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 kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by smtp.lore.kernel.org (Postfix) with ESMTP id 6E34BC54EE9 for ; Tue, 20 Sep 2022 02:23:12 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id ED04880008; Mon, 19 Sep 2022 22:23:11 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id E592680007; Mon, 19 Sep 2022 22:23:11 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id CD25B80008; Mon, 19 Sep 2022 22:23:11 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0014.hostedemail.com [216.40.44.14]) by kanga.kvack.org (Postfix) with ESMTP id B84D880007 for ; Mon, 19 Sep 2022 22:23:11 -0400 (EDT) Received: from smtpin17.hostedemail.com (a10.router.float.18 [10.200.18.1]) by unirelay05.hostedemail.com (Postfix) with ESMTP id 91311405AF for ; Tue, 20 Sep 2022 02:23:11 +0000 (UTC) X-FDA: 79930866582.17.C824B7F Received: from loongson.cn (mail.loongson.cn [114.242.206.163]) by imf03.hostedemail.com (Postfix) with ESMTP id E0B63200A5 for ; Tue, 20 Sep 2022 02:23:09 +0000 (UTC) Received: from [10.180.13.185] (unknown [10.180.13.185]) by localhost.localdomain (Coremail) with SMTP id AQAAf8AxFeIJJCljTkweAA--.49621S3; Tue, 20 Sep 2022 10:23:06 +0800 (CST) Subject: Re: [PATCH] mm/vmscan: don't scan adjust too much if current is not kswapd From: Hongchen Zhang To: Mel Gorman Cc: Matthew Wilcox , Andrew Morton , linux-mm@kvack.org, linux-kernel@vger.kernel.org References: <20220914023318.549118-1-zhanghongchen@loongson.cn> <20220914155142.bf388515a39fb45bae987231@linux-foundation.org> <6bcb4883-03d0-88eb-4c42-84fff0a9a141@loongson.cn> <54813a74-cc0e-e470-c632-78437a0d0ad4@loongson.cn> <4bd0012e-77ff-9d0d-e295-800471994aeb@loongson.cn> Message-ID: Date: Tue, 20 Sep 2022 10:23:05 +0800 User-Agent: Mozilla/5.0 (X11; Linux loongarch64; rv:68.0) Gecko/20100101 Thunderbird/68.7.0 MIME-Version: 1.0 In-Reply-To: <4bd0012e-77ff-9d0d-e295-800471994aeb@loongson.cn> Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: en-US Content-Transfer-Encoding: 8bit X-CM-TRANSID:AQAAf8AxFeIJJCljTkweAA--.49621S3 X-Coremail-Antispam: 1UD129KBjvJXoW3XF43Gw1xAw1UXr1UKr1rZwb_yoW3Gry5pF 15tFsrKr48Jr1Utr47Kw4qqr18Kr1DC3W5Wry8Jr17CF1qvrn8Jw48Gr4YkF1DGr18ury2 qrW5Xr12vr17Jw7anT9S1TB71UUUUUUqnTZGkaVYY2UrUUUUjbIjqfuFe4nvWSU5nxnvy2 9KBjDU0xBIdaVrnRJUUUvYb7Iv0xC_Kw4lb4IE77IF4wAFF20E14v26r1j6r4UM7CY07I2 0VC2zVCF04k26cxKx2IYs7xG6rWj6s0DM7CIcVAFz4kK6r1j6r18M28lY4IEw2IIxxk0rw A2F7IY1VAKz4vEj48ve4kI8wA2z4x0Y4vE2Ix0cI8IcVAFwI0_Xr0_Ar1l84ACjcxK6xII jxv20xvEc7CjxVAFwI0_Cr0_Gr1UM28EF7xvwVC2z280aVAFwI0_Gr1j6F4UJwA2z4x0Y4 vEx4A2jsIEc7CjxVAFwI0_Cr1j6rxdM2AIxVAIcxkEcVAq07x20xvEncxIr21l5I8CrVAC Y4xI64kE6c02F40Ex7xfMcIj6xIIjxv20xvE14v26r1j6r18McIj6I8E87Iv67AKxVWUJV W8JwAm72CE4IkC6x0Yz7v_Jr0_Gr1lF7xvr2IY64vIr41lc7I2V7IY0VAS07AlzVAYIcxG 8wCY02Avz4vE-syl42xK82IYc2Ij64vIr41l4I8I3I0E4IkC6x0Yz7v_Jr0_Gr1lx2IqxV