Re: [patch resend] mm: page_alloc: fix zone allocation fairness on UP

[Johannes Weiner <hannes@cmpxchg.org>]

On Wed, Sep 10, 2014 at 07:32:20AM +0300, Leon Romanovsky wrote:
> Hi Johaness,
> 
> 
> On Tue, Sep 9, 2014 at 4:15 PM, Johannes Weiner <hannes@cmpxchg.org> wrote:
> 
> > The zone allocation batches can easily underflow due to higher-order
> > allocations or spills to remote nodes.  On SMP that's fine, because
> > underflows are expected from concurrency and dealt with by returning
> > 0.  But on UP, zone_page_state will just return a wrapped unsigned
> > long, which will get past the <= 0 check and then consider the zone
> > eligible until its watermarks are hit.
> >
> > 3a025760fc15 ("mm: page_alloc: spill to remote nodes before waking
> > kswapd") already made the counter-resetting use atomic_long_read() to
> > accomodate underflows from remote spills, but it didn't go all the way
> > with it.  Make it clear that these batches are expected to go negative
> > regardless of concurrency, and use atomic_long_read() everywhere.
> >
> > Fixes: 81c0a2bb515f ("mm: page_alloc: fair zone allocator policy")
> > Reported-by: Vlastimil Babka <vbabka@suse.cz>
> > Reported-by: Leon Romanovsky <leon@leon.nu>
> > Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
> > Acked-by: Mel Gorman <mgorman@suse.de>
> > Cc: "3.12+" <stable@kernel.org>
> > ---
> >  mm/page_alloc.c | 7 +++----
> >  1 file changed, 3 insertions(+), 4 deletions(-)
> >
> > Sorry I forgot to CC you, Leon.  Resend with updated Tags.
> >
> > diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> > index 18cee0d4c8a2..eee961958021 100644
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -1612,7 +1612,7 @@ again:
> >         }
> >
> >         __mod_zone_page_state(zone, NR_ALLOC_BATCH, -(1 << order));
> > -       if (zone_page_state(zone, NR_ALLOC_BATCH) == 0 &&
> > +       if (atomic_long_read(&zone->vm_stat[NR_ALLOC_BATCH]) <= 0 &&
> >             !zone_is_fair_depleted(zone))
> >                 zone_set_flag(zone, ZONE_FAIR_DEPLETED);
> >
> > @@ -5701,9 +5701,8 @@ static void __setup_per_zone_wmarks(void)
> >                 zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp
> > >> 1);
> >
> >                 __mod_zone_page_state(zone, NR_ALLOC_BATCH,
> > -                                     high_wmark_pages(zone) -
> > -                                     low_wmark_pages(zone) -
> > -                                     zone_page_state(zone,
> > NR_ALLOC_BATCH));
> > +                       high_wmark_pages(zone) - low_wmark_pages(zone) -
> > +                       atomic_long_read(&zone->vm_stat[NR_ALLOC_BATCH]));
> >
> >                 setup_zone_migrate_reserve(zone);
> >                 spin_unlock_irqrestore(&zone->lock, flags);
> >
> 
> I think the better way will be to apply Mel's patch
> https://lkml.org/lkml/2014/9/8/214 which fix zone_page_state shadow casting
> issue and convert all atomic_long_read(&zone->vm_stat[NR_ALLOC_BATCH])) to
> zone_page__state(zone, NR_ALLOC_BATCH). This move will unify access to
> vm_stat.

It's not as simple.  The counter can go way negative and we need that
negative number, not 0, to calculate the reset delta.  As I said in
response to Mel's patch, we could make the vmstat API signed but I'm
not convinced that is reasonable, given the 99% majority of usecases.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>