On Wed, Feb 26, 2014 at 03:13:33PM -0500, Johannes Weiner wrote:
> On Wed, Feb 26, 2014 at 12:12:06PM -0500, Johannes Weiner wrote:
> > On Wed, Feb 26, 2014 at 09:54:22AM +0000, Mel Gorman wrote:
> > > How about special casing the (alloc_flags & ALLOC_WMARK_LOW) check in
> > > get_page_from_freelist to also ignore GFP_THISNODE? The NR_ALLOC_BATCH
> > > will go further negative if there are storms of GFP_THISNODE allocations
> > > forcing other allocations into the slow path doing multiple calls to
> > > prepare_slowpath but it would be closer to current behaviour and avoid
> > > weirdness with kswapd.
> >
> > I think the result would be much uglier. The allocations wouldn't
> > participate in the fairness protocol, and they'd create work for
> > kswapd without waking it up, diminishing the latency reduction for
> > which we have kswapd in the first place.
> >
> > If kswapd wakeups should be too aggressive, I'd rather we ratelimit
> > them in some way rather than exempting random order-0 allocation types
> > as a moderation measure. Exempting higher order wakeups, like THP
> > does is one thing, but we want order-0 watermarks to be met at all
> > times anyway, so it would make sense to me to nudge kswapd for every
> > failing order-0 request.
>
> So I'd still like to fix this and wake kswapd even for GFP_THISNODE
> allocations, but let's defer it for now in favor of a minimal bugfix
> that can be ported to -stable.
>
> Would this be an acceptable replacement for 1/2?
>
> ---
>
> From: Johannes Weiner <hannes@xxxxxxxxxxx>
> Subject: [patch 1/2] mm: page_alloc: exempt GFP_THISNODE allocations from zone
> fairness
>
> Jan Stancek reports manual page migration encountering allocation
> failures after some pages when there is still plenty of memory free,
> and bisected the problem down to 81c0a2bb515f ("mm: page_alloc: fair
> zone allocator policy").
>
> The problem is that GFP_THISNODE obeys the zone fairness allocation
> batches on one hand, but doesn't reset them and wake kswapd on the
> other hand. After a few of those allocations, the batches are
> exhausted and the allocations fail.
>
> Fixing this means either having GFP_THISNODE wake up kswapd, or
> GFP_THISNODE not participating in zone fairness at all. The latter
> seems safer as an acute bugfix, we can clean up later.
>
> Reported-by: Jan Stancek <jstancek@xxxxxxxxxx>
> Signed-off-by: Johannes Weiner <hannes@xxxxxxxxxxx>
> Cc: <stable@xxxxxxxxxx> # 3.12+
Acked-by: Mel Gorman <mgorman@xxxxxxx>
--
Mel Gorman
SUSE Labs
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>