From: Peter Zijlstra <peterz@infradead.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Chris Mason <chris.mason@oracle.com>,
Steven Rostedt <rostedt@goodmis.org>, Ingo Molnar <mingo@elte.hu>,
paulmck@linux.vnet.ibm.com, Gregory Haskins <ghaskins@novell.com>,
Matthew Wilcox <matthew@wil.cx>, Andi Kleen <andi@firstfloor.org>,
Andrew Morton <akpm@linux-foundation.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
linux-fsdevel <linux-fsdevel@vger.kernel.org>,
linux-btrfs <linux-btrfs@vger.kernel.org>,
Thomas Gleixner <tglx@linutronix.de>,
Nick Piggin <npiggin@suse.de>,
Peter Morreale <pmorreale@novell.com>,
Sven Dietrich <SDietrich@novell.com>
Subject: Re: [PATCH -v7][RFC]: mutex: implement adaptive spinning
Date: Fri, 09 Jan 2009 16:06:54 +0100 [thread overview]
Message-ID: <1231513614.442.11.camel@twins> (raw)
In-Reply-To: <1231498062.11687.608.camel@twins>
On Fri, 2009-01-09 at 11:47 +0100, Peter Zijlstra wrote:
> > So I think the bug is still there, we just hid it better by breaking out
> > of the loop with that "if (need_resched())" always eventually triggering.
> > And it would be ok if it really is guaranteed to _eventually_ trigger, and
> > I guess with timeslices it eventually always will, but I suspect we could
> > have some serious latency spikes.
>
> Yes, the owner getting preempted after acquiring the lock, but before
> setting the owner can give some nasties :-(
>
> I initially did that preempt_disable/enable around the fast path, but I
> agree that slowing down the fast path is unwelcome.
>
> Alternatively we could go back to block on !owner, with the added
> complexity of not breaking out of the spin on lock->owner != owner
> when !lock->owner, so that the premature owner clearing of the unlock
> fast path will not force a schedule right before we get a chance to
> acquire the lock.
>
> Let me do that..
Ok a few observations..
Adding that need_resched() in the outer loop utterly destroys the
performance gain for PREEMPT=y. Voluntary preemption is mostly good, but
somewhat unstable results.
Adding that blocking on !owner utterly destroys everything.
Going to look into where that extra preemption comes from.
next prev parent reply other threads:[~2009-01-09 15:06 UTC|newest]
Thread overview: 362+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-12-31 11:28 Btrfs for mainline Chris Mason
2008-12-31 18:45 ` Andrew Morton
2008-12-31 23:19 ` Chris Mason
2009-01-02 16:37 ` Ryusuke Konishi
2009-01-02 18:06 ` Chris Mason
2009-01-02 19:38 ` Chris Mason
2009-01-03 9:44 ` Ryusuke Konishi
2009-01-05 14:14 ` Chris Mason
2009-01-05 16:43 ` Ryusuke Konishi
