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=-13.1 required=3.0 tests=BAYES_00,DKIMWL_WL_HIGH, DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS, HTML_MESSAGE,INCLUDES_CR_TRAILER,INCLUDES_PATCH,MAILING_LIST_MULTI, MSGID_FROM_MTA_HEADER,SPF_HELO_NONE,SPF_PASS 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 36D0BC4743C for ; Mon, 21 Jun 2021 14:50:28 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 8C46E60E0B for ; Mon, 21 Jun 2021 14:50:27 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 8C46E60E0B Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=nvidia.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=owner-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix) id EE89A6B006C; Mon, 21 Jun 2021 10:50:26 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id EBF4D6B0070; Mon, 21 Jun 2021 10:50:26 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id CC46C6B0072; Mon, 21 Jun 2021 10:50:26 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0171.hostedemail.com [216.40.44.171]) by kanga.kvack.org (Postfix) with ESMTP id 893836B006C for ; Mon, 21 Jun 2021 10:50:26 -0400 (EDT) Received: from smtpin13.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay02.hostedemail.com (Postfix) with ESMTP id 197F81263F for ; Mon, 21 Jun 2021 14:50:26 +0000 (UTC) X-FDA: 78278016852.13.9FB1109 Received: from NAM10-MW2-obe.outbound.protection.outlook.com (mail-mw2nam10on2078.outbound.protection.outlook.com [40.107.94.78]) by imf18.hostedemail.com (Postfix) with ESMTP id 5FB6120015E7 for ; Mon, 21 Jun 2021 14:50:25 +0000 (UTC) ARC-Seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=OKC/dFJUg1VK+msiIEuDpJGTEkJlDnaYAuSTdRG/zD4pmq1M/UlJ3H2LKP7dZu2rOs70peX96u1pjBmKaF6C0eAWjBQH3C76M6T/V2XhZ1Bn+qpROEe1p6ZxqKvg8ovQ7yMaVIZRwE7gB8z7l7rTrCFQBgTCpOCW47wpfbg4EL+fNdpp63CQmaeTR5X1V6RQ6idX6cBgEDOg1u4iKWsvGEiBIMgn545BO0CMtLhFnAdk8Z8VuiAj1bN52yPCFoQJu2oRAZcLdQ6nroAs6T7W62y13cEdRzfosEm5Q2+vH8XqINwXib3ibfpU2Qr688ny60hTqGPYnoRUgc6DfCnoTA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=YmGYS/kOe7O5n5yEK83G/ePg1q3FcobzrJeaJEFd7Y0=; b=NeM/dzCZHJaO81aM2ylmZSZiLnOjH8OfqvnACJdw9L6rgngFr1AQRnBMQhsw2i3bSio67E0WxQnMURDiK/0P2YpBMn/nkXF2UmxNw9tKjvrb5TeFIDPmR36QanEYXAvvTX6eiRqoVePF5NtrllmMb6zkxvLXutQsu8+Cq99NRgHbza/p1T0luXn8wSDc5ZVL3hXE0gWn7mCYGOX2vp5BtIf9k2f3mIbTZcPY3FJDTJea9JsHwdo2hsF9nDB82AQ9J3IL/poukIYDIJU+mhmsqAjoOMgbXsQG7o3elDUB/0TB8OIu7KJMduPLrJOUKGeKeUkl8BKviEx5Ifxga/FCsA== ARC-Authentication-Results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=nvidia.com; dmarc=pass action=none header.from=nvidia.com; dkim=pass header.d=nvidia.com; arc=none DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=Nvidia.com; s=selector2; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-SenderADCheck; bh=YmGYS/kOe7O5n5yEK83G/ePg1q3FcobzrJeaJEFd7Y0=; b=e0qkpJWO6EF2VqZCBhFmt75mJY42eC41geD0hudkYly533J3HNo+I/xC+NEkI6pbOz+vz3WNes4c+F8q06O0dHYh5ZgaBf5wT9ysRgSuaX40awNRfL3K9GTLtipoQgjRmjvu7uvvFRzWBT/SnT6BIkqC8RxAjpvHBYSgaE5DKYWvxaQg334uU7FdmqjGKiMuD37uQE2GH92zqcdWUX20Y1v67z01vcpaGXijCKT9sen+x3HeBTrowFjP68cyGbLahQSKKaGh41s6WOWhZkNixCTFD78CCF6EBu8GA5O4nYuwhVLEDUbZrKc4TGYqEG8I8BiYFVA+kbdi6iv4MAvZXg== Received: from MN2PR12MB3823.namprd12.prod.outlook.com (2603:10b6:208:168::26) by BL0PR12MB4724.namprd12.prod.outlook.com (2603:10b6:208:87::23) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4242.21; Mon, 21 Jun 2021 14:50:21 +0000 Received: from MN2PR12MB3823.namprd12.prod.outlook.com ([fe80::dcee:535c:30e:95f4]) by MN2PR12MB3823.namprd12.prod.outlook.com ([fe80::dcee:535c:30e:95f4%6]) with mapi id 15.20.4242.023; Mon, 21 Jun 2021 14:50:21 +0000 From: Zi Yan To: "Huang, Ying" Cc: Dave Hansen , linux-mm@kvack.org, linux-kernel@vger.kernel.org, Yang Shi , Michal Hocko , Wei Xu , David Rientjes , Dan Williams , David Hildenbrand , osalvador