From: Johannes Weiner <hannes@cmpxchg.org> To: Michal Hocko <mhocko@suse.cz> Cc: linux-mm@kvack.org, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, Linus Torvalds <torvalds@linux-foundation.org>, Andrew Morton <akpm@linux-foundation.org>, Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>, Huang Ying <ying.huang@intel.com>, Andrea Arcangeli <aarcange@redhat.com>, Dave Chinner <david@fromorbit.com>, "Theodore Ts'o" <tytso@mit.edu> Subject: Re: [patch 09/12] mm: page_alloc: private memory reserves for OOM-killing allocations Date: Fri, 24 Apr 2015 15:13:53 -0400 [thread overview] Message-ID: <20150424191353.GA5293@cmpxchg.org> (raw) In-Reply-To: <20150414164939.GJ17160@dhcp22.suse.cz> On Tue, Apr 14, 2015 at 06:49:40PM +0200, Michal Hocko wrote: > On Wed 25-03-15 02:17:13, Johannes Weiner wrote: > > @@ -5747,17 +5765,18 @@ static void __setup_per_zone_wmarks(void) > > > > min_pages = zone->managed_pages / 1024; > > min_pages = clamp(min_pages, SWAP_CLUSTER_MAX, 128UL); > > - zone->watermark[WMARK_MIN] = min_pages; > > + zone->watermark[WMARK_OOM] = min_pages; > > } else { > > /* > > * If it's a lowmem zone, reserve a number of pages > > * proportionate to the zone's size. > > */ > > - zone->watermark[WMARK_MIN] = tmp; > > + zone->watermark[WMARK_OOM] = tmp; > > } > > > > - zone->watermark[WMARK_LOW] = min_wmark_pages(zone) + (tmp >> 2); > > - zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp >> 1); > > + zone->watermark[WMARK_MIN] = oom_wmark_pages(zone) + (tmp >> 3); > > + zone->watermark[WMARK_LOW] = oom_wmark_pages(zone) + (tmp >> 2); > > + zone->watermark[WMARK_HIGH] = oom_wmark_pages(zone) + (tmp >> 1); > > This will basically elevate the min watermark, right? And that might lead > to subtle performance differences even when OOM killer is not invoked > because the direct reclaim will start sooner. It will move the min watermark a bit closer to the kswapd watermarks, so I guess the risk of entering direct reclaim when kswapd won't wake up fast enough before concurrent allocator slowpaths deplete the zone from low to min is marginally increased. That seems like a farfetched worry, especially given that waking up a sleeping kswapd is not a high frequency event in the first place. > Shouldn't we rather give WMARK_OOM half of WMARK_MIN instead? I guess conceptually that would work as well, since an OOM killing task is technically reclaiming memory and this reserve is meant to help reclaiming tasks make forward progress. That being said, the separate OOM reserve was designed for when the allocation can actually fail: deplete our own little reserve before returning failure. But it looks like neither the low-order nor the GFP_NOFS deadlock fixes got any traction, and so right now all OOM killing allocations still have the potential to deadlock. Is there a reason we shouldn't just let them do an ALLOC_NO_WATERMARK allocation after the OOM victim exited (or timed out)? Otherwise, I'll just do that in the next iteration.
WARNING: multiple messages have this Message-ID (diff)
From: Johannes Weiner <hannes@cmpxchg.org> To: Michal Hocko <mhocko@suse.cz> Cc: linux-mm@kvack.org, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, Linus Torvalds <torvalds@linux-foundation.org>, Andrew Morton <akpm@linux-foundation.org>, Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>, Huang Ying <ying.huang@intel.com>, Andrea Arcangeli <aarcange@redhat.com>, Dave Chinner <david@fromorbit.com>, Theodore Ts'o <tytso@mit.edu> Subject: Re: [patch 09/12] mm: page_alloc: private memory reserves for OOM-killing allocations Date: Fri, 24 Apr 2015 15:13:53 -0400 [thread overview] Message-ID: <20150424191353.GA5293@cmpxchg.org> (raw) In-Reply-To: <20150414164939.GJ17160@dhcp22.suse.cz> On Tue, Apr 14, 2015 at 06:49:40PM +0200, Michal Hocko wrote: > On Wed 25-03-15 02:17:13, Johannes Weiner wrote: > > @@ -5747,17 +5765,18 @@ static void __setup_per_zone_wmarks(void) > > > > min_pages = zone->managed_pages / 1024; > > min_pages = clamp(min_pages, SWAP_CLUSTER_MAX, 128UL); > > - zone->watermark[WMARK_MIN] = min_pages; > > + zone->watermark[WMARK_OOM] = min_pages; > > } else { > > /* > > * If it's a lowmem zone, reserve a number of pages > > * proportionate to the zone's size. > > */ > > - zone->watermark[WMARK_MIN] = tmp; > > + zone->watermark[WMARK_OOM] = tmp; > > } > > > > - zone->watermark[WMARK_LOW] = min_wmark_pages(zone) + (tmp >> 2); > > - zone->watermark[WMARK_HIGH] = min_wmark_pages(zone) + (tmp >> 1); > > + zone->watermark[WMARK_MIN] = oom_wmark_pages(zone) + (tmp >> 3); > > + zone->watermark[WMARK_LOW] = oom_wmark_pages(zone) + (tmp >> 2); > > + zone->watermark[WMARK_HIGH] = oom_wmark_pages(zone) + (tmp >> 1); > > This will basically elevate the min watermark, right? And that might lead > to subtle performance differences even when OOM killer is not invoked > because the direct reclaim will start sooner. It will move the min watermark a bit closer to the kswapd watermarks, so I guess the risk of entering direct reclaim when kswapd won't wake up fast enough before concurrent allocator slowpaths deplete the zone from low to min is marginally increased. That seems like a farfetched worry, especially given that waking up a sleeping kswapd is not a high frequency event in the first place. > Shouldn't we rather give WMARK_OOM half of WMARK_MIN instead? I guess conceptually that would work as well, since an OOM killing task is technically reclaiming memory and this reserve is meant to help reclaiming tasks make forward progress. That being said, the separate OOM reserve was designed for when the allocation can actually fail: deplete our own little reserve before returning failure. But it looks like neither the low-order nor the GFP_NOFS deadlock fixes got any traction, and so right now all OOM killing allocations still have the potential to deadlock. Is there a reason we shouldn't just let them do an ALLOC_NO_WATERMARK allocation after the OOM victim exited (or timed out)? Otherwise, I'll just do that in the next iteration. -- 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>
next prev parent reply other threads:[~2015-04-24 19:14 UTC|newest] Thread overview: 138+ messages / expand[flat|nested] mbox.gz Atom feed top 2015-03-25 6:17 [patch 00/12] mm: page_alloc: improve OOM mechanism and policy Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-25 6:17 ` [patch 01/12] mm: oom_kill: remove unnecessary locking in oom_enable() Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 0:51 ` David Rientjes 2015-03-26 0:51 ` David Rientjes 2015-03-26 11:51 ` Michal Hocko 2015-03-26 11:51 ` Michal Hocko 2015-03-26 13:18 ` Michal Hocko 2015-03-26 13:18 ` Michal Hocko 2015-03-26 19:30 ` David Rientjes 2015-03-26 19:30 ` David Rientjes 2015-03-26 11:43 ` Michal Hocko 2015-03-26 11:43 ` Michal Hocko 2015-03-26 20:05 ` David Rientjes 2015-03-26 20:05 ` David Rientjes 2015-03-25 6:17 ` [patch 02/12] mm: oom_kill: clean up victim marking and exiting interfaces Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 3:34 ` David Rientjes 2015-03-26 3:34 ` David Rientjes 2015-03-26 11:54 ` Michal Hocko 2015-03-26 11:54 ` Michal Hocko 2015-03-25 6:17 ` [patch 03/12] mm: oom_kill: switch test-and-clear of known TIF_MEMDIE to clear Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 3:31 ` David Rientjes 2015-03-26 3:31 ` David Rientjes 2015-03-26 11:05 ` Johannes Weiner 2015-03-26 11:05 ` Johannes Weiner 2015-03-26 19:50 ` David Rientjes 2015-03-26 19:50 ` David Rientjes 2015-03-30 14:48 ` Michal Hocko 2015-03-30 14:48 ` Michal Hocko 2015-04-02 23:01 ` [patch] android, lmk: avoid setting TIF_MEMDIE if process has already exited David Rientjes 2015-04-02 23:01 ` David Rientjes 2015-04-28 22:50 ` [patch resend] " David Rientjes 2015-04-28 22:50 ` David Rientjes 2015-03-26 11:57 ` [patch 03/12] mm: oom_kill: switch test-and-clear of known TIF_MEMDIE to clear Michal Hocko 2015-03-26 11:57 ` Michal Hocko 2015-03-25 6:17 ` [patch 04/12] mm: oom_kill: remove unnecessary locking in exit_oom_victim() Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 12:53 ` Michal Hocko 2015-03-26 12:53 ` Michal Hocko 2015-03-26 13:01 ` Michal Hocko 2015-03-26 13:01 ` Michal Hocko 2015-03-26 15:10 ` Johannes Weiner 2015-03-26 15:10 ` Johannes Weiner 2015-03-26 15:04 ` Johannes Weiner 2015-03-26 15:04 ` Johannes Weiner 2015-03-25 6:17 ` [patch 05/12] mm: oom_kill: generalize OOM progress