From: Mel Gorman <mel@csn.ul.ie>
To: Gilad Ben-Yossef <gilad@benyossef.com>
Cc: linux-kernel@vger.kernel.org,
KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>,
Christoph Lameter <cl@linux.com>,
Chris Metcalf <cmetcalf@tilera.com>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
Frederic Weisbecker <fweisbec@gmail.com>,
Russell King <linux@arm.linux.org.uk>,
linux-mm@kvack.org, Pekka Enberg <penberg@kernel.org>,
Matt Mackall <mpm@selenic.com>,
Sasha Levin <levinsasha928@gmail.com>,
Rik van Riel <riel@redhat.com>, Andi Kleen <andi@firstfloor.org>,
Andrew Morton <akpm@linux-foundation.org>,
Alexander Viro <viro@zeniv.linux.org.uk>,
linux-fsdevel@vger.kernel.org, Avi Kivity <avi@redhat.com>,
Michal Nazarewicz <mina86@mina86.com>,
Milton Miller <miltonm@bga.com>
Subject: Re: [v7 7/8] mm: only IPI CPUs to drain local pages if they exist
Date: Mon, 30 Jan 2012 14:59:00 +0000 [thread overview]
Message-ID: <20120130145900.GR25268@csn.ul.ie> (raw)
In-Reply-To: <1327572121-13673-8-git-send-email-gilad@benyossef.com>
On Thu, Jan 26, 2012 at 12:02:00PM +0200, Gilad Ben-Yossef wrote:
> Calculate a cpumask of CPUs with per-cpu pages in any zone
> and only send an IPI requesting CPUs to drain these pages
> to the buddy allocator if they actually have pages when
> asked to flush.
>
> This patch saves 85%+ of IPIs asking to drain per-cpu
> pages in case of severe memory preassure that leads
> to OOM since in these cases multiple, possibly concurrent,
> allocation requests end up in the direct reclaim code
> path so when the per-cpu pages end up reclaimed on first
> allocation failure for most of the proceeding allocation
> attempts until the memory pressure is off (possibly via
> the OOM killer) there are no per-cpu pages on most CPUs
> (and there can easily be hundreds of them).
>
> This also has the side effect of shortening the average
> latency of direct reclaim by 1 or more order of magnitude
> since waiting for all the CPUs to ACK the IPI takes a
> long time.
>
> Tested by running "hackbench 400" on a 8 CPU x86 VM and
> observing the difference between the number of direct
> reclaim attempts that end up in drain_all_pages() and
> those were more then 1/2 of the online CPU had any per-cpu
> page in them, using the vmstat counters introduced
> in the next patch in the series and using proc/interrupts.
>
> In the test sceanrio, this was seen to save around 3600 global
> IPIs after trigerring an OOM on a concurrent workload:
>
> $ cat /proc/vmstat | tail -n 2
> pcp_global_drain 0
> pcp_global_ipi_saved 0
>
> $ cat /proc/interrupts | grep CAL
> CAL: 1 2 1 2
> 2 2 2 2 Function call interrupts
>
> $ hackbench 400
> [OOM messages snipped]
>
> $ cat /proc/vmstat | tail -n 2
> pcp_global_drain 3647
> pcp_global_ipi_saved 3642
>
> $ cat /proc/interrupts | grep CAL
> CAL: 6 13 6 3
> 3 3 1 2 7 Function call interrupts
>
> Please note that if the global drain is removed from the
> direct reclaim path as a patch from Mel Gorman currently
> suggests this should be replaced with an on_each_cpu_cond
> invocation.
