Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932245AbcJNIFn (ORCPT ); Fri, 14 Oct 2016 04:05:43 -0400 Received: from mx1.redhat.com ([209.132.183.28]:10670 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751423AbcJNIF3 (ORCPT ); Fri, 14 Oct 2016 04:05:29 -0400 Date: Fri, 14 Oct 2016 10:05:24 +0200 From: Andrew Jones To: Laszlo Ersek Cc: Zhen Lei , Will Deacon , main kernel list , linux-arm-kernel@lists.infradead.org, Ard Biesheuvel , Shannon Zhao , Wei Huang Subject: Re: aarch64 ACPI boot regressed by commit 7ba5f605f3a0 ("arm64/numa: remove the limitation that cpu0 must bind to node0") Message-ID: <20161014080524.4hm2b4p373r7rhel@hawk.localdomain> References: <4a64cd93-5ead-aad6-1057-f42224d65b43@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <4a64cd93-5ead-aad6-1057-f42224d65b43@redhat.com> User-Agent: Mutt/1.6.0.1 (2016-04-01) X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.31]); Fri, 14 Oct 2016 08:05:29 +0000 (UTC) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2270 Lines: 62 On Fri, Oct 14, 2016 at 12:50:29AM +0200, Laszlo Ersek wrote: > (4) Analysis (well, a lame attempt at that, because I have zero > familiarity with this code). Let me quote the patch: > > > commit 7ba5f605f3a0d9495aad539eeb8346d726dfc183 > > Author: Zhen Lei > > Date: Thu Sep 1 14:55:04 2016 +0800 > > > > arm64/numa: remove the limitation that cpu0 must bind to node0 > > > > 1. Remove the old binding code. > > 2. Read the nid of cpu0 from dts. > > 3. Fallback the nid of cpu0 to 0 when numa=off is set in bootargs. > > > > Signed-off-by: Zhen Lei > > Signed-off-by: Will Deacon > > > > diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c > > index c3c08368a685..8b048e6ec34a 100644 > > --- a/arch/arm64/kernel/smp.c > > +++ b/arch/arm64/kernel/smp.c > > @@ -624,6 +624,7 @@ static void __init of_parse_and_init_cpus(void) > > } > > > > bootcpu_valid = true; > > + early_map_cpu_to_node(0, of_node_to_nid(dn)); > > > > /* > > * cpu_logical_map has already been > > diff --git a/arch/arm64/mm/numa.c b/arch/arm64/mm/numa.c > > index 0a15f010b64a..778a985c8a70 100644 > > --- a/arch/arm64/mm/numa.c > > +++ b/arch/arm64/mm/numa.c > > @@ -116,16 +116,24 @@ static void __init setup_node_to_cpumask_map(void) > > */ > > void numa_store_cpu_info(unsigned int cpu) > > { > > - map_cpu_to_node(cpu, numa_off ? 0 : cpu_to_node_map[cpu]); > > + map_cpu_to_node(cpu, cpu_to_node_map[cpu]); > > } > > > > void __init early_map_cpu_to_node(unsigned int cpu, int nid) > > { > > /* fallback to node 0 */ > > - if (nid < 0 || nid >= MAX_NUMNODES) > > + if (nid < 0 || nid >= MAX_NUMNODES || numa_off) > > nid = 0; The ACPI equivalent code must be missing (at least) the above, because, even with DT, mach-virt won't have cpu to node mappings unless numa is configured on the command line. Can you try adding something like -m 512 -smp 4 \ -numa node,mem=256M,cpus=0-1,nodeid=0 \ -numa node,mem=256M,cpus=2-3,nodeid=1 to your QEMU command line? Then when you boot with ACPI you'll get a SRAT. If that works, then we're just missing the "no SRAT, nid = 0" code (that should have been added with this patch) Thanks, drew