From: Eric Sandeen <sandeen@redhat.com> Date: Wed, 19 Dec 2007 12:53:06 -0600 Subject: [fs] ext3: error in ext3_lookup if corruption found Message-id: 47696892.3020001@redhat.com O-Subject: [RHEL 5.2 PATCH, updated] - error in ext3_lookup if corruption is found Bugzilla: 181662 Bugzilla Bug 181662: Filesystem errors should be logged and not silently ignored I have a hand-crafted bad filesystem image which has corruption replicating that reported by the customer: [root@inode ~]# ls mnt/dir file1 file2 file3 file4 file5 [root@inode ~]# ls mnt/dir/file4 ls: cannot access mnt/dir/file4: No such file or directory [root@inode ~]# ls -l mnt/dir ls: cannot access mnt/dir/file4: No such file or directory total 8 drwxr-xr-x 2 root root 1024 2007-09-04 13:36 file1 drwxr-xr-x 2 root root 1024 2007-09-04 13:36 file2 drwxr-xr-x 2 root root 1024 2007-09-04 13:36 file3 d????????? ? ? ? ? ? file4 drwxr-xr-x 2 root root 1024 2007-09-04 13:36 file5 e2fsck also knows it's corrupted: Pass 2: Checking directory structure Entry 'file4' in /dir (2049) has deleted/unused inode 13. Clear? no Entry 'file4' in /dir (2049) has an incorrect filetype (was 2, should be 1). Fix? no Pass 3: Checking directory connectivity Unconnected directory inode 2053 (/dir/???) BUT there are no kernel messages logged anywhere because ext3_read_inode silently makes a bad_inode in this case, so that stale NFS filehandles aren't noisy. However, when we encounter such a problem after a by-name lookup, I think an ext3_error is appropriate, as it indicates filesystem corruption. I discussed the patch w/ Andreas Dilger, who though it looked reasonable, and I've sent it upstream to -mm, where Andrew says "It's in my backlog somewhere - I'll get to it." Signed-off-by: Eric Sandeen <sandeen@redhat.com> Acked-by: Peter Staubach <staubach@redhat.com> Acked-by: Jeff Layton <jlayton@redhat.com> diff --git a/fs/ext3/namei.c b/fs/ext3/namei.c index e97619c..35590db 100644 --- a/fs/ext3/namei.c +++ b/fs/ext3/namei.c @@ -1027,6 +1027,11 @@ static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry, str return ERR_PTR(-EACCES); if (is_bad_inode(inode)) { + /* if bad because unlinked, something has gone wrong */ + if (!inode->i_nlink && printk_ratelimit()) + ext3_error(inode->i_sb, __FUNCTION__, + "unlinked inode %lu in dir #%lu", + inode->i_ino, dir->i_ino); iput(inode); return ERR_PTR(-ENOENT); }