linux-kernel - Re: [PATCH 29/35] nfs: in-commit pages accounting and wait queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20101214154026.GA8959@localhost>
Date:	Tue, 14 Dec 2010 23:40:26 +0800
From:	Wu Fengguang <fengguang.wu@...el.com>
To:	Trond Myklebust <Trond.Myklebust@...app.com>
Cc:	Andrew Morton <akpm@...ux-foundation.org>, Jan Kara <jack@...e.cz>,
	Christoph Hellwig <hch@....de>,
	Dave Chinner <david@...morbit.com>,
	Theodore Ts'o <tytso@....edu>,
	Chris Mason <chris.mason@...cle.com>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Mel Gorman <mel@....ul.ie>, Rik van Riel <riel@...hat.com>,
	KOSAKI Motohiro <kosaki.motohiro@...fujitsu.com>,
	Greg Thelen <gthelen@...gle.com>,
	Minchan Kim <minchan.kim@...il.com>,
	linux-mm <linux-mm@...ck.org>,
	"linux-fsdevel@...r.kernel.org" <linux-fsdevel@...r.kernel.org>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 29/35] nfs: in-commit pages accounting and wait queue

On Tue, Dec 14, 2010 at 05:15:51AM +0800, Trond Myklebust wrote:
> On Mon, 2010-12-13 at 22:47 +0800, Wu Fengguang wrote:
> > plain text document attachment (writeback-nfs-in-commit.patch)
> > When doing 10+ concurrent dd's, I observed very bumpy commits submission
> > (partly because the dd's are started at the same time, and hence reached
> > 4MB to-commit pages at the same time). Basically we rely on the server
> > to complete and return write/commit requests, and want both to progress
> > smoothly and not consume too many pages. The write request wait queue is
> > not enough as it's mainly network bounded. So add another commit request
> > wait queue. Only async writes need to sleep on this queue.
> > 
> 
> I'm not understanding the above reasoning. Why should we serialise
> commits at the per-filesystem level (and only for non-blocking flushes
> at that)?

I did the commit wait queue after seeing this graph, where there is
very bursty pattern of commit submission and hence completion:

http://www.kernel.org/pub/linux/kernel/people/wfg/writeback/tests/3G/nfs-100dd-1M-8p-2953M-2.6.37-rc3+-2010-12-03-01/nfs-commit-1000.png

leading to big fluctuations, eg. the almost straight up/straight down
lines below
http://www.kernel.org/pub/linux/kernel/people/wfg/writeback/tests/3G/nfs-100dd-1M-8p-2953M-2.6.37-rc3+-2010-12-03-01/vmstat-dirty-300.png
http://www.kernel.org/pub/linux/kernel/people/wfg/writeback/tests/3G/nfs-100dd-1M-8p-2953M-2.6.37-rc3+-2010-12-03-01/dirty-pages.png
http://www.kernel.org/pub/linux/kernel/people/wfg/writeback/tests/3G/nfs-100dd-1M-8p-2953M-2.6.37-rc3+-2010-12-03-01/dirty-pages-200.png

A commit wait queue will help wipe out the "peaks". The "fixed" graph
is
http://www.kernel.org/pub/linux/kernel/people/wfg/writeback/tests/3G/nfs-100dd-1M-8p-2952M-2.6.37-rc5+-2010-12-09-03-23/vmstat-dirty-300.png
http://www.kernel.org/pub/linux/kernel/people/wfg/writeback/tests/3G/nfs-100dd-1M-8p-2952M-2.6.37-rc5+-2010-12-09-03-23/dirty-pages.png

Blocking flushes don't need to wait on this queue because they already
throttle themselves by waiting on the inode commit lock before/after
the commit.  They actually should not wait on this queue, to prevent
sync requests being unnecessarily blocked by async ones.

Thanks,
Fengguang
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/