From: Alexander Viro <viro@math.psu.edu>
To: Linus Torvalds <torvalds@transmeta.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>, linux-kernel@vger.kernel.org
Subject: [PATCH] SMP race in ext2 - metadata corruption.
Date: Thu, 26 Apr 2001 11:45:47 -0400 (EDT) [thread overview]
Message-ID: <Pine.GSO.4.21.0104261137410.15385-100000@weyl.math.psu.edu> (raw)
Ext2 does getblk+wait_on_buffer for new metadata blocks before
filling them with zeroes. While that is enough for single-processor,
on SMP we have the following race:
getblk gives us unlocked, non-uptodate bh
wait_on_buffer() does nothing
read from device locks it and starts IO
we zero it out.
on-disk data overwrites our zeroes.
we mark it dirty
bdflush writes the old data (_not_ zeroes) back to disk.
Result: crap in metadata block. Proposed fix: lock_buffer()/unlock_buffer()
around memset()/mark_buffer_uptodate() instead of wait_on_buffer() before
them.
Patch against 2.4.4-pre7 follows. Please, apply.
Al
--- S4-pre7/fs/ext2/inode.c Wed Apr 25 20:43:08 2001
+++ S4-pre7-ext2/fs/ext2/inode.c Thu Apr 26 11:36:11 2001
@@ -397,13 +397,13 @@
* the pointer to new one, then send parent to disk.
*/
bh = getblk(inode->i_dev, parent, blocksize);
- if (!buffer_uptodate(bh))
- wait_on_buffer(bh);
+ lock_buffer(bh);
memset(bh->b_data, 0, blocksize);
branch[n].bh = bh;
branch[n].p = (u32*) bh->b_data + offsets[n];
*branch[n].p = branch[n].key;
mark_buffer_uptodate(bh, 1);
+ unlock_buffer(bh);
mark_buffer_dirty_inode(bh, inode);
if (IS_SYNC(inode) || inode->u.ext2_i.i_osync) {
ll_rw_block (WRITE, 1, &bh);
@@ -587,10 +587,10 @@
struct buffer_head *bh;
bh = getblk(dummy.b_dev, dummy.b_blocknr, inode->i_sb->s_blocksize);
if (buffer_new(&dummy)) {
- if (!buffer_uptodate(bh))
- wait_on_buffer(bh);
+ lock_buffer(bh);
memset(bh->b_data, 0, inode->i_sb->s_blocksize);
mark_buffer_uptodate(bh, 1);
+ unlock_buffer(bh);
mark_buffer_dirty_inode(bh, inode);
}
return bh;
next reply other threads:[~2001-04-26 15:47 UTC|newest]
Thread overview: 90+ messages / expand[flat|nested] mbox.gz Atom feed top
2001-04-26 15:45 Alexander Viro [this message]
2001-04-26 18:12 ` [PATCH] SMP race in ext2 - metadata corruption Andrea Arcangeli
2001-04-26 18:24 ` Alexander Viro
2001-04-26 19:00 ` Chris Mason
2001-04-26 18:49 ` Linus Torvalds
2001-04-26 19:08 ` Alexander Viro
2001-04-26 19:17 ` Alexander Viro
2001-04-26 20:06 ` Andrea Arcangeli
2001-04-26 19:15 ` Andrea Arcangeli
2001-04-26 19:34 ` Alexander Viro
2001-04-26 19:44 ` Andrea Arcangeli
2001-04-26 19:55 ` Alexander Viro
2001-04-26 20:08 ` Linus Torvalds
2001-04-26 20:21 ` Andrea Arcangeli
2001-04-26 20:49 ` Alexander Viro
2001-04-26 21:07 ` Andrea Arcangeli
2001-04-26 23:25 ` Alexander Viro
2001-04-27 0:05 ` Andrea Arcangeli
2001-04-27 0:19 ` Linus Torvalds
2001-04-27 0:35 ` Andrea Arcangeli