waitqueue Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 13:03 ` Michal Hocko 2015-03-26 13:03 ` Michal Hocko 2015-03-25 6:17 ` [patch 06/12] mm: oom_kill: simplify OOM killer locking Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 13:31 ` Michal Hocko 2015-03-26 13:31 ` Michal Hocko 2015-03-26 15:17 ` Johannes Weiner 2015-03-26 15:17 ` Johannes Weiner 2015-03-26 16:07 ` Michal Hocko 2015-03-26 16:07 ` Michal Hocko 2015-03-25 6:17 ` [patch 07/12] mm: page_alloc: inline should_alloc_retry() Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 14:11 ` Michal Hocko 2015-03-26 14:11 ` Michal Hocko 2015-03-26 15:18 ` Johannes Weiner 2015-03-26 15:18 ` Johannes Weiner 2015-03-25 6:17 ` [patch 08/12] mm: page_alloc: wait for OOM killer progress before retrying Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-25 14:15 ` Tetsuo Handa 2015-03-25 14:15 ` Tetsuo Handa 2015-03-25 17:01 ` Vlastimil Babka 2015-03-25 17:01 ` Vlastimil Babka 2015-03-26 11:28 ` Johannes Weiner 2015-03-26 11:28 ` Johannes Weiner 2015-03-26 11:24 ` Johannes Weiner 2015-03-26 11:24 ` Johannes Weiner 2015-03-26 14:32 ` Michal Hocko 2015-03-26 14:32 ` Michal Hocko 2015-03-26 15:23 ` Johannes Weiner 2015-03-26 15:23 ` Johannes Weiner 2015-03-26 15:38 ` Michal Hocko 2015-03-26 15:38 ` Michal Hocko 2015-03-26 18:17 ` Johannes Weiner 2015-03-26 18:17 ` Johannes Weiner 2015-03-27 14:01 ` [patch 08/12] mm: page_alloc: wait for OOM killer progressbefore retrying Tetsuo Handa 2015-03-27 14:01 ` Tetsuo Handa 2015-03-26 15:58 ` [patch 08/12] mm: page_alloc: wait for OOM killer progress before retrying Michal Hocko 2015-03-26 15:58 ` Michal Hocko 2015-03-26 18:23 ` Johannes Weiner 2015-03-26 18:23 ` Johannes Weiner 2015-03-25 6:17 ` [patch 09/12] mm: page_alloc: private memory reserves for OOM-killing allocations Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-04-14 16:49 ` Michal Hocko 2015-04-14 16:49 ` Michal Hocko 2015-04-24 19:13 ` Johannes Weiner [this message] 2015-04-24 19:13 ` Johannes Weiner 2015-03-25 6:17 ` [patch 10/12] mm: page_alloc: emergency reserve access for __GFP_NOFAIL allocations Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-04-14 16:55 ` Michal Hocko 2015-04-14 16:55 ` Michal Hocko 2015-03-25 6:17 ` [patch 11/12] mm: page_alloc: do not lock up GFP_NOFS allocations upon OOM Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 14:50 ` Michal Hocko 2015-03-26 14:50 ` Michal Hocko 2015-03-25 6:17 ` [patch 12/12] mm: page_alloc: do not lock up low-order " Johannes Weiner 2015-03-25 6:17 ` Johannes Weiner 2015-03-26 15:32 ` Michal Hocko 2015-03-26 15:32 ` Michal Hocko 2015-03-26 19:58 ` [patch 00/12] mm: page_alloc: improve OOM mechanism and policy Dave Chinner 2015-03-26 19:58 ` Dave Chinner 2015-03-27 15:05 ` Johannes Weiner 2015-03-27 15:05 ` Johannes Weiner 2015-03-30 0:32 ` Dave Chinner 2015-03-30 0:32 ` Dave Chinner 2015-03-30 19:31 ` Johannes Weiner 2015-03-30 19:31 ` Johannes Weiner 2015-04-01 15:19 ` Michal Hocko 2015-04-01 15:19 ` Michal Hocko 2015-04-01 21:39 ` Dave Chinner 2015-04-01 21:39 ` Dave Chinner 2015-04-02 7:29 ` Michal Hocko 2015-04-02 7:29 ` Michal Hocko 2015-04-07 14:18 ` Johannes Weiner 2015-04-07 14:18 ` Johannes Weiner 2015-04-11 7:29 ` Tetsuo Handa 2015-04-11 7:29 ` Tetsuo Handa 2015-04-13 12:49 ` Michal Hocko 2015-04-13 12:49 ` Michal Hocko 2015-04-13 12:46 ` Michal Hocko 2015-04-13 12:46 ` Michal Hocko 2015-04-14 0:11 ` Dave Chinner 2015-04-14 0:11 ` Dave Chinner 2015-04-14 7:20 ` Michal Hocko 2015-04-14 7:20 ` Michal Hocko 2015-04-14 10:36 ` Johannes Weiner 2015-04-14 10:36 ` Johannes Weiner 2015-04-14 14:23 ` Michal Hocko 2015-04-14 14:23 ` Michal Hocko
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=20150424191353.GA5293@cmpxchg.org \ --to=hannes@cmpxchg.org \ --cc=aarcange@redhat.com \ --cc=akpm@linux-foundation.org \ --cc=david@fromorbit.com \ --cc=linux-fsdevel@vger.kernel.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-mm@kvack.org \ --cc=mhocko@suse.cz \ --cc=penguin-kernel@I-love.SAKURA.ne.jp \ --cc=torvalds@linux-foundation.org \ --cc=tytso@mit.edu \ --cc=ying.huang@intel.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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.