ceph: preserve I_COMPLETE across rename
d_move puts the renamed dentry at the end of d_subdirs, screwing with our
cached dentry directory offsets. We were just clearing I_COMPLETE to avoid
any possibility of trouble. However, assigning the renamed dentry an
offset at the end of the directory (to match it's new d_subdirs position)
is sufficient to maintain correct behavior and hold onto I_COMPLETE.
This is especially important for workloads like rsync, which renames files
into place. Before, we would lose I_COMPLETE and do MDS lookups for each
file. With this patch we only talk to the MDS on create and rename.
Signed-off-by: Sage Weil <sage@newdream.net>
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 193bfa5..6045636 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1030,9 +1030,6 @@
dout("fill_trace doing d_move %p -> %p\n",
req->r_old_dentry, dn);
- /* d_move screws up d_subdirs order */
- ceph_i_clear(dir, CEPH_I_COMPLETE);
-
d_move(req->r_old_dentry, dn);
dout(" src %p '%.*s' dst %p '%.*s'\n",
req->r_old_dentry,
@@ -1044,12 +1041,15 @@
rehashing bug in vfs_rename_dir */
ceph_invalidate_dentry_lease(dn);
- /* take overwritten dentry's readdir offset */
- dout("dn %p gets %p offset %lld (old offset %lld)\n",
- req->r_old_dentry, dn, ceph_dentry(dn)->offset,
+ /*
+ * d_move() puts the renamed dentry at the end of
+ * d_subdirs. We need to assign it an appropriate
+ * directory offset so we can behave when holding
+ * I_COMPLETE.
+ */
+ ceph_set_dentry_offset(req->r_old_dentry);
+ dout("dn %p gets new offset %lld\n", req->r_old_dentry,
ceph_dentry(req->r_old_dentry)->offset);
- ceph_dentry(req->r_old_dentry)->offset =
- ceph_dentry(dn)->offset;
dn = req->r_old_dentry; /* use old_dentry */
in = dn->d_inode;