From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from merlin.infradead.org ([205.233.59.134]:34382 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S964946Ab3IDRwy (ORCPT ); Wed, 4 Sep 2013 13:52:54 -0400 Message-ID: <52277371.5070004@kernel.dk> Date: Wed, 04 Sep 2013 11:52:49 -0600 From: Jens Axboe MIME-Version: 1.0 Subject: Re: How to loop a timebased job References: <522639CE.2030705@enovance.com> <20130903223727.GB31170@kernel.dk> <52273BF5.90109@enovance.com> In-Reply-To: <52273BF5.90109@enovance.com> Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: fio-owner@vger.kernel.org List-Id: fio@vger.kernel.org To: Erwan Velu Cc: "fio@vger.kernel.org" On 09/04/2013 07:56 AM, Erwan Velu wrote: > On 04/09/2013 00:37, Jens Axboe wrote: >> It should work, if it doesn't, I agree it's a bug. It's not certain >> than anyone has considered this specific use case before. Feel free to >> fix! Or I can take a look at it tomorrow. > I think I found the issue but my main question is about priority. > > If both time & loops are set, what shall be considered as the key > element to stop the benchmark. > > Today, time have the priority over loops as keep_running() returns if > time_based while time is handled in do_io(). > > If I do remove this test in keep_running, the first of the two getting > the limit will stop the job. > On one hand, we have timebased option to insure time over end of device. > Shall we put a loopbased on the other hand ? Good point, yes, that is a problem. Adding a loopbased option would seem to be the best way to solve the dependency issue. -- Jens Axboe