ceph: don't set req->r_locked_dir in ceph_d_revalidate
authorJeff Layton <jlayton@redhat.com>
Wed, 30 Nov 2016 20:56:46 +0000 (15:56 -0500)
committerIlya Dryomov <idryomov@gmail.com>
Thu, 8 Dec 2016 13:32:16 +0000 (14:32 +0100)
commitc3f4688a08fd86f1bf8e055724c84b7a40a09733
treeebcb5ab3dc9517a0a644a16421dcfc90b5b28454
parent3e5de27e940d00d8d504dfb96625fb654f641509
ceph: don't set req->r_locked_dir in ceph_d_revalidate

This function sets req->r_locked_dir which is supposed to indicate to
ceph_fill_trace that the parent's i_rwsem is locked for write.
Unfortunately, there is no guarantee that the dir will be locked when
d_revalidate is called, so we really don't want ceph_fill_trace to do
any dcache manipulation from this context. Clear req->r_locked_dir since
it's clearly not safe to do that.

What we really want to know with d_revalidate is whether the dentry
still points to the same inode. ceph_fill_trace installs a pointer to
the inode in req->r_target_inode, so we can just compare that to
d_inode(dentry) to see if it's the same one after the lookup.

Also, since we aren't generally interested in the parent here, we can
switch to using a GETATTR to hint that to the MDS, which also means that
we only need to reserve one cap.

Finally, just remove the d_unhashed check. That's really outside the
purview of a filesystem's d_revalidate. If the thing became unhashed
while we're checking it, then that's up to the VFS to handle anyway.

Fixes: 200fd27c8fa2 ("ceph: use lookup request to revalidate dentry")
Link: http://tracker.ceph.com/issues/18041
Reported-by: Donatas Abraitis <donatas.abraitis@gmail.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
fs/ceph/dir.c