Messages in this thread | | | From | Mike Waychison <> | Date | Wed, 14 Dec 2011 07:37:22 -0800 | Subject | Re: [PATCH] mm: Fix kswapd livelock on single core, no preempt kernel |
| |
On Wed, Dec 14, 2011 at 4:20 AM, Mel Gorman <mgorman@suse.de> wrote: > On Tue, Dec 13, 2011 at 08:36:43PM -0800, Mike Waychison wrote: >> FYI, this was seen with a 2.6.39-based kernel with no-numa, no-memcg >> and swap-enabled. >> > > If this is 2.6.39, can you try applying the commit > [f06590bd: mm: vmscan: correctly check if reclaimer should schedule during shrink_slab] > > There have been a few fixes around kswapd hogging the CPU since 2.6.39.
In this particular case, I didn't see any problem acquiring shrinker_rwsem (the shrinkers should up in the cpu profile I gathered). I think this patch would fix my issue though as it happens to drop in a cond_resched() into the path. It isn't obvious that this cond_resched() really belongs in shrink_slab() though. Thanks :)
> > -- > Mel Gorman > SUSE Labs
| |