From nobody Tue Sep 9 22:23:41 2025 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 vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 620ABC678D4 for ; Fri, 3 Mar 2023 03:02:37 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229718AbjCCDCg (ORCPT ); Thu, 2 Mar 2023 22:02:36 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:51496 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229699AbjCCDC1 (ORCPT ); Thu, 2 Mar 2023 22:02:27 -0500 Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id C0F2857D38 for ; Thu, 2 Mar 2023 19:02:25 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1677812545; x=1709348545; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=dadG4erEiBTCMMwG4DB9rDq0meqMywzEolW1Kh5p2sI=; b=FeYqUOWOmNd9ADW/CaEFc385WgAIrdC+NVFwAK7skxohgis/PliRKHib dfdaVmMoekOKx3z7TvObrlkSsJ4IlR5QB/UHbL4U/NDxhRvqU4e1U9hnP racPqaAZ3XS37LQseGZzW6rEWY0oVYjoD3o3CL4iWdoE83stpyFKydomt KKt6RzvSzD5eU6K/YbvK+fJ0cTLp294qHR5HKy35A9rfnJmrYLATbRvLX bakIk0/AK51mIc3OujFzvnTbOTFBjK65VzZZxhuYwisX1WsjS7P0wwgIc 6rPQXsQ09tDGk7KnCOKTDhoSbrANf3u3vFRbgRr80JkOFNbd9xRjXpim3 g==; X-IronPort-AV: E=McAfee;i="6500,9779,10637"; a="421207103" X-IronPort-AV: E=Sophos;i="5.98,229,1673942400"; d="scan'208";a="421207103" Received: from orsmga007.jf.intel.com ([10.7.209.58]) by fmsmga105.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 02 Mar 2023 19:02:25 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6500,9779,10637"; a="668497684" X-IronPort-AV: E=Sophos;i="5.98,229,1673942400"; d="scan'208";a="668497684" Received: from xinyumao-mobl.ccr.corp.intel.com (HELO yhuang6-mobl2.ccr.corp.intel.com) ([10.255.31.231]) by orsmga007-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 02 Mar 2023 19:02:21 -0800 From: Huang Ying To: Andrew Morton Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Huang Ying , Hugh Dickins , "Xu, Pengfei" , Jan Kara , Baolin Wang , Christoph Hellwig , Stefan Roesch , Tejun Heo , Xin Hao , Zi Yan , Yang Shi , Matthew Wilcox , Mike Kravetz Subject: [PATCH -V2 1/3] migrate_pages: fix deadlock in batched migration Date: Fri, 3 Mar 2023 11:01:53 +0800 Message-Id: <20230303030155.160983-2-ying.huang@intel.com> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230303030155.160983-1-ying.huang@intel.com> References: <20230303030155.160983-1-ying.huang@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" Two deadlock bugs were reported for the migrate_pages() batching series. Thanks Hugh and Pengfei! For example, in the following deadlock trace snippet, INFO: task kworker/u4:0:9 blocked for more than 147 seconds. Not tainted 6.2.0-rc4-kvm+ #1314 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. task:kworker/u4:0 state:D stack:0 pid:9 ppid:2 flags:0x000= 04000 Workqueue: loop4 loop_rootcg_workfn Call Trace: __schedule+0x43b/0xd00 schedule+0x6a/0xf0 io_schedule+0x4a/0x80 folio_wait_bit_common+0x1b5/0x4e0 ? __pfx_wake_page_function+0x10/0x10 __filemap_get_folio+0x73d/0x770 shmem_get_folio_gfp+0x1fd/0xc80 shmem_write_begin+0x91/0x220 generic_perform_write+0x10e/0x2e0 __generic_file_write_iter+0x17e/0x290 ? generic_write_checks+0x12b/0x1a0 generic_file_write_iter+0x97/0x180 ? __sanitizer_cov_trace_const_cmp4+0x1a/0x20 do_iter_readv_writev+0x13c/0x210 ? __sanitizer_cov_trace_const_cmp4+0x1a/0x20 do_iter_write+0xf6/0x330 vfs_iter_write+0x46/0x70 loop_process_work+0x723/0xfe0 loop_rootcg_workfn+0x28/0x40 process_one_work+0x3cc/0x8d0 worker_thread+0x66/0x630 ? __pfx_worker_thread+0x10/0x10 kthread+0x153/0x190 ? __pfx_kthread+0x10/0x10 ret_from_fork+0x29/0x50 INFO: task repro:1023 blocked for more than 147 seconds. Not tainted 6.2.0-rc4-kvm+ #1314 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. task:repro state:D stack:0 pid:1023 