Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758240AbcDEOfA (ORCPT ); Tue, 5 Apr 2016 10:35:00 -0400 Received: from mx1.redhat.com ([209.132.183.28]:50611 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753752AbcDEOe7 (ORCPT ); Tue, 5 Apr 2016 10:34:59 -0400 Date: Tue, 5 Apr 2016 09:34:56 -0500 From: Josh Poimboeuf To: Miroslav Benes Cc: Jiri Kosina , Jessica Yu , linux-kernel@vger.kernel.org, live-patching@vger.kernel.org, Vojtech Pavlik Subject: Re: [RFC PATCH v1.9 00/14] livepatch: hybrid consistency model Message-ID: <20160405143456.jyr4iol4jw233xll@treble.redhat.com> References: <20160404170310.bpbb5uhaz423hd6f@treble.redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.23.1 (2014-03-12) X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.38]); Tue, 05 Apr 2016 14:34:58 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1443 Lines: 34 On Tue, Apr 05, 2016 at 04:24:33PM +0200, Miroslav Benes wrote: > On Mon, 4 Apr 2016, Josh Poimboeuf wrote: > > > > I'd add a fake signal facility for sleeping non-migrated tasks. This > > > would accelerate a migration to a new universe. We have it in kgraft for > > > quite some time and it worked out great. See > > > lkml.kernel.org/r/1430739625-4658-9-git-send-email-jslaby@suse.cz which > > > went with Jiri's kgraft-on-klp patch set. See also Oleg's reply as it is > > > important (I changed kgraft implementation according to that). > > > > Ok, I'll look into sending a fake signal to remaining tasks. > > Oh, do not worry about this one. I can prepare a patch once we discuss and > review the rest. I think you have a lot on your plate even now. > > There are several options how to do it. I think it would be nice not send > a fake signal immediately. So we can... > > 1. do it after some time of waiting, or > > 2. we can have a knob in sysfs for an admin to send a fake signal if he > does not want to wait anymore, or > > 3. (and this is crazy one) we can trace the progress of a migration and > when there is none for several iterations we can send a signal Ok, thanks, I'll skip it for now. We do need to think about the options and I do have a full plate already. I might also procrastinate with the patch module removal patch, since that's another separate piece which might need more discussion. -- Josh