This patch appears to have been missed for ext3, while the ext4 version was
merged back in March -- see [email protected].
A customer of mine has been running into this bug on their production
servers on 3.4-stable. It applies cleanly to the 3.3 through 3.8 trees.
-ben
commit e643692138cfa33528f054b071ba2583509bb217
Author: Jan Kara <[email protected]>
Date: Wed Mar 20 14:39:05 2013 +0100
ext3: fix data=journal fast mount/umount hang
In data=journal mode, if we unmount the file system before a
transaction has a chance to complete, when the journal inode is being
evicted, we can end up calling into log_wait_commit() for the
last transaction, after the journalling machinery has been shut down.
That triggers the WARN_ONCE in __log_start_commit().
Arguably we should adjust ext3_should_journal_data() to return FALSE
for the journal inode, but the only place it matters is
ext3_evict_inode(), and so it's to save a bit of CPU time, and to make
the patch much more obviously correct by inspection(tm), we'll fix it
by explicitly not trying to waiting for a journal commit when we are
evicting the journal inode, since it's guaranteed to never succeed in
this case.
This can be easily replicated via:
mount -t ext3 -o data=journal /dev/vdb /vdb ; umount /vdb
This is a port of ext4 fix from Ted Ts'o.
Signed-off-by: Jan Kara <[email protected]>
Signed-off-by: Benjamin LaHaise <[email protected]>
diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
index d512c4b..d706dbf 100644
--- a/fs/ext3/inode.c
+++ b/fs/ext3/inode.c
@@ -218,7 +218,8 @@ void ext3_evict_inode (struct inode *inode)
*/
if (inode->i_nlink && ext3_should_journal_data(inode) &&
EXT3_SB(inode->i_sb)->s_journal &&
- (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode))) {
+ (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode)) &&
+ inode->i_ino != EXT3_JOURNAL_INO) {
tid_t commit_tid = atomic_read(&ei->i_datasync_tid);
journal_t *journal = EXT3_SB(inode->i_sb)->s_journal;
Benjamin LaHaise <[email protected]> writes:
> This patch appears to have been missed for ext3, while the ext4 version was
> merged back in March -- see [email protected].
> A customer of mine has been running into this bug on their production
> servers on 3.4-stable. It applies cleanly to the 3.3 through 3.8 trees.
>
> -ben
Thanks Ben, I'm queuing it for the 3.5 kernel.
Cheers,
--
Luis
>
> commit e643692138cfa33528f054b071ba2583509bb217
> Author: Jan Kara <[email protected]>
> Date: Wed Mar 20 14:39:05 2013 +0100
>
> ext3: fix data=journal fast mount/umount hang
>
> In data=journal mode, if we unmount the file system before a
> transaction has a chance to complete, when the journal inode is being
> evicted, we can end up calling into log_wait_commit() for the
> last transaction, after the journalling machinery has been shut down.
> That triggers the WARN_ONCE in __log_start_commit().
>
> Arguably we should adjust ext3_should_journal_data() to return FALSE
> for the journal inode, but the only place it matters is
> ext3_evict_inode(), and so it's to save a bit of CPU time, and to make
> the patch much more obviously correct by inspection(tm), we'll fix it
> by explicitly not trying to waiting for a journal commit when we are
> evicting the journal inode, since it's guaranteed to never succeed in
> this case.
>
> This can be easily replicated via:
>
> mount -t ext3 -o data=journal /dev/vdb /vdb ; umount /vdb
>
> This is a port of ext4 fix from Ted Ts'o.
>
> Signed-off-by: Jan Kara <[email protected]>
> Signed-off-by: Benjamin LaHaise <[email protected]>
>
> diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
> index d512c4b..d706dbf 100644
> --- a/fs/ext3/inode.c
> +++ b/fs/ext3/inode.c
> @@ -218,7 +218,8 @@ void ext3_evict_inode (struct inode *inode)
> */
> if (inode->i_nlink && ext3_should_journal_data(inode) &&
> EXT3_SB(inode->i_sb)->s_journal &&
> - (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode))) {
> + (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode)) &&
> + inode->i_ino != EXT3_JOURNAL_INO) {
> tid_t commit_tid = atomic_read(&ei->i_datasync_tid);
> journal_t *journal = EXT3_SB(inode->i_sb)->s_journal;
>
> --
> To unsubscribe from this list: send the line "unsubscribe stable" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html