2009-01-05 10:32 ` Nick Piggin
2009-01-05 13:21 ` Chris Mason
2009-01-05 14:39 ` generic pagecache to block mapping layer (was Re: Btrfs for mainline) Nick Piggin
2009-01-05 16:37 ` Chris Mason
2009-01-05 17:10 ` Nick Piggin
2009-01-06 17:20 ` Nick Piggin
2009-01-02 19:05 ` Btrfs for mainline Andi Kleen
2009-01-02 19:32 ` Chris Mason
2009-01-02 21:01 ` Andi Kleen
2009-01-02 21:35 ` Chris Mason
2009-01-02 22:26 ` Roland Dreier
2009-01-04 13:28 ` KOSAKI Motohiro
2009-01-04 15:56 ` Ed Tomlinson
2009-01-05 10:07 ` Chris Samuel
2009-01-05 13:18 ` Chris Mason
2009-01-05 16:33 ` J. Bruce Fields
2009-01-06 22:09 ` Jamie Lokier
2009-01-03 19:17 ` Matthew Wilcox
2009-01-03 19:50 ` Christoph Hellwig
2009-01-03 20:17 ` Chris Mason
2009-01-04 21:52 ` Arnd Bergmann
2009-01-05 14:01 ` Chris Mason
2009-01-03 21:12 ` Matthew Wilcox
2009-01-04 18:21 ` Peter Zijlstra
2009-01-04 18:41 ` Matthew Wilcox
2009-01-05 14:47 ` Nick Piggin
2009-01-05 16:23 ` Matthew Wilcox
2009-01-05 16:30 ` Chris Mason
2009-01-07 13:07 ` Ingo Molnar
2009-01-07 13:24 ` Matthew Wilcox
2009-01-07 14:56 ` Ingo Molnar
2009-01-06 11:40 ` [PATCH][RFC]: mutex: adaptive spin Peter Zijlstra
2009-01-06 12:10 ` Ingo Molnar
2009-01-06 12:21 ` Peter Zijlstra
2009-01-06 15:01 ` Matthew Wilcox
2009-01-06 13:10 ` Gregory Haskins
2009-01-06 13:16 ` Ingo Molnar
2009-01-06 13:20 ` Peter Zijlstra
2009-01-06 14:15 ` Gregory Haskins
2009-01-06 21:42 ` Paul E. McKenney
2009-01-06 21:44 ` Peter Zijlstra
2009-01-06 21:50 ` Linus Torvalds
2009-01-06 22:07 ` Peter Zijlstra
2009-01-06 22:22 ` Linus Torvalds
2009-01-06 22:43 ` Peter Zijlstra
2009-01-06 22:56 ` Linus Torvalds
2009-01-06 23:00 ` Linus Torvalds
2009-01-06 23:09 ` Matthew Wilcox
2009-01-07 0:06 ` Linus Torvalds
2009-01-06 23:16 ` Peter Zijlstra
2009-01-07 12:03 ` [PATCH -v4][RFC]: mutex: implement adaptive spinning Peter Zijlstra
2009-01-07 14:50 ` Frédéric Weisbecker
2009-01-07 14:58 ` Peter Zijlstra
2009-01-07 15:22 ` Steven Rostedt
2009-01-07 15:29 ` Steven Rostedt
2009-01-07 15:54 ` Peter Zijlstra
2009-01-07 16:25 ` Linus Torvalds
2009-01-07 16:57 ` [PATCH -v5][RFC]: " Peter Zijlstra
2009-01-07 18:55 ` Linus Torvalds
2009-01-07 20:40 ` Steven Rostedt
2009-01-07 20:55 ` Linus Torvalds
2009-01-07 21:09 ` Matthew Wilcox
2009-01-07 21:24 ` Linus Torvalds
2009-01-07 21:32 ` Ingo Molnar
2009-01-07 21:47 ` Andrew Morton
2009-01-07 21:57 ` Ingo Molnar
2009-01-07 21:39 ` Linus Torvalds
2009-01-07 21:39 ` Andi Kleen
2009-01-07 22:28 ` Gregory Haskins
2009-01-07 22:33 ` Ingo Molnar
2009-01-07 22:51 ` Peter W. Morreale
2009-01-07 23:14 ` Peter W. Morreale
