From: Mel Gorman <mgorman@suse.de>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>,
Oleg Nesterov <oleg@redhat.com>,
Johannes Weiner <hannes@cmpxchg.org>,
Vlastimil Babka <vbabka@suse.cz>, Jan Kara <jack@suse.cz>,
Michal Hocko <mhocko@suse.cz>, Hugh Dickins <hughd@google.com>,
Dave Hansen <dave.hansen@intel.com>,
Linux Kernel <linux-kernel@vger.kernel.org>,
Linux-MM <linux-mm@kvack.org>,
Linux-FSDevel <linux-fsdevel@vger.kernel.org>,
Paul McKenney <paulmck@linux.vnet.ibm.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
David Howells <dhowells@redhat.com>
Subject: Re: [PATCH] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath v5
Date: Thu, 22 May 2014 20:53:13 +0100 [thread overview]
Message-ID: <20140522195313.GN23991@suse.de> (raw)
In-Reply-To: <20140522104722.f76b5b8dc0ec28510687be2e@linux-foundation.org>
On Thu, May 22, 2014 at 10:47:22AM -0700, Andrew Morton wrote:
> On Thu, 22 May 2014 09:46:43 +0100 Mel Gorman <mgorman@suse.de> wrote:
>
> > > > If I'm still on track here, what happens if we switch to wake-all so we
> > > > can avoid the dangling flag? I doubt if there are many collisions on
> > > > that hash table?
> > >
> > > Wake-all will be ugly and loose a herd of waiters, all racing to
> > > acquire, all but one of whoem will loose the race. It also looses the
> > > fairness, its currently a FIFO queue. Wake-all will allow starvation.
> > >
> >
> > And the cost of the thundering herd of waiters may offset any benefit of
> > reducing the number of calls to page_waitqueue and waker functions.
>
> Well, none of this has been demonstrated.
>
True, but it's also the type of thing that would deserve a patch of its
own with some separation in case bisection fingerpoints to a patch that
is doing too much on its own.
> As I speculated earlier, hash chain collisions will probably be rare,
They are meant to be (well, they're documented to be). It's the primary
reason why I'm not concerned about "dangling waiters" being that common
a case.
> except for the case where a bunch of processes are waiting on the same
> page. And in this case, perhaps wake-all is the desired behavior.
>
> Take a look at do_read_cache_page(). It does lock_page(), but it
> doesn't actually *need* to. It checks ->mapping and PG_uptodate and
> then... unlocks the page! We could have used wait_on_page_locked()
> there and permitted concurrent threads to run concurrently.
>
It does that later when it calls wait_on_page_read but the flow is weird. It
looks like the first lock_page was to serialise against any IO and double
check it was not racing against a parallel reclaim although the elevated
reference count should have prevented that. Historical artifact maybe?
It looks like there could be some improvement there but also would deserve
a patch on its own.
--
Mel Gorman
SUSE Labs
--
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>
WARNING: multiple messages have this Message-ID (diff)
From: Mel Gorman <mgorman@suse.de>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>,
Oleg Nesterov <oleg@redhat.com>,
Johannes Weiner <hannes@cmpxchg.org>,
Vlastimil Babka <vbabka@suse.cz>, Jan Kara <jack@suse.cz>,
Michal Hocko <mhocko@suse.cz>, Hugh Dickins <hughd@google.com>,
Dave Hansen <dave.hansen@intel.com>,
Linux Kernel <linux-kernel@vger.kernel.org>,
Linux-MM <linux-mm@kvack.org>,
Linux-FSDevel <linux-fsdevel@vger.kernel.org>,
Paul McKenney <paulmck@linux.vnet.ibm.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
David Howells <dhowells@redhat.com>
Subject: Re: [PATCH] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath v5
Date: Thu, 22 May 2014 20:53:13 +0100 [thread overview]
Message-ID: <20140522195313.GN23991@suse.de> (raw)
In-Reply-To: <20140522104722.f76b5b8dc0ec28510687be2e@linux-foundation.org>
On Thu, May 22, 2014 at 10:47:22AM -0700, Andrew Morton wrote:
> On Thu, 22 May 2014 09:46:43 +0100 Mel Gorman <mgorman@suse.de> wrote:
>
> > > > If I'm still on track here, what happens if we switch to wake-all so we
> > > > can avoid the dangling flag? I doubt if there are many collisions on
> > > > that hash table?