2001-04-26 21:13 ` Andrzej Krzysztofowicz
2001-04-27 18:02 ` LA Walsh
2001-04-27 18:17 ` dek_ml
2001-04-27 7:58 ` Vojtech Pavlik
2001-04-27 13:23 ` Alexander Viro
2001-04-27 14:32 ` Ville Herva
2001-04-27 16:52 ` Linus Torvalds
2001-04-27 19:22 ` Shane Wegner
2001-04-28 4:55 ` Albert D. Cahalan
2001-04-28 10:05 ` Jens Axboe
2001-04-27 14:29 ` Andi Kleen
2001-05-03 6:23 ` volodya
2001-05-03 9:29 ` Alan Cox
2001-05-03 17:21 ` Linus Torvalds
2001-05-04 11:40 ` Rogier Wolff
2001-05-04 11:56 ` Jens Axboe
2001-05-04 15:29 ` Andrea Arcangeli
2001-05-04 17:40 ` Linus Torvalds
2001-05-05 3:18 ` Chris Wedgwood
2001-05-06 0:37 ` Andrea Arcangeli
2001-05-06 2:14 ` Chris Wedgwood
2001-05-06 2:50 ` Andrea Arcangeli
2001-05-06 3:00 ` Chris Wedgwood
2001-05-06 3:15 ` Alexander Viro
2001-05-06 3:32 ` Chris Wedgwood
2001-05-06 3:59 ` Alexander Viro
2001-05-06 12:47 ` Alan Cox
2001-05-06 19:46 ` Andreas Dilger
2001-05-06 20:58 ` Alan Cox
2001-05-07 4:08 ` Andreas Dilger
2001-05-07 18:42 ` Pavel Machek
2001-05-11 14:54 ` Daniel Phillips
2001-05-18 14:46 ` Stephen C. Tweedie
2001-05-11 20:00 ` Alexander Viro
2001-05-06 3:38 ` Andreas Dilger
2001-05-06 3:51 ` Andrea Arcangeli
2001-05-04 12:15 ` Marc SCHAEFER
2001-05-04 17:28 ` Linus Torvalds
2001-05-04 17:39 ` Alexander Viro
2001-05-04 17:55 ` Linus Torvalds
2001-05-04 18:29 ` Alexander Viro
2001-05-04 19:13 ` Linus Torvalds
2001-05-05 11:20 ` Albert D. Cahalan
2001-05-05 17:11 ` Alexander Viro
2001-05-04 21:11 ` Alan Cox
2001-05-04 21:50 ` Linus Torvalds
2001-05-04 18:20 ` Richard Gooch
2001-05-04 18:35 ` Alexander Viro
2001-05-04 18:49 ` Richard Gooch
2001-05-04 19:15 ` Alexander Viro
2001-05-04 19:36 ` Richard Gooch
2001-05-04 19:58 ` Alexander Viro
2001-05-04 20:04 ` Richard Gooch
2001-05-04 19:04 ` Jens Axboe
2001-05-05 13:52 ` Rogier Wolff
2001-05-05 18:12 ` Richard Gooch
2001-05-08 13:46 ` volodya
2001-05-09 10:30 ` Helge Hafting
2001-04-26 20:11 ` Andrea Arcangeli
2001-04-26 20:14 ` Andrea Arcangeli
2001-04-26 20:26 ` Linus Torvalds
2001-04-26 21:16 ` Andrea Arcangeli
2001-04-26 20:42 ` Richard B. Johnson
2001-04-26 20:55 ` Alexander Viro
2001-04-27 11:43 ` Richard B. Johnson
2001-04-26 19:39 ` Andrea Arcangeli
[not found] <Pine.LNX.4.21.0104270953280.2067-100000@penguin.transmeta.com>
2001-04-27 17:36 ` Jeff Garzik
[not found] ` <3AE9A69B.D11F0BBD@evision-ventures.com>
2001-04-28 8:31 ` Matthias Urlichs
2001-04-28 13:20 ` Olaf Titz
2001-04-30 8:47 ` Neil Conway
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=Pine.GSO.4.21.0104261137410.15385-100000@weyl.math.psu.edu \
--to=viro@math.psu.edu \
--cc=alan@lxorguk.ukuu.org.uk \
--cc=linux-kernel@vger.kernel.org \
--cc=torvalds@transmeta.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: 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).