ppid:360 flags:0x000= 04004 Call Trace: __schedule+0x43b/0xd00 schedule+0x6a/0xf0 io_schedule+0x4a/0x80 folio_wait_bit_common+0x1b5/0x4e0 ? compaction_alloc+0x77/0x1150 ? __pfx_wake_page_function+0x10/0x10 folio_wait_bit+0x30/0x40 folio_wait_writeback+0x2e/0x1e0 migrate_pages_batch+0x555/0x1ac0 ? __pfx_compaction_alloc+0x10/0x10 ? __pfx_compaction_free+0x10/0x10 ? __this_cpu_preempt_check+0x17/0x20 ? lock_is_held_type+0xe6/0x140 migrate_pages+0x100e/0x1180 ? __pfx_compaction_free+0x10/0x10 ? __pfx_compaction_alloc+0x10/0x10 compact_zone+0xe10/0x1b50 ? lock_is_held_type+0xe6/0x140 ? check_preemption_disabled+0x80/0xf0 compact_node+0xa3/0x100 ? __sanitizer_cov_trace_const_cmp8+0x1c/0x30 ? _find_first_bit+0x7b/0x90 sysctl_compaction_handler+0x5d/0xb0 proc_sys_call_handler+0x29d/0x420 proc_sys_write+0x2b/0x40 vfs_write+0x3a3/0x780 ksys_write+0xb7/0x180 __x64_sys_write+0x26/0x30 do_syscall_64+0x3b/0x90 entry_SYSCALL_64_after_hwframe+0x72/0xdc RIP: 0033:0x7f3a2471f59d RSP: 002b:00007ffe567f7288 EFLAGS: 00000217 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f3a2471f59d RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000005 RBP: 00007ffe567f72a0 R08: 0000000000000010 R09: 0000000000000010 R10: 0000000000000010 R11: 0000000000000217 R12: 00000000004012e0 R13: 00007ffe567f73e0 R14: 0000000000000000 R15: 0000000000000000 The page migration task has held the lock of the shmem folio A, and is waiting the writeback of the folio B of the file system on the loop block device to complete. While the loop worker task which writes back the folio B is waiting to lock the shmem folio A, because the folio A backs the folio B in the loop device. Thus deadlock is triggered. In general, if we have locked some other folios except the one we are migrating, it's not safe to wait synchronously, for example, to wait the writeback to complete or wait to lock the buffer head. To fix the deadlock, in this patch, we avoid to batch the page migration except for MIGRATE_ASYNC mode. In MIGRATE_ASYNC mode, synchronous waiting is avoided. The fix can be improved further. We will do that as soon as possible. Link: https://lore.kernel.org/linux-mm/87a6c8c-c5c1-67dc-1e32-eb30831d6e3d@= google.com/ Link: https://lore.kernel.org/linux-mm/874jrg7kke.fsf@yhuang6-desk2.ccr.cor= p.intel.com/ Link: https://lore.kernel.org/linux-mm/20230227110614.dngdub2j3exr6dfp@quac= k3/ Fixes: 5dfab109d519 ("migrate_pages: batch _unmap and _move") Signed-off-by: "Huang, Ying" Reported-by: Hugh Dickins Reported-by: "Xu, Pengfei" Cc: Jan Kara Cc: Baolin Wang Cc: Christoph Hellwig Cc: Stefan Roesch Cc: Tejun Heo Cc: Xin Hao Cc: Zi Yan Cc: Yang Shi Cc: Matthew Wilcox Cc: Mike Kravetz --- mm/migrate.c | 69 ++++++++++++++++++++-------------------------------- 1 file changed, 26 insertions(+), 43 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 04b9b3eeecef..7d751d6a2c15 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1115,7 +1115,7 @@ static void migrate_folio_done(struct folio *src, /* Obtain the lock on page, remove all ptes. */ static int migrate_folio_unmap(new_page_t get_new_page, free_page_t put_ne= w_page, unsigned long private, struct folio *src, - struct folio **dstp, int force, bool avoid_force_lock, + struct folio **dstp, int force, enum migrate_mode mode, enum migrate_reason reason, struct list_head *ret) { @@ -1166,17 +1166,6 @@ static int migrate_folio_unmap(new_page_t get_new_pa= ge, free_page_t put_new_page if (current->flags & PF_MEMALLOC) goto out; =20 - /* - * We have locked some folios and are going to wait to lock - * this folio. To avoid a potential deadlock, let's bail - * out and not do that. The locked folios will be moved and - * unlocked, then we can wait to lock