Subject: Re: [PATCH -V8 02/10] mm/numa: automatically generate node migration order Date: Mon, 21 Jun 2021 10:50:14 -0400 X-Mailer: MailMate (1.14r5812) Message-ID: <2AA3D792-7F14-4297-8EDD-3B5A7B31AECA@nvidia.com> In-Reply-To: <87v96anu6o.fsf@yhuang6-desk2.ccr.corp.intel.com> References: <20210618061537.434999-1-ying.huang@intel.com> <20210618061537.434999-3-ying.huang@intel.com> <79397FE3-4B08-4DE5-8468-C5CAE36A3E39@nvidia.com> <87v96anu6o.fsf@yhuang6-desk2.ccr.corp.intel.com> Content-Type: multipart/signed; boundary="=_MailMate_B213593D-221A-421A-B151-2B8E71B15048_="; micalg=pgp-sha512; protocol="application/pgp-signature" X-Originating-IP: [216.228.112.21] X-ClientProxiedBy: BL1PR13CA0008.namprd13.prod.outlook.com (2603:10b6:208:256::13) To MN2PR12MB3823.namprd12.prod.outlook.com (2603:10b6:208:168::26) MIME-Version: 1.0 X-MS-Exchange-MessageSentRepresentingType: 1 Received: from [10.2.58.56] (216.228.112.21) by BL1PR13CA0008.namprd13.prod.outlook.com (2603:10b6:208:256::13) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id 15.20.4264.7 via Frontend Transport; Mon, 21 Jun 2021 14:50:18 +0000 X-MS-PublicTrafficType: Email X-MS-Office365-Filtering-Correlation-Id: 05597fdc-4b24-472f-e501-08d934c3e4d2 X-MS-TrafficTypeDiagnostic: BL0PR12MB4724: X-MS-Exchange-Transport-Forked: True X-Microsoft-Antispam-PRVS: X-MS-Oob-TLC-OOBClassifiers: OLM:10000; X-MS-Exchange-SenderADCheck: 1 X-Microsoft-Antispam: BCL:0; X-Microsoft-Antispam-Message-Info: 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 X-Forefront-Antispam-Report: CIP:255.255.255.255;CTRY:;LANG:en;SCL:1;SRV:;IPV:NLI;SFV:NSPM;H:MN2PR12MB3823.namprd12.prod.outlook.com;PTR:;CAT:NONE;SFS:(4636009)(366004)(86362001)(33656002)(7416002)(53546011)(186003)(2616005)(66556008)(26005)(16526019)(4326008)(6666004)(36756003)(956004)(66946007)(6486002)(38100700002)(2906002)(83380400001)(66476007)(6916009)(30864003)(54906003)(5660300002)(16576012)(33964004)(21480400003)(8676002)(498600001)(235185007)(8936002)(45980500001)(72826004);DIR:OUT;SFP:1101; X-MS-Exchange-AntiSpam-MessageData-ChunkCount: 1 X-MS-Exchange-AntiSpam-MessageData-0: =?utf-8?B?dWxyMkZzcHBFeWljVVhidWZ3VjlkUHFrN0crVDBKeUpLQzZMalhBcHF5aG1F?= =?utf-8?B?OWFqeUZ6Z2FKaCtkVlpiL0t0c04zTDltR2kwZVpFMGpDelVlRW8yNHhQZ0FI?= =?utf-8?B?dGpXckIyeFJsaGFCV0dlWHR3ZXhHaGtPQkg3RWxNVSttVUpkOUdZYXcvcDVV?= =?utf-8?B?TzdHZHJZSFJFZExFSE1MWDNoU0VZUGlSUm1KYUVqR3g3Q0ZZcTZIVjVOdTY5?= =?utf-8?B?ZWZmN2RqY3NXbmliWUp2dVJENlFycG5nemNtZnpTOGhhZlJMN2xNTWJzLy9G?= =?utf-8?B?QVN5eUZ6K3N5aThUbHVFQ2x2Z1lSZnZrRDU5Z0RubCtBaWlNSDdYc0wvVzNO?= =?utf-8?B?NHpaWXhpbnVCcUllWGtnZW9yb2xsUVJXVllWM1RFUGpPMloyZVNndEZNU0ly?= =?utf-8?B?WUZ0TUtmS01nOTNBNndFYVJ0QzdkU1RwWVZ1Qit2RFZpdXk4ZlNHNkRUUzhy?= =?utf-8?B?RXVsQzhFbGg0ZFBtMWJCb1RmcHB3WUpSczdHRU1vTjlrc29GZ05Xd3JMR2M1?= =?utf-8?B?NjZJbFN0ZHk4U2tNUXp3ODVkOVJNLzBLa1ZLejlRU29sUW1jSGtMb3FyYnlT?= =?utf-8?B?aExDTkdlYkFaMEFXbU5PUm9laE8waDVwVWxEbGlVSkVvcFVRcGVUcHRZNUpN?= =?utf-8?B?dld3OUtKZmhLNEhQVWRPeWVpTnF5dU9EelAzZlZLcjNvUUlHZGlQU0xHaEh4?= =?utf-8?B?K0trRjNDeU9Zeit3VEVjMFVxNi90aFUxb250dERwUGt4ZXVwNnp4bjJLSXAz?= =?utf-8?B?SjVDWVJ4MlpFbTlBNlFENkRoNGZCRGpuY3RvTDIwbm9OeVh5Y3diUDVqbHRo?= =?utf-8?B?M2N6ZzFYZWdIc0pLckVVUFdoN0x6OHFrN3VuQXRVK3l4NUZoZkN3RXFQNGdr?= =?utf-8?B?Vkd0cGxaR2RXNm5ST2Z3WkFyaEtINCswTkljMTRGdWl0WGZqQ1l3SmRlZ2xT?= =?utf-8?B?TTJUaUV5N1E3c3Q4UVROSlJUb0RhZWpKT1c1U3JQZlRQSFprL2NoU29yMVh5?= =?utf-8?B?R1hrRW5GTkd1VVR4aXVac0Vqbi9rbTlzODA2YWRPcjhYcGlFdWJTQzh0c3Rh?= =?utf-8?B?Q1EvbS9RVUJVKzQ0d2RCTnlWcTZNOUtqUkZzTlp5OGhhbHhvdSsyWmhMU2JC?= =?utf-8?B?dFVZWWI3T1NsM3MzVWczb1plM0lwdWc1dlNLb0x3QnZoM2RmTnFOSGlnajVZ?= =?utf-8?B?QzdqUEFRL0dTZ1l6YkFaQXE5aFJlV0ZWU0pueEF3cDRLaS9QR3ZhTzlaWk9Q?= =?utf-8?B?bTh3aW1TTnp1K0R1djc5a3R0dHh0dlh2K0o3NWJ4TUlKVGNYSHhYTW5YRUlr?= =?utf-8?B?Y01rUk9RNkx6Yk1reDdUbmg1T3pSdm15WElYVnhNTEN3b0pUUnQ1RXlvUGpD?