From: Andrew Morton <akpm@linux-foundation.org>
To: Mel Gorman <mgorman@techsingularity.net>
Cc: Nicolas Saenz Julienne <nsaenzju@redhat.com>,
Marcelo Tosatti <mtosatti@redhat.com>,
Vlastimil Babka <vbabka@suse.cz>,
Michal Hocko <mhocko@kernel.org>,
LKML <linux-kernel@vger.kernel.org>,
Linux-MM <linux-mm@kvack.org>
Subject: Re: [PATCH 6/6] mm/page_alloc: Remotely drain per-cpu lists
Date: Thu, 12 May 2022 12:37:43 -0700 [thread overview]
Message-ID: <20220512123743.5be26b3ad4413f20d5f46564@linux-foundation.org> (raw)
In-Reply-To: <20220512085043.5234-7-mgorman@techsingularity.net>
On Thu, 12 May 2022 09:50:43 +0100 Mel Gorman <mgorman@techsingularity.net> wrote:
> From: Nicolas Saenz Julienne <nsaenzju@redhat.com>
>
> Some setups, notably NOHZ_FULL CPUs, are too busy to handle the per-cpu
> drain work queued by __drain_all_pages(). So introduce a new mechanism to
> remotely drain the per-cpu lists. It is made possible by remotely locking
> 'struct per_cpu_pages' new per-cpu spinlocks. A benefit of this new scheme
> is that drain operations are now migration safe.
>
> There was no observed performance degradation vs. the previous scheme.
> Both netperf and hackbench were run in parallel to triggering the
> __drain_all_pages(NULL, true) code path around ~100 times per second.
> The new scheme performs a bit better (~5%), although the important point
> here is there are no performance regressions vs. the previous mechanism.
> Per-cpu lists draining happens only in slow paths.
>
> Minchan Kim tested this independently and reported;
>
> My workload is not NOHZ CPUs but run apps under heavy memory
> pressure so they goes to direct reclaim and be stuck on
> drain_all_pages until work on workqueue run.
>
> unit: nanosecond
> max(dur) avg(dur) count(dur)
> 166713013 487511.77786438033 1283
>
> From traces, system encountered the drain_all_pages 1283 times and
> worst case was 166ms and avg was 487us.
>
> The other problem was alloc_contig_range in CMA. The PCP draining
> takes several hundred millisecond sometimes though there is no
> memory pressure or a few of pages to be migrated out but CPU were
> fully booked.
>
> Your patch perfectly removed those wasted time.
I'm not getting a sense here of the overall effect upon userspace
performance. As Thomas said last year in
https://lkml.kernel.org/r/87v92sgt3n.ffs@tglx
: The changelogs and the cover letter have a distinct void vs. that which
: means this is just another example of 'scratch my itch' changes w/o
: proper justification.
Is there more to all of this than itchiness and if so, well, you know
the rest ;)
next prev parent reply other threads:[~2022-05-12 19:37 UTC|newest]
Thread overview: 46+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-12 8:50 [PATCH 0/6] Drain remote per-cpu directly v3 Mel Gorman
2022-05-12 8:50 ` [PATCH 1/6] mm/page_alloc: Add page->buddy_list and page->pcp_list Mel Gorman
2022-05-13 11:59 ` Nicolas Saenz Julienne
2022-05-19 9:36 ` Vlastimil Babka
2022-05-12 8:50 ` [PATCH 2/6] mm/page_alloc: Use only one PCP list for THP-sized allocations Mel Gorman
2022-05-19 9:45 ` Vlastimil Babka
2022-05-12 8:50 ` [PATCH 3/6] mm/page_alloc: Split out buddy removal code from rmqueue into separate helper Mel Gorman
2022-05-13 12:01 ` Nicolas Saenz Julienne
2022-05-19 9:52 ` Vlastimil Babka
2022-05-23 16:09 ` Qais Yousef
2022-05-24 11:55 ` Mel Gorman
2022-05-25 11:23 ` Qais Yousef
2022-05-12 8:50 ` [PATCH 4/6] mm/page_alloc: Remove unnecessary page == NULL check in rmqueue Mel Gorman
2022-05-13 12:03 ` Nicolas Saenz Julienne
2022-05-19 10:57 ` Vlastimil Babka
2022-05-19 12:13 ` Mel Gorman
2022-05-19 12:26 ` Vlastimil Babka
2022-05-12 8:50 ` [PATCH 5/6] mm/page_alloc: Protect PCP lists with a spinlock Mel Gorman
2022-05-13 12:22 ` Nicolas Saenz Julienne
2022-05-12 8:50 ` [PATCH 6/6] mm/page_alloc: Remotely drain per-cpu lists Mel Gorman
2022-05-12 19:37 ` Andrew Morton [this message]
2022-05-13 15:04 ` Mel Gorman
2022-05-13 15:19 ` Nicolas Saenz Julienne
2022-05-13 18:23 ` Mel Gorman
2022-05-17 12:57 ` Mel Gorman
2022-05-12 19:43 ` [PATCH 0/6] Drain remote per-cpu directly v3 Andrew Morton
2022-05-13 14:23 ` Mel Gorman
2022-05-13 19:38 ` Andrew Morton
2022-05-16 10:53 ` Mel Gorman
2022-05-13 12:24 ` Nicolas Saenz Julienne
2022-05-17 23:35 ` Qian Cai
2022-05-18 12:51 ` Mel Gorman
2022-05-18 16:27 ` Qian Cai
2022-05-18 17:15 ` Paul E. McKenney
2022-05-19 13:29 ` Qian Cai
2022-05-19 19:15 ` Paul E. McKenney
2022-05-19 21:05 ` Qian Cai
2022-05-19 21:29 ` Paul E. McKenney
2022-05-18 17:26 ` Marcelo Tosatti
2022-05-18 17:44 ` Marcelo Tosatti
2022-05-18 18:01 ` Nicolas Saenz Julienne
2022-05-26 17:19 ` Qian Cai
2022-05-27 8:39 ` Mel Gorman
2022-05-27 12:58 ` Qian Cai
-- strict thread matches above, loose matches on Subject: below --
2022-05-09 13:07 [RFC PATCH 0/6] Drain remote per-cpu directly v2 Mel Gorman
2022-05-09 13:08 ` [PATCH 6/6] mm/page_alloc: Remotely drain per-cpu lists Mel Gorman
2022-04-20 9:59 [RFC PATCH 0/6] Drain remote per-cpu directly Mel Gorman
2022-04-20 9:59 ` [PATCH 6/6] mm/page_alloc: Remotely drain per-cpu lists 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=20220512123743.5be26b3ad4413f20d5f46564@linux-foundation.org \
--to=akpm@linux-foundation.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@techsingularity.net \
--cc=mhocko@kernel.org \
--cc=mtosatti@redhat.com \
--cc=nsaenzju@redhat.com \
--cc=vbabka@suse.cz \
/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).