this folio. - */ - if (avoid_force_lock) { - rc =3D -EDEADLOCK; - goto out; - } - folio_lock(src); } locked =3D true; @@ -1256,7 +1245,7 @@ static int migrate_folio_unmap(new_page_t get_new_pag= e, free_page_t put_new_page /* Establish migration ptes */ VM_BUG_ON_FOLIO(folio_test_anon(src) && !folio_test_ksm(src) && !anon_vma, src); - try_to_migrate(src, TTU_BATCH_FLUSH); + try_to_migrate(src, mode =3D=3D MIGRATE_ASYNC ? TTU_BATCH_FLUSH : 0); page_was_mapped =3D 1; } =20 @@ -1270,7 +1259,7 @@ static int migrate_folio_unmap(new_page_t get_new_pag= e, free_page_t put_new_page * A folio that has not been unmapped will be restored to * right list unless we want to retry. */ - if (rc =3D=3D -EAGAIN || rc =3D=3D -EDEADLOCK) + if (rc =3D=3D -EAGAIN) ret =3D NULL; =20 migrate_folio_undo_src(src, page_was_mapped, anon_vma, locked, ret); @@ -1621,6 +1610,11 @@ static int migrate_hugetlbs(struct list_head *from, = new_page_t get_new_page, /* * migrate_pages_batch() first unmaps folios in the from list as many as * possible, then move the unmapped folios. + * + * We only batch migration if mode =3D=3D MIGRATE_ASYNC to avoid to wait a + * lock or bit when we have locked more than one folio. Which may cause + * deadlock (e.g., for loop device). So, if mode !=3D MIGRATE_ASYNC, the + * length of the from list must be <=3D 1. */ static int migrate_pages_batch(struct list_head *from, new_page_t get_new_= page, free_page_t put_new_page, unsigned long private, @@ -1643,11 +1637,11 @@ static int migrate_pages_batch(struct list_head *fr= om, new_page_t get_new_page, LIST_HEAD(dst_folios); bool nosplit =3D (reason =3D=3D MR_NUMA_MISPLACED); bool no_split_folio_counting =3D false; - bool avoid_force_lock; =20 + VM_WARN_ON_ONCE(mode !=3D MIGRATE_ASYNC && + !list_empty(from) && !list_is_singular(from)); retry: rc_saved =3D 0; - avoid_force_lock =3D false; retry =3D 1; for (pass =3D 0; pass < NR_MAX_MIGRATE_PAGES_RETRY && (retry || large_retry); @@ -1692,15 +1686,14 @@ static int migrate_pages_batch(struct list_head *fr= om, new_page_t get_new_page, } =20 rc =3D migrate_folio_unmap(get_new_page, put_new_page, private, - folio, &dst, pass > 2, avoid_force_lock, - mode, reason, ret_folios); + folio, &dst, pass > 2, mode, + reason, ret_folios); /* * The rules are: * Success: folio will be freed * Unmap: folio will be put on unmap_folios list, * dst folio put on dst_folios list * -EAGAIN: stay on the from list - * -EDEADLOCK: stay on the from list * -ENOMEM: stay on the from list * Other errno: put on ret_folios list */ @@ -1752,14 +1745,6 @@ static int migrate_pages_batch(struct list_head *fro= m, new_page_t get_new_page, goto out; else goto move; - case -EDEADLOCK: - /* - * The folio cannot be locked for potential deadlock. - * Go move (and unlock) all locked folios. Then we can - * try again. - */ - rc_saved =3D rc; - goto move; case -EAGAIN: if (is_large) { large_retry++; @@ -1774,11 +1759,6 @@ static int migrate_pages_batch(struct list_head *fro= m, new_page_t get_new_page, stats->nr_thp_succeeded +=3D is_thp; break; case MIGRATEPAGE_UNMAP: - /* - * We have locked some folios, don't force lock - * to avoid deadlock. - */ - avoid_force_lock =3D true; list_move_tail(&folio->lru, &unmap_folios); list_add_tail(&dst->lru, &dst_folios); break; @@ -1903,17 +1883,15 @@ static int migrate_pages_batch(struct list_head *fr= om, new_page_t get_new_page, */ list_splice_init(from, ret_folios); list_splice_init(&split_folios, from); + /* + * Force async mode to avoid to wait lock or bit when we have + * locked more than one folios. + */ + mode =3D MIGRATE_ASYNC; no_split_folio_counting =3D true; goto retry; } =20 - /* - * We have unlocked all locked folios, so we can force lock now, let's - * try again. - */ - if (rc =3D=3D -EDEADLOCK) - goto retry; - return rc; } =20 @@ -1948,7 +1926,7 @@ int migrate_pages(struct list_head *from, new_page_t = get_new_page, enum migrate_mode mode, int reason, unsigned int *ret_succeeded) { int rc, rc_gather; - int nr_pages; + int nr_pages, batch; struct folio *folio, *folio2; LIST_HEAD(folios); LIST_HEAD(ret_folios); @@ -1962,6 +1940,11 @@ int migrate_pages(struct list_head *from, new_page_t= get_new_page, mode, reason, &stats, &ret_folios); if (rc_gather < 0) goto out; + + if (mode =3D=3D MIGRATE_ASYNC) + batch =3D NR_MAX_BATCHED_MIGRATION; + else + batch =3D 1; again: nr_pages =3D 0; list_for_each_entry_safe(folio, folio2, from, lru) { @@ -1972,11 +1955,11 @@ int migrate_pages(struct list_head *from, new_page_= t get_new_page, } =20 nr_pages +=3D folio_nr_pages(folio); - if (nr_pages > NR_MAX_BATCHED_MIGRATION) + if (nr_pages >=3D batch) break; } - if (nr_pages > NR_MAX_BATCHED_MIGRATION) - list_cut_before(&folios, from, &folio->lru); + if (nr_pages >=3D batch) + list_cut_before(&folios, from, &folio2->lru); else list_splice_init(from, &folios); rc =3D migrate_pages_batch(&folios, get_new_page, put_new_page, private, --=20 2.39.2 From nobody Tue Sep 9 22:23:41 2025 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 vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id E2D4AC6FA8E for ; Fri, 3 Mar 2023 03:02:41 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229797AbjCCDCk (ORCPT ); Thu, 2 Mar 2023 22:02:40 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:51358 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229756AbjCCDCe (ORCPT ); Thu, 2 Mar 2023 22:02:34 -0500 Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id C5D3357D0F for ; Thu, 2 Mar 2023 19:02:29 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1677812549; x=1709348549; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=CQPo+dMFAQNly6q5kcMf4N2WZveYtr21szEchM1lhiI=; b=Za/Iy2LWazIztYhCqUk9X1W9wEk0adKLkM7Isbh9Tl4KtCJDBz6x8NYP Qi/LzWdU3HNYU4Kna+7/0M3e7c/1eLY+PQHCFzGiPLLYYTWLehkg7Y5gu I9+85PzMf7Q0G5/a7wX71iSMc24o00iRJChPMBv8p1juDprbOZ6YfQCKJ 7DoCv1oBlwlm4AYPPf2qk1g2DH8nstAcl0aMIciH9RIB8rp4a0ATDzaKg wQLklcxovVCFLcIzevRXzLhL3Ahzk1M3JkCdzMTKANU49NKDRQ9r6BArB MMSvf77nhn19ZcjhcY6LitKraxOFgLFD6dsY7ctDZwkkHkmxnfnK/xVM4 A==; X-IronPort-AV: E=McAfee;i="6500,9779,10637"; a="421207125" X-IronPort-AV: E=Sophos;i="5.98,229,1673942400"; d="scan'208";a="421207125" Received: from orsmga007.jf.intel.com ([10.7.209.58]) by fmsmga105.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 02 Mar 2023 19:02:29 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6500,9779,10637"; a="668497693" X-IronPort-AV: E=Sophos;i="5.98,229,1673942400"; d="scan'208";a="668497693" Received: from xinyumao-mobl.ccr.corp.intel.com (HELO yhuang6-mobl2.ccr.corp.intel.com) ([10.255.31.231]) by orsmga007-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 02 Mar 2023 19:02:25 -0800 From: Huang Ying To: Andrew Morton Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Huang Ying , Baolin Wang , Hugh Dickins , "Xu, Pengfei" , Christoph Hellwig , Stefan Roesch , Tejun Heo , Xin Hao , Zi Yan , Yang Shi , Matthew Wilcox , Mike Kravetz Subject: [PATCH -V2 2/3] migrate_pages: move split folios processing out of migrate_pages_batch() Date: Fri, 3 Mar 2023 11:01:54 +0800 Message-Id: <20230303030155.160983-3-ying.huang@intel.com> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230303030155.160983-1-ying.huang@intel.com> References: <20230303030155.160983-1-ying.huang@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" To simplify the code logic and reduce the line number. Fixes: 5dfab109d519 ("migrate_pages: batch _unmap and _move") Signed-off-by: "Huang, Ying" Reviewed-by: Baolin Wang Cc: Hugh Dickins Cc: "Xu, Pengfei" Cc: Christoph Hellwig Cc: Stefan Roesch Cc: Tejun Heo Cc: Xin Hao Cc: Zi Yan Cc: Yang Shi Cc: Matthew Wilcox Cc: Mike Kravetz --- mm/migrate.c | 78 +++++++++++++++++++--------------------------------- 1 file changed, 28 insertions(+), 50 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 7d751d6a2c15..61009e2ae2be 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1619,9 +1619,10 @@ static int migrate_hugetlbs(struct list_head *from, = new_page_t get_new_page, static int migrate_pages_batch(struct list_head *from, new_page_t get_new_= page, free_page_t put_new_page, unsigned long private, enum migrate_mode mode, int reason, struct list_head *ret_folios, - struct migrate_pages_stats *stats) + struct list_head *split_folios, struct migrate_pages_stats *stats, + int nr_pass) { - int retry; + int retry =3D 1; int large_retry =3D 1; int thp_retry =3D 1; int nr_failed =3D 0; @@ -1631,21 +1632,15 @@ static int migrate_pages_batch(struct list_head *fr= om, new_page_t get_new_page, bool is_large =3D false; bool is_thp =3D false; struct folio *folio, *folio2, *dst =3D NULL, *dst2; - int rc, rc_saved, nr_pages; - LIST_HEAD(split_folios); + int rc, rc_saved =3D 0, nr_pages; LIST_HEAD(unmap_folios); LIST_HEAD(dst_folios); bool nosplit =3D (reason =3D=3D MR_NUMA_MISPLACED); - bool no_split_folio_counting =3D false; =20 VM_WARN_ON_ONCE(mode !=3D MIGRATE_ASYNC && !list_empty(from) && !list_is_singular(from)); -retry: - rc_saved =3D 0; - retry =3D 1; - for (pass =3D 0; - pass < NR_MAX_MIGRATE_PAGES_RETRY && (retry || large_retry); - pass++) { + + for (pass =3D 0; pass < nr_pass && (retry || large_retry); pass++) { retry =3D 0; large_retry =3D 0; thp_retry =3D 0; @@ -1676,7 +1671,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, if (!thp_migration_supported() && is_thp) { nr_large_failed++; stats->nr_thp_failed++; - if (!try_split_folio(folio, &split_folios)) { + if (!try_split_folio(folio, split_folios)) { stats->nr_thp_split++; continue; } @@ -1708,7 +1703,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, stats->nr_thp_failed +=3D is_thp; /* Large folio NUMA faulting doesn't split to retry. */ if (!nosplit) { - int ret =3D try_split_folio(folio, &split_folios); + int ret =3D try_split_folio(folio, split_folios); =20 if (!ret) { stats->nr_thp_split +=3D is_thp; @@ -1725,18 +1720,11 @@ static int migrate_pages_batch(struct list_head *fr= om, new_page_t get_new_page, break; } } - } else if (!no_split_folio_counting) { + } else { nr_failed++; } =20 stats->nr_failed_pages +=3D nr_pages + nr_retry_pages; - /* - * There might be some split folios of fail-to-migrate large - * folios left in split_folios list. Move them to ret_folios - * list so that they could be put back to the right list by - * the caller otherwise the folio refcnt will be leaked. - */ - list_splice_init(&split_folios, ret_folios); /* nr_failed isn't updated for not used */ nr_large_failed +=3D large_retry; stats->nr_thp_failed +=3D thp_retry; @@ -1749,7 +1737,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, if (is_large) { large_retry++; thp_retry +=3D is_thp; - } else if (!no_split_folio_counting) { + } else { retry++; } nr_retry_pages +=3D nr_pages; @@ -1772,7 +1760,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, if (is_large) { nr_large_failed++; stats->nr_thp_failed +=3D is_thp; - } else if (!no_split_folio_counting) { + } else { nr_failed++; } =20 @@ -1790,9 +1778,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, try_to_unmap_flush(); =20 retry =3D 1; - for (pass =3D 0; - pass < NR_MAX_MIGRATE_PAGES_RETRY && (retry || large_retry); - pass++) { + for (pass =3D 0; pass < nr_pass && (retry || large_retry); pass++) { retry =3D 0; large_retry =3D 0; thp_retry =3D 0; @@ -1821,7 +1807,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, if (is_large) { large_retry++; thp_retry +=3D is_thp; - } else if (!no_split_folio_counting) { + } else { retry++; } nr_retry_pages +=3D nr_pages; @@ -1834,7 +1820,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, if (is_large) { nr_large_failed++; stats->nr_thp_failed +=3D is_thp; - } else if (!no_split_folio_counting) { + } else { nr_failed++; } =20 @@ -1871,27 +1857,6 @@ static int migrate_pages_batch(struct list_head *fro= m, new_page_t get_new_page, dst2 =3D list_next_entry(dst, lru); } =20 - /* - * Try to migrate split folios of fail-to-migrate large folios, no - * nr_failed counting in this round, since all split folios of a - * large folio is counted as 1 failure in the first round. - */ - if (rc >=3D 0 && !list_empty(&split_folios)) { - /* - * Move non-migrated folios (after NR_MAX_MIGRATE_PAGES_RETRY - * retries) to ret_folios to avoid migrating them again. - */ - list_splice_init(from, ret_folios); - list_splice_init(&split_folios, from); - /* - * Force async mode to avoid to wait lock or bit when we have - * locked more than one folios. - */ - mode =3D MIGRATE_ASYNC; - no_split_folio_counting =3D true; - goto retry; - } - return rc; } =20 @@ -1930,6 +1895,7 @@ int migrate_pages(struct list_head *from, new_page_t = get_new_page, struct folio *folio, *folio2; LIST_HEAD(folios); LIST_HEAD(ret_folios); + LIST_HEAD(split_folios); struct migrate_pages_stats stats; =20 trace_mm_migrate_pages_start(mode, reason); @@ -1963,12 +1929,24 @@ int migrate_pages(struct list_head *from, new_page_= t get_new_page, else list_splice_init(from, &folios); rc =3D migrate_pages_batch(&folios, get_new_page, put_new_page, private, - mode, reason, &ret_folios, &stats); + mode, reason, &ret_folios, &split_folios, &stats, + NR_MAX_MIGRATE_PAGES_RETRY); list_splice_tail_init(&folios, &ret_folios); if (rc < 0) { rc_gather =3D rc; + list_splice_tail(&split_folios, &ret_folios); goto out; } + if (!list_empty(&split_folios)) { + /* + * Failure isn't counted since all split folios of a large folio + * is counted as 1 failure already. And, we only try to migrate + * with minimal effort, force MIGRATE_ASYNC mode and retry once. + */ + migrate_pages_batch(&split_folios, get_new_page, put_new_page, private, + MIGRATE_ASYNC, reason, &ret_folios, NULL, &stats, 1); + list_splice_tail_init(&split_folios, &ret_folios); + } rc_gather +=3D rc; if (!list_empty(from)) goto again; --=20 2.39.2 From nobody Tue Sep 9 22:23:41 2025 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 vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id E5CEFC6FA8E for ; Fri, 3 Mar 2023 03:02:50 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229804AbjCCDCt (ORCPT ); Thu, 2 Mar 2023 22:02:49 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:52036 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229775AbjCCDCj (ORCPT ); Thu, 2 Mar 2023 22:02:39 -0500 Received: from mga05.intel.com (mga05.intel.com [192.55.52.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id F240258B61 for ; Thu, 2 Mar 2023 19:02:33 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1677812553; x=1709348553; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=6dOm2+hfvH1d/Mz6q4OoqdRr+NjQWevzknZQQER47XU=; b=CtOZoioXJ7ASFdPyhQbdxb7bWJykIGEU8SMPqp+ZB4xvPIqIA50DgwVS 9WEjCgimanjZmsRs7yziHcSSzN8ZacUj3zhwaFkZNtxfA74z0lSnVRYEq d0YifGQaqnlBJzmohpkJi+GjhdRe0TE0qR+tUndEWXFFieW4QRZlUx9Iu ULHo+cOFAtRU25duC9urve/K3oWw09Td8RuT+nvsUnXpJSdzQLWl+sXj6 /nfoSHI1nT3mXuMdB0bEXWSqi4tdwSJTTSzc0a05DLEThVQai5Sh2pjN7 L51f5H7ity70JA6DIjIP0nmTvC87jd2SWzTGUkH42lMREGoq9DAGY0CHo A==; X-IronPort-AV: E=McAfee;i="6500,9779,10637"; a="421207150" X-IronPort-AV: E=Sophos;i="5.98,229,1673942400"; d="scan'208";a="421207150" Received: from orsmga007.jf.intel.com ([10.7.209.58]) by fmsmga105.