From: "Theodore Ts'o" Subject: Re: [PATCH] jbd2: fix ocfs2 corrupt when updating journal superblock fails Date: Fri, 19 Jun 2015 10:48:31 -0400 Message-ID: <20150619144831.GL4076@thunk.org> References: <5574F060.2080503@huawei.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Cc: Mark Fasheh , linux-ext4@vger.kernel.org, "ocfs2-devel@oss.oracle.com" To: Joseph Qi Return-path: Content-Disposition: inline In-Reply-To: <5574F060.2080503@huawei.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: ocfs2-devel-bounces@oss.oracle.com Errors-To: ocfs2-devel-bounces@oss.oracle.com List-Id: linux-ext4.vger.kernel.org This patch caused test ext4/306 to fail, because it caused resize2fs to fail. The problem is that jbd2_cleanup_journal_tail() will return 1 if there is nothing to cleanup, and a negative error number if there is an error. Unfortunately, this patch hunk: On Mon, Jun 08, 2015 at 09:31:12AM +0800, Joseph Qi wrote: > diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c > index b96bd80..6b33a42 100644 > --- a/fs/jbd2/journal.c > +++ b/fs/jbd2/journal.c > @@ -1950,7 +1966,13 @@ int jbd2_journal_flush(journal_t *journal) > return -EIO; > > mutex_lock(&journal->j_checkpoint_mutex); > - jbd2_cleanup_journal_tail(journal); > + if (!err) { > + err = jbd2_cleanup_journal_tail(journal); > + if (err < 0) { > + mutex_unlock(&journal->j_checkpoint_mutex); > + goto out; > + } > + } ... would let the non-negative return value leak out to jbd2_journal_flush(), and its callers are *not* prepared to handle the non-negative return value (since jbd2_journal_flush wasn't doing this before.) I've fixed this by adding a "err = 0;" after the if statement. - Ted