From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Sandeen Subject: [PATCH] limit minixfs dir_pages on corrupted dir i_size, CVE-2006-6058 Date: Thu, 09 Aug 2007 13:46:25 -0500 Message-ID: <46BB6101.4030307@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit To: linux-kernel Mailing List , linux-fsdevel Return-path: Received: from mx1.redhat.com ([66.187.233.31]:53537 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752176AbXHISqZ (ORCPT ); Thu, 9 Aug 2007 14:46:25 -0400 Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org This attempts to address CVE-2006-6058 http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-6058 first reported at http://projects.info-pull.com/mokb/MOKB-17-11-2006.html Essentially a corrupted minix dir inode reporting a very large i_size will loop for a very long time in minix_readdir, minix_find_entry, etc, because on EIO they just move on to try the next page. This is under the BKL, printk'ing as well. This can lock up the machine for a very long time. A simple approach is to at least limit the nr. of pages attempted to no more than s_max_size. (s_max_size is about 256MB for V1, but 2GB for V2; this could still result in a lot of EIO reads in the V2 case, should the retry loops in minix_readdir & friends be short-circuited somehow instead? A simple "break" rather than "continue" on error would certainly resolve it, too...) Signed-off-by: Eric Sandeen --- linux-2.6.22-rc4.orig/fs/minix/dir.c +++ linux-2.6.22-rc4/fs/minix/dir.c @@ -42,7 +42,15 @@ minix_last_byte(struct inode *inode, uns static inline unsigned long dir_pages(struct inode *inode) { - return (inode->i_size+PAGE_CACHE_SIZE-1)>>PAGE_CACHE_SHIFT; + loff_t size = inode->i_size; + + if (size > minix_sb(inode->i_sb)->s_max_size) { + printk("%s: inode %lld i_size > s_max_size\n", + __FUNCTION__, inode->i_size); + size = minix_sb(inode->i_sb)->s_max_size; + } + + return (size+PAGE_CACHE_SIZE-1)>>PAGE_CACHE_SHIFT; } static int dir_commit_chunk(struct page *page, unsigned from, unsigned to)