2009-01-07 22:56 ` Gregory Haskins
2009-01-07 23:09 ` Steven Rostedt
2009-01-07 23:32 ` Linus Torvalds
2009-01-07 23:46 ` Steven Rostedt
2009-01-07 23:47 ` Steven Rostedt
2009-01-07 23:52 ` Linus Torvalds
2009-01-08 3:28 ` Gregory Haskins
2009-01-08 7:10 ` Peter Zijlstra
2009-01-07 23:15 ` Linus Torvalds
2009-01-07 23:18 ` Linus Torvalds
2009-01-08 3:27 ` Gregory Haskins
2009-01-08 3:38 ` Steven Rostedt
2009-01-08 4:00 ` Gregory Haskins
2009-01-07 23:23 ` Paul E. McKenney
2009-01-07 21:28 ` Ingo Molnar
2009-01-07 21:51 ` Peter Zijlstra
2009-01-07 21:58 ` Linus Torvalds
2009-01-07 22:06 ` Linus Torvalds
2009-01-07 22:18 ` Peter Zijlstra
2009-01-07 23:10 ` Linus Torvalds
2009-01-08 9:58 ` [PATCH -v6][RFC]: " Peter Zijlstra
2009-01-08 14:18 ` Ingo Molnar
2009-01-08 14:33 ` Gregory Haskins
2009-01-08 14:46 ` [PATCH -v7][RFC]: " Peter Zijlstra
2009-01-08 15:09 ` Steven Rostedt
2009-01-08 15:23 ` Peter Zijlstra
2009-01-08 15:28 ` Steven Rostedt
2009-01-08 15:30 ` Peter Zijlstra
2009-01-08 15:30 ` Steven Rostedt
2009-01-08 16:11 ` Linus Torvalds
2009-01-08 16:58 ` Linus Torvalds
2009-01-08 17:08 ` Chris Mason
2009-01-08 17:33 ` Steven Rostedt
2009-01-08 17:52 ` Linus Torvalds
2009-01-08 18:03 ` Steven Rostedt
[not found] ` <alpine.DEB.1.10.0901081300370.24688@gandalf.stny.rr.com>
2009-01-08 18:14 ` Steven Rostedt
2009-01-08 19:17 ` Chris Mason
2009-01-08 19:45 ` Steven Rostedt
2009-01-08 18:16 ` Linus Torvalds
2009-01-08 18:27 ` Chris Mason
2009-01-08 19:02 ` Chris Mason
2009-01-08 19:13 ` Linus Torvalds
2009-01-08 19:23 ` Peter Zijlstra
2009-01-09 17:40 ` Jiri Kosina
2009-01-09 9:28 ` Peter Zijlstra
2009-01-10 23:59 ` Jeremy Fitzhardinge
2009-01-08 19:54 ` Linus Torvalds
2009-01-08 20:12 ` Steven Rostedt
2009-01-09 10:47 ` Peter Zijlstra
2009-01-09 15:06 ` Peter Zijlstra [this message]
2009-01-09 15:11 ` Chris Mason
2009-01-09 15:59 ` Steven Rostedt
2009-01-09 16:03 ` Peter Zijlstra
2009-01-09 16:34 ` Linus Torvalds
2009-01-09 16:44 ` Steven Rostedt
2009-01-09 18:16 ` Peter Zijlstra
2009-01-09 18:24 ` Linus Torvalds
2009-01-08 18:00 ` Linus Torvalds
2009-01-08 18:33 ` Ingo Molnar
2009-01-08 18:41 ` H. Peter Anvin
2009-01-09 3:46 ` Linus Torvalds
2009-01-09 4:59 ` David Miller
2009-01-09 5:00 ` H. Peter Anvin
2009-01-09 13:00 ` [patch] measurements, numbers about CONFIG_OPTIMIZE_INLINING=y impact Ingo Molnar
2009-01-09 14:03 ` jim owens
2009-01-09 15:35 ` Ingo Molnar
2009-01-09 16:28 ` Linus Torvalds
2009-01-09 19:56 ` H. Peter Anvin
2009-01-10 8:55 ` Chris Samuel
2009-01-09 20:17 ` Nicholas Miell
2009-01-09 20:29 ` Linus Torvalds
2009-01-09 23:28 ` Nicholas Miell
2009-01-10 0:05 ` Linus Torvalds
