autofs4 - fix expire check
authorIan Kent <raven@themaw.net>
Fri, 17 Aug 2012 03:09:04 +0000 (11:09 +0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 17 Aug 2012 13:56:39 +0000 (06:56 -0700)
In some cases when an autofs indirect mount is contained in a file
system that is marked as shared (such as when systemd does the
equivalent of "mount --make-rshared /" early in the boot), mounts
stop expiring.

When this happens the first expiry check on a mountpoint dentry in
autofs_expire_indirect() sees a mountpoint dentry with a higher
than minimal reference count. Consequently the dentry is condidered
busy and the actual expiry check is never done.

This particular check was originally meant as an optimisation to
detect a path walk in progress but with the addition of rcu-walk
it can be ineffective anyway.

Removing the test allows automounts to expire again since the
actual expire check doesn't rely on the dentry reference count.

Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/autofs4/expire.c

index 8c0e56d92938c49ad61fef0bfb3888f47429ebc9..842d00048a652f8cb9ebda7eeea40d62e9cb6720 100644 (file)
@@ -399,11 +399,6 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
                        DPRINTK("checking mountpoint %p %.*s",
                                dentry, (int)dentry->d_name.len, dentry->d_name.name);
 
-                       /* Path walk currently on this dentry? */
-                       ino_count = atomic_read(&ino->count) + 2;
-                       if (dentry->d_count > ino_count)
-                               goto next;
-
                        /* Can we umount this guy */
                        if (autofs4_mount_busy(mnt, dentry))
                                goto next;