ("workqueue: schedule WORK_CPU_UNBOUND
work on wq_unbound_cpumask CPUs") there is no guarantee the future
iterations will run on the originally intended cpu, although it's still
preferred.  I was able to demonstrate this with
/sys/module/workqueue/parameters/debug_force_rr_cpu.  IIUC, it may also
happen due to migrating timers in nohz context.  As a result, some cpu's
would be calling cache_reap() more frequently and others never.

This patch uses schedule_delayed_work_on() with the current cpu when
scheduling the next iteration.

Link: http://lkml.kernel.org/r/20180411070007.32225-1-vbabka@suse.cz
Fixes: ef557180447f ("workqueue: schedule WORK_CPU_UNBOUND work on wq_unbound_cpumask CPUs")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Pekka Enberg <penberg@kernel.org>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Stephen Boyd <sboyd@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/slab.c

index e3a9b8e2330642e98755ad12b6b722957fbea78c..2f308253c3d7689c677dbdc0b8dd7a686689bb7b 100644 (file)
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -4086,7 +4086,8 @@ next:
        next_reap_node();
 out:
        /* Set up the next iteration */
-       schedule_delayed_work(work, round_jiffies_relative(REAPTIMEOUT_AC));
+       schedule_delayed_work_on(smp_processor_id(), work,
+                               round_jiffies_relative(REAPTIMEOUT_AC));
 }
 
 void get_slabinfo(struct kmem_cache *cachep, struct slabinfo *sinfo)