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 02 Mar 2023 19:02:33 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6500,9779,10637"; a="668497706" X-IronPort-AV: E=Sophos;i="5.98,229,1673942400"; d="scan'208";a="668497706" Received: from xinyumao-mobl.ccr.corp.intel.com (HELO yhuang6-mobl2.ccr.corp.intel.com) ([10.255.31.231]) by orsmga007-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 02 Mar 2023 19:02:29 -0800 From: Huang Ying To: Andrew Morton Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Huang Ying , Hugh Dickins , Baolin Wang , "Xu, Pengfei" , Christoph Hellwig , Stefan Roesch , Tejun Heo , Xin Hao , Zi Yan , Yang Shi , Matthew Wilcox , Mike Kravetz Subject: [PATCH -V2 3/3] migrate_pages: try migrate in batch asynchronously firstly Date: Fri, 3 Mar 2023 11:01:55 +0800 Message-Id: <20230303030155.160983-4-ying.huang@intel.com> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230303030155.160983-1-ying.huang@intel.com> References: <20230303030155.160983-1-ying.huang@intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Type: text/plain; charset="utf-8" When we have locked more than one folios, we cannot wait the lock or bit (e.g., page lock, buffer head lock, writeback bit) synchronously. Otherwise deadlock may be triggered. This make it hard to batch the synchronous migration directly. This patch re-enables batching synchronous migration via trying to migrate in batch asynchronously firstly. And any folios that are failed to be migrated asynchronously will be migrated synchronously one by one. Test shows that this can restore the TLB flushing batching performance for synchronous migration effectively. Fixes: 5dfab109d519 ("migrate_pages: batch _unmap and _move") Signed-off-by: "Huang, Ying" Tested-by: Hugh Dickins Reviewed-by: Baolin Wang Cc: "Xu, Pengfei" Cc: Christoph Hellwig Cc: Stefan Roesch Cc: Tejun Heo Cc: Xin Hao Cc: Zi Yan Cc: Yang Shi Cc: Matthew Wilcox Cc: Mike Kravetz --- mm/migrate.c | 80 ++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 62 insertions(+), 18 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 61009e2ae2be..ae86345e12a5 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -1115,9 +1115,8 @@ static void migrate_folio_done(struct folio *src, /* Obtain the lock on page, remove all ptes. */ static int migrate_folio_unmap(new_page_t get_new_page, free_page_t put_ne= w_page, unsigned long private, struct folio *src, - struct folio **dstp, int force, - enum migrate_mode mode, enum migrate_reason reason, - struct list_head *ret) + struct folio **dstp, enum migrate_mode mode, + enum migrate_reason reason, struct list_head *ret) { struct folio *dst; int rc =3D -EAGAIN; @@ -1147,7 +1146,7 @@ static int migrate_folio_unmap(new_page_t get_new_pag= e, free_page_t put_new_page dst->private =3D NULL; =20 if (!folio_trylock(src)) { - if (!force || mode =3D=3D MIGRATE_ASYNC) + if (mode =3D=3D MIGRATE_ASYNC) goto out; =20 /* @@ -1185,8 +1184,6 @@ static int migrate_folio_unmap(new_page_t get_new_pag= e, free_page_t put_new_page rc =3D -EBUSY; goto out; } - if (!force) - goto out; folio_wait_writeback(src); } =20 @@ -1500,6 +1497,9 @@ static inline int try_split_folio(struct folio *folio= , struct list_head *split_f #define NR_MAX_BATCHED_MIGRATION 512 #endif #define NR_MAX_MIGRATE_PAGES_RETRY 10 +#define NR_MAX_MIGRATE_ASYNC_RETRY 3 +#define NR_MAX_MIGRATE_SYNC_RETRY \ + (NR_MAX_MIGRATE_PAGES_RETRY - NR_MAX_MIGRATE_ASYNC_RETRY) =20 struct migrate_pages_stats { int nr_succeeded; /* Normal and large folios migrated