Re: mm, vmstat: kernel BUG at mm/vmstat.c:1408!

[Christoph Lameter]

On Thu, 21 Jan 2016, Michal Hocko wrote:

> It goes like this:
> CPU0:						CPU1
> vmstat_update
>   cpumask_test_and_set_cpu (0->1)
> [...]
> 						vmstat_shepherd
> <enter idle>					  cpumask_test_and_clear_cpu(CPU0) (1->0)
> quiet_vmstat
>   cpumask_test_and_set_cpu (0->1)
>   						  queue_delayed_work_on(CPU0)
> refresh_cpu_vm_stats()
> [...]
> vmstat_update
>   nothing_to_do
>   cpumask_test_and_set_cpu (1->1)
>   VM_BUG_ON
>
> Or am I missing something?

Ok then the following should fix it:



Subject: vmstat: Queue work before clearing cpu_stat_off

There is a race between vmstat_shepherd and quiet_vmstat() because
the responsibility for checking for counter updates changes depending
on the state of teh bit in cpu_stat_off. So queue the work before
changing state of the bit in vmstat_shepherd. That way quiet_vmstat
is guaranteed to remove the work request when clearing the bit and the
bug in vmstat_update wont trigger anymore.

Signed-off-by: Christoph Lameter <cl@xxxxxxxxx>

Index: linux/mm/vmstat.c
===================================================================
--- linux.orig/mm/vmstat.c
+++ linux/mm/vmstat.c
@@ -1480,12 +1480,14 @@ static void vmstat_shepherd(struct work_
 	get_online_cpus();
 	/* Check processors whose vmstat worker threads have been disabled */
 	for_each_cpu(cpu, cpu_stat_off)
-		if (need_update(cpu) &&
-			cpumask_test_and_clear_cpu(cpu, cpu_stat_off))
+		if (need_update(cpu)) {

 			queue_delayed_work_on(cpu, vmstat_wq,
 				&per_cpu(vmstat_work, cpu), 0);

+			cpumask_clear_cpu(smp_processor_id(), cpu_stat_off);
+		}
+
 	put_online_cpus();

 	schedule_delayed_work(&shepherd,