From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932227AbWFGMNQ (ORCPT ); Wed, 7 Jun 2006 08:13:16 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932228AbWFGMNQ (ORCPT ); Wed, 7 Jun 2006 08:13:16 -0400 Received: from cantor2.suse.de ([195.135.220.15]:481 "EHLO mx2.suse.de") by vger.kernel.org with ESMTP id S932227AbWFGMNQ (ORCPT ); Wed, 7 Jun 2006 08:13:16 -0400 From: Andi Kleen To: Keith Owens Subject: Re: NMI problems with Dell SMP Xeons Date: Wed, 7 Jun 2006 14:13:00 +0200 User-Agent: KMail/1.9.3 Cc: Ashok Raj , linux-kernel@vger.kernel.org, "Brendan Trotter" References: <6799.1149680860@ocs3.ocs.com.au> In-Reply-To: <6799.1149680860@ocs3.ocs.com.au> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200606071413.00990.ak@suse.de> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Wednesday 07 June 2006 13:47, Keith Owens wrote: > Andi Kleen (on Wed, 7 Jun 2006 10:01:40 +0200) wrote: > > > >> > >> Two ways: > >> > >> (1) Boot with a kernel with CONFIG_ACPI=n, so the OS only finds 2 cpus > >> in the MPT instead of the 4 listed by ACPI. > >> > >> (2) The kernel has ACPI=y, but is booted with maxcpus=2. > >> > >> In both cases, send_IPI_allbutself() with IPI 2 or an NMI will result > >> in a hard reset. > > > >Sounds both like a "Don't do that when it hurts" . I know some people > >have religious issues with ACPI, but it's simple a fact that many > >modern boxes don't work correctly in obvious or subtle ways without it. > > Building a kernel without ACPI is silly nowadays. But even with ACPI, > booting with a restricted maxcpus and sending IPI 2 or NMI as broadcast > will kill these boxes. maxcpus is a valid option. Yes, I'm not opposed to using *_sequence on cpu_online_map here. Please send a patch. -Andi