On Wed, 2016-02-03 at 11:24 -0500, Tejun Heo wrote:
> On Wed, Feb 03, 2016 at 01:28:56PM +0100, Michal Hocko wrote:
> > > The CPU was 168, and that one was offlined in the meantime. So
> > > __queue_work fails at:
> > > if (!(wq->flags & WQ_UNBOUND))
> > > pwq = per_cpu_ptr(wq->cpu_pwqs, cpu);
> > > else
> > > pwq = unbound_pwq_by_node(wq, cpu_to_node(cpu));
> > > ^^^ ^^^^ NODE is -1
> > > \ pwq is NULL
> > >
> > > if (last_pool && last_pool != pwq->pool) { <--- BOOM
>
> So, the proper fix here is keeping cpu <-> node mapping stable across
> cpu on/offlining which has been being worked on for a long time now.
> The patchst is pending and it fixes other issues too.
>
> > So I think 874bbfe600a6 is really bogus. It should be reverted. We
> > already have a proper fix for vmstat 176bed1de5bf ("vmstat:
> > explicitly
> > schedule per-cpu work on the CPU we need it to run on"). This which
> > should be used for the stable trees as a replacement.
>
> It's not bogus. We can't flip a property that has been guaranteed
> without any provision for verification. Why do you think vmstat blow
> up in the first place? vmstat would be the canary case as it runs
> frequently on all systems. It's exactly the sign that we can't break
> this guarantee willy-nilly.
If the intent of the below is to fulfill a guarantee...
+ /* timer isn't guaranteed to run in this cpu, record earlier */
+ if (cpu == WORK_CPU_UNBOUND)
+ cpu = raw_smp_processor_id();
dwork->cpu = cpu;
timer->expires = jiffies + delay;
- if (unlikely(cpu != WORK_CPU_UNBOUND))
- add_timer_on(timer, cpu);
- else
- add_timer(timer);
+ add_timer_on(timer, cpu);
...it appears to be incomplete. Hotplug aside, when adding a timer
with the expectation that it stay put, should it not also be pinned?
-Mike