> > >
> > > Wake-all will be ugly and loose a herd of waiters, all racing to
> > > acquire, all but one of whoem will loose the race. It also looses the
> > > fairness, its currently a FIFO queue. Wake-all will allow starvation.
> > >
> >
> > And the cost of the thundering herd of waiters may offset any benefit of
> > reducing the number of calls to page_waitqueue and waker functions.
>
> Well, none of this has been demonstrated.
>
True, but it's also the type of thing that would deserve a patch of its
own with some separation in case bisection fingerpoints to a patch that
is doing too much on its own.
> As I speculated earlier, hash chain collisions will probably be rare,
They are meant to be (well, they're documented to be). It's the primary
reason why I'm not concerned about "dangling waiters" being that common
a case.
> except for the case where a bunch of processes are waiting on the same
> page. And in this case, perhaps wake-all is the desired behavior.
>
> Take a look at do_read_cache_page(). It does lock_page(), but it
> doesn't actually *need* to. It checks ->mapping and PG_uptodate and
> then... unlocks the page! We could have used wait_on_page_locked()
> there and permitted concurrent threads to run concurrently.
>
It does that later when it calls wait_on_page_read but the flow is weird. It
looks like the first lock_page was to serialise against any IO and double
check it was not racing against a parallel reclaim although the elevated
reference count should have prevented that. Historical artifact maybe?
It looks like there could be some improvement there but also would deserve
a patch on its own.
--
Mel Gorman
SUSE Labs
next prev parent reply other threads:[~2014-05-22 19:53 UTC|newest]
Thread overview: 195+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-05-13 9:45 [PATCH 00/19] Misc page alloc, shmem, mark_page_accessed and page_waitqueue optimisations v3r33 Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 01/19] mm: page_alloc: Do not update zlc unless the zlc is active Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 02/19] mm: page_alloc: Do not treat a zone that cannot be used for dirty pages as "full" Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 03/19] jump_label: Expose the reference count Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 04/19] mm: page_alloc: Use jump labels to avoid checking number_of_cpusets Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 10:58 ` Peter Zijlstra
2014-05-13 12:28 ` Mel Gorman
2014-05-13 12:28 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 05/19] mm: page_alloc: Calculate classzone_idx once from the zonelist ref Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 22:25 ` Andrew Morton
2014-05-13 22:25 ` Andrew Morton
2014-05-14 6:32 ` Mel Gorman
2014-05-14 6:32 ` Mel Gorman
2014-05-14 20:29 ` Mel Gorman
2014-05-14 20:29 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 06/19] mm: page_alloc: Only check the zone id check if pages are buddies Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 07/19] mm: page_alloc: Only check the alloc flags and gfp_mask for dirty once Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 08/19] mm: page_alloc: Take the ALLOC_NO_WATERMARK check out of the fast path Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 09/19] mm: page_alloc: Use word-based accesses for get/set pageblock bitmaps Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-22 9:24 ` Vlastimil Babka
2014-05-22 9:24 ` Vlastimil Babka
2014-05-22 18:23 ` Andrew Morton
2014-05-22 18:23 ` Andrew Morton
2014-05-22 18:45 ` Vlastimil Babka
2014-05-22 18:45 ` Vlastimil Babka
