Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753176AbcCAKgX (ORCPT ); Tue, 1 Mar 2016 05:36:23 -0500 Received: from cn.fujitsu.com ([59.151.112.132]:22323 "EHLO heian.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org with ESMTP id S1752358AbcCAKgW (ORCPT ); Tue, 1 Mar 2016 05:36:22 -0500 X-IronPort-AV: E=Sophos;i="5.22,518,1449504000"; d="scan'208";a="4067614" From: Zhao Lei To: "'Tejun Heo'" CC: , , , References: <1450696483-2864-1-git-send-email-yangds.fnst@cn.fujitsu.com> <1450696483-2864-2-git-send-email-yangds.fnst@cn.fujitsu.com> <20151221213328.GP4026@mtj.duckdns.org> In-Reply-To: <20151221213328.GP4026@mtj.duckdns.org> Subject: RE: [RESEND PATCH 2/2] cpuacct: split usage into user_usage and sys_usage. Date: Tue, 1 Mar 2016 18:35:56 +0800 Message-ID: <010801d173a6$237a2d70$6a6e8850$@cn.fujitsu.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit X-Mailer: Microsoft Outlook 15.0 Thread-Index: AQG2OOuHd/TAzameKRX2rB6rJtQPagLJmph0AnZuU6GfUNQ6QA== Content-Language: zh-cn X-yoursite-MailScanner-ID: C1A5242B4AFB.AF95A X-yoursite-MailScanner: Found to be clean X-yoursite-MailScanner-From: zhaolei@cn.fujitsu.com Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2319 Lines: 78 Hi, Tejun Heo > From: linux-kernel-owner@vger.kernel.org > [mailto:linux-kernel-owner@vger.kernel.org] On Behalf Of Tejun Heo > Sent: Tuesday, December 22, 2015 5:33 AM > To: Dongsheng Yang > Cc: cgroups@vger.kernel.org; linux-kernel@vger.kernel.org; > mingo@redhat.com; peterz@infradead.org > Subject: Re: [RESEND PATCH 2/2] cpuacct: split usage into user_usage and > sys_usage. > > On Mon, Dec 21, 2015 at 07:14:43PM +0800, Dongsheng Yang wrote: > > Sometimes, cpuacct.usage is not detialed enough to user > > to see how much usage a group used. We want to know how > > much time it used in user mode and how much in kernel mode. > Yang gave this patch to me :) I'm plan to continue this function. > cpuusage is being phased out. If you need these stats, please > implement it on cpu side. > Currently, old stats are put in cpuacct, if we put new stats into cpu side, we have following way: 1: Only put new stats into cpu side, and don't change old interface, we'll see following content in cgroup dir: /cgroup/cpuacct: cpuacct.usage cpuacct.usage_percpu /cgroup/cpu: cpu.usage_sys cpu.usage_percpu_sys cpu.usage_user cpu.usage_percpu_user It is obviously not unify. 2: Move old stats into cpu side, and add new stats: we'll see following content in cgroup dir: /cgroup/cpuacct: [NONE] /cgroup/cpu: cpu.usage cpu.usage_percpu cpu.usage_sys cpu.usage_percpu_sys cpu.usage_user cpu.usage_percpu_user It have compatibility issues with old system. 3: Clone old stats into cpu side(share one copy of code), and add new stats Into cpu side, we'll see following content in cgroup dir: /cgroup/cpuacct: cpuacct.usage cpuacct.usage_percpu /cgroup/cpu: cpu.usage cpu.usage_percpu cpu.usage_sys cpu.usage_percpu_sys cpu.usage_user cpu.usage_percpu_user It can support both old and new systems, and we can remove old interface in future, but we if mount both cpu and cpuacct into same dir, we'll see following duplicated files: /cgroup/cpu,cpuacct cpu.usage cpuacct.usage cpu.usage_percpu cpuacct.usage_percpu ... Maybe above way3 looks having less defect, cound you talk about your opinion on it. Thanks Zhaolei