From: Ingo Molnar Subject: Re: [PATCH] vfs: Add a trace point in the mark_inode_dirty function Date: Fri, 20 Nov 2009 11:51:54 +0100 Message-ID: <20091120105154.GB12634@elte.hu> References: <20091025225342.007138f5@infradead.org> <20091111020108.GA11423@localhost> <20091110223456.01ef355f@infradead.org> <4AFA6AEF.5060306@garzik.org> <20091111081905.270a4e55@infradead.org> <4AFB4AC7.1090405@intel.com> <20091112072217.GA31719@elte.hu> <20091120104335.GB29143@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "Kok, Auke" , "Frank Ch. Eigler" , Arjan van de Ven , Jeff Garzik , "Wu, Fengguang" , "linux-kernel@vger.kernel.org" , "linux-fsdevel@vger.kernel.org" , Al Viro , Frederic Weisbecker To: Christoph Hellwig Return-path: Received: from mx2.mail.elte.hu ([157.181.151.9]:36018 "EHLO mx2.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751887AbZKTKwD (ORCPT ); Fri, 20 Nov 2009 05:52:03 -0500 Content-Disposition: inline In-Reply-To: <20091120104335.GB29143@infradead.org> Sender: linux-fsdevel-owner@vger.kernel.org List-ID: * Christoph Hellwig wrote: > Guys, I think both the inode number and name do have a use case. For > file system developers observing the filesystem the inode number is > very useful, and if you look at the ext4 tracing already in tree or > the xfs tracing going in in the next window they use the inode number > all over. > > Which btw brings up another good argument - to make the tracing really > useful we need to have conventions. While the inode number seems to > be a realtively easy one printing the device is more difficult. XFS > just prints the raw major/minor to stay simple, ext4 has a complicated > ad-hoc cache of device names, and this one just prints the superblock > id string. Agreed. > Of course for a user the name is a lot more meaninful, but also > relatively expensive to generate. Then again I'm not even sure how > the last pathname component only here is all that useful - it can't be > used to easily find the file. That's not the main point though - the point is for app developers (and users) being able to see 'oh, _that_ file it is, we need to fix that'. In the context of a specific app, the last component filename carries 95% of the useful information. Look at how PowerTOP does it, for a real-life usecase. Ingo