From mboxrd@z Thu Jan 1 00:00:00 1970 From: Linus Torvalds Subject: Re: [PATCH 2.6.27-rc8 5/6] e1000e: debug contention on NVM SWFLAG Date: Fri, 3 Oct 2008 08:16:16 -0700 (PDT) Message-ID: References: <20081002233314.12556.49143.stgit@jbrandeb-bw.jf.intel.com> <20081002233340.12556.33137.stgit@jbrandeb-bw.jf.intel.com> Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Cc: Jesse Brandeburg , Thomas Gleixner , linux-kernel@vger.kernel.org, netdev@vger.kernel.org, arjan@linux.intel.com, airlied@gmail.com, davem@davemloft.net, Jeff Garzik To: Jiri Kosina Return-path: Received: from smtp1.linux-foundation.org ([140.211.169.13]:43846 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752294AbYJCPRO (ORCPT ); Fri, 3 Oct 2008 11:17:14 -0400 In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Fri, 3 Oct 2008, Jiri Kosina wrote: > > The debugging message is racy anyway with respect to accessing nvm_owner, > right? It should be done after the mutex has been succesfully acquired. It's done that way on purpose - to see who _could_ be racing. After the mutex, it could never trigger, because the mutex is the thing that guarantees non-racy-ness. IOW, it's a debugging message just to see that the old bug (the "before the fix") really did happen. We can/will remove it, but I think people are still looking at the e1000e driver and probably want to see the paths that can cause problems. Of course, it's entirely possible that we should remove it in mainline already, and just let the people inside intel/suse/xyzzy who are trying to reproduce it have it. Jesse? Thomas? Linus