ext4: avoid setting directory i_nlink to zero
authorAndreas Dilger <adilger@whamcloud.com>
Wed, 26 Oct 2011 07:22:31 +0000 (03:22 -0400)
committerTheodore Ts'o <tytso@mit.edu>
Wed, 26 Oct 2011 07:22:31 +0000 (03:22 -0400)
If a directory with more than EXT4_LINK_MAX subdirectories, the nlink
count is set to 1.  Subsequently, if any subdirectories are deleted,
ext4_dec_count() decrements the i_nlink count, which may go to 0
temporarily before being incremented back to 1.

While this is done under i_mutex, which prevents races for directory
and inode operations that check i_nlink, the temporary i_nlink == 0
case is exposed to userspace via stat() and similar calls that do not
hold i_mutex.

Instead, change the code to not decrement i_nlink count for any
directories that do not already have i_nlink larger than 2.

Reported-by: Cliff White <cliffw@whamcloud.com>
Reviewed-by: Johann Lombardi <johann@whamcloud.com>
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
fs/ext4/namei.c

index a067835..4a550aa 100644 (file)
@@ -1706,9 +1706,8 @@ static void ext4_inc_count(handle_t *handle, struct inode *inode)
  */
 static void ext4_dec_count(handle_t *handle, struct inode *inode)
 {
-       drop_nlink(inode);
-       if (S_ISDIR(inode->i_mode) && inode->i_nlink == 0)
-               inc_nlink(inode);
+       if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 2)
+               drop_nlink(inode);
 }