= =?utf-8?B?a2RFcWsyMVRuSGdGMUx1VDlGMGRnaHR0T0xTOWw4Yksya1RQejRoL29seE9q?= =?utf-8?B?dUlqa1NEbW81UjNSbTRRZURpWGtHUmdBWDJ0bnpaV3gxWEFjano0VUxJbnNy?= =?utf-8?B?aG9qL205eFdPY2hTU2dRYkZLcWJjYTNIQUFTeHc0TnRBa2wrS2ljSzJsdmpN?= =?utf-8?B?OVVGdHpGL3JNQS9uOHFmWnlsbXhDUWlLVVNZSDlQMUhZaG5ISWJVY2pnaXNF?= =?utf-8?B?WXpDdmpBWHlZZ0hsWmZaM21Cam42SVJsQlFzMUtnSTZPc2l3QjI4QTlSakJp?= =?utf-8?B?VVJnRFFQd2NkNytydGFLTlhpOHMxWVpqYWc1MEFodmN1S0pHcVppeUs5V0dW?= =?utf-8?Q?kkO6YXEl8c81DP48Ca37Rk9ITIIuLAOSiE+AirA?= X-OriginatorOrg: Nvidia.com X-MS-Exchange-CrossTenant-Network-Message-Id: 05597fdc-4b24-472f-e501-08d934c3e4d2 X-MS-Exchange-CrossTenant-AuthSource: MN2PR12MB3823.namprd12.prod.outlook.com X-MS-Exchange-CrossTenant-AuthAs: Internal X-MS-Exchange-CrossTenant-OriginalArrivalTime: 21 Jun 2021 14:50:21.7186 (UTC) X-MS-Exchange-CrossTenant-FromEntityHeader: Hosted X-MS-Exchange-CrossTenant-Id: 43083d15-7273-40c1-b7db-39efd9ccc17a X-MS-Exchange-CrossTenant-MailboxType: HOSTED X-MS-Exchange-CrossTenant-UserPrincipalName: Ule+ORHCaoaeqBQB/GkSIkqH21ALpgOignOaphBU2LBMI2oyVwNKE/f9U2/7ekg5 X-MS-Exchange-Transport-CrossTenantHeadersStamped: BL0PR12MB4724 X-Rspamd-Server: rspam01 X-Rspamd-Queue-Id: 5FB6120015E7 Authentication-Results: imf18.hostedemail.com; dkim=pass header.d=Nvidia.com header.s=selector2 header.b=e0qkpJWO; spf=none (imf18.hostedemail.com: domain of ziy@nvidia.com has no SPF policy when checking 40.107.94.78) smtp.mailfrom=ziy@nvidia.com; dmarc=pass (policy=none) header.from=nvidia.com X-Stat-Signature: he6arcs1uw55ba9chdkenpihcth6pyzi X-HE-Tag: 1624287025-824446 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: --=_MailMate_B213593D-221A-421A-B151-2B8E71B15048_= Content-Type: multipart/alternative; boundary="=_MailMate_EE747415-A08F-42A4-80A0-0ACA9A83928A_=" --=_MailMate_EE747415-A08F-42A4-80A0-0ACA9A83928A_= Content-Type: text/plain; charset=UTF-8; markup=markdown Content-Transfer-Encoding: quoted-printable On 19 Jun 2021, at 4:18, Huang, Ying wrote: > Zi Yan writes: > >> On 18 Jun 2021, at 2:15, Huang Ying wrote: >> >>> From: Dave Hansen >>> >>> When memory fills up on a node, memory contents can be >>> automatically migrated to another node. The biggest problems are >>> knowing when to migrate and to where the migration should be >>> targeted. >>> >>> The most straightforward way to generate the "to where" list would >>> be to follow the page allocator fallback lists. Those lists >>> already tell us if memory is full where to look next. It would >>> also be logical to move memory in that order. >>> >>> But, the allocator fallback lists have a fatal flaw: most nodes >>> appear in all the lists. This would potentially lead to migration >>> cycles (A->B, B->A, A->B, ...). >>> >>> Instead of using the allocator fallback lists directly, keep a >>> separate node migration ordering. But, reuse the same data used >>> to generate page allocator fallback in the first place: >>> find_next_best_node(). >>> >>> This means that the firmware data used to populate node distances >>> essentially dictates the ordering for now. It should also be >>> architecture-neutral since all NUMA architectures have a working >>> find_next_best_node(). >>> >>> The protocol for node_demotion[] access and writing is not >>> standard. It has no specific locking and is intended to be read >>> locklessly. Readers must take care to avoid observing changes >>> that appear incoherent. This was done so that node_demotion[] >>> locking has no chance of becoming a bottleneck on large systems >>> with lots of CPUs in direct reclaim. >>> >>> This code is unused for now. It will be called later in the >>> series. >>> >>> Signed-off-by: Dave Hansen >>> Signed-off-by: "Huang, Ying" >>> Reviewed-by: Yang Shi >>> Cc: Michal Hocko >>> Cc: Wei Xu >>> Cc: David Rientjes >>> Cc: Dan