locks: close potential race between setlease and open
authorJeff Layton <jlayton@redhat.com>
Mon, 3 Feb 2014 17:13:06 +0000 (12:13 -0500)
committerJeff Layton <jlayton@redhat.com>
Mon, 31 Mar 2014 12:24:42 +0000 (08:24 -0400)
commit24cbe7845ea50b636ab2218b9d648270ff55f148
tree89769bf1271ff804c830d0ba7898294a5762b7a6
parent18156e7e66e5147d2483382a1e3817d2401ca1a8
locks: close potential race between setlease and open

As Al Viro points out, there is an unlikely, but possible race between
opening a file and setting a lease on it. generic_add_lease is done with
the i_lock held, but the inode->i_flock check in break_lease is
lockless. It's possible for another task doing an open to do the entire
pathwalk and call break_lease between the point where generic_add_lease
checks for a conflicting open and adds the lease to the list. If this
occurs, we can end up with a lease set on the file with a conflicting
open.

To guard against that, check again for a conflicting open after adding
the lease to the i_flock list. If the above race occurs, then we can
simply unwind the lease setting and return -EAGAIN.

Because we take dentry references and acquire write access on the file
before calling break_lease, we know that if the i_flock list is empty
when the open caller goes to check it then the necessary refcounts have
already been incremented. Thus the additional check for a conflicting
open will see that there is one and the setlease call will fail.

Cc: Bruce Fields <bfields@fieldses.org>
Cc: David Howells <dhowells@redhat.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@fieldses.org>
fs/locks.c
include/linux/fs.h