From: Gilad Ben-Yossef <gilad@benyossef.com>
To: Mel Gorman <mgorman@suse.de>
Cc: Linux-MM <linux-mm@kvack.org>,
Linux-FSDevel <linux-fsdevel@vger.kernel.org>,
LKML <linux-kernel@vger.kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
"Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com>,
Russell King - ARM Linux <linux@arm.linux.org.uk>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Miklos Szeredi <mszeredi@novell.com>,
"Eric W. Biederman" <ebiederm@xmission.com>,
Greg KH <gregkh@suse.de>, Gong Chen <gong.chen@intel.com>
Subject: Re: [PATCH 2/2] mm: page allocator: Do not drain per-cpu lists via IPI from page allocator context
Date: Thu, 12 Jan 2012 17:08:10 +0200 [thread overview]
Message-ID: <CAOtvUMeL19942umZH22THGU-TQbnjRfxz-JLdY_wdWPN7ZNy5Q@mail.gmail.com> (raw)
In-Reply-To: <CAOtvUMfmSrotCGn-51SC3eiQU=xK4C_Trh+8FEfTGOJcGUgVag@mail.gmail.com>
On Thu, Jan 12, 2012 at 4:51 PM, Gilad Ben-Yossef <gilad@benyossef.com> wrote:
> On Wed, Jan 11, 2012 at 12:11 PM, Mel Gorman <mgorman@suse.de> wrote:
> <SNIP>
>> Rather than making it safe to call get_online_cpus() from the page
>> allocator, this patch simply removes the page allocator call to
>> drain_all_pages(). To avoid impacting high-order allocation success
>> rates, it still drains the local per-cpu lists for high-order
>> allocations that failed. As a side effect, this reduces the number
>> of IPIs sent during low memory situations.
>>
>> Signed-off-by: Mel Gorman <mgorman@suse.de>
>> ---
>> mm/page_alloc.c | 16 ++++++++++++----
>> 1 files changed, 12 insertions(+), 4 deletions(-)
>>
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index 2b8ba3a..b6df6fc 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -1119,7 +1119,9 @@ void drain_local_pages(void *arg)
>> */
>> void drain_all_pages(void)
>> {
>> + get_online_cpus();
>> on_each_cpu(drain_local_pages, NULL, 1);
>> + put_online_cpus();
>> }
>>
>> #ifdef CONFIG_HIBERNATION
>> @@ -1982,11 +1984,17 @@ retry:
>> migratetype);
>>
>> /*
>> - * If an allocation failed after direct reclaim, it could be because
>> - * pages are pinned on the per-cpu lists. Drain them and try again
>> + * If a high-order allocation failed after direct reclaim, there is a
>> + * possibility that it is because the necessary buddies have been
>> + * freed to the per-cpu list. Drain the local list and try again.
>> + * drain_all_pages is not used because it is unsafe to call
>> + * get_online_cpus from this context as it is possible that kthreadd
>> + * would block during thread creation and the cost of sending storms
>> + * of IPIs in low memory conditions is quite high.
>> */
>> - if (!page && !drained) {
>> - drain_all_pages();
>> + if (!page && order && !drained) {
>> + drain_pages(get_cpu());
>> + put_cpu();
>> drained = true;
>> goto retry;
>> }
>> --
>> 1.7.3.4
>>
>
> I very much like the judo like quality of relying on the fact that in
> memory pressure conditions most
> of the cpus will end up in the direct reclaim path to drain them all
> without IPIs.
>
> What I can't figure out is why we don't need get/put_online_cpus()
> pair around each and every call
> to on_each_cpu everywhere? and if we do, perhaps making it a part of
> on_each_cpu is the way to go?
>
> Something like:
>
> diff --git a/kernel/smp.c b/kernel/smp.c
> index f66a1b2..cfa3882 100644
> --- a/kernel/smp.c
> +++ b/kernel/smp.c
> @@ -691,11 +691,15 @@ void on_each_cpu(void (*func) (void *info), void
> *info, int wait)
> {
> unsigned long flags;
>
> + BUG_ON(in_atomic());
> +
> + get_online_cpus();
> preempt_disable();
> smp_call_function(func, info, wait);
> local_irq_save(flags);
> func(info);
> local_irq_restore(flags);
> preempt_enable();
> + put_online_cpus();
> }
> EXPORT_SYMBOL(on_each_cpu);
>
> Does that makes?
Well, it dies on boot (after adding the needed include file), so it's
obviously wrong, but I'm still guessing
other users of on_each_cpu will need an get/put_online_cpus() wrapper too.
Maybe -
on_each_cpu(...)
{
get_online_cpus();
__on_each_cpu(...);
put_online_cpus();
}
We'll need to audit all callers.
Gilad
--
Gilad Ben-Yossef
Chief Coffee Drinker
gilad@benyossef.com
Israel Cell: +972-52-8260388
US Cell: +1-973-8260388
http://benyossef.com
"Unfortunately, cache misses are an equal opportunity pain provider."
-- Mike Galbraith, LKML
next prev parent reply other threads:[~2012-01-12 15:08 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-01-11 10:11 [RFC PATCH 0/2] Improve reliability of CPU hotplug Mel Gorman
2012-01-11 10:11 ` [PATCH 1/2] fs: sysfs: Do dcache-related updates to sysfs dentries under sysfs_mutex Mel Gorman
2012-01-11 17:11 ` Eric W. Biederman
2012-01-11 18:07 ` Mel Gorman
2012-01-11 19:02 ` Eric W. Biederman
2012-01-11 10:11 ` [PATCH 2/2] mm: page allocator: Do not drain per-cpu lists via IPI from page allocator context Mel Gorman
2012-01-12 14:51 ` Gilad Ben-Yossef
2012-01-12 15:08 ` Peter Zijlstra
2012-01-12 15:13 ` Gilad Ben-Yossef
2012-01-12 15:08 ` Gilad Ben-Yossef [this message]
2012-01-12 15:18 ` Peter Zijlstra
2012-01-12 15:37 ` Mel Gorman
2012-01-12 15:52 ` Peter Zijlstra
2012-01-12 17:18 ` Mel Gorman
2012-01-12 19:14 ` Gilad Ben-Yossef
2012-01-13 20:58 ` Milton Miller
2012-01-15 13:53 ` Gilad Ben-Yossef
2012-01-13 20:58 ` Milton Miller
2012-01-19 16:20 ` Mel Gorman
2012-01-19 21:46 ` Srivatsa S. Bhat
2012-01-20 8:48 ` Mel Gorman
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=CAOtvUMeL19942umZH22THGU-TQbnjRfxz-JLdY_wdWPN7ZNy5Q@mail.gmail.com \
--to=gilad@benyossef.com \
--cc=a.p.zijlstra@chello.nl \
--cc=akpm@linux-foundation.org \
--cc=ebiederm@xmission.com \
--cc=gong.chen@intel.com \
--cc=gregkh@suse.de \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux@arm.linux.org.uk \
--cc=mgorman@suse.de \
--cc=mszeredi@novell.com \
--cc=paulmck@linux.vnet.ibm.com \
--cc=srivatsa.bhat@linux.vnet.ibm.com \
/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).