Re: Kernel falls apart under light memory pressure (i.e. linking vmlinux)

[Minchan Kim <minchan.kim@gmail.com>]

On Fri, May 20, 2011 at 12:01:12PM -0400, Andrew Lutomirski wrote:
> On Fri, May 20, 2011 at 11:33 AM, Minchan Kim <minchan.kim@gmail.com> wrote:
> 
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 8bfd450..a5c01e9 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -1430,7 +1430,10 @@ shrink_inactive_list(unsigned long nr_to_scan, struct zone *zone,
> >
> >        /* Check if we should syncronously wait for writeback */
> >        if (should_reclaim_stall(nr_taken, nr_reclaimed, priority, sc)) {
> > +               unsigned long nr_active;
> >                set_reclaim_mode(priority, sc, true);
> > +               nr_active = clear_active_flags(&page_list, NULL);
> > +               count_vm_events(PGDEACTIVATE, nr_active);
> >                nr_reclaimed += shrink_page_list(&page_list, zone, sc);
> >        }
> >
> > --
> 
> I'm now running that patch *without* the pgdat_balanced fix or the
> need_resched check.  The VM_BUG_ON doesn't happen but I still get

Please forget need_resched.
Instead of it, could you test shrink_slab patch with !pgdat_balanced?

@@ -231,8 +231,11 @@ unsigned long shrink_slab(struct shrink_control *shrink,
       if (scanned == 0)
               scanned = SWAP_CLUSTER_MAX;

-       if (!down_read_trylock(&shrinker_rwsem))
-               return 1;       /* Assume we'll be able to shrink next time */
+       if (!down_read_trylock(&shrinker_rwsem)) {
+               /* Assume we'll be able to shrink next time */
+               ret = 1;
+               goto out;
+       }

       list_for_each_entry(shrinker, &shrinker_list, list) {
               unsigned long long delta;
@@ -286,6 +289,8 @@ unsigned long shrink_slab(struct shrink_control *shrink,
               shrinker->nr += total_scan;
       }
       up_read(&shrinker_rwsem);
+out:
+       cond_resched();
       return ret;
 }

> incorrect OOM kills.
> 
> However, if I replace the check with:
> 
> 	if (false &&should_reclaim_stall(nr_taken, nr_reclaimed, priority, sc)) {
> 
> then my system lags under bad memory pressure but recovers without
> OOMs or oopses.

I agree you can see OOM but oops? Did you see any oops?

> 
> Is that expected?


No..  :(

It's totally opposite.
That routine is for getting the memory althought we lose latency
It's another issue. :(

> 
> --Andy
> 
> > 1.7.1
> >
> > --
> > Kind regards,
> > Minchan Kim
> >

-- 
Kind regards,
Minchan Kim