From: Mark Tinguely <tinguely@sgi.com>
To: xfs@oss.sgi.com
Subject: [PATCH] xfs: logsunit rounding causes iclog corruption/crash
Date: Sat, 02 Mar 2013 14:14:47 -0600 [thread overview]
Message-ID: <20130302201452.236378200@sgi.com> (raw)
In-Reply-To: 20130302201446.854313570@sgi.com
[-- Attachment #1: xfs-limit-round-log-buffer-size-to-lsunit.patch --]
[-- Type: text/plain, Size: 2544 bytes --]
When the iclog buffer size and log stripe unit are both defined and
the log stripe unit is less the log buffer size then the buffer is
rounded up to the log stripe unit size during the xlog_sync().
This rounding can exceed the iclog buffer length and in xlog_data_pack():
1) Cause corruption inside the iclog buffer because there will not be
enough space for the headers in the front of the iclog buffer for
the rounding.
2) Cause corruption in memory that follows the iclog buffer when
stamping the lsn in each of the rounded blocks.
3) If CONFIG_XFS_DEBUG is defined will cause a crash in xlog_verify_iclog().
4) Cause page fault crash if the memory after the buffer is not mapped.
This has been found in XFS versions at least as far back as
Linux 2.6.32.
This patch forces the iclog buffer to be a multiple of the log stripe
unit when they are both defined.
Example:
# mkfs.xfs -l su=192k -f /dev/sda2
# mount -o logbsize=256k /dev/sda3 /scratch
# io such as fsstress in /scratch will immediately crash a debug xfs
kernel and most like a non-debug xfs kernel.
Signed-off-by: Mark Tinguely <tinguely@sgi.com>
---
fs/xfs/xfs_super.c | 20 ++++++++++++++++----
1 file changed, 16 insertions(+), 4 deletions(-)
Index: b/fs/xfs/xfs_super.c
===================================================================
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1357,11 +1357,23 @@ xfs_finish_flags(
if (mp->m_logbsize <= 0 &&
mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
mp->m_logbsize = mp->m_sb.sb_logsunit;
- } else if (mp->m_logbsize > 0 &&
- mp->m_logbsize < mp->m_sb.sb_logsunit) {
- xfs_warn(mp,
+ } else if (mp->m_logbsize > 0) {
+ if (mp->m_logbsize > mp->m_sb.sb_logsunit) {
+ int logbsize;
+ /* round up to the next multiple of logsunit */
+ logbsize = roundup(mp->m_logbsize,
+ mp->m_sb.sb_logsunit);
+ if (logbsize > XLOG_MAX_RECORD_BSIZE)
+ /* buffer size too large. round down. */
+ logbsize -= mp->m_sb.sb_logsunit;
+ xfs_warn(mp, "log bufsize rounded from %d to %d",
+ mp->m_logbsize, logbsize);
+ mp->m_logbsize = logbsize;
+ } else if (mp->m_logbsize < mp->m_sb.sb_logsunit) {
+ xfs_warn(mp,
"logbuf size must be greater than or equal to log stripe size");
- return XFS_ERROR(EINVAL);
+ return XFS_ERROR(EINVAL);
+ }
}
} else {
/* Fail a mount if the logbuf is larger than 32K */
_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs
next parent reply other threads:[~2013-03-02 20:20 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <20130302201446.854313570@sgi.com>
2013-03-02 20:14 ` Mark Tinguely [this message]
2013-03-02 23:05 ` [PATCH] xfs: logsunit rounding causes iclog corruption/crash Dave Chinner
2013-03-03 19:04 ` Mark Tinguely
2013-03-04 0:31 ` Dave Chinner
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=20130302201452.236378200@sgi.com \
--to=tinguely@sgi.com \
--cc=xfs@oss.sgi.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