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 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id BBBC3C433EF for ; Wed, 10 Nov 2021 10:44:37 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 6358B61027 for ; Wed, 10 Nov 2021 10:44:37 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.1 mail.kernel.org 6358B61027 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.alibaba.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=kvack.org Received: by kanga.kvack.org (Postfix) id CBEBE6B0072; Wed, 10 Nov 2021 05:44:36 -0500 (EST) Received: by kanga.kvack.org (Postfix, from userid 40) id C6E946B0073; Wed, 10 Nov 2021 05:44:36 -0500 (EST) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id B848D6B0074; Wed, 10 Nov 2021 05:44:36 -0500 (EST) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0123.hostedemail.com [216.40.44.123]) by kanga.kvack.org (Postfix) with ESMTP id A9D216B0072 for ; Wed, 10 Nov 2021 05:44:36 -0500 (EST) Received: from smtpin07.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay01.hostedemail.com (Postfix) with ESMTP id 61ECD180CCF40 for ; Wed, 10 Nov 2021 10:44:36 +0000 (UTC) X-FDA: 78792686952.07.9540ACC Received: from out30-42.freemail.mail.aliyun.com (out30-42.freemail.mail.aliyun.com [115.124.30.42]) by imf28.hostedemail.com (Postfix) with ESMTP id 1BF8090000A8 for ; Wed, 10 Nov 2021 10:44:34 +0000 (UTC) X-Alimail-AntiSpam:AC=PASS;BC=-1|-1;BR=01201311R181e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=e01e04394;MF=baolin.wang@linux.alibaba.com;NM=1;PH=DS;RN=10;SR=0;TI=SMTPD_---0UvvHA6u_1636541070; Received: from 30.21.164.34(mailfrom:baolin.wang@linux.alibaba.com fp:SMTPD_---0UvvHA6u_1636541070) by smtp.aliyun-inc.com(127.0.0.1); Wed, 10 Nov 2021 18:44:30 +0800 Subject: Re: [PATCH] mm: migrate: Support multiple target nodes demotion To: "Huang, Ying" Cc: akpm@linux-foundation.org, dave.hansen@linux.intel.com, ziy@nvidia.com, osalvador@suse.de, shy828301@gmail.com, zhongjiang-ali@linux.alibaba.com, xlpang@linux.alibaba.com, linux-mm@kvack.org, linux-kernel@vger.kernel.org References: <8850612186ea23eb5d328d84e4008a6b60f418e1.1636514506.git.baolin.wang@linux.alibaba.com> <87ee7opf4p.fsf@yhuang6-desk2.ccr.corp.intel.com> From: Baolin Wang Message-ID: Date: Wed, 10 Nov 2021 18:45:17 +0800 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101 Thunderbird/78.14.0 MIME-Version: 1.0 In-Reply-To: <87ee7opf4p.fsf@yhuang6-desk2.ccr.corp.intel.com> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit X-Rspamd-Server: rspam01 X-Rspamd-Queue-Id: 1BF8090000A8 X-Stat-Signature: we6dc8nf8tknma6uhahnq77ngfk6j6rt Authentication-Results: imf28.hostedemail.com; dkim=none; spf=pass (imf28.hostedemail.com: domain of baolin.wang@linux.alibaba.com designates 115.124.30.42 as permitted sender) smtp.mailfrom=baolin.wang@linux.alibaba.com; dmarc=pass (policy=none) header.from=alibaba.com X-HE-Tag: 1636541074-444259 X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: On 2021/11/10 16:51, Huang, Ying writes: > Baolin Wang writes: > >> We have some machines with multiple memory types like below, which >> have one fast (DRAM) memory node and two slow (persistent memory) memory >> nodes. According to current node demotion, if node 0 fills up, > ~~~~~~~~~~~~~ > > node demotion policy? Yes, will fix in next version. > > >> its memory should be migrated to node 1, when node 1 fills up, its >> memory will be migrated to node 2: node 0 -> node 1 -> node 2 ->stop. >> >> But this is not efficient and suitbale memory migration route >> for our machine with multiple slow memory nodes. Since the distance >> between node 0 to node 1 and node 0 to node 2 is equal, and memory >> migration between slow memory nodes will increase persistent memory >> bandwidth greatly, which will hurt the whole system's performance. >> >> Thus for this