Return-Path: From: Arnd Bergmann To: Benjamin Herrenschmidt Subject: Re: [Cbe-oss-dev] [PATCH 12/16] cell: add temperature to SPU and CPU sysfs entries Date: Thu, 26 Oct 2006 09:35:00 +0200 References: <20061024163113.694643000@arndb.de> <20061025080048.GB7090@osiris.boeblingen.de.ibm.com> <1161818364.22582.145.camel@localhost.localdomain> In-Reply-To: <1161818364.22582.145.camel@localhost.localdomain> MIME-Version: 1.0 Message-Id: <200610260935.01801.arnd@arndb.de> Content-Type: text/plain; charset="iso-8859-15" Cc: Andrew Morton , linuxppc-dev@ozlabs.org, Heiko Carstens , linux-kernel@vger.kernel.org, cbe-oss-dev@ozlabs.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Thursday 26 October 2006 01:19, Benjamin Herrenschmidt wrote: > > > > Will crash if cpu_add_sysdev_attr_group failed... > > > Which is a total PITA. If this is the case, then we should modify the > add calls to at least initialize enough fields before they can fail for > the remove calls not to crash. You don't want to keep track precisely of > what file was added and what not and test all of that in your exit code > path, it's just insane. Heiko suggested that earlied in http://lkml.org/lkml/2006/10/9/22, but Andrew didn't like it. Currently, the worst is that sysfs_remove_file can be used on a nonexisting file, but sysfs_remove_group cannot, which is inconsistent. Either sysfs_remove_file should WARN_ON or sysfs_remove_group should silently return, and I'd prefer the latter, as it makes users simpler. Arnd <><