Re: [PATCH] mm: fix hanging shrinker management on long do_shrink_slab

[Michal Hocko]

On Fri 06-12-19 13:09:53, Dave Chinner wrote:
> [please cc me on future shrinker infrastructure modifications]
> 
> On Mon, Dec 02, 2019 at 07:36:03PM +0300, Andrey Ryabinin wrote:
> > 
> > On 11/30/19 12:45 AM, Pavel Tikhomirov wrote:
> > > We have a problem that shrinker_rwsem can be held for a long time for
> > > read in shrink_slab, at the same time any process which is trying to
> > > manage shrinkers hangs.
> > > 
> > > The shrinker_rwsem is taken in shrink_slab while traversing shrinker_list.
> > > It tries to shrink something on nfs (hard) but nfs server is dead at
> > > these moment already and rpc will never succeed. Generally any shrinker
> > > can take significant time to do_shrink_slab, so it's a bad idea to hold
> > > the list lock here.
> 
> registering/unregistering a shrinker is not a performance critical
> task.

We have had a bug report from production system which stumbled over a
long [u]nmount which was stuck on a shrinker {un}register path waiting for
the lock. This wouldn't be a big deal most of the time but [u]mount were
part of the login session in this case. This was on an older
distribution kernel and e496612c5130 ("mm: do not stall register_shrinker()")
helped a bit but not really for shrinkers that take a lot of time.

> If a shrinker is blocking for a long time, then we need to
> work to fix the shrinker implementation because blocking is a much
> bigger problem than just register/unregister.

Absolutely agreed here.
-- 
Michal Hocko
SUSE Labs