From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Chinner Subject: Re: [RESEND][PATCH 2/2 v2] ext4: let us fully support punching hole feature in fallocate Date: Sat, 26 Jan 2013 11:47:07 +1100 Message-ID: <20130126004707.GG2667@dastard> References: <1358493381-20150-1-git-send-email-wenqing.lz@taobao.com> <1358493381-20150-2-git-send-email-wenqing.lz@taobao.com> <20130125033201.GD28908@thunk.org> <20130125035912.GA15806@gmail.com> <20130125042122.GG28908@thunk.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-ext4@vger.kernel.org, Zheng Liu , xfs@oss.sgi.com To: Theodore Ts'o Return-path: Received: from ipmail07.adl2.internode.on.net ([150.101.137.131]:14947 "EHLO ipmail07.adl2.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753441Ab3AZArM (ORCPT ); Fri, 25 Jan 2013 19:47:12 -0500 Content-Disposition: inline In-Reply-To: <20130125042122.GG28908@thunk.org> Sender: linux-ext4-owner@vger.kernel.org List-ID: [cc xfs@oss.sgi.com] On Thu, Jan 24, 2013 at 11:21:22PM -0500, Theodore Ts'o wrote: > On Fri, Jan 25, 2013 at 11:59:12AM +0800, Zheng Liu wrote: > > > > I wonder that maybe we need to submit a patch to let xfstest understand > > that a filesystem supports extents or not because after applied this > > patch indirect-based file in ext4 has supported seek_data/hole and hole > > punching. I usually run xfstest automatically, and every time I need > > to check the result of #255 and #285 manually. That is annoying for me. > > I would think the right thing to do is to have xfstests make sure it > understands that fallocate working with FALLOC_FL_PUNCH_HOLE does not > imply that fallocate without the FALLOC_FL_PUNCH_HOLE flag OR'ed in > will work. We already have this capabiity in xfstests via _require_xfs_io_falloc_punch and _require_xfs_io_falloc. That, however, doesn't mean the tests that use these calls do the correct requirement checks. That's the problem with 255 - it doesn't call _require_xfs_io_falloc. As to 285, the seek_sanity_test does it's own check for seek hole/data support, and error out if it fails. This needs to be turned into an equivalent _require_seek_hole_data (e.g. by running "seek_sanity_test -t" to test for support) and 285 needs to the call the _require_seek_hole_data before running the test proper. Please send patches to xfs@oss.sgi.com.... Cheers, Dave. -- Dave Chinner david@fromorbit.com