Aqx4xG67AKxVWUJVWUGwC20s026x8GjcxK67AKxVWUGVWUWwC2zVAF1VAY17CE14v26r12 6r1DMIIYrxkI7VAKI48JMIIF0xvE2Ix0cI8IcVAFwI0_Jr0_JF4lIxAIcVC0I7IYx2IY6x kF7I0E14v26r1j6r4UMIIF0xvE42xK8VAvwI8IcIk0rVWUJVWUCwCI42IY6I8E87Iv67AK xVWUJVW8JwCI42IY6I8E87Iv6xkF7I0E14v26r1j6r4UYxBIdaVFxhVjvjDU0xZFpf9x07 bOoGdUUUUU= X-CM-SenderInfo: x2kd0w5krqwupkhqwqxorr0wxvrqhubq/ ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1663640591; a=rsa-sha256; cv=none; b=Q1ZuRqgumHPYGdlX4+nguRTw2BM9Tq2lIKD0R74RLshjXkDfxk74eM+UHZ1gb1NVA4Cv6O e1VE8cGMdue+aTJUhziPaNbBWLiIOFXwAaRtp/RcOX9CF3a0mZKYo4uO5XdzriphFkNTtG NhBZjg6UUMnKLbxdLL++GvmlnR87s3M= ARC-Authentication-Results: i=1; imf03.hostedemail.com; dkim=none; spf=pass (imf03.hostedemail.com: domain of zhanghongchen@loongson.cn designates 114.242.206.163 as permitted sender) smtp.mailfrom=zhanghongchen@loongson.cn; dmarc=none ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1663640591; h=from:from:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references; bh=n9KNq6sHReKZaEkyi7++u3TAtfM/pj61q+xzfIhbcPU=; b=4wpuWDyoCfAt+q3t3qJhXn/ucpv27gQIh8HxfVxo2wHJgQk2wJx2swFG2cXfjXrw6pP78f RtqzBLJl/gtApUVarmbCSlHAnx5EIzpV+z1hEYEiB6evSMu7I2/2mFHIn7bBl4XKtUaO+t GzeF9B8rO4w2TQgQ08x/k9I6LK3HJdA= X-Stat-Signature: xu6dszzenndpbjusehdw7ibxd6h97jtu X-Rspamd-Queue-Id: E0B63200A5 Authentication-Results: imf03.hostedemail.com; dkim=none; spf=pass (imf03.hostedemail.com: domain of zhanghongchen@loongson.cn designates 114.242.206.163 as permitted sender) smtp.mailfrom=zhanghongchen@loongson.cn; dmarc=none X-Rspam-User: X-Rspamd-Server: rspam01 X-HE-Tag: 1663640589-225484 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: Hi Mel, The scan adjust algorithm was originally introduced by you from commmit e82e0561dae9 ("mm: vmscan: obey proportional scanning requirements for kswapd"), any suggestion about this fix patch? In short, proportional scanning is not friendly to processes other than kswapd. Thanks Hongchen Zhang On 2022/9/16 pm 6:19, Hongchen Zhang wrote: > Hi Andrew and Matthew, > > On 2022/9/16 pm 4:40, Matthew Wilcox wrote: >> On Fri, Sep 16, 2022 at 08:57:50AM +0800, Hongchen Zhang wrote: >>> Hi Andrew , >>> >>> On 2022/9/15 pm 5:00, Matthew Wilcox wrote: >>>> On Thu, Sep 15, 2022 at 04:02:41PM +0800, Hongchen Zhang wrote: >>>>> Hi Matthew, >>>>> On 2022/9/15 pm 3:28, Matthew Wilcox wrote: >>>>>> On Thu, Sep 15, 2022 at 09:19:48AM +0800, Hongchen Zhang wrote: >>>>>>> [ 3748.453561] INFO: task float_bessel:77920 blocked for more >>>>>>> than 120 >>>>>>> seconds. >>>>>>> [ 3748.460839]       Not tainted 5.15.0-46-generic #49-Ubuntu >>>>>>> [ 3748.466490] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" >>>>>>> disables >>>>>>> this message. >>>>>>> [ 3748.474618] task:float_bessel    state:D stack:    0 pid:77920 >>>>>>> ppid: >>>>>>> 77327 flags:0x00004002 >>>>>>> [ 3748.483358] Call Trace: >>>>>>> [ 3748.485964]  >>>>>>> [ 3748.488150]  __schedule+0x23d/0x590 >>>>>>> [ 3748.491804]  schedule+0x4e/0xc0 >>>>>>> [ 3748.495038]  rwsem_down_read_slowpath+0x336/0x390 >>>>>>> [ 3748.499886]  ? copy_user_enhanced_fast_string+0xe/0x40 >>>>>>> [ 3748.505181]  down_read+0x43/0xa0 >>>>>>> [ 3748.508518]  do_user_addr_fault+0x41c/0x670 >>>>>>> [ 3748.512799]  exc_page_fault+0x77/0x170 >>>>>>> [ 3748.516673]  asm_exc_page_fault+0x26/0x30 >>>>>>> [ 3748.520824] RIP: 0010:copy_user_enhanced_fast_string+0xe/0x40 >>>>>>> [ 3748.526764] Code: 89 d1 c1 e9 03 83 e2 07 f3 48 a5 89 d1 f3 a4 >>>>>>> 31 c0 0f >>>>>>> 01 ca c3 cc cc cc cc 0f 1f 00 0f 01 cb 83 fa 40 0f 82 70 ff ff ff >>>>>>> 89 d1 >>>>>>> a4 31 c0 0f 01 ca c3 cc cc cc cc 66 08 >>>>>>> [ 3748.546120] RSP: 0018:ffffaa9248fffb90 EFLAGS: 00050206 >>>>>>> [ 3748.551495] RAX: 00007f99faa1a010 RBX: ffffaa9248fffd88 RCX: >>>>>>> 0000000000000010 >>>>>>> [ 3748.558828] RDX: 0000000000001000 RSI: ffff9db397ab8ff0 RDI: >>>>>>> 00007f99faa1a000 >>>>>>> [ 3748.566160] RBP: ffffaa9248fffbf0 R08: ffffcc2fc2965d80 R09: >>>>>>> 0000000000000014 >>>>>>> [ 3748.573492] R10: 0000000000000000 R11: 0000000000000014 R12: >>>>>>> 0000000000001000 >>>>>>> [ 3748.580858] R13: 0000000000001000 R14: 0000000000000000 R15: >>>>>>> ffffaa9248fffd98 >>>>>>> [ 3748.588196]  ? copy_page_to_iter+0x10e/0x400 >>>>>>> [ 3748.592614]  filemap_read+0x174/0x3e0 >>>>>> >>>>>> Interesting; it wasn't the process itself which triggered the page >>>>>> fault; the process called read() and the kernel took the page >>>>>> fault to >>>>>> satisfy the read() call. >>>>>> >>>>>>> [ 3748.596354]  ? ima_file_check+0x6a/0xa0 >>>>>>> [ 3748.600301]  generic_file_read_iter+0xe5/0x150 >>>>>>> [ 3748.604884]  ext4_file_read_iter+0x5b/0x190 >>>>>>> [ 3748.609164]  ? aa_file_perm+0x102/0x250 >>>>>>> [ 3748.613125]  new_sync_read+0x10d/0x1a0 >>>>>>> [ 3748.617009]  vfs_read+0x103/0x1a0 >>>>>>> [ 3748.620423]  ksys_read+0x67/0xf0 >>>>>>> [ 3748.623743]  __x64_sys_read+0x19/0x20 >>>>>>> [ 3748.627511]  do_syscall_64+0x59/0xc0 >>>>>>> [ 3748.631203]  ? syscall_exit_to_user_mode+0x27/0x50 >>>>>>> [ 3748.636144]  ? do_syscall_64+0x69/0xc0 >>>>>>> [ 3748.639992]  ? exit_to_user_mode_prepare+0x96/0xb0 >>>>>>> [ 3748.644931]  ? irqentry_exit_to_user_mode+0x9/0x20 >>>>>>> [ 3748.649872]  ? irqentry_exit+0x1d/0x30 >>>>>>> [ 3748.653737]  ? exc_page_fault+0x89/0x170 >>>>>>> [ 3748.657795]  entry_SYSCALL_64_after_hwframe+0x61/0xcb >>>>>>> [ 3748.663030] RIP: 0033:0x7f9a852989cc >>>>>>> [ 3748.666713] RSP: 002b:00007f9a8497dc90 EFLAGS: 00000246 ORIG_RAX: >>>>>>> 0000000000000000 >>>>>>> [ 3748.674487] RAX: ffffffffffffffda RBX: 00007f9a8497f5c0 RCX: >>>>>>> 00007f9a852989cc >>>>>>> [ 3748.681817] RDX: 0000000000027100 RSI: 00007f99faa18010 RDI: >>>>>>> 0000000000000061 >>>>>>> [ 3748.689150] RBP: 00007f9a8497dd60 R08: 0000000000000000 R09: >>>>>>> 00007f99faa18010 >>>>>>> [ 3748.696493] R10: 0000000000000000 