>
> Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
> CC: Mel Gorman <mel@csn.ul.ie>
> CC: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
> CC: Christoph Lameter <cl@linux.com>
> CC: Chris Metcalf <cmetcalf@tilera.com>
> CC: Peter Zijlstra <a.p.zijlstra@chello.nl>
> CC: Frederic Weisbecker <fweisbec@gmail.com>
> CC: Russell King <linux@arm.linux.org.uk>
> CC: linux-mm@kvack.org
> CC: Pekka Enberg <penberg@kernel.org>
> CC: Matt Mackall <mpm@selenic.com>
> CC: Sasha Levin <levinsasha928@gmail.com>
> CC: Rik van Riel <riel@redhat.com>
> CC: Andi Kleen <andi@firstfloor.org>
> CC: Andrew Morton <akpm@linux-foundation.org>
> CC: Alexander Viro <viro@zeniv.linux.org.uk>
> CC: linux-fsdevel@vger.kernel.org
> CC: Avi Kivity <avi@redhat.com>
> CC: Michal Nazarewicz <mina86@mina86.com>
> CC: Milton Miller <miltonm@bga.com>
> ---
> mm/page_alloc.c | 31 ++++++++++++++++++++++++++++++-
> 1 files changed, 30 insertions(+), 1 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index d2186ec..4135983 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -1165,7 +1165,36 @@ void drain_local_pages(void *arg)
> */
> void drain_all_pages(void)
> {
> - on_each_cpu(drain_local_pages, NULL, 1);
> + int cpu;
> + struct per_cpu_pageset *pcp;
> + struct zone *zone;
> +
> + /* Allocate in the BSS so we wont require allocation in
> + * direct reclaim path for CONFIG_CPUMASK_OFFSTACK=y
> + */
> + static cpumask_t cpus_with_pcps;
> +
> + /*
> + * We don't care about racing with CPU hotplug event
> + * as offline notification will cause the notified
> + * cpu to drain that CPU pcps and on_each_cpu_mask
> + * disables preemption as part of its processing
> + */
> + for_each_online_cpu(cpu) {
> + bool has_pcps = false;
> + for_each_populated_zone(zone) {
> + pcp = per_cpu_ptr(zone->pageset, cpu);
> + if (pcp->pcp.count) {
> + has_pcps = true;
> + break;
> + }
> + }
> + if (has_pcps)
> + cpumask_set_cpu(cpu, &cpus_with_pcps);
> + else
> + cpumask_clear_cpu(cpu, &cpus_with_pcps);
> + }
Lets take two CPUs running this code at the same time. CPU 1 has per-cpu
pages in all zones. CPU 2 has no per-cpu pages in any zone. If both run
at the same time, CPU 2 can be clearing the mask for CPU 1 before it has
had a chance to send the IPI. This means we'll miss sending IPIs to CPUs
that we intended to. As I was willing to send no IPI at all;
Acked-by: Mel Gorman <mel@csn.ul.ie>
But if this gets another revision, add a comment saying that two CPUs
can interfere with each other running at the same time but we don't
care.
> + on_each_cpu_mask(&cpus_with_pcps, drain_local_pages, NULL, 1);
> }
>
> #ifdef CONFIG_HIBERNATION
> --
> 1.7.0.4
>
--
Mel Gorman
SUSE Labs
next prev parent reply other threads:[~2012-01-30 14:59 UTC|newest]
Thread overview: 77+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-01-26 10:01 [v7 0/8] Reduce cross CPU IPI interference Gilad Ben-Yossef
2012-01-26 10:01 ` [v7 1/8] smp: introduce a generic on_each_cpu_mask function Gilad Ben-Yossef
2012-01-29 12:24 ` Gilad Ben-Yossef
2012-01-30 21:52 ` Andrew Morton
2012-01-31 6:33 ` Gilad Ben-Yossef
2012-01-26 10:01 ` [v7 2/8] arm: move arm over to generic on_each_cpu_mask Gilad Ben-Yossef
2012-01-26 10:01 ` [v7 3/8] tile: move tile to use " Gilad Ben-Yossef
2012-01-26 10:01 ` [v7 4/8] smp: add func to IPI cpus based on parameter func Gilad Ben-Yossef
