Re: [PATCH 1/2] mm,vmscan: Kill global shrinker lock.

From: Minchan Kim
Date: Wed Nov 15 2017 - 19:46:30 EST


On Tue, Nov 14, 2017 at 10:28:10PM -0800, Shakeel Butt wrote:
> On Tue, Nov 14, 2017 at 4:56 PM, Minchan Kim <minchan@xxxxxxxxxx> wrote:
> > On Tue, Nov 14, 2017 at 06:37:42AM +0900, Tetsuo Handa wrote:
> >> When shrinker_rwsem was introduced, it was assumed that
> >> register_shrinker()/unregister_shrinker() are really unlikely paths
> >> which are called during initialization and tear down. But nowadays,
> >> register_shrinker()/unregister_shrinker() might be called regularly.
> >> This patch prepares for allowing parallel registration/unregistration
> >> of shrinkers.
> >>
> >> Since do_shrink_slab() can reschedule, we cannot protect shrinker_list
> >> using one RCU section. But using atomic_inc()/atomic_dec() for each
> >> do_shrink_slab() call will not impact so much.
> >>
> >> This patch uses polling loop with short sleep for unregister_shrinker()
> >> rather than wait_on_atomic_t(), for we can save reader's cost (plain
> >> atomic_dec() compared to atomic_dec_and_test()), we can expect that
> >> do_shrink_slab() of unregistering shrinker likely returns shortly, and
> >> we can avoid khungtaskd warnings when do_shrink_slab() of unregistering
> >> shrinker unexpectedly took so long.
> >>
> >> Signed-off-by: Tetsuo Handa <penguin-kernel@xxxxxxxxxxxxxxxxxxx>
> >
> > Before reviewing this patch, can't we solve the problem with more
> > simple way? Like this.
> >
> > Shakeel, What do you think?
> >
>
> Seems simple enough. I will run my test (running fork bomb in one
> memcg and separately time a mount operation) and update if numbers
> differ significantly.

Thanks.

>
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 13d711dd8776..cbb624cb9baa 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -498,6 +498,14 @@ static unsigned long shrink_slab(gfp_t gfp_mask, int nid,
> > sc.nid = 0;
> >
> > freed += do_shrink_slab(&sc, shrinker, nr_scanned, nr_eligible);
> > + /*
> > + * bail out if someone want to register a new shrinker to prevent
> > + * long time stall by parallel ongoing shrinking.
> > + */
> > + if (rwsem_is_contended(&shrinker_rwsem)) {
> > + freed = 1;
>
> freed = freed ?: 1;

Yub.