From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753964AbaEMSgY (ORCPT ); Tue, 13 May 2014 14:36:24 -0400 Received: from mx1.redhat.com ([209.132.183.28]:53286 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752364AbaEMSgP (ORCPT ); Tue, 13 May 2014 14:36:15 -0400 Date: Tue, 13 May 2014 20:36:01 +0200 From: Radim =?utf-8?B?S3LEjW3DocWZ?= To: David Vrabel Cc: linux-kernel@vger.kernel.org, Andrew Jones , Konrad Rzeszutek Wilk , Boris Ostrovsky , Thomas Gleixner , Ingo Molnar , "H. Peter Anvin" , x86@kernel.org, xen-devel@lists.xenproject.org Subject: Re: [PATCH] xen/x86: set panic notifier priority to minimum Message-ID: <20140513183600.GA10992@potion.brq.redhat.com> References: <1400000185-15148-1-git-send-email-rkrcmar@redhat.com> <537253C8.7000401@citrix.com> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <537253C8.7000401@citrix.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org 2014-05-13 18:18+0100, David Vrabel: > On 13/05/14 17:56, Radim Krčmář wrote: > > Execution is not going to continue after telling Xen about the crash. > > Let other panic notifiers run by postponing the final hypercall as much > > as possible. > > I can't tell how important this fix is. Sorry, I could have put reasoning in the commit: Not very important, depends on long linux+xen is going to last. This patch is thinking about users/developers that would eventually find a problem with it, and questioned our inteligence -- it makes no sense to use the default priority for a notifier that is (can be) final. INT_MIN is set because I don't think that Xen indended to resume domains after crash. (We should reason any other value as well.) > What notifiers does this allow to be run? How important are they? Priority 0 (default) or lower. Registration is stable and paravirt does it early, so there won't be many priority 0 users before it. (None now.) Hardware that can be passed through (most notably graphic cards), and few debug markers (dump_kernel_offset, hung_task_panic) are affected at the moment.