From: Jan Kara Subject: [PATCH 0/4] Fix fsync on ext3 and ext4 (v2) Date: Tue, 27 Oct 2009 13:48:45 +0100 Message-ID: <1256647729-29834-1-git-send-email-jack@suse.cz> Cc: tytso@mit.edu To: linux-ext4@vger.kernel.org Return-path: Received: from cantor.suse.de ([195.135.220.2]:53090 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753957AbZJ0Msq (ORCPT ); Tue, 27 Oct 2009 08:48:46 -0400 Sender: linux-ext4-owner@vger.kernel.org List-ID: Hi, this is a second try for a patchset which makes ext3 and ext4 properly force a transaction commit when needed. We now do not rely on buffer dirty bits (which does not work as pdflush can just clear them without forcing a transaction commit) but rather keep transaction ids that need to be committed for each inode. Since last version, I've implemented Aneesh's and Curt's comments and also fixed a missing initialization of the fields. I've tested that now the patch works correctly for uninitialized extents as well as for standard writes. If noone objects, would you merge the ext4 part Ted? I'll take care of the ext3 patch. Honza