case, we can treat the slow memory node 1 and node 2 >> as a whole slow memory region, and we should migrate memory from >> node 0 to node 1 and node 2 if node 0 fills up. >> >> This patch changes the node_demotion data structure to support multiple >> target nodes, and establishes the migration path to support multiple >> target nodes with validating if the node distance is the best or not. >> >> available: 3 nodes (0-2) >> node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 >> node 0 size: 62153 MB >> node 0 free: 55135 MB >> node 1 cpus: >> node 1 size: 127007 MB >> node 1 free: 126930 MB >> node 2 cpus: >> node 2 size: 126968 MB >> node 2 free: 126878 MB >> node distances: >> node 0 1 2 >> 0: 10 20 20 >> 1: 20 10 20 >> 2: 20 20 10 >> >> Signed-off-by: Baolin Wang >> --- >> Changes from RFC v2: >> - Change to 'short' type for target nodes array. >> - Remove nodemask instead selecting target node directly. >> - Add WARN_ONCE() if the target nodes exceed the maximum value. >> >> Changes from RFC v1: >> - Re-define the node_demotion structure. >> - Set up multiple target nodes by validating the node distance. >> - Add more comments. >> --- >> mm/migrate.c | 138 +++++++++++++++++++++++++++++++++++++++++++---------------- >> 1 file changed, 102 insertions(+), 36 deletions(-) >> >> diff --git a/mm/migrate.c b/mm/migrate.c >> index cf25b00..7f1d745 100644 >> --- a/mm/migrate.c >> +++ b/mm/migrate.c >> @@ -50,6 +50,7 @@ >> #include >> #include >> #include >> +#include >> >> #include >> >> @@ -1119,12 +1120,25 @@ static int __unmap_and_move(struct page *page, struct page *newpage, >> * >> * This is represented in the node_demotion[] like this: >> * >> - * { 1, // Node 0 migrates to 1 >> - * 2, // Node 1 migrates to 2 >> - * -1, // Node 2 does not migrate >> - * 4, // Node 3 migrates to 4 >> - * 5, // Node 4 migrates to 5 >> - * -1} // Node 5 does not migrate >> + * { nr=1, nodes[0]=1 }, // Node 0 migrates to 1 >> + * { nr=1, nodes[0]=2 }, // Node 1 migrates to 2 >> + * { nr=0, nodes[0]=-1 }, // Node 2 does not migrate >> + * { nr=1, nodes[0]=4 }, // Node 3 migrates to 4 >> + * { nr=1, nodes[0]=5 }, // Node 4 migrates to 5 >> + * { nr=0, nodes[0]=-1} // Node 5 does not migrate >> + * >> + * Moreover some systems may have multiple same class memory >> + * types. Suppose a system has one socket with 3 memory nodes, > > s/same class memory types/slow memory nodes/ > > ? > > We don't support multiple fast memory types, right? Until now we have no machines with multiple fast memory types. OK, I will change the words. > >> + * node 0 is fast memory type, and node 1/2 both are slow memory >> + * type, and the distance between fast memory node and slow >> + * memory node is same. So the migration path should be: >> + * >> + * 0 -> 1/2 -> stop >> + * >> + * This is represented in the node_demotion[] like this: >> + * { nr=2, {nodes[0]=1, nodes[1]=2} }, // Node 0 migrates to node 1 and node 2 >> + * { nr=0, nodes[0]=-1, }, // Node 1 dose not migrate >> + * { nr=0, nodes[0]=-1, }, // Node 2 does not migrate >> */ >> >> /* >> @@ -1135,8 +1149,13 @@ static int __unmap_and_move(struct page *page, struct page *newpage, >> * must be held over all reads to ensure that no cycles are >> * observed. >> */ >> -static int node_demotion[MAX_NUMNODES] __read_mostly = >> - {[0 ... MAX_NUMNODES - 1] = NUMA_NO_NODE}; >> +#define DEMOTION_TARGET_NODES 15 >> +struct demotion_nodes { >> + unsigned short nr; >> + short nodes[DEMOTION_TARGET_NODES]; >> +}; >> + >> +static struct demotion_nodes node_demotion[MAX_NUMNODES] __read_mostly; > > If MAX_NUMNODES is 1024, the total size will be (16 * 2 * 1024) = 32K > bytes. That appears too large. We may consider to allocate > node_demotion[] dynamically. Sure. I'd like to optimize it in a separate patch to keep current patch easy to review. Thanks.