Williams >>> Cc: David Hildenbrand >>> Cc: osalvador >>> >>> -- >>> >>> Changes from 20200122: >>> * Add big node_demotion[] comment >>> Changes from 20210302: >>> * Fix typo in node_demotion[] comment >>> --- >>> mm/internal.h | 5 ++ >>> mm/migrate.c | 175 ++++++++++++++++++++++++++++++++++++++++++++++= +- >>> mm/page_alloc.c | 2 +- >>> 3 files changed, 180 insertions(+), 2 deletions(-) >>> >>> diff --git a/mm/internal.h b/mm/internal.h >>> index 2f1182948aa6..0344cd78e170 100644 >>> --- a/mm/internal.h >>> +++ b/mm/internal.h >>> @@ -522,12 +522,17 @@ static inline void mminit_validate_memmodel_lim= its(unsigned long *start_pfn, >>> >>> #ifdef CONFIG_NUMA >>> extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); >>> +extern int find_next_best_node(int node, nodemask_t *used_node_mask)= ; >>> #else >>> static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask= , >>> unsigned int order) >>> { >>> return NODE_RECLAIM_NOSCAN; >>> } >>> +static inline int find_next_best_node(int node, nodemask_t *used_nod= e_mask) >>> +{ >>> + return NUMA_NO_NODE; >>> +} >>> #endif >>> >>> extern int hwpoison_filter(struct page *p); >>> diff --git a/mm/migrate.c b/mm/migrate.c >>> index 6cab668132f9..111f8565f75d 100644 >>> --- a/mm/migrate.c >>> +++ b/mm/migrate.c >>> @@ -1136,6 +1136,44 @@ static int __unmap_and_move(struct page *page,= struct page *newpage, >>> return rc; >>> } >>> >>> + >>> +/* >>> + * node_demotion[] example: >>> + * >>> + * Consider a system with two sockets. Each socket has >>> + * three classes of memory attached: fast, medium and slow. >>> + * Each memory class is placed in its own NUMA node. The >>> + * CPUs are placed in the node with the "fast" memory. The >>> + * 6 NUMA nodes (0-5) might be split among the sockets like >>> + * this: >>> + * >>> + * Socket A: 0, 1, 2 >>> + * Socket B: 3, 4, 5 >>> + * >>> + * When Node 0 fills up, its memory should be migrated to >>> + * Node 1. When Node 1 fills up, it should be migrated to >>> + * Node 2. The migration path start on the nodes with the >>> + * processors (since allocations default to this node) and >>> + * fast memory, progress through medium and end with the >>> + * slow memory: >>> + * >>> + * 0 -> 1 -> 2 -> stop >>> + * 3 -> 4 -> 5 -> stop >>> + * >>> + * 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 >>> + */ >>> + >>> +/* >>> + * Writes to this array occur without locking. READ_ONCE() >>> + * is recommended for readers to ensure consistent reads. >>> + */ >>> static int node_demotion[MAX_NUMNODES] __read_mostly =3D >>> {[0 ... MAX_NUMNODES - 1] =3D NUMA_NO_NODE}; >>> >>> @@ -1150,7 +1188,13 @@ static int node_demotion[MAX_NUMNODES] __read_= mostly =3D >>> */ >>> int next_demotion_node(int node) >>> { >>> - return node_demotion[node]; >>> + /* >>> + * node_demotion[] is updated without excluding >>> + * this function from running. READ_ONCE() avoids >>> + * reading multiple, inconsistent 'node' values >>> + * during an update. >>> + */ >>> + return READ_ONCE(node_demotion[node]); >>> } >> >> Is it necessary to have two separate patches to add node_demotion and >> next_demotion_node() then modify it immediately? Maybe merge Patch 1 i= nto 2? >> >> Hmm, I just checked Patch 3 and it changes node_demotion again and use= s RCU. >> I guess it might be much simpler to just introduce node_demotion with = RCU >> in this patch and Patch 3 only takes care of hotplug events. > > Hi, Dave, > > What do you think about this? > >>> >>> /* >>> @@ -3144,3 +3188,132 @@ void migrate_vma_finalize(struct migrate_vma = *migrate) >>> } >>> EXPORT_SYMBOL(migrate_vma_finalize); >>> #endif /* CONFIG_DEVICE_PRIVATE */ >>> + >>> +/* Disable reclaim-based migration. */ >>> +static void disable_all_migrate_targets(void) >>> +{ >>> + int node; >>> + >>> + for_each_online_node(node) >>> + node_demotion[node] =3D