From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1161009AbWJOOVv (ORCPT ); Sun, 15 Oct 2006 10:21:51 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1161008AbWJOOVv (ORCPT ); Sun, 15 Oct 2006 10:21:51 -0400 Received: from outpipe-village-512-1.bc.nu ([81.2.110.250]:3730 "EHLO lxorguk.ukuu.org.uk") by vger.kernel.org with ESMTP id S1161007AbWJOOVu (ORCPT ); Sun, 15 Oct 2006 10:21:50 -0400 Subject: Re: + drivers-ide-fix-error-return-bugs-interface.patch added to -mm tree From: Alan Cox To: akpm@osdl.org Cc: linux-kernel@vger.kernel.org, jeff@garzik.org In-Reply-To: <200610102047.k9AKl53O022518@shell0.pdx.osdl.net> References: <200610102047.k9AKl53O022518@shell0.pdx.osdl.net> Content-Type: text/plain Content-Transfer-Encoding: 7bit Date: Sun, 15 Oct 2006 15:46:46 +0100 Message-Id: <1160923606.5732.61.camel@localhost.localdomain> Mime-Version: 1.0 X-Mailer: Evolution 2.6.2 (2.6.2-1.fc5.5) Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Ar Maw, 2006-10-10 am 13:47 -0700, ysgrifennodd akpm@osdl.org: > This fixes the following error handling problems: > > * The init_chipset API function is defined to return 'unsigned int', but > - the caller code tests the return value for '< 0' > - drivers sometimes return a negative value Dropped for good. The core code doesn't support failing in init_chipset. Instead I've adjusted via82cxxx to do the check earlier in init_one(). Also avoids an API change and all the noise. > * cs5530: handle pci_set_mwi() failure with a printk()... shouldn't kill driver NAK. Fix is to remove bogus must_check from pci_set_mwi (and some of the other functions) > * sc1200: handle pci_enable_device() failure during resume NAK: if the pci_enable_device fails here the best we can do is attempt to get it back and hope the pci_enable_device failure is bogus. This is a can't happen case anyway so its a waste of memory. If you insist on checking pci_enable_device returns everywhere then please move the printk into the pci_enable_device function so that we don't bloat the kernel with a load of pointless identically messages for an event that never happen, one per driver. Alan