2014-05-13 9:45 ` [PATCH 10/19] mm: page_alloc: Reduce number of times page_to_pfn is called Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 13:27 ` Vlastimil Babka
2014-05-13 13:27 ` Vlastimil Babka
2014-05-13 14:09 ` Mel Gorman
2014-05-13 14:09 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 11/19] mm: page_alloc: Lookup pageblock migratetype with IRQs enabled during free Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 13:36 ` Vlastimil Babka
2014-05-13 13:36 ` Vlastimil Babka
2014-05-13 14:23 ` Mel Gorman
2014-05-13 14:23 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 12/19] mm: page_alloc: Use unsigned int for order in more places Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 13/19] mm: page_alloc: Convert hot/cold parameter and immediate callers to bool Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 14/19] mm: shmem: Avoid atomic operation during shmem_getpage_gfp Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 15/19] mm: Do not use atomic operations when releasing pages Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 16/19] mm: Do not use unnecessary atomic operations when adding pages to the LRU Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 17/19] fs: buffer: Do not use unnecessary atomic operations when discarding buffers Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 11:09 ` Peter Zijlstra
2014-05-13 12:50 ` Mel Gorman
2014-05-13 12:50 ` Mel Gorman
2014-05-13 13:49 ` Jan Kara
2014-05-13 13:49 ` Jan Kara
2014-05-13 14:30 ` Mel Gorman
2014-05-13 14:30 ` Mel Gorman
2014-05-13 14:01 ` Peter Zijlstra
2014-05-13 14:01 ` Peter Zijlstra
2014-05-13 14:46 ` Mel Gorman
2014-05-13 14:46 ` Mel Gorman
2014-05-13 13:50 ` Jan Kara
2014-05-13 13:50 ` Jan Kara
2014-05-13 22:29 ` Andrew Morton
2014-05-13 22:29 ` Andrew Morton
2014-05-14 6:12 ` Mel Gorman
2014-05-14 6:12 ` Mel Gorman
2014-05-13 9:45 ` [PATCH 18/19] mm: Non-atomically mark page accessed during page cache allocation where possible Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 14:29 ` Theodore Ts'o
2014-05-13 14:29 ` Theodore Ts'o
2014-05-20 15:49 ` [PATCH] mm: non-atomically mark page accessed during page cache allocation where possible -fix Mel Gorman
2014-05-20 15:49 ` Mel Gorman
2014-05-20 19:34 ` Andrew Morton
2014-05-20 19:34 ` Andrew Morton
2014-05-21 12:09 ` Mel Gorman
2014-05-21 12:09 ` Mel Gorman
2014-05-21 22:11 ` Andrew Morton
2014-05-21 22:11 ` Andrew Morton
2014-05-22 0:07 ` Mel Gorman
2014-05-22 0:07 ` Mel Gorman
2014-05-22 5:35 ` Prabhakar Lad
2014-05-22 5:35 ` Prabhakar Lad
2014-05-13 9:45 ` [PATCH 19/19] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath Mel Gorman
2014-05-13 9:45 ` Mel Gorman
2014-05-13 12:53 ` Mel Gorman
2014-05-13 12:53 ` Mel Gorman
2014-05-13 14:17 ` Peter Zijlstra
2014-05-13 14:17 ` Peter Zijlstra
2014-05-13 15:27 ` Paul E. McKenney
2014-05-13 15:27 ` Paul E. McKenney
2014-05-13 15:44 ` Peter Zijlstra
2014-05-13 15:44 ` Peter Zijlstra
2014-05-13 16:14 ` Paul E. McKenney
2014-05-13 16:14 ` Paul E. McKenney
2014-05-13 18:57 ` Oleg Nesterov
2014-05-13 18:57 ` Oleg Nesterov
2014-05-13 20:24 ` Paul E. McKenney
2014-05-13 20:24 ` Paul E. McKenney
2014-05-14 14:25 ` Oleg Nesterov
2014-05-14 14:25 ` Oleg Nesterov
2014-05-13 18:22 ` Oleg Nesterov
2014-05-13 18:22 ` Oleg Nesterov
2014-05-13 18:18 ` Oleg Nesterov
2014-05-13 18:18 ` Oleg Nesterov
2014-05-13 18:24 ` Peter Zijlstra
2014-05-13 18:24 ` Peter Zijlstra
2014-05-13 18:52 ` Paul E. McKenney
2014-05-13 18:52 ` Paul E. McKenney
2014-05-13 19:31 ` Oleg Nesterov
2014-05-13 19:31 ` Oleg Nesterov
2014-05-13 20:32 ` Paul E. McKenney
2014-05-13 20:32 ` Paul E. McKenney