NUMA_NO_NODE; >>> +} >>> + >>> +/* >>> + * Find an automatic demotion target for 'node'. >>> + * Failing here is OK. It might just indicate >>> + * being at the end of a chain. >>> + */ >>> +static int establish_migrate_target(int node, nodemask_t *used) >>> +{ >>> + int migration_target; >>> + >>> + /* >>> + * Can not set a migration target on a >>> + * node with it already set. >>> + * >>> + * No need for READ_ONCE() here since this >>> + * in the write path for node_demotion[]. >>> + * This should be the only thread writing. >>> + */ >>> + if (node_demotion[node] !=3D NUMA_NO_NODE) >>> + return NUMA_NO_NODE; >>> + >>> + migration_target =3D find_next_best_node(node, used); >>> + if (migration_target =3D=3D NUMA_NO_NODE) >>> + return NUMA_NO_NODE; >>> + >>> + node_demotion[node] =3D migration_target; >>> + >>> + return migration_target; >>> +} >>> + >>> +/* >>> + * When memory fills up on a node, memory contents can be >>> + * automatically migrated to another node instead of >>> + * discarded at reclaim. >>> + * >>> + * Establish a "migration path" which will start at nodes >>> + * with CPUs and will follow the priorities used to build the >>> + * page allocator zonelists. >>> + * >>> + * The difference here is that cycles must be avoided. If >>> + * node0 migrates to node1, then neither node1, nor anything >>> + * node1 migrates to can migrate to node0. >>> + * >>> + * This function can run simultaneously with readers of >>> + * node_demotion[]. However, it can not run simultaneously >>> + * with itself. Exclusion is provided by memory hotplug events >>> + * being single-threaded. >>> + */ >>> +static void __set_migration_target_nodes(void) >>> +{ >>> + nodemask_t next_pass =3D NODE_MASK_NONE; >>> + nodemask_t this_pass =3D NODE_MASK_NONE; >>> + nodemask_t used_targets =3D NODE_MASK_NONE; >>> + int node; >>> + >>> + /* >>> + * Avoid any oddities like cycles that could occur >>> + * from changes in the topology. This will leave >>> + * a momentary gap when migration is disabled. >>> + */ >>> + disable_all_migrate_targets(); >>> + >>> + /* >>> + * Ensure that the "disable" is visible across the system. >>> + * Readers will see either a combination of before+disable >>> + * state or disable+after. They will never see before and >>> + * after state together. >>> + * >>> + * The before+after state together might have cycles and >>> + * could cause readers to do things like loop until this >>> + * function finishes. This ensures they can only see a >>> + * single "bad" read and would, for instance, only loop >>> + * once. >>> + */ >>> + smp_wmb(); >>> + >>> + /* >>> + * Allocations go close to CPUs, first. Assume that >>> + * the migration path starts at the nodes with CPUs. >>> + */ >>> + next_pass =3D node_states[N_CPU]; >> >> Is there a plan of allowing user to change where the migration >> path starts? Or maybe one step further providing an interface >> to allow user to specify the demotion path. Something like >> /sys/devices/system/node/node*/node_demotion. > > I don't think that's necessary at least for now. Do you know any real > world use case for this? In our P9+volta system, GPU memory is exposed as a NUMA node. For the GPU workloads with data size greater than GPU memory size, it will be very helpful to allow pages in GPU memory to be migrated/demot= ed to CPU memory. With your current assumption, GPU memory -> CPU memory demotion seems not possible, right? This should also apply to any system with a device memory exposed as a NUMA node and workloads running on the device and using CPU memory as a lower tier memory than the device= memory. =E2=80=94 Best Regards, Yan, Zi --=_MailMate_EE747415-A08F-42A4-80A0-0ACA9A83928A_= Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable

