From: "Darrick J. Wong" Subject: [PATCH] ext4: Always verify extent tree blocks Date: Thu, 11 Aug 2011 14:13:41 -0700 Message-ID: <20110811211348.GK20655@tux1.beaverton.ibm.com> Reply-To: djwong@us.ibm.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-kernel , linux-ext4 To: "Theodore Ts'o" Return-path: Received: from e9.ny.us.ibm.com ([32.97.182.139]:56822 "EHLO e9.ny.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753840Ab1HKVO7 (ORCPT ); Thu, 11 Aug 2011 17:14:59 -0400 Content-Disposition: inline Sender: linux-ext4-owner@vger.kernel.org List-ID: It turns out that ext4_ext_check only verifies the validity of the extent block it's processing if the block has to be read in from the disk. Unfortunately, this means that the check is NOT done if the block is already in memory, which means that if a file has a corrupted extent block, then the first IO peformed on the file will find the corrupt block and fail, but a second IO will see that the extent block is in memory, bypass the corruption check, and use garbage data as if they were extent data. A simple testcase is to allocate a file with enough extents to overflow the inode i_block, umount, overwrite the extent block magic with garbage, then mount the filesystem and try to access the file. The first access causes the kernel to spit out an error, but subsequent accesses seem to succeed. Signed-off-by: Darrick J. Wong --- fs/ext4/extents.c | 6 +----- 1 files changed, 1 insertions(+), 5 deletions(-) diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index ee4b391..bb07b79 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c @@ -744,8 +744,6 @@ ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block, i = depth; /* walk through the tree */ while (i) { - int need_to_validate = 0;