From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751413AbWB0Qba (ORCPT ); Mon, 27 Feb 2006 11:31:30 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751485AbWB0Qba (ORCPT ); Mon, 27 Feb 2006 11:31:30 -0500 Received: from newmail.sw.starentnetworks.com ([12.33.234.78]:30220 "EHLO mail.sw.starentnetworks.com") by vger.kernel.org with ESMTP id S1751413AbWB0Qb3 (ORCPT ); Mon, 27 Feb 2006 11:31:29 -0500 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <17411.10591.927433.619327@zeus.sw.starentnetworks.com> Date: Mon, 27 Feb 2006 11:31:27 -0500 From: Dave Johnson To: Olaf Hering Cc: linux-kernel@vger.kernel.org Subject: Re: cramfs mounts provide corrupted content since 2.6.15 In-Reply-To: <20060225220130.GA2748@suse.de> References: <20060225110844.GA18221@suse.de> <20060225125551.GA21203@suse.de> <17408.34808.422077.518881@zeus.sw.starentnetworks.com> <20060225220130.GA2748@suse.de> X-Mailer: VM 7.17 under 21.4 (patch 17) "Jumbo Shrimp" XEmacs Lucid Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Olaf Hering writes: > On Sat, Feb 25, Dave Johnson wrote: > > > Looking at your output it's definitely getting inodes confused with > > each other so the checks in cramfs_iget5_test() aren't working. > > > > Can you stat the files in question to make sure they are actually > > inode #1 on a working as well as non-working kernel? If your mkcramfs > > isn't using #1 for empty files/links/dirs that'd be the problem. > > Another try, and different results again: Is it the same files every time you mount/umount the image? I think I've spotted an issue. Both ifind() and find_inode() will call the test function on inodes that still have I_LOCK|I_NEW set. This means everything that the test function needs _must_ be set in the set function (which is called while the inode_lock is still held). This could cause issues for inodes of 1 (only i_ino is getting set right now). However since you're seeing issues for inodes != 1, it could indicate code elsewhere that isn't checking for I_LOCK|I_NEW. Anyway, can you give the following patch a try? -- Dave Johnson Starent Networks ====================================== Fill out inode contents in cramfs_iget5_set() instead of get_cramfs_inode() to prevent issues if cramfs_iget5_test() is called with I_LOCK|I_NEW still set. Signed-off-by: Dave Johnson diff -Naur linux-2.6.15.4.orig/fs/cramfs/inode.c linux-2.6.15.4/fs/cramfs/inode.c --- linux-2.6.15.4.orig/fs/cramfs/inode.c 2006-02-10 07:22:48.000000000 +0000 +++ linux-2.6.15.4/fs/cramfs/inode.c 2006-02-27 15:16:52.000000000 +0000 @@ -66,8 +66,36 @@ static int cramfs_iget5_set(struct inode *inode, void *opaque) { + static struct timespec zerotime; struct cramfs_inode *cramfs_inode = opaque; + inode->i_mode = cramfs_inode->mode; + inode->i_uid = cramfs_inode->uid; + inode->i_size = cramfs_inode->size; + inode->i_blocks = (cramfs_inode->size - 1) / 512 + 1; + inode->i_blksize = PAGE_CACHE_SIZE; + inode->i_gid = cramfs_inode->gid; + /* Struct copy intentional */ + inode->i_mtime = inode->i_atime = inode->i_ctime = zerotime; inode->i_ino = CRAMINO(cramfs_inode); + /* inode->i_nlink is left 1 - arguably wrong for directories, + but it's the best we can do without reading the directory + contents. 1 yields the right result in GNU find, even + without -noleaf option. */ + if (S_ISREG(inode->i_mode)) { + inode->i_fop = &generic_ro_fops; + inode->i_data.a_ops = &cramfs_aops; + } else if (S_ISDIR(inode->i_mode)) { + inode->i_op = &cramfs_dir_inode_operations; + inode->i_fop = &cramfs_directory_operations; + } else if (S_ISLNK(inode->i_mode)) { + inode->i_op = &page_symlink_inode_operations; + inode->i_data.a_ops = &cramfs_aops; + } else { + inode->i_size = 0; + inode->i_blocks = 0; + init_special_inode(inode, inode->i_mode, + old_decode_dev(cramfs_inode->size)); + } return 0; } @@ -77,37 +105,7 @@ struct inode *inode = iget5_locked(sb, CRAMINO(cramfs_inode), cramfs_iget5_test, cramfs_iget5_set, cramfs_inode); - static struct timespec zerotime; - if (inode && (inode->i_state & I_NEW)) { - inode->i_mode = cramfs_inode->mode; - inode->i_uid = cramfs_inode->uid; - inode->i_size = cramfs_inode->size; - inode->i_blocks = (cramfs_inode->size - 1) / 512 + 1; - inode->i_blksize = PAGE_CACHE_SIZE; - inode->i_gid = cramfs_inode->gid; - /* Struct copy intentional */ - inode->i_mtime = inode->i_atime = inode->i_ctime = zerotime; - inode->i_ino = CRAMINO(cramfs_inode); - /* inode->i_nlink is left 1 - arguably wrong for directories, - but it's the best we can do without reading the directory - contents. 1 yields the right result in GNU find, even - without -noleaf option. */ - if (S_ISREG(inode->i_mode)) { - inode->i_fop = &generic_ro_fops; - inode->i_data.a_ops = &cramfs_aops; - } else if (S_ISDIR(inode->i_mode)) { - inode->i_op = &cramfs_dir_inode_operations; - inode->i_fop = &cramfs_directory_operations; - } else if (S_ISLNK(inode->i_mode)) { - inode->i_op = &page_symlink_inode_operations; - inode->i_data.a_ops = &cramfs_aops; - } else { - inode->i_size = 0; - inode->i_blocks = 0; - init_special_inode(inode, inode->i_mode, - old_decode_dev(cramfs_inode->size)); - } unlock_new_inode(inode); } return inode;