luck/tmp_suning_uos_patched
1
0
Fork 1
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[XFS] Fix race in xfs_write() b/w dmapi callout and direct I/O checks.

Browse Source 
In xfs_write() the iolock is dropped and reacquired in XFS_SEND_DATA()
which means that the file could change from not-cached to cached and we
need to redo the direct I/O checks. We should also redo the direct I/O
checks when the file size changes regardless if O_APPEND is set or not.

SGI-PV: 963483
SGI-Modid: xfs-linux-melb:xfs-kern:28440a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
This commit is contained in:
Lachlan McIlroy 2007-05-08 13:50:12 +10:00 committed by Tim Shimmin
parent 3a02ee1828
commit 71dfd5a396
1 changed files with 28 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

57
fs/xfs/linux-2.6/xfs_lrw.c
View File

@ -724,6 +724,34 @@ xfs_write(
goto out_unlock_mutex; goto out_unlock_mutex;
} }
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
int dmflags = FILP_DELAY_FLAG(file);
if (need_i_mutex)
dmflags |= DM_FLAGS_IMUX;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
pos, count,
dmflags, &locktype);
if (error) {
goto out_unlock_internal;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reacquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) && pos != xip->i_size)
goto start;
}
if (ioflags & IO_ISDIRECT) { if (ioflags & IO_ISDIRECT) {
xfs_buftarg_t *target = xfs_buftarg_t *target =
(xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ? (xip->i_d.di_flags & XFS_DIFLAG_REALTIME) ?
@ -749,35 +777,6 @@ xfs_write(
if (new_size > xip->i_size) if (new_size > xip->i_size)
io->io_new_size = new_size; io->io_new_size = new_size;
if ((DM_EVENT_ENABLED(vp->v_vfsp, xip, DM_EVENT_WRITE) &&
!(ioflags & IO_INVIS) && !eventsent)) {
loff_t savedsize = pos;
int dmflags = FILP_DELAY_FLAG(file);
if (need_i_mutex)
dmflags |= DM_FLAGS_IMUX;
xfs_iunlock(xip, XFS_ILOCK_EXCL);
error = XFS_SEND_DATA(xip->i_mount, DM_EVENT_WRITE, vp,
pos, count,
dmflags, &locktype);
if (error) {
goto out_unlock_internal;
}
xfs_ilock(xip, XFS_ILOCK_EXCL);
eventsent = 1;
/*
* The iolock was dropped and reacquired in XFS_SEND_DATA
* so we have to recheck the size when appending.
* We will only "goto start;" once, since having sent the
* event prevents another call to XFS_SEND_DATA, which is
* what allows the size to change in the first place.
*/
if ((file->f_flags & O_APPEND) && savedsize != xip->i_size)
goto start;
}
if (likely(!(ioflags & IO_INVIS))) { if (likely(!(ioflags & IO_INVIS))) {
file_update_time(file); file_update_time(file);
xfs_ichgtime_fast(xip, inode, xfs_ichgtime_fast(xip, inode,