2014-05-14 16:11 ` Oleg Nesterov
2014-05-14 16:11 ` Oleg Nesterov
2014-05-14 16:17 ` Peter Zijlstra
2014-05-16 13:51 ` [PATCH 0/1] ptrace: task_clear_jobctl_trapping()->wake_up_bit() needs mb() Oleg Nesterov
2014-05-16 13:51 ` Oleg Nesterov
2014-05-16 13:51 ` [PATCH 1/1] " Oleg Nesterov
2014-05-16 13:51 ` Oleg Nesterov
2014-05-21 9:29 ` Peter Zijlstra
2014-05-21 19:19 ` Andrew Morton
2014-05-21 19:19 ` Andrew Morton
2014-05-21 19:18 ` [PATCH 0/1] " Andrew Morton
2014-05-21 19:18 ` Andrew Morton
2014-05-14 19:29 ` [PATCH 19/19] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath Oleg Nesterov
2014-05-14 19:29 ` Oleg Nesterov
2014-05-14 20:53 ` Mel Gorman
2014-05-14 20:53 ` Mel Gorman
2014-05-15 10:48 ` [PATCH] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath v4 Mel Gorman
2014-05-15 10:48 ` Mel Gorman
2014-05-15 13:20 ` Peter Zijlstra
2014-05-15 13:29 ` Peter Zijlstra
2014-05-15 15:34 ` Oleg Nesterov
2014-05-15 15:34 ` Oleg Nesterov
2014-05-15 15:45 ` Peter Zijlstra
2014-05-15 16:18 ` Mel Gorman
2014-05-15 16:18 ` Mel Gorman
2014-05-15 15:03 ` Oleg Nesterov
2014-05-15 15:03 ` Oleg Nesterov
2014-05-15 21:24 ` Andrew Morton
2014-05-15 21:24 ` Andrew Morton
2014-05-21 12:15 ` [PATCH] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath v5 Mel Gorman
2014-05-21 12:15 ` Mel Gorman
2014-05-21 13:02 ` Peter Zijlstra
2014-05-21 13:02 ` Peter Zijlstra
2014-05-21 15:33 ` Mel Gorman
2014-05-21 15:33 ` Mel Gorman
2014-05-21 16:08 ` Peter Zijlstra
2014-05-21 16:08 ` Peter Zijlstra
2014-05-21 21:26 ` Andrew Morton
2014-05-21 21:26 ` Andrew Morton
2014-05-21 21:33 ` Peter Zijlstra
2014-05-21 21:33 ` Peter Zijlstra
2014-05-21 21:50 ` Andrew Morton
2014-05-21 21:50 ` Andrew Morton
2014-05-22 0:07 ` Mel Gorman
2014-05-22 0:07 ` Mel Gorman
2014-05-22 7:20 ` Peter Zijlstra
2014-05-22 10:40 ` [PATCH] mm: filemap: Avoid unnecessary barriers and waitqueue lookups in unlock_page fastpath v7 Mel Gorman
2014-05-22 10:40 ` Mel Gorman
2014-05-22 10:56 ` Peter Zijlstra
2014-05-22 13:00 ` Mel Gorman
2014-05-22 13:00 ` Mel Gorman
2014-05-22 14:40 ` Mel Gorman
2014-05-22 14:40 ` Mel Gorman
2014-05-22 15:04 ` Peter Zijlstra
2014-05-22 15:36 ` Mel Gorman
2014-05-22 15:36 ` Mel Gorman
2014-05-22 16:58 ` [PATCH] mm: filemap: Avoid unnecessary barriers and waitqueue lookups in unlock_page fastpath v8 Mel Gorman
2014-05-22 16:58 ` Mel Gorman
2014-05-22 6:45 ` [PATCH] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath v5 Peter Zijlstra
2014-05-22 8:46 ` Mel Gorman
2014-05-22 8:46 ` Mel Gorman
2014-05-22 17:47 ` Andrew Morton
2014-05-22 17:47 ` Andrew Morton
2014-05-22 19:53 ` Mel Gorman [this message]
2014-05-22 19:53 ` Mel Gorman
2014-05-21 23:35 ` Mel Gorman
2014-05-21 23:35 ` Mel Gorman
2014-05-13 16:52 ` [PATCH 19/19] mm: filemap: Avoid unnecessary barries and waitqueue lookups in unlock_page fastpath Peter Zijlstra
2014-05-13 16:52 ` Peter Zijlstra
2014-05-14 7:31 ` Mel Gorman
2014-05-14 7:31 ` Mel Gorman
2014-05-19 8:57 ` [PATCH] mm: Avoid unnecessary atomic operations during end_page_writeback Mel Gorman
2014-05-19 8:57 ` 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=20140522195313.GN23991@suse.de \
--to=mgorman@suse.de \
--cc=akpm@linux-foundation.org \
--cc=dave.hansen@intel.com \
--cc=dhowells@redhat.com \
--cc=hannes@cmpxchg.org \
--cc=hughd@google.com \
--cc=jack@suse.cz \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mhocko@suse.cz \
--cc=oleg@redhat.com \
--cc=paulmck@linux.vnet.ibm.com \
--cc=peterz@infradead.org \
--cc=torvalds@linux-foundation.org \
--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 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.