[PATCH] autofs4: check for directory re-create in lookup
authorIan Kent <raven@themaw.net>
Tue, 20 Feb 2007 21:58:10 +0000 (13:58 -0800)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Wed, 21 Feb 2007 01:10:15 +0000 (17:10 -0800)
This problem was identified and fixed some time ago by Jeff Moyer but it fell
through the cracks somehow.

It is possible that a user space application could remove and re-create a
directory during a request.  To avoid returning a failure from lookup
incorrectly when our current dentry is unhashed we need to check if another
positive, hashed dentry matching this one exists and if so return it instead
of a fail.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/autofs4/root.c

index 47adf270106a656ea98b54cf6875a66cfae70ab5..b4631046867e30a92e2bf5b866d9bc8a05e0c375 100644 (file)
@@ -655,14 +655,29 @@ static struct dentry *autofs4_lookup(struct inode *dir, struct dentry *dentry, s
 
        /*
         * If this dentry is unhashed, then we shouldn't honour this
-        * lookup even if the dentry is positive.  Returning ENOENT here
-        * doesn't do the right thing for all system calls, but it should
-        * be OK for the operations we permit from an autofs.
+        * lookup.  Returning ENOENT here doesn't do the right thing
+        * for all system calls, but it should be OK for the operations
+        * we permit from an autofs.
         */
        if (dentry->d_inode && d_unhashed(dentry)) {
+               /*
+                * A user space application can (and has done in the past)
+                * remove and re-create this directory during the callback.
+                * This can leave us with an unhashed dentry, but a
+                * successful mount!  So we need to perform another
+                * cached lookup in case the dentry now exists.
+                */
+               struct dentry *parent = dentry->d_parent;
+               struct dentry *new = d_lookup(parent, &dentry->d_name);
+               if (new != NULL)
+                       dentry = new;
+               else
+                       dentry = ERR_PTR(-ENOENT);
+
                if (unhashed)
                        dput(unhashed);
-               return ERR_PTR(-ENOENT);
+
+               return dentry;
        }
 
        if (unhashed)