2012-01-27 23:57 ` Andrew Morton
2012-01-29 12:04 ` Gilad Ben-Yossef
2012-01-26 10:01 ` [v7 5/8] slub: only IPI CPUs that have per cpu obj to flush Gilad Ben-Yossef
2012-01-26 15:09 ` Christoph Lameter
2012-01-26 10:01 ` [v7 6/8] fs: only send IPI to invalidate LRU BH when needed Gilad Ben-Yossef
2012-01-26 10:02 ` [v7 7/8] mm: only IPI CPUs to drain local pages if they exist Gilad Ben-Yossef
2012-01-26 15:13 ` Christoph Lameter
2012-01-28 0:12 ` Andrew Morton
2012-01-29 12:18 ` Gilad Ben-Yossef
2012-01-30 21:49 ` Andrew Morton
2012-01-31 6:32 ` Gilad Ben-Yossef
2012-01-30 14:59 ` Mel Gorman [this message]
2012-01-30 15:14 ` Gilad Ben-Yossef
2012-01-30 15:44 ` Mel Gorman
2012-01-26 10:02 ` [v7 8/8] mm: add vmstat counters for tracking PCP drains Gilad Ben-Yossef
2012-01-26 15:19 ` [v7 0/8] Reduce cross CPU IPI interference Peter Zijlstra
2012-01-29 8:25 ` Gilad Ben-Yossef
2012-02-01 17:04 ` Frederic Weisbecker
2012-02-02 8:46 ` Gilad Ben-Yossef
2012-02-02 15:41 ` Chris Metcalf
2012-02-05 11:46 ` Gilad Ben-Yossef
2012-02-10 18:39 ` Peter Zijlstra
2012-02-10 20:13 ` Gilad Ben-Yossef
2012-02-10 20:29 ` Peter Zijlstra
2012-02-10 20:39 ` Gilad Ben-Yossef
2012-02-10 18:33 ` Peter Zijlstra
2012-02-10 20:33 ` Gilad Ben-Yossef
2012-02-15 21:50 ` Chris Metcalf
2012-02-15 22:15 ` Christoph Lameter
2012-02-15 23:44 ` Chris Metcalf
2012-02-21 1:34 ` Frederic Weisbecker
2012-03-01 18:27 ` Chris Metcalf
2012-02-10 18:38 ` Peter Zijlstra
2012-02-10 20:24 ` Gilad Ben-Yossef
2012-02-15 15:11 ` Peter Zijlstra
2012-02-15 15:19 ` Gilad Ben-Yossef
2012-02-15 21:51 ` Chris Metcalf
2012-02-02 16:24 ` Frederic Weisbecker
2012-02-02 16:29 ` Christoph Lameter
2012-02-09 15:52 ` Frederic Weisbecker
2012-02-09 15:59 ` Chris Metcalf
2012-02-09 18:11 ` Frederic Weisbecker
2012-02-09 16:26 ` Christoph Lameter
2012-02-09 18:32 ` Frederic Weisbecker
2012-02-01 17:35 ` Peter Zijlstra
2012-02-01 17:57 ` Peter Zijlstra
2012-02-02 9:42 ` Gilad Ben-Yossef
2012-02-01 18:40 ` Paul E. McKenney
2012-02-01 20:06 ` Christoph Lameter
2012-02-01 20:13 ` Paul E. McKenney
2012-02-02 9:34 ` Avi Kivity
2012-02-02 15:34 ` Paul E. McKenney
2012-02-02 16:14 ` Avi Kivity
2012-02-02 17:01 ` Paul E. McKenney
2012-02-02 17:23 ` Avi Kivity
2012-02-02 17:51 ` Paul E. McKenney
2012-02-05 12:16 ` Avi Kivity
2012-02-05 16:59 ` Paul E. McKenney
2012-02-09 15:22 ` Frederic Weisbecker
2012-02-09 16:05 ` Avi Kivity
2012-02-09 18:22 ` Frederic Weisbecker
2012-02-09 23:41 ` Paul E. McKenney
2012-02-10 1:39 ` Frederic Weisbecker
2012-02-14 13:18 ` Avi Kivity
2012-02-21 0:02 ` Frederic Weisbecker
2012-02-02 17:25 ` Christoph Lameter
2012-02-05 12:06 ` Gilad Ben-Yossef
2012-02-06 18:19 ` Christoph Lameter
2012-02-09 15:37 ` Frederic Weisbecker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20120130145900.GR25268@csn.ul.ie \
--to=mel@csn.ul.ie \
--cc=a.p.zijlstra@chello.nl \
--cc=akpm@linux-foundation.org \
--cc=andi@firstfloor.org \
--cc=avi@redhat.com \
--cc=cl@linux.com \
--cc=cmetcalf@tilera.com \
--cc=fweisbec@gmail.com \
--cc=gilad@benyossef.com \
--cc=kosaki.motohiro@jp.fujitsu.com \
--cc=levinsasha928@gmail.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux@arm.linux.org.uk \
--cc=miltonm@bga.com \
--cc=mina86@mina86.com \
--cc=mpm@selenic.com \
--cc=penberg@kernel.org \
--cc=riel@redhat.com \
--cc=viro@zeniv.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).