[PATCH] fix deadlock in ext2
Peter Staubach [Fri, 17 Feb 2006 21:52:36 +0000 (13:52 -0800)]
Fix a deadlock possible in the ext2 file system implementation.  This
deadlock occurs when a file is removed from an ext2 file system which was
mounted with the "sync" mount option.

The problem is that ext2_xattr_delete_inode() was invoking the routine,
sync_dirty_buffer(), using a buffer head which was previously locked via
lock_buffer().  The first thing that sync_dirty_buffer() does is to lock
the buffer head that it was passed.  It does this via lock_buffer().  Oops.

The solution is to unlock the buffer head in ext2_xattr_delete_inode()
before invoking sync_dirty_buffer().  This makes the code in
ext2_xattr_delete_inode() obey the same locking rules as all other callers
of sync_dirty_buffer() in the ext2 file system implementation.

Signed-off-by: Peter Staubach <staubach@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/ext2/xattr.c

index a2ca310..86ae8e9 100644 (file)
@@ -792,18 +792,20 @@ ext2_xattr_delete_inode(struct inode *inode)
                ext2_free_blocks(inode, EXT2_I(inode)->i_file_acl, 1);
                get_bh(bh);
                bforget(bh);
+               unlock_buffer(bh);
        } else {
                HDR(bh)->h_refcount = cpu_to_le32(
                        le32_to_cpu(HDR(bh)->h_refcount) - 1);
                if (ce)
                        mb_cache_entry_release(ce);
+               ea_bdebug(bh, "refcount now=%d",
+                       le32_to_cpu(HDR(bh)->h_refcount));
+               unlock_buffer(bh);
                mark_buffer_dirty(bh);
                if (IS_SYNC(inode))
                        sync_dirty_buffer(bh);
                DQUOT_FREE_BLOCK(inode, 1);
        }
-       ea_bdebug(bh, "refcount now=%d", le32_to_cpu(HDR(bh)->h_refcount) - 1);
-       unlock_buffer(bh);
        EXT2_I(inode)->i_file_acl = 0;
 
 cleanup: