From: Li Dongyang Date: Fri, 9 Apr 2010 10:32:42 +0800 Subject: [Ocfs2-devel] [PATCH] ocfs2: alloc orphaned inode in ocfs2_symlink In-Reply-To: <20100408211222.GC20661@mail.oracle.com> References: <201004080913.08494.lidongyang@novell.com> <20100408211222.GC20661@mail.oracle.com> Message-ID: <201004091032.42548.lidongyang@novell.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ocfs2-devel@oss.oracle.com On Friday 09 April 2010 05:12:22 Joel Becker wrote: > On Thu, Apr 08, 2010 at 09:13:08AM +0800, Li Dongyang wrote: > > alloc the inode orphaned in ocfs2_symlink so we can delete the inode > > in the iput if we meet errors with the inode, e.g. ocfs2_add_inode_data. > > or ocfs2_query_inode_wipe will complain the inode is not orphaned on > > disk. > > Huh. At first I was sure this was ok, as the iput is sanely > killing the inode. But you're right, the wipe code expects it to be > orphaned. > Do you have a testcase for this? > > Joel > I found it while trying to reproduce the bug fixed by the patch ocfs2: avoid direct write if we fall back to buffered. In my case, just run fsstress on a single node, and when the ocfs2 partition is out of space, ocfs2_add_inode_data will return -ENOSPC and then will trigger the error in ocfs2_query_inode_wipe