Fix recalim handling in xfs_iget_core
Signature · 52 lines
--
Dave Chinner
Principal Engineer
SGI Australian Software Group
Fix the xfs_iget_core() handling of the XFS_IRECLAIMABLE flag so it
doesn't violate the guarantee we need to provide to xfs_iunpin()
w.r.t. the existence of a linux inode.
---
fs/xfs/xfs_iget.c | 28 ++++++++++++++++++++++++++++
1 file changed, 28 insertions(+)
Index: 2.6.x-xfs-new/fs/xfs/xfs_iget.c
===================================================================
--- 2.6.x-xfs-new.orig/fs/xfs/xfs_iget.c 2006-10-19 10:25:07.000000000 +1000
+++ 2.6.x-xfs-new/fs/xfs/xfs_iget.c 2006-10-19 10:29:26.460972850 +1000
@@ -238,6 +238,34 @@ again:
goto again;
}
+ /*
+ * If IRECLAIMABLE is set on this inode and lookup is
+ * racing with unlink, then we should return an error
+ * immediately so we don't remove it from the reclaim
+ * list and potentially leak the inode.
+ *
+ * Also, there may be transactions sitting in the
+ * incore log buffers or being flushed to disk at this
+ * time. We can't clear the XFS_IRECLAIMABLE flag
+ * until these transactions have hit the disk,
+ * otherwise we will void the guarantee the flag
+ * provides xfs_iunpin()
+ */
+ if (xfs_iflags_test(ip, XFS_IRECLAIMABLE)) {
+ if ((ip->i_d.di_mode == 0) &&
+ !(flags & XFS_IGET_CREATE)) {
+ read_unlock(&ih->ih_lock);
+ return ENOENT;
+ }
+ if (xfs_ipincount(ip)) {
+ read_unlock(&ih->ih_lock);
+ xfs_log_force(mp, 0,
+ XFS_LOG_FORCE|XFS_LOG_SYNC);
+ XFS_STATS_INC(xs_ig_frecycle);
+ goto again;
+ }
+ }
+
vn_trace_exit(vp, "xfs_iget.alloc",
(inst_t *)__return_address);Hi David,
I can't really see why we need this test:
if (xfs_iflags_test(ip, XFS_IRECLAIMABLE))
I think, An inode with no VP can be possibly in only 3
states (NEW, RECLAIM or RECLAIMABLE). This check is being made inside (inode_vp == NULL) check. If I am correct, may be we can omit an extra instruction here.
It appears that you can see inode with XFS_ISTALE can
potentially reach. I am not sure how it should reach that path. Following code just after this assumes that it must be in reclaimable path:
XFS_MOUNT_ILOCK(mp);
list_del_init(&ip->i_reclaim);
XFS_MOUNT_IUNLOCK(mp);Regards,
Shailendra
David Chinner wrote:
Quoted 7 lines · Shailendra Tripathi
Hi David,
I can't really see why we need this test:
if (xfs_iflags_test(ip, XFS_IRECLAIMABLE))
I think, An inode with no VP can be possibly in only 3
states (NEW, RECLAIM or RECLAIMABLE). This check is being made inside (inode_vp== NULL) check. If I am correct, may be we can omit an extra instruction here.
We can't have an I_NEW here - that is checked even before we we try to get the vnode. Hence we can be in either RECLAIM or RECLAIMABLE state. Then we check RECLAIm and retry. So only reclaimable can get here.
Yes, I think you're right. Good catch. I'll replace the test with an ASSERT.
Quoted 7 lines
It appears that you can see inode with
XFS_ISTALEcan
potentially reach. I am not sure how it should reach that path. Following code just after this assumes that it must be in reclaimable path:XFS_MOUNT_ILOCK(mp); list_del_init(&ip->i_reclaim); XFS_MOUNT_IUNLOCK(mp);
That's a no-op if it's not on the reclaim list (next and prev both point to itself) so it should be safe even on non-IRECLAIMABLE inodes.
Thanks, Shailendra.
Cheers,
Dave.
Signature · 3 lines
Dave Chinner Principal Engineer SGI Australian Software Group