On Tue 05-09-17 10:19:13, Thomas Gleixner wrote:
> On Mon, 4 Sep 2017, Michal Hocko wrote:
>
> > Thomas, Johannes,
> > could you double check my thinking here? I will repost the patch to
> > Andrew if you are OK with this.
> > > + /*
> > > + * The only protection from memory hotplug vs. drain_stock races is
> > > + * that we always operate on local CPU stock here with IRQ disabled
> > > + */
> > > local_irq_save(flags);
> > >
> > > stock = this_cpu_ptr(&memcg_stock);
> > > @@ -1807,26 +1811,27 @@ static void drain_all_stock(struct mem_cgroup
> > > *root_memcg)
> > > if (!mutex_trylock(&percpu_charge_mutex))
> > > return;
> > > /* Notify other cpus that system-wide "drain" is running */
> > > - get_online_cpus();
> > > curcpu = get_cpu();
>
> The problem here is that this does only protect you against a CPU being
> unplugged, but not against a CPU coming online concurrently.
Yes but same as the drain_all_pages we do not have any cpu up specific
intialization so there is no specific action to race against AFAICS.
> I have no idea
> whether that might be a problem, but at least you should put a comment in
> which explains why it is not.
What about this?
---
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 5c70f47abb3d..ff9b0979ccc3 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -1810,7 +1810,12 @@ static void drain_all_stock(struct mem_cgroup
*root_memcg)
/* If someone's already draining, avoid adding running more workers. */
if (!mutex_trylock(&percpu_charge_mutex))
return;
- /* Notify other cpus that system-wide "drain" is running */
+ /*
+ * Notify other cpus that system-wide "drain" is running
+ * We do not care about races with the cpu hotplug because cpu down
+ * as well as workers from this path always operate on the local
+ * per-cpu data. CPU up doesn't touch memcg_stock at all.
+ */
curcpu = get_cpu();
for_each_online_cpu(cpu) {
struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
--
Michal Hocko
SUSE Labs