On 19 Jun 2021, at 4:18, Huang, Ying wrote:

Zi Yan ziy@nvidia.com writes:

On 18 Jun 2021, at 2:15, Huang Ying wrote:

From: Dave Hansen dave.hansen@linux.intel.com

When memory fills up on a node, memory contents can be automatically migrated to another node. The biggest problems are
knowing when to migrate and to where the migration should be
targeted.

The most straightforward way to generate the "to where" l= ist would
be to follow the page allocator fallback lists. Those lists
already tell us if memory is full where to look next. It would
also be logical to move memory in that order.

But, the allocator fallback lists have a fatal flaw: most= nodes
appear in all the lists. This would potentially lead to migration
cycles (A->B, B->A, A->B, ...).

Instead of using the allocator fallback lists directly, k= eep a
separate node migration ordering. But, reuse the same data used
to generate page allocator fallback in the first place:
find_next_best_node().

This means that the firmware data used to populate node d= istances
essentially dictates the ordering for now. It should also be
architecture-neutral since all NUMA architectures have a working
find_next_best_node().

The protocol for node_demotion[] access and writing is no= t
standard. It has no specific locking and is intended to be read
locklessly. Readers must take care to avoid observing changes
that appear incoherent. This was done so that node_demotion[]
locking has no chance of becoming a bottleneck on large systems
with lots of CPUs in direct reclaim.

This code is unused for now. It will be called later in = the
series.