successfully, in @@ -1681,8 +1681,7 @@ static int migrate_pages_batch(struct list_head *from= , new_page_t get_new_page, } =20 rc =3D migrate_folio_unmap(get_new_page, put_new_page, private, - folio, &dst, pass > 2, mode, - reason, ret_folios); + folio, &dst, mode, reason, ret_folios); /* * The rules are: * Success: folio will be freed @@ -1860,6 +1859,51 @@ static int migrate_pages_batch(struct list_head *fro= m, new_page_t get_new_page, return rc; } =20 +static int migrate_pages_sync(struct list_head *from, new_page_t get_new_p= age, + free_page_t put_new_page, unsigned long private, + enum migrate_mode mode, int reason, struct list_head *ret_folios, + struct list_head *split_folios, struct migrate_pages_stats *stats) +{ + int rc, nr_failed =3D 0; + LIST_HEAD(folios); + struct migrate_pages_stats astats; + + memset(&astats, 0, sizeof(astats)); + /* Try to migrate in batch with MIGRATE_ASYNC mode firstly */ + rc =3D migrate_pages_batch(from, get_new_page, put_new_page, private, MIG= RATE_ASYNC, + reason, &folios, split_folios, &astats, + NR_MAX_MIGRATE_ASYNC_RETRY); + stats->nr_succeeded +=3D astats.nr_succeeded; + stats->nr_thp_succeeded +=3D astats.nr_thp_succeeded; + stats->nr_thp_split +=3D astats.nr_thp_split; + if (rc < 0) { + stats->nr_failed_pages +=3D astats.nr_failed_pages; + stats->nr_thp_failed +=3D astats.nr_thp_failed; + list_splice_tail(&folios, ret_folios); + return rc; + } + stats->nr_thp_failed +=3D astats.nr_thp_split; + nr_failed +=3D astats.nr_thp_split; + /* + * Fall back to migrate all failed folios one by one synchronously. All + * failed folios except split THPs will be retried, so their failure + * isn't counted + */ + list_splice_tail_init(&folios, from); + while (!list_empty(from)) { + list_move(from->next, &folios); + rc =3D migrate_pages_batch(&folios, get_new_page, put_new_page, + private, mode, reason, ret_folios, + split_folios, stats, NR_MAX_MIGRATE_SYNC_RETRY); + list_splice_tail_init(&folios, ret_folios); + if (rc < 0) + return rc; + nr_failed +=3D rc; + } + + return nr_failed; +} + /* * migrate_pages - migrate the folios specified in a list, to the free fol= ios * supplied as the target for the page migration @@ -1891,7 +1935,7 @@ int migrate_pages(struct list_head *from, new_page_t = get_new_page, enum migrate_mode mode, int reason, unsigned int *ret_succeeded) { int rc, rc_gather; - int nr_pages, batch; + int nr_pages; struct folio *folio, *folio2; LIST_HEAD(folios); LIST_HEAD(ret_folios); @@ -1907,10 +1951,6 @@ int migrate_pages(struct list_head *from, new_page_t= get_new_page, if (rc_gather < 0) goto out; =20 - if (mode =3D=3D MIGRATE_ASYNC) - batch =3D NR_MAX_BATCHED_MIGRATION; - else - batch =3D 1; again: nr_pages =3D 0; list_for_each_entry_safe(folio, folio2, from, lru) { @@ -1921,16 +1961,20 @@ int migrate_pages(struct list_head *from, new_page_= t get_new_page, } =20 nr_pages +=3D folio_nr_pages(folio); - if (nr_pages >=3D batch) + if (nr_pages >=3D NR_MAX_BATCHED_MIGRATION) break; } - if (nr_pages >=3D batch) + if (nr_pages >=3D NR_MAX_BATCHED_MIGRATION) list_cut_before(&folios, from, &folio2->lru); else list_splice_init(from, &folios); - rc =3D migrate_pages_batch(&folios, get_new_page, put_new_page, private, - mode, reason, &ret_folios, &split_folios, &stats, - NR_MAX_MIGRATE_PAGES_RETRY); + if (mode =3D=3D MIGRATE_ASYNC) + rc =3D migrate_pages_batch(&folios, get_new_page, put_new_page, private, + mode, reason, &ret_folios, &split_folios, &stats, + NR_MAX_MIGRATE_PAGES_RETRY); + else + rc =3D migrate_pages_sync(&folios, get_new_page, put_new_page, private, + mode, reason, &ret_folios, &split_folios, &stats); list_splice_tail_init(&folios, &ret_folios); if (rc < 0) { rc_gather =3D rc; --=20 2.39.2