Once commit 920b4530fb80430ff30ef83efe21ba1fa5623731 "NFS: nfs_rename()
handle -ERESTARTSYS dentry left behind" moved the local d_move() into the
RPC asyncronous context, d_move() could be called without holding the
directories' i_mutex.
Let's revert that commit, and a follow-up fix for it in 1/2, and then fix
the original problem once more by forcing a revalidation of the old and new
directories if we notice that the rename was interrupted in 2/2.
v2: Add memory barrier, hold directory i_locks while revalidating, specify
single-bit field width for cancelled flag.
v3: Add some comments, change cancelled flag to bool.
v4: Where old_dir == new_dir, don't bump revalidation counter twice.
Benjamin Coddington (2):
Revert "NFS: nfs_rename() handle -ERESTARTSYS dentry left behind"
NFS: nfs_rename() - revalidate directories on -ERESTARTSYS
fs/nfs/dir.c | 57 +++++++++++++++++++++++++------------------------
fs/nfs/unlink.c | 13 +++++++++++
include/linux/nfs_xdr.h | 1 +
3 files changed, 43 insertions(+), 28 deletions(-)
--
2.9.3
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