Signed-off-by: Dave Hansen dave.hansen@linux.intel.com
= Signed-off-by: "Huang, Ying" ying.huang@intel.com
Reviewed-by: Yang Shi shy828301@gmail.com
Cc: Michal Hocko = mhocko@suse.com
Cc: Wei Xu wei= xugc@google.com
Cc: David Rientjes rientjes@google.com
Cc: Dan Williams dan.j.williams@intel.com
Cc: David Hildenbrand david@redhat.com
Cc: osalvador o= salvador@suse.de

--

Changes from 20200122:

  • Add big node_demotion[] comment

Changes from 20210302:

  • Fix typo in node_demotion[] comment

mm/internal.h | 5 ++
mm/migrate.c | 175 +++++++++++++++++++++++++++++++++++++++++++++++- mm/page_alloc.c | 2 +-
3 files changed, 180 insertions(+), 2 deletions(-)

diff --git a/mm/internal.h b/mm/internal.h
index 2f1182948aa6..0344cd78e170 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -522,12 +522,17 @@ static inline void mminit_validate_memmodel_limits(= unsigned long *start_pfn,

#ifdef CONFIG_NUMA
extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
+extern int find_next_best_node(int node, nodemask_t *used_node_mask); #else
static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
= unsigned int order)
{
return NODE_RECLAIM_NOSCAN;
}
+static inline int find_next_best_node(int node, nodemask_t *used_node_ma= sk)
+{

  • return NUMA_NO_NODE;

+}
#endif

extern int hwpoison_filter(struct page *p);
diff --git a/mm/migrate.c b/mm/migrate.c
index 6cab668132f9..111f8565f75d 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -1136,6 +1136,44 @@ static int __unmap_and_move(struct page *page, str= uct page *newpage,
return rc;
}

+/*

    • node_demotion[] example:
    • Consider a system with two sockets. Each socket has
    • three classes of memory attached: fast, medium and slow.
    • Each memory class is placed in its own NUMA node. The
    • CPUs are placed in the node with the "fast" memory. The
    • 6 NUMA nodes (0-5) might be split among the sockets like
    • this:
    • Socket A: 0, 1, 2
    • Socket B: 3, 4, 5
    • When Node 0 fills up, its memory should be migrated to
    • Node 1. When Node 1 fills up, it should be migrated to
    • Node 2. The migration path start on the nodes with the
    • processors (since allocations default to this node) and
    • fast memory, progress through medium and end with the
    • slow memory:
    • 0 -> 1 -> 2 -> stop
    • 3 -> 4 -> 5 -> stop
    • 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
  • */

+/*

    • Writes to this array occur without locking. READ_ONCE()
    • is recommended for readers to ensure consistent reads.
  • */

static int node_demotion[MAX_NUMNODES] __read_mostly =3D<= br> {[0 ... MAX_NUMNODES - 1] =3D NUMA_NO_NODE};

@@ -1150,7 +1188,13 @@ static int node_demotion[MAX_NUMNO= DES] __read_mostly =3D
*/
int next_demotion_node(int node)
{

  • return node_demotion[node];
  • /*
    • node_demotion[] is updated without excluding
    • this function from running. READ_ONCE() avoids
    • reading multiple, inconsistent 'node' values
    • during an update.
  • */
  • return READ_ONCE(node_demotion[node]);

}

Is it necessary to have two separate patches to add node_= demotion and
next_demotion_node() then modify it immediately? Maybe merge Patch 1 into= 2?

Hmm, I just checked Patch 3 and it changes node_demotion = again and uses RCU.
I guess it might be much simpler to just introduce node_demotion with RCU=
in this patch and Patch 3 only takes care of hotplug events.

Hi, Dave,

What do you think about this?

/*
@@ -3144,3 +3188,132 @@ void migrate_vma_finalize(struct migrate_vma = migrate)
}
EXPORT_SYMBOL(migrate_vma_finalize);
#endif / CONFIG_DEVICE_PRIVATE /
+
+/ Disable reclaim-based migration. */
+static void disable_all_migrate_targets(void)
+{

  • int node;
  • for_each_online_node(node)
  •   node_demotion[n=
ode] =3D NUMA_NO_NODE;

+}
+
+/*

    • Find an automatic demotion target for 'node'.
    • Failing here is OK. It might just indicate
    • being at the end of a chain.
  • */

+static int establish_migrate_target(int node, nodemask_t= *used)
+{

  • int migration_target;
  • /*
    • Can not set a migration target on a
    • node with it already set.
    • No need for READ_ONCE() here since this
    • in the write path for node_demotion[].
    • This should be the only thread writing.
  • */
  • if (node_demotion[node] !=3D NUMA_NO_NODE)
  •   return NUMA_NO_=
NODE;
  • migration_target =3D find_next_best_node(node, used);
  • if (migration_target =3D=3D NUMA_NO_NODE)
  •   return NUMA_NO_=
NODE;
  • node_demotion[node] =3D migration_target;
  • return migration_target;

