From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-4.0 required=3.0 tests=DKIM_SIGNED,DKIM_VALID, DKIM_VALID_AU,FREEMAIL_FORGED_FROMDOMAIN,FREEMAIL_FROM, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_PATCH,MAILING_LIST_MULTI,SPF_PASS, T_MIXED_ES autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id ABA79C65BAE for ; Thu, 13 Dec 2018 09:04:21 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 67F772086D for ; Thu, 13 Dec 2018 09:04:21 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="kB1pv8ng" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 67F772086D Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=gmail.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727472AbeLMJEU (ORCPT ); Thu, 13 Dec 2018 04:04:20 -0500 Received: from mail-ed1-f68.google.com ([209.85.208.68]:47071 "EHLO mail-ed1-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726416AbeLMJER (ORCPT ); Thu, 13 Dec 2018 04:04:17 -0500 Received: by mail-ed1-f68.google.com with SMTP id o10so1334295edt.13 for ; Thu, 13 Dec 2018 01:04:15 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20161025; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=5Jau0KZIbSwhBmgsfQIHq6Kyerd1wbla26kBo3eWxvE=; b=kB1pv8ng7UFTuT2P707SCwz+dOi9JTc++agFI2b9+4WiyDw1TfYcosrIzR7JL4pYRh JLRknpub+4c7xjklWVgbLL5UB4zMru/izGmN09pyza+tkm4/94ovqo6UhxaAZsHZwjxD 513y8ink/xMwyv36n9j3K13TebAjHqIcIMftPzwCY53l3wKCzM3/bIwOD07DcUxde9Jk 0L7G46Uk0cUx4ur1BrryT0U1Dy3jm7sx6Rg4A+f0Hi9RR0Mq5zOImQ+4Dt0aA70qtgUQ zcuQiOnid7FwMfpg61lfSQUh8599fhhJZfscY0Gv9ApGDHMCjeMcxGmvs1jxx5HVyqdh B4Fg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=5Jau0KZIbSwhBmgsfQIHq6Kyerd1wbla26kBo3eWxvE=; b=rjqQMg6G8xvG24iWLnGD+R4ooJ316KWWtWoS4RGFxamB+CWbojNNgWJx92gQ6cMiuF 8Bh3uF2vJp8n3ff439+71D7G6Krgovkp1gMohOEBXAYb3iqDF04a9m/YLq9KGEgEaa0p Cy+yC5oLgnH23ce0jeWkjC9JgzU3V1BdOrpZNeLwfiqaovszM3loMOVVjluUUeDW9BLg oPvMwoO/Fuo+1XEXtgQS8tg7FRJ+kAZcgmtpQYUDHOpvBenX2Xud2tN8I4bWwL0IAW/N wdECNJP0kZuyEAOSBREGlryjby4jxsKi7TNIJjvKOD/8oKp8pxwY7a5oDnbUc1BDWqrd 5HOQ== X-Gm-Message-State: AA+aEWYL5sKQ5IlOm7gOZQxtJbkgMYEzTR4m7E7RLh4YWhSUigTktb3p NwXA4byuFKZgmaHXbS+hx9p8iAwp43rSngAhiw== X-Google-Smtp-Source: AFSGD/WMusQ8CF5w83puUAMvcs+v/aZCJwFa2MNPzDUEvK0/3TTOI0JVfrgFMkN+heLswTxA0Ef0XRGdUJK3AvVUR+o= X-Received: by 2002:a17:906:2acf:: with SMTP id m15-v6mr17601300eje.180.1544691854549; Thu, 13 Dec 2018 01:04:14 -0800 (PST) MIME-Version: 1.0 References: <20181207075322.GS1286@dhcp22.suse.cz> <20181207113044.GB1286@dhcp22.suse.cz> <20181207142240.GC1286@dhcp22.suse.cz> <20181207155627.GG1286@dhcp22.suse.cz> <20181210123738.GN1286@dhcp22.suse.cz> <20181212115340.GQ1286@dhcp22.suse.cz> In-Reply-To: From: Pingfan Liu Date: Thu, 13 Dec 2018 17:04:01 +0800 Message-ID: Subject: Re: [PATCH] mm/alloc: fallback to first node if the wanted node offline To: mhocko@kernel.org Cc: Vlastimil Babka , linux-mm@kvack.org, linux-kernel@vger.kernel.org, Andrew Morton , Mike Rapoport , Bjorn Helgaas , Jonathan Cameron Content-Type: text/plain; charset="UTF-8" Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Dec 13, 2018 at 4:37 PM Pingfan Liu wrote: > > On Wed, Dec 12, 2018 at 7:53 PM Michal Hocko wrote: > > > > On Wed 12-12-18 16:31:35, Pingfan Liu wrote: > > > On Mon, Dec 10, 2018 at 8:37 PM Michal Hocko wrote: > > > > > > > [...] > > > > > > > > In other words. Does the following work? I am sorry to wildguess this > > > > way but I am not able to recreate your setups to play with this myself. > > > > > > > > diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c > > > > index 1308f5408bf7..d51643e10d00 100644 > > > > --- a/arch/x86/mm/numa.c > > > > +++ b/arch/x86/mm/numa.c > > > > @@ -216,8 +216,6 @@ static void __init alloc_node_data(int nid) > > > > > > > > node_data[nid] = nd; > > > > memset(NODE_DATA(nid), 0, sizeof(pg_data_t)); > > > > - > > > > - node_set_online(nid); > > > > } > > > > > > > > /** > > > > @@ -527,6 +525,19 @@ static void __init numa_clear_kernel_node_hotplug(void) > > > > } > > > > } > > > > > > > > +static void __init init_memory_less_node(int nid) > > > > +{ > > > > + unsigned long zones_size[MAX_NR_ZONES] = {0}; > > > > + unsigned long zholes_size[MAX_NR_ZONES] = {0}; > > > > + > > > > + free_area_init_node(nid, zones_size, 0, zholes_size); > > > > + > > > > + /* > > > > + * All zonelists will be built later in start_kernel() after per cpu > > > > + * areas are initialized. > > > > + */ > > > > +} > > > > + > > > > static int __init numa_register_memblks(struct numa_meminfo *mi) > > > > { > > > > unsigned long uninitialized_var(pfn_align); > > > > @@ -570,7 +581,7 @@ static int __init numa_register_memblks(struct numa_meminfo *mi) > > > > return -EINVAL; > > > > > > > > /* Finally register nodes. */ > > > > - for_each_node_mask(nid, node_possible_map) { > > > > + for_each_node(nid) { > > > > u64 start = PFN_PHYS(max_pfn); > > > > u64 end = 0; > > > > > > > > @@ -592,6 +603,10 @@ static int __init numa_register_memblks(struct numa_meminfo *mi) > > > > continue; > > > > > > > > alloc_node_data(nid); > > > > + if (!end) > > > > > > Here comes the bug, since !end can not reach here. > > > > You are right. I am dumb. I've just completely missed that. Sigh. > > Anyway, I think the code is more complicated than necessary and we can > > simply drop the check. I do not think we really have to worry about > > the start overflowing end. So the end patch should look as follows. > > Btw. I believe it is better to pull alloc_node_data out of init_memory_less_node > > because a) there is no need to duplicate the call and moreover we want > > to pull node_set_online as well. The code also seems cleaner this way. > > > I have no strong opinion here. > > Thanks for your testing and your patience with me here. > Np. > > > > diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c > > index 1308f5408bf7..a5548fe668fb 100644 > > --- a/arch/x86/mm/numa.c > > +++ b/arch/x86/mm/numa.c > > @@ -216,8 +216,6 @@ static void __init alloc_node_data(int nid) > > > > node_data[nid] = nd; > > memset(NODE_DATA(nid), 0, sizeof(pg_data_t)); > > - > > - node_set_online(nid); > > } > > > > /** > > @@ -527,6 +525,19 @@ static void __init numa_clear_kernel_node_hotplug(void) > > } > > } > > > > +static void __init init_memory_less_node(int nid) > > +{ > > + unsigned long zones_size[MAX_NR_ZONES] = {0}; > > + unsigned long zholes_size[MAX_NR_ZONES] = {0}; > > + > > + free_area_init_node(nid, zones_size, 0, zholes_size); > > + > > + /* > > + * All zonelists will be built later in start_kernel() after per cpu > > + * areas are initialized. > > + */ > > +} > > + > > static int __init numa_register_memblks(struct numa_meminfo *mi) > > { > > unsigned long uninitialized_var(pfn_align); > > @@ -570,7 +581,7 @@ static int __init numa_register_memblks(struct numa_meminfo *mi) > > return -EINVAL; > > > > /* Finally register nodes. */ > > - for_each_node_mask(nid, node_possible_map) { > > + for_each_node(nid) { > > u64 start = PFN_PHYS(max_pfn); > > u64 end = 0; > > > > @@ -581,9 +592,6 @@ static int __init numa_register_memblks(struct numa_meminfo *mi) > > end = max(mi->blk[i].end, end); > > } > > > > - if (start >= end) > > - continue; > > - > > /* > > * Don't confuse VM with a node that doesn't have the > > * minimum amount of memory: > > @@ -592,6 +600,10 @@ static int __init numa_register_memblks(struct numa_meminfo *mi) > > continue; > > > > alloc_node_data(nid); > > + if (!end) > > + init_memory_less_node(nid); Just have some opinion on this. Here is two issue. First, is this node online? I do not see node_set_online() is called in this patch. Second, if node is online here, then init_memory_less_node-> free_area_init_node is called duplicated when free_area_init_nodes(). This should be a critical design issue. Thanks, Pingfan > > + else > > + node_set_online(nid); > > } > > > > /* Dump memblock with node info and return. */ > > @@ -721,21 +733,6 @@ void __init x86_numa_init(void) > > numa_init(dummy_numa_init); > > } > > > > -static void __init init_memory_less_node(int nid) > > -{ > > - unsigned long zones_size[MAX_NR_ZONES] = {0}; > > - unsigned long zholes_size[MAX_NR_ZONES] = {0}; > > - > > - /* Allocate and initialize node data. Memory-less node is now online.*/ > > - alloc_node_data(nid); > > - free_area_init_node(nid, zones_size, 0, zholes_size); > > - > > - /* > > - * All zonelists will be built later in start_kernel() after per cpu > > - * areas are initialized. > > - */ > > -} > > - > > /* > > * Setup early cpu_to_node. > > * > > @@ -763,9 +760,6 @@ void __init init_cpu_to_node(void) > > if (node == NUMA_NO_NODE) > > continue; > > > > - if (!node_online(node)) > > - init_memory_less_node(node); > > - > > numa_set_node(cpu, node); > > } > > } > > -- > Regret, it still has bug, and I got panic. Attached log. > > Thanks, > Pingfan