From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH 2/3] sysfs: remove error messages for -EEXIST case Date: Wed, 14 May 2008 18:26:37 -0700 (PDT) Message-ID: <20080514.182637.211478588.davem@davemloft.net> References: <20080514170316.672f809d@extreme> <20080514181257.74fbb5aa@extreme> <20080514181603.411e834f@extreme> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: greg@kroah.com, fubar@us.ibm.com, netdev@vger.kernel.org, bonding-devel@lists.sourceforge.net, linux-kernel@vger.kernel.org To: shemminger@vyatta.com Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:33881 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1751284AbYEOB0m (ORCPT ); Wed, 14 May 2008 21:26:42 -0400 In-Reply-To: <20080514181603.411e834f@extreme> Sender: netdev-owner@vger.kernel.org List-ID: From: Stephen Hemminger Date: Wed, 14 May 2008 18:16:03 -0700 > It is possible that the entry in sysfs already exists, one case of this is > when a network device is renamed to bonding_masters. Anyway, in this case > the proper error path is for device_rename to return an error code, not to > generate bogus backtrace and errors. > > Also, to avoid possible races, the create link should be done before the > remove link. This makes a device rename atomic operation like other renames. > > Signed-off-by: Stephen Hemminger I definitely agree with this change. We have several cases where device names are user configurable, yet the devices live in a directory which also has subdirectories created by other subsystems. It's pointless to require the top-level guy to look for any purge out any subdirectory cases, that's none of it's business. I realize the backtrace is useful for finding bugs, but in this case it's definitely not appropriate.