From: Theodore Tso <tytso@mit.edu>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jens Axboe <jens.axboe@oracle.com>,
Linux Kernel Developers List <linux-kernel@vger.kernel.org>,
Ext4 Developers List <linux-ext4@vger.kernel.org>
Subject: Re: [GIT PULL] Ext3 latency fixes
Date: Fri, 3 Apr 2009 15:54:53 -0400 [thread overview]
Message-ID: <20090403195453.GC11661@mit.edu> (raw)
In-Reply-To: <alpine.LFD.2.00.0904031110420.19690@localhost.localdomain>
On Fri, Apr 03, 2009 at 11:24:50AM -0700, Linus Torvalds wrote:
> But at the same time, I now suspect that we could actually have solved
> this problem more easily by just doing things the other way around: make
> the default "WRITE" be the high-priority one (to match "READ"), and then
> just explicitly marking the data writes with "WRITE_ASYNC".
>
> Why? Because I think that with all the writes sprinkled around in random
> places, it's probably _easier_ to find the bulk writes that cause the
> biggest issues, and just fix _those_ to be WRITE_ASYNC. They may be bulk,
> they may be the common case, but they also tend to be the case where we
> write with generic routines (eg the whole "do_writepages()" thing).
>
> So the VFS layer tends to already do much of the bulk writeout, and maybe
> we would have been better off just changing those to ASYNC and leaving any
> more specialized cases as the SYNC case? That would have avoided a lot of
> this effort at the filesystem level. We'd just assume that the default
> filesystem-specific writes tend to all be SYNC.
Well, most filesystem-specific writes tend all to be ASYNC; it's only
those related to commits and fsync() which are SYNC. Ext3 is unusual
in that data=ordered and the physical-block journalling design of the
jbd layer means that we actually have a much larger number of blocks
that need to be written out synchronously than most other filesystems.
But even so, the number of callsites that I needed to change weren't
that large; in fact, over half of them weren't in the filesystem at
all, but in the page writeback code, since fsync() and data=ordered
both need to wait for the inodes's pages to be flushed out to disk,
and that's all done in common code. The other 40% was in the jbd's
commit code, while we are writing out the journal buffers.
I suspect the more important thing to address is the fact that
WRITE_SYNC unplugs the block device queue, and we would be better off
separating marking a particular I/O as "a user is waiting for this"
from "unplug the device queue now". That will hopefully improve
things even more.
- Ted
next prev parent reply other threads:[~2009-04-03 19:56 UTC|newest]
Thread overview: 69+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-04-03 7:01 [GIT PULL] Ext3 latency fixes Theodore Ts'o
2009-04-03 7:01 ` [PATCH 1/4] block_write_full_page: Use synchronous writes for WBC_SYNC_ALL writebacks Theodore Ts'o
2009-04-03 7:01 ` [PATCH 2/4] ext3: Use WRITE_SYNC for commits which are caused by fsync() Theodore Ts'o
2009-04-03 7:01 ` [PATCH 3/4] ext3: Add replace-on-truncate hueristics for data=writeback mode Theodore Ts'o
2009-04-03 7:01 ` [PATCH 4/4] ext3: Add replace-on-rename " Theodore Ts'o
2009-04-03 18:24 ` [GIT PULL] Ext3 latency fixes Linus Torvalds
2009-04-03 18:47 ` Jens Axboe
2009-04-03 19:13 ` Theodore Tso
2009-04-03 21:01 ` Chris Mason
2009-04-03 19:02 ` Linus Torvalds
2009-04-03 20:41 ` Linus Torvalds
2009-04-04 13:57 ` Theodore Tso
2009-04-04 15:16 ` Jens Axboe
2009-04-04 15:57 ` Linus Torvalds
2009-04-04 16:06 ` Linus Torvalds
2009-04-04 17:36 ` Jens Axboe
2009-04-04 17:34 ` Jens Axboe
2009-04-04 17:44 ` Linus Torvalds
2009-04-04 18:00 ` Trenton D. Adams
2009-04-04 18:01 ` Jens Axboe
2009-04-04 18:10 ` Linus Torvalds
2009-04-04 23:22 ` Theodore Tso
2009-04-04 23:33 ` Arjan van de Ven
2009-04-05 0:10 ` Theodore Tso
2009-04-05 15:05 ` Arjan van de Ven
2009-04-05 17:01 ` Linus Torvalds
2009-04-05 17:15 ` Mark Lord
2009-04-05 20:57 ` Jeff Garzik
2009-04-05 23:48 ` Arjan van de Ven
2009-04-06 2:32 ` Mark Lord
2009-04-06 5:47 ` Jeff Garzik
2009-04-07 18:18 ` Linus Torvalds
2009-04-07 18:22 ` Linus Torvalds
2009-04-07 19:40 ` [PATCH libata: add SSD detection hueristic; move SSD setup to ata_dev_configure (was Re: [GIT PULL] Ext3 latency fixes) Jeff Garzik
2009-04-09 18:21 ` Tejun Heo
2009-04-18 3:02 ` George Spelvin
2009-04-06 8:13 ` [GIT PULL] Ext3 latency fixes Jens Axboe
2009-04-05 18:56 ` Arjan van de Ven
2009-04-05 19:34 ` Linus Torvalds
2009-04-05 20:06 ` Arjan van de Ven
2009-04-06 6:25 ` Jens Axboe
2009-04-06 6:05 ` Theodore Tso
2009-04-06 6:23 ` Jens Axboe
2009-04-06 8:16 ` Jens Axboe
2009-04-06 14:48 ` Linus Torvalds
2009-04-06 15:09 ` Jens Axboe
2009-04-06 6:15 ` Jens Axboe
2009-04-04 20:18 ` Ingo Molnar
2009-04-06 21:50 ` Lennart Sorensen
2009-04-07 13:31 ` Mark Lord
2009-04-07 14:48 ` Lennart Sorensen
2009-04-07 19:21 ` Mark Lord
2009-04-07 19:57 ` Lennart Sorensen
2009-04-04 20:56 ` Arjan van de Ven
2009-04-06 7:06 ` Jens Axboe
2009-04-07 15:39 ` Indan Zupancic
2009-04-04 19:18 ` Theodore Tso
2009-04-06 8:12 ` Jens Axboe
2009-04-04 22:13 ` Linus Torvalds
2009-04-04 22:19 ` Linus Torvalds
2009-04-05 0:20 ` Theodore Tso
2009-04-03 19:54 ` Theodore Tso [this message]
2009-04-08 23:40 Theodore Ts'o
2009-04-09 15:49 ` Linus Torvalds
2009-04-09 16:23 ` Chris Mason
2009-04-09 17:49 ` Jan Kara
2009-04-09 18:10 ` Chris Mason
2009-04-09 19:04 ` Jan Kara
2009-04-09 17:36 ` Jan Kara
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=20090403195453.GC11661@mit.edu \
--to=tytso@mit.edu \
--cc=jens.axboe@oracle.com \
--cc=linux-ext4@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--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).