From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261686AbVE0GhP (ORCPT ); Fri, 27 May 2005 02:37:15 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261741AbVE0GhP (ORCPT ); Fri, 27 May 2005 02:37:15 -0400 Received: from fmr23.intel.com ([143.183.121.15]:60825 "EHLO scsfmr003.sc.intel.com") by vger.kernel.org with ESMTP id S261686AbVE0GhH (ORCPT ); Fri, 27 May 2005 02:37:07 -0400 Date: Thu, 26 May 2005 23:36:59 -0700 From: "Siddha, Suresh B" To: Andrew Morton Cc: "Siddha, Suresh B" , linux-kernel@vger.kernel.org, venkatesh.pallipadi@intel.com, shaohua.li@intel.com, ashok.raj@intel.com Subject: Re: [Patch] x86: fix smp_num_siblings on buggy BIOSes Message-ID: <20050526233658.A28476@unix-os.sc.intel.com> References: <20050525173456.A11038@unix-os.sc.intel.com> <20050526221737.7fc42984.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5.1i In-Reply-To: <20050526221737.7fc42984.akpm@osdl.org>; from akpm@osdl.org on Thu, May 26, 2005 at 10:17:37PM -0700 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, May 26, 2005 at 10:17:37PM -0700, Andrew Morton wrote: > "Siddha, Suresh B" wrote: > > > > Appended patch will fix 'smp_num_siblings' value on the systems with > > buggy bios, which sets number of siblings to '2' even when HT is disabled. > > (more details are at http://bugzilla.kernel.org/show_bug.cgi?id=4359) > > > > I am planning to do more cleanup in this area (like moving smp_num_siblings > > to per cpuinfo) shortly. > > > > But meanwhile, please pickup this patch for 2.6.12. > > > > Signed-off-by: Suresh Siddha > > Signed-off-by: Venkatesh Pallipadi > > > > --- linux-2.6.12-rc5/arch/i386/kernel/smpboot.c~ 2005-05-25 16:47:56.462416368 -0700 > > +++ linux-2.6.12-rc5/arch/i386/kernel/smpboot.c 2005-05-25 16:48:14.556665624 -0700 > > @@ -1074,8 +1074,10 @@ > > cpu_set(cpu, cpu_sibling_map[cpu]); > > } > > > > - if (siblings != smp_num_siblings) > > + if (siblings != smp_num_siblings) { > > printk(KERN_WARNING "WARNING: %d siblings found for CPU%d, should be %d\n", siblings, cpu, smp_num_siblings); > > + smp_num_siblings = siblings; > > + } > > > > if (c->x86_num_cores > 1) { > > for (i = 0; i < NR_CPUS; i++) { > > OK, but be aware that -mm's sibling-map-initializing-rework.patch > cheerfully deletes all the surrounding code. So -mm will not have this > fix. > > There's a completely loony amount of x86 stuff in -mm. Thanks for this info. I will send a fix for -mm also. BTW, when do you plan to integrate x86 cpu hotplug patches to mainline? Depending on that I will plan my "smp_num_siblings" global cleanup. thanks, suresh