[PATCH] ext3/EA: Race in ext[23] xattr sharing code

Andrew Tridgell and Stephen C.  Tweedie have reported two different Oopses
caused by a race condition in the mbcache, which is responsible for
extended attribute sharing in ext2 and ext3.  Stephen tracked down the bug;
I did the fix.

Explanation:

The mbcache caches the locations and content hashes of xattr blocks.  There
are two access strategies: [1] xattr block disposal via
mb_cache_entry_get(), [2] xattr block reuse (sharing) via
mb_cache_entry_find_{first,next}().  There is no locking between the two
methods, so between one mb_cache_entry_find_x and the next, a
mb_cache_entry_get might come in, unhash the cache entry, and change the
journaling state of the xattr buffer.  Subsequently, two things can happen:
[a] the next mb_cache_entry_find_x may try to follow the mbcache hash chain
starting from the entry that has become unhashed, which now is a stale
pointer, [b] the block may have become deallocated, and then we try to
reuse it.

Fix this by converting the mbcache into a readers-writer style lock, and
protect all block accesses in ext2/ext3 by the mbcache entry lock.  This
ensures that destroying blocks is an exclusive operation that may not
overlap xattr block reuse, while allowing multiple "re-users".  Write
access to the xattr block's buffer is protected by the buffer lock.  

Signed-off-by: Andreas Gruenbacher <agruen@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

1 files changed, 31 insertions, 24 deletions

diff --git a/fs/ext2/xattr.c b/fs/ext2/xattr.c
index d3cbbdd8b726b2..7560d9eec5c162 100644
--- a/fs/ext2/xattr.c
+++ b/fs/ext2/xattr.c
@@ -95,7 +95,6 @@ static int ext2_xattr_set2(struct inode *, struct buffer_head *,
 static int ext2_xattr_cache_insert(struct buffer_head *);
 static struct buffer_head *ext2_xattr_cache_find(struct inode *,
 						 struct ext2_xattr_header *);
-static void ext2_xattr_cache_remove(struct buffer_head *);
 static void ext2_xattr_rehash(struct ext2_xattr_header *,
 			      struct ext2_xattr_entry *);
 
@@ -494,15 +493,22 @@ bad_block:		ext2_error(sb, "ext2_xattr_set",
 	/* Here we know that we can set the new attribute. */
 
 	if (header) {
+		struct mb_cache_entry *ce;
+
 		/* assert(header == HDR(bh)); */
+		ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_bdev,
+					bh->b_blocknr);
 		lock_buffer(bh);
 		if (header->h_refcount == cpu_to_le32(1)) {
 			ea_bdebug(bh, "modifying in-place");
-			ext2_xattr_cache_remove(bh);
+			if (ce)
+				mb_cache_entry_free(ce);
 			/* keep the buffer locked while modifying it. */
 		} else {
 			int offset;
 
+			if (ce)
+				mb_cache_entry_release(ce);
 			unlock_buffer(bh);
 			ea_bdebug(bh, "cloning");
 			header = kmalloc(bh->b_size, GFP_KERNEL);
@@ -707,13 +713,19 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 
 	error = 0;
 	if (old_bh && old_bh != new_bh) {
+		struct mb_cache_entry *ce;
+
 		/*
 		 * If there was an old block and we are no longer using it,
 		 * release the old block.
 		 */
+		ce = mb_cache_entry_get(ext2_xattr_cache, old_bh->b_bdev,
+					old_bh->b_blocknr);
 		lock_buffer(old_bh);
 		if (HDR(old_bh)->h_refcount == cpu_to_le32(1)) {
 			/* Free the old block. */
+			if (ce)
+				mb_cache_entry_free(ce);
 			ea_bdebug(old_bh, "freeing");
 			ext2_free_blocks(inode, old_bh->b_blocknr, 1);
 			/* We let our caller release old_bh, so we
@@ -724,6 +736,8 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 			/* Decrement the refcount only. */
 			HDR(old_bh)->h_refcount = cpu_to_le32(
 				le32_to_cpu(HDR(old_bh)->h_refcount) - 1);
+			if (ce)
+				mb_cache_entry_release(ce);
 			DQUOT_FREE_BLOCK(inode, 1);
 			mark_buffer_dirty(old_bh);
 			ea_bdebug(old_bh, "refcount now=%d",
@@ -748,6 +762,7 @@ void
 ext2_xattr_delete_inode(struct inode *inode)
 {
 	struct buffer_head *bh = NULL;
+	struct mb_cache_entry *ce;
 
 	down_write(&EXT2_I(inode)->xattr_sem);
 	if (!EXT2_I(inode)->i_file_acl)
@@ -767,15 +782,19 @@ ext2_xattr_delete_inode(struct inode *inode)
 			EXT2_I(inode)->i_file_acl);
 		goto cleanup;
 	}
+	ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_bdev, bh->b_blocknr);
 	lock_buffer(bh);
 	if (HDR(bh)->h_refcount == cpu_to_le32(1)) {
-		ext2_xattr_cache_remove(bh);
+		if (ce)
+			mb_cache_entry_free(ce);
 		ext2_free_blocks(inode, EXT2_I(inode)->i_file_acl, 1);
 		get_bh(bh);
 		bforget(bh);
 	} else {
 		HDR(bh)->h_refcount = cpu_to_le32(
 			le32_to_cpu(HDR(bh)->h_refcount) - 1);
+		if (ce)
+			mb_cache_entry_release(ce);
 		mark_buffer_dirty(bh);
 		if (IS_SYNC(inode))
 			sync_dirty_buffer(bh);
@@ -892,11 +911,19 @@ ext2_xattr_cache_find(struct inode *inode, struct ext2_xattr_header *header)
 	if (!header->h_hash)
 		return NULL;  /* never share */
 	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
+again:
 	ce = mb_cache_entry_find_first(ext2_xattr_cache, 0,
 				       inode->i_sb->s_bdev, hash);
 	while (ce) {
-		struct buffer_head *bh = sb_bread(inode->i_sb, ce->e_block);
+		struct buffer_head *bh;
 
+		if (IS_ERR(ce)) {
+			if (PTR_ERR(ce) == -EAGAIN)
+				goto again;
+			break;
+		}
+
+		bh = sb_bread(inode->i_sb, ce->e_block);
 		if (!bh) {
 			ext2_error(inode->i_sb, "ext2_xattr_cache_find",
 				"inode %ld: block %ld read error",
@@ -923,26 +950,6 @@ ext2_xattr_cache_find(struct inode *inode, struct ext2_xattr_header *header)
 	return NULL;
 }
 
-/*
- * ext2_xattr_cache_remove()
- *
- * Remove the cache entry of a block from the cache. Called when a
- * block becomes invalid.
- */
-static void
-ext2_xattr_cache_remove(struct buffer_head *bh)
-{
-	struct mb_cache_entry *ce;
-
-	ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_bdev, bh->b_blocknr);
-	if (ce) {
-		ea_bdebug(bh, "removing (%d cache entries remaining)",
-			  atomic_read(&ext2_xattr_cache->c_entry_count)-1);
-		mb_cache_entry_free(ce);
-	} else 
-		ea_bdebug(bh, "no cache entry");
-}
-
 #define NAME_HASH_SHIFT 5
 #define VALUE_HASH_SHIFT 16