From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753802Ab1EVCVd (ORCPT ); Sat, 21 May 2011 22:21:33 -0400 Received: from zeniv.linux.org.uk ([195.92.253.2]:57651 "EHLO ZenIV.linux.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753430Ab1EVCVb (ORCPT ); Sat, 21 May 2011 22:21:31 -0400 Date: Sun, 22 May 2011 03:21:26 +0100 From: Al Viro To: Andi Kleen Cc: linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org, Andi Kleen , npiggin@kernel.dk, tim.c.chen@intel.com Subject: Re: [PATCH] VFS: Add VFS event counter infrastructure Message-ID: <20110522022126.GJ19987@ZenIV.linux.org.uk> References: <1305332052-15633-1-git-send-email-andi@firstfloor.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1305332052-15633-1-git-send-email-andi@firstfloor.org> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, May 13, 2011 at 05:14:12PM -0700, Andi Kleen wrote: > From: Andi Kleen > > The recently discovered problem with RCU walks not working for absolute path > motivated me to add some counters for these events to the VFS. Networking > and VM has had such counters for a long time and they were always useful > to diagnose performance problems. An advantage of counters over tracepoints > is that they are always collected and are low enough overhead that > they can be always enabled (unlike tracing) > > This patch implements a simple per CPU counter framework for the VFS. > The counters are per CPU and are very little overhead. The counters > are output in debugfs (/sys/kernel/fs/vfsstat) WTF? Since when had debugfs become mandatory? > + fs = debugfs_create_dir("fs", NULL); > + if (fs) > + debugfs_create_file("vfsstat", 0444, fs, NULL, &vfsstat_fops); > + return 0; Note that if that FPOS is configured away debugfs_create_dir() does _NOT_ return NULL - you get ERR_PTR(-ENODEV).