From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751499AbeB0A0L (ORCPT ); Mon, 26 Feb 2018 19:26:11 -0500 Received: from out30-132.freemail.mail.aliyun.com ([115.124.30.132]:54534 "EHLO out30-132.freemail.mail.aliyun.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751094AbeB0A0J (ORCPT ); Mon, 26 Feb 2018 19:26:09 -0500 X-Alimail-AntiSpam: AC=PASS;BC=-1|-1;BR=01201311R171e4;CH=green;FP=0|-1|-1|-1|0|-1|-1|-1;HT=e01e01355;MF=yang.shi@linux.alibaba.com;NM=1;PH=DS;RN=6;SR=0;TI=SMTPD_---0SyYb1iR_1519691153; From: Yang Shi To: akpm@linux-foundation.org, mingo@kernel.org, adobriyan@gmail.com Cc: yang.shi@linux.alibaba.com, linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: [RFC PATCH 0/4 v2] Define killable version for access_remote_vm() and use it in fs/proc Date: Tue, 27 Feb 2018 08:25:47 +0800 Message-Id: <1519691151-101999-1-git-send-email-yang.shi@linux.alibaba.com> X-Mailer: git-send-email 1.8.3.1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Background: When running vm-scalability with large memory (> 300GB), the below hung task issue happens occasionally. INFO: task ps:14018 blocked for more than 120 seconds. Tainted: G E 4.9.79-009.ali3000.alios7.x86_64 #1 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. ps D 0 14018 1 0x00000004 ffff885582f84000 ffff885e8682f000 ffff880972943000 ffff885ebf499bc0 ffff8828ee120000 ffffc900349bfca8 ffffffff817154d0 0000000000000040 00ffffff812f872a ffff885ebf499bc0 024000d000948300 ffff880972943000 Call Trace: [] ? __schedule+0x250/0x730 [] schedule+0x36/0x80 [] rwsem_down_read_failed+0xf0/0x150 [] call_rwsem_down_read_failed+0x18/0x30 [] down_read+0x20/0x40 [] proc_pid_cmdline_read+0xd9/0x4e0 [] ? do_filp_open+0xa5/0x100 [] __vfs_read+0x37/0x150 [] ? security_file_permission+0x9b/0xc0 [] vfs_read+0x96/0x130 [] SyS_read+0x55/0xc0 [] entry_SYSCALL_64_fastpath+0x1a/0xc5 When manipulating a large mapping, the process may hold the mmap_sem for long time, so reading /proc//cmdline may be blocked in uninterruptible state for long time. We already have killable version APIs for semaphore, here use down_read_killable() to improve the responsiveness. When reviewing the v1 patch (https://patchwork.kernel.org/patch/10230809/), Alexey pointed out access_remote_vm() need to be killable too. And, /proc/*/environ reading may suffer from the same issue, so it should be converted to killable version for both down_read and access_remote_vm too. With reading the code, both access_remote_vm() and access_process_vm() calls __access_remote_vm() which acquires mmap_sem by down_read(). access_remote_vm() is only used by fs/proc/base.c, but access_process_vm() is used by other subsystems too, i.e. ptrace, audit, etc. So, it sounds not that safe to convert both access_remote_vm() and access_process_vm() to killable. Instead of doing so, extract command part of __access_remote_vm() (gup part) to a new static function, called raw_access_remote_vm(), then define __access_remote_vm() and __access_remote_vm_killable(), which acquire mmap_sem by down_read() and _killable() respectively. Then define access_remote_vm() and access_remote_vm_killable() to call them respectively. Keep access_process_vm() calls __access_remote_vm(). So far fs/proc/base.c is the only user of access_remote_vm_killable(), but there might be other users in the future. There are 4 patches in this revision: #1 define access_remote_vm_killable() APIs #2 convert /proc/*/cmdline reading to down_read_killable() and access_remote_vm_killable() #3 convert /proc/*/environ reading to down_read_killable() and access_remote_vm_killable() #4 replace access_process_vm() to access_remote_vm() in get_cmdline to save one mm reference count inc (please see the commit log for the details). This change makes get_cmdline() is the only user of access_remote_vm() Yang Shi (4): mm: add access_remote_vm_killable APIs fs: proc: use down_read_killable in proc_pid_cmdline_read() fs: proc: use down_read_killable() in environ_read() mm: use access_remote_vm() in get_cmdline() fs/proc/base.c | 21 +++++++++++++++------ include/linux/mm.h | 5 +++++ mm/memory.c | 44 +++++++++++++++++++++++++++++++++++++------- mm/nommu.c | 36 ++++++++++++++++++++++++++++++++---- mm/util.c | 4 ++-- 5 files changed, 91 insertions(+), 19 deletions(-)