2009-01-10 0:37 ` Andi Kleen
2009-01-10 0:41 ` Linus Torvalds
2009-01-10 1:08 ` Andi Kleen
2009-01-10 2:12 ` Nicholas Miell
2009-01-10 4:05 ` Linus Torvalds
2009-01-10 6:44 ` Nicholas Miell
2009-01-09 20:29 ` Steven Rostedt
2009-01-09 16:34 ` H. Peter Anvin
2009-01-09 16:44 ` Linus Torvalds
2009-01-09 16:47 ` Dirk Hohndel
2009-01-09 16:46 ` Dirk Hohndel
2009-01-09 16:51 ` H. Peter Anvin
2009-01-09 17:07 ` Steven Rostedt
2009-01-09 17:14 ` Linus Torvalds
2009-01-09 17:32 ` Andi Kleen
2009-01-09 17:20 ` Andi Kleen
2009-01-09 17:11 ` Linus Torvalds
2009-01-09 17:46 ` Andi Kleen
2009-01-09 17:28 ` Matthew Wilcox
2009-01-09 17:47 ` Andi Kleen
2009-01-09 17:39 ` Matthew Wilcox
2009-01-09 17:54 ` Linus Torvalds
2009-01-09 18:07 ` H. Peter Anvin
2009-01-09 18:19 ` Andi Kleen
2009-01-09 18:59 ` Richard Guenther
2009-01-09 19:13 ` H. Peter Anvin
2009-01-09 19:21 ` Andi Kleen
2009-01-09 19:10 ` Richard Guenther
2009-01-09 19:17 ` H. Peter Anvin
2009-01-09 19:40 ` Andi Kleen
2009-01-09 19:32 ` Richard Guenther
2009-01-09 19:09 ` Richard Guenther
2009-01-09 19:44 ` Linus Torvalds
2009-01-09 20:14 ` Richard Guenther
2009-01-09 20:26 ` Linus Torvalds
2009-01-09 20:37 ` Richard Guenther
2009-01-09 21:23 ` Theodore Tso
2009-01-09 21:33 ` Steven Rostedt
2009-01-09 20:58 ` Sam Ravnborg
2009-01-09 17:41 ` Dirk Hohndel
2009-01-09 18:02 ` Andi Kleen
2009-01-09 17:57 ` Linus Torvalds
2009-01-09 18:55 ` Andi Kleen
2009-01-09 19:00 ` Linus Torvalds
2009-01-09 19:35 ` Andi Kleen
2009-01-09 19:29 ` Matthew Wilcox
2009-01-09 19:48 ` Linus Torvalds
2009-01-09 19:52 ` Theodore Tso
2009-01-09 22:07 ` Arjan van de Ven
2009-01-09 22:44 ` Andi Kleen
[not found] ` <20090109224410.GQ26290@one.firstfloor.org>
2009-01-09 22:35 ` H. Peter Anvin
2009-01-09 22:55 ` Arjan van de Ven
2009-01-09 23:32 ` Sam Ravnborg
2009-01-10 5:29 ` Arjan van de Ven
2009-01-10 5:57 ` H. Peter Anvin
2009-01-11 22:45 ` Valdis.Kletnieks
2009-01-09 17:13 ` Christoph Hellwig
2009-01-09 15:25 ` Ingo Molnar
2009-01-08 19:00 ` [PATCH -v7][RFC]: mutex: implement adaptive spinning Andi Kleen
2009-01-09 1:42 ` H. Peter Anvin
2009-01-09 13:37 ` Ingo Molnar
2009-01-09 16:09 ` Linus Torvalds
2009-01-09 16:23 ` H. Peter Anvin
2009-01-09 16:37 ` Linus Torvalds
2009-01-09 20:41 ` Ingo Molnar
2009-01-09 20:56 ` Ingo Molnar
2009-01-09 20:56 ` Linus Torvalds
2009-01-09 21:34 ` Ingo Molnar
2009-01-09 21:41 ` Harvey Harrison
2009-01-09 21:50 ` Linus Torvalds
2009-01-09 21:59 ` Harvey Harrison
2009-01-09 22:09 ` Linus Torvalds
2009-01-09 22:13 ` Harvey Harrison
2009-01-09 22:25 ` Harvey Harrison
2009-01-10 0:53 ` Jamie Lokier
