hfsplus: fix oops on mount with corrupted btree extent records
authorJeff Mahoney <jeffm@suse.com>
Thu, 14 Oct 2010 13:53:37 +0000 (09:53 -0400)
committerChristoph Hellwig <hch@lst.de>
Thu, 14 Oct 2010 13:53:37 +0000 (09:53 -0400)
A particular fsfuzzer run caused an hfs file system to crash on mount. This
is due to a corrupted MDB extent record causing a miscalculation of
HFSPLUS_I(inode)->first_blocks for the extent tree. If the extent records
are zereod out, then it won't trigger the first_blocks special case and
instead falls through to the extent code, which we're in the middle
of initializing.

This patch catches the 0 size extent records, reports the corruption,
and fails the mount.

[hch: ported of commit 47f365eb575735c6b2edf5d08e0d16d26a9c23bd from hfs]

Reported-by: Ramon de Carvalho Valle <rcvalle@linux.vnet.ibm.com>
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Christoph Hellwig <hch@tuxera.com>
fs/hfsplus/btree.c

index 8306479f6b3ae7bf73df9921790af4b8e969be83..d4bd864ce1088205ab9c32706d163aa925323154 100644 (file)
@@ -39,10 +39,16 @@ struct hfs_btree *hfs_btree_open(struct super_block *sb, u32 id)
                goto free_tree;
        tree->inode = inode;
 
+       if (!HFSPLUS_I(tree->inode)->first_blocks) {
+               printk(KERN_ERR
+                      "hfs: invalid btree extent records (0 size).\n");
+               goto free_inode;
+       }
+
        mapping = tree->inode->i_mapping;
        page = read_mapping_page(mapping, 0, NULL);
        if (IS_ERR(page))
-               goto free_tree;
+               goto free_inode;
 
        /* Load the header */
        head = (struct hfs_btree_header_rec *)(kmap(page) + sizeof(struct hfs_bnode_desc));
@@ -89,8 +95,9 @@ struct hfs_btree *hfs_btree_open(struct super_block *sb, u32 id)
  fail_page:
        tree->inode->i_mapping->a_ops = &hfsplus_aops;
        page_cache_release(page);
- free_tree:
+ free_inode:
        iput(tree->inode);
+ free_tree:
        kfree(tree);
        return NULL;
 }