2008-09-22 21:11:42

by Andrew Morton

[permalink] [raw]
Subject: [patch for 2.6.27? 1/2] nfs: BUG_ON in nfs_follow_mountpoint

From: "Denis V. Lunev" <[email protected]>

Unfortunately, BUG_ON(IS_ROOT(dentry)) can happen inside
nfs_follow_mountpoint with NFS running Fedora 8 using a specific setup.

See https://bugzilla.redhat.com/show_bug.cgi?id=458622

So the situation should be handled on NFS client gracefully.


I'm not convinced that ESTALE is the correct error to return to the
user. ESTALE has a very precise meaning in the NFS protocol: it tells
you that access to the object has been revoked.

That again means either that either the object has been deleted, or it
is no longer exported by the server to this client.

In the case referred to by Denis, the source of the bug is a broken
server, and so I'd expect that EIO is probably more appropriate.)

Signed-off-by: Denis V. Lunev <[email protected]>
Cc: Trond Myklebust <[email protected]>
Cc: J. Bruce Fields <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

fs/nfs/namespace.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff -puN fs/nfs/namespace.c~nfs-bug_on-in-nfs_follow_mountpoint fs/nfs/namespace.c
--- a/fs/nfs/namespace.c~nfs-bug_on-in-nfs_follow_mountpoint
+++ a/fs/nfs/namespace.c
@@ -105,7 +105,10 @@ static void * nfs_follow_mountpoint(stru

dprintk("--> nfs_follow_mountpoint()\n");

- BUG_ON(IS_ROOT(dentry));
+ err = -ESTALE;
+ if (IS_ROOT(dentry))
+ goto out_err;
dprintk("%s: enter\n", __func__);
nd->path.dentry = dget(dentry);