2009-01-10 1:04 ` Linus Torvalds
2009-01-10 1:13 ` Andi Kleen
2009-01-09 23:12 ` Ingo Molnar
2009-01-09 23:24 ` Linus Torvalds
2009-01-10 1:01 ` Ingo Molnar
2009-01-10 1:06 ` Linus Torvalds
2009-01-10 1:20 ` Ingo Molnar
2009-01-10 1:34 ` Linus Torvalds
2009-01-10 1:08 ` Harvey Harrison
2009-01-10 1:30 ` Linus Torvalds
2009-01-10 5:03 ` H. Peter Anvin
2009-01-10 5:28 ` Linus Torvalds
2009-01-10 5:57 ` H. Peter Anvin
2009-01-11 0:54 ` Ingo Molnar
2009-01-12 1:56 ` Jamie Lokier
2009-01-12 8:40 ` Ingo Molnar
2009-01-10 1:18 ` Steven Rostedt
2009-01-10 1:39 ` Linus Torvalds
2009-01-10 1:41 ` Andrew Morton
2009-01-10 3:02 ` Andi Kleen
2009-01-11 12:26 ` David Woodhouse
2009-01-11 18:13 ` Andi Kleen
2009-01-11 19:25 ` Linus Torvalds
2009-01-11 20:14 ` gcc inlining heuristics was " Andi Kleen
2009-01-11 20:15 ` David Woodhouse
2009-01-11 20:34 ` Andi Kleen
2009-01-11 20:51 ` Linus Torvalds
2009-01-11 23:05 ` Linus Torvalds
2009-01-12 0:12 ` Andi Kleen
2009-01-12 0:21 ` Linus Torvalds
2009-01-12 0:52 ` Andi Kleen
2009-01-12 1:20 ` H. Peter Anvin
2009-01-12 18:06 ` Bernd Schmidt
2009-01-12 19:02 ` Linus Torvalds
2009-01-12 19:32 ` Andi Kleen
2009-01-12 19:22 ` H. Peter Anvin
2009-01-12 19:45 ` Linus Torvalds
2009-01-12 23:01 ` Jamie Lokier
2009-01-12 23:19 ` Linus Torvalds
2009-01-12 19:42 ` Linus Torvalds
[not found] ` <alpine.LFD.2.00.0901121128080.6528@localhost.localdomain>
2009-01-12 20:08 ` Linus Torvalds
2009-01-12 22:03 ` Bernd Schmidt
2009-01-13 0:21 ` Linus Torvalds
2009-01-13 1:30 ` Steven Rostedt
2009-01-19 0:13 ` Ingo Molnar
2009-01-19 6:22 ` Nick Piggin
2009-01-19 21:01 ` Linus Torvalds
2009-01-20 0:51 ` Nick Piggin
2009-01-20 12:38 ` Ingo Molnar
2009-01-20 19:49 ` David Woodhouse
2009-01-20 21:05 ` Ingo Molnar
2009-01-20 21:23 ` Linus Torvalds
2009-01-20 22:05 ` Ingo Molnar
2009-01-21 1:22 ` H. Peter Anvin
2009-01-21 8:54 ` Andi Kleen
2009-01-21 8:52 ` Nick Piggin
2009-01-21 9:20 ` Andi Kleen
2009-01-21 9:25 ` Nick Piggin
2009-01-21 9:54 ` Andi Kleen
2009-01-21 10:14 ` Nick Piggin
2009-01-20 4:20 ` Andi Kleen
2009-01-20 17:43 ` Miguel F Mascarenhas Sousa Filipe
2009-01-12 19:55 ` Bernd Schmidt
2009-01-12 20:11 ` Linus Torvalds
2009-01-12 7:59 ` Hard to debug kernel issues (was Re: [PATCH -v7][RFC]: mutex: implement adaptive spinning) Chris Samuel
2009-01-10 6:44 ` [PATCH -v7][RFC]: mutex: implement adaptive spinning H. Peter Anvin
2009-01-09 21:46 ` Linus Torvalds
2009-01-09 21:58 ` Ingo Molnar
2009-01-09 21:15 ` Andi Kleen
2009-01-08 19:59 ` Harvey Harrison
2009-01-09 1:44 ` H. Peter Anvin
2009-01-09 2:24 ` Harvey Harrison
2009-01-09 5:05 ` H. Peter Anvin
2009-01-09 3:35 ` Andi Kleen
2009-01-09 3:42 ` Linus Torvalds
2009-01-09 6:57 ` Andi Kleen
2009-01-09 6:00 ` Andrew Morton
2009-01-09 6:58 ` Andi Kleen
2009-01-09 13:05 ` Chris Mason
2009-01-08 17:23 ` Peter Zijlstra
2009-01-07 22:54 ` [PATCH -v5][RFC]: " Dave Kleikamp
2009-01-07 23:19 ` Linus Torvalds
2009-01-07 23:49 ` Steven Rostedt
2009-01-07 23:57 ` Linus Torvalds
2009-01-08 2:18 ` KAMEZAWA Hiroyuki
2009-01-08 2:33 ` Steven Rostedt
2009-01-08 2:49 ` KAMEZAWA Hiroyuki
2009-01-08 6:52 ` Andi Kleen
2009-01-08 14:24 ` Steven Rostedt
2009-01-08 14:45 ` Andi Kleen
2009-01-07 21:37 ` Andi Kleen
2009-01-07 21:35 ` Andrew Morton
2009-01-07 17:20 ` [PATCH -v4][RFC]: " Chris Mason
2009-01-07 17:50 ` Linus Torvalds
2009-01-07 18:00 ` Chris Mason
2009-01-07 3:57 ` [PATCH][RFC]: mutex: adaptive spin Lai Jiangshan
2009-01-07 6:32 ` Peter Zijlstra
2009-01-07 7:34 ` Lai Jiangshan
2009-01-07 13:35 ` Matthew Wilcox
2009-01-06 22:31 ` Paul E. McKenney
2009-01-06 15:55 ` Linus Torvalds
2009-01-06 16:11 ` Linus Torvalds
2009-01-06 16:28 ` Steven Rostedt
2009-01-06 16:40 ` Linus Torvalds
2009-01-06 16:42 ` Linus Torvalds
2009-01-06 17:02 ` Ingo Molnar
2009-01-06 16:54 ` Ingo Molnar
2009-01-06 16:56 ` Ingo Molnar
2009-01-06 16:59 ` Linus Torvalds
2009-01-06 16:19 ` Chris Mason
2009-01-10 17:52 ` Pavel Machek
2009-01-06 13:29 ` Nick Piggin
2009-01-06 16:23 ` Chris Mason
2009-01-06 17:08 ` Andrew Morton
2009-01-06 17:37 ` Steven Rostedt
2009-01-06 18:02 ` Linus Torvalds
2009-01-06 18:20 ` Steven Rostedt
2009-01-06 18:28 ` Linus Torvalds
2009-01-06 18:25 ` Linus Torvalds
2009-01-06 19:03 ` Steven Rostedt
2009-01-05 14:34 ` Btrfs for mainline Chris Mason
2009-01-04 8:32 ` Gabor MICSKO
2009-01-05 14:35 ` Chris Mason
2009-01-05 15:01 ` Tomasz Torcz
2009-01-05 16:30 ` Chris Mason
2009-01-06 19:41 ` Chris Mason
2009-01-07 9:33 ` David Woodhouse
2009-01-07 18:45 ` Chris Mason
2009-01-08 20:15 ` jim owens
2009-01-11 23:34 ` Andrew Morton
2009-01-12 13:58 ` Chris Mason
2009-01-12 15:14 ` Miguel Figueiredo Mascarenhas Sousa Filipe
2009-01-12 15:17 ` Chris Mason
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=1231513614.442.11.camel@twins \
--to=peterz@infradead.org \
--cc=SDietrich@novell.com \
--cc=akpm@linux-foundation.org \
--cc=andi@firstfloor.org \
--cc=chris.mason@oracle.com \
--cc=ghaskins@novell.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=matthew@wil.cx \
--cc=mingo@elte.hu \
--cc=npiggin@suse.de \
--cc=paulmck@linux.vnet.ibm.com \
--cc=pmorreale@novell.com \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/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).