+}
+
+/*

    • When memory fills up on a node, memory contents can be
    • automatically migrated to another node instead of
    • discarded at reclaim.
    • Establish a "migration path" which will start at nodes
    • with CPUs and will follow the priorities used to build the
    • page allocator zonelists.
    • The difference here is that cycles must be avoided. If
    • node0 migrates to node1, then neither node1, nor anything
    • node1 migrates to can migrate to node0.
    • This function can run simultaneously with readers of
    • node_demotion[]. However, it can not run simultaneously
    • with itself. Exclusion is provided by memory hotplug events
    • being single-threaded.
  • */

+static void __set_migration_target_nodes(void)
+{

  • nodemask_t next_pass =3D NODE_MASK_NONE;
  • nodemask_t this_pass =3D NODE_MASK_NONE;
  • nodemask_t used_targets =3D NODE_MASK_NONE;
  • int node;
  • /*
    • Avoid any oddities like cycles that could occur
    • from changes in the topology. This will leave
    • a momentary gap when migration is disabled.
  • */
  • disable_all_migrate_targets();
  • /*
    • Ensure that the "disable" is visible across the system.
    • Readers will see either a combination of before+disable
    • state or disable+after. They will never see before and
    • after state together.
    • The before+after state together might have cycles and
    • could cause readers to do things like loop until this
    • function finishes. This ensures they can only see a
    • single "bad" read and would, for instance, only loop
    • once.
  • */
  • smp_wmb();
  • /*
    • Allocations go close to CPUs, first. Assume that
    • the migration path starts at the nodes with CPUs.
  • */
  • next_pass =3D node_states[N_CPU];

Is there a plan of allowing user to change where the migr= ation
path starts? Or maybe one step further providing an interface
to allow user to specify the demotion path. Something like
/sys/devices/system/node/node*/node_demotion.

I don't think that's necessary at least for now. Do you = know any real
world use case for this?

In our P9+volta system, GPU memory is exposed as a NUMA n= ode.
For the GPU workloads with data size greater than GPU memory size,
it will be very helpful to allow pages in GPU memory to be migrated/demot= ed
to CPU memory. With your current assumption, GPU memory -> CPU memory<= br> demotion seems not possible, right? This should also apply to any
system with a device memory exposed as a NUMA node and workloads running<= br> on the device and using CPU memory as a lower tier memory than the device=
memory.

=E2=80=94
Best Regards,
Yan, Zi

--=_MailMate_EE747415-A08F-42A4-80A0-0ACA9A83928A_=-- --=_MailMate_B213593D-221A-421A-B151-2B8E71B15048_= Content-Description: OpenPGP digital signature Content-Disposition: attachment; filename="signature.asc" Content-Type: application/pgp-signature; name="signature.asc" -----BEGIN PGP SIGNATURE----- iQJDBAEBCgAtFiEEh7yFAW3gwjwQ4C9anbJR82th+ooFAmDQpycPHHppeUBudmlk aWEuY29tAAoJEJ2yUfNrYfqKXBwQAInh/fb9llQpNNIsGWQ5OWXGNw7Om2QVNRUX ftMx0eQWXCNo3jqP7Fn5kNHeop6vlIILVooCu+Q6AynoPHIEsEMuS4McFNY+s+zT NKeAY+3FsdWWMJKr8n1SPAV0XZeVuQP2cyDeJGA6R8bP51+oNbt4WdV0NJqr1Vvw bA8eLja4JVqqTbrcpKFttMKo+kqU8ooFfXvxbespFfnxofokXMV+os2fleVfbDHH Wc6vmyD4hJ/FuZw5oONq1i4e7DQ9scnUYbj/zy2JLcoLSbkbWzjo5DiuyWs2+G3p wm8y/5M1TgbJ1q5BUmwE7FtjVT5U2Kt6vbdKsy9TIIdiJ7nbcSRlhCqXe6f20CWH 4SN4ocGXWNLelAcGpFbsmR5Un8D9dseOKIs0MkdztlfG6rrLq6xGUg/LcSD4z9vb GaWCzZm2/NMN7SjRRbrgmJXHeyk/1f4z9MvyaR54MjbS4QG7RQ+SaK8Kb2w57ay3 a7h5METrg47rpJMsAip80sp40JLeM92ns3+gpH0QgbUNL8fM5xqqSLJnlS6MgPqp qOX8Y9Fl5HCYVKaGkpJXJARBAWzlX2DzP3cNdooTenNjOw0SlbnjLqZo4e4gCAjc cNUhOhl4ACoyWZupeLmc6I7WmKZxeYWEdhimhxEUz4LIGx7GAVG7KwhIiQ7b7/ld H+4gp2VS =5Sfx -----END PGP SIGNATURE----- --=_MailMate_B213593D-221A-421A-B151-2B8E71B15048_=--