R11: 0000000000000246 R12: >>>>>>> 00007f99faa18010 >>>>>>> [ 3748.703841] R13: 00005605e11c406f R14: 0000000000000001 R15: >>>>>>> 0000000000027100 >>>>>> >>>>>> ORIG_RAX is 0, which matches sys_read. >>>>>> RDI is file descriptor 0x61 >>>>>> RSI is plausibly a userspace pointer, 0x7f99faa18010 >>>>>> RDX is the length, 0x27100 or 160kB. >>>>>> >>>>>> That all seems reasonable. >>>>>> >>>>>> What I really want to know is who is _holding_ the lock.  We stash >>>>>> a pointer to the task_struct in 'owner', so we could clearly find >>>>>> this >>>>>> out in the 'blocked for too long' report, and print their stack >>>>>> trace. >>>>>> >>>>> As described in the comment for __rwsem_set_reader_owned,it is hard >>>>> to track >>>>> read owners.So we could not clearly find out who blocked the >>>>> process,it was >>>>> caused by multiple tasks. >>>> >>>> Readers don't block readers.  You have a reader here, so it's being >>>> blocked by a writer.  And that writer's task_struct is stashed in >>>> rwsem->owner.  It would be nice if we dumped that information >>>> automatically ... but we don't do that today.  Perhaps you could >>>> grab that information from a crash dump if you have one. >>>> >>>>>> You must have done something like this already in order to deduce >>>>>> that >>>>>> it was the direct reclaim path that was the problem? >>>>>> >>>>> The method we used is to track the direct reclaim using the >>>>> trace_mm_vmscan_direct_reclaim_{begin,end} interface.When the problem >>>>> occurred,we could get a very large "nr_reclaimed" which is not a >>>>> desirable >>>>> value for process except kswapd. >>>> >>>> I disagree.  If a process needs to allocate memory then it should be >>>> paying the cost of reclaiming that memory itself.  kswapd is a last >>>> resort to reclaim memory when we have a workload (eg a network router) >>>> that does its memory allocation primarily in interrupt context. >>>> >>> What's your opinion about this scan adjust issue? Is there a better >>> way to >>> fix this issue? >> >> Yes, but we need you to gather more information about what's causing >> the issue before we can suggest what that is. >> > I think the following scenery triggers the scan adjust issue: > In function shrink_lruvec, we call get_scan_count and get the following > values: > targets[LRU_INACTIVE_ANON]=50000 > targets[LRU_ACTIVE_ANON]=50000 > targets[LRU_INACTIVE_FILE]=128 > targets[LRU_ACTIVE_FILE]=129 > > After the first scan, we get more than nr_to_reclaim pages, but the > percentage of scanning nr[LRU_INACTIVE_FILE+LRU_ACTIVE_FILE] is 256/257, > Then when we scan adjust, we must scan(possibly reclaim) > 256*(50000+50000)/257-256=99354 pages, which is too large and would > waste too many time. > If it is not kswapd, it is unacceptable to reclaim so many pages. > So we should limit the number of pages of scan adjust. > > Thanks > Hongchen Zhang