From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mika Westerberg Subject: Re: [PULL REQUEST] i2c for 3.13 Date: Tue, 19 Nov 2013 17:00:01 +0200 Message-ID: <20131119150001.GM2281@intel.com> References: <20131118193802.GA14925@katana> <3690707.5664qBQcL8@vostro.rjw.lan> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <3690707.5664qBQcL8-sKB8Sp2ER+y1GS7QM15AGw@public.gmane.org> Sender: linux-i2c-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: "Rafael J. Wysocki" Cc: Linus Torvalds , Wolfram Sang , "Rafael J. Wysocki" , Jean Delvare , linux-i2c-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Linux Kernel Mailing List List-Id: linux-i2c@vger.kernel.org On Tue, Nov 19, 2013 at 01:11:25AM +0100, Rafael J. Wysocki wrote: > On Monday, November 18, 2013 03:46:34 PM Linus Torvalds wrote: > > On Mon, Nov 18, 2013 at 11:38 AM, Wolfram Sang wrote: > > > > > > here is the pull request from the i2c subsystem for 3.13: > > > > So while resolving some fairly trivial conflicts here, I noticed that > > commit a76e9bd89ae7 ("i2c: attach/detach I2C client device to the ACPI > > power domain") that I got earlier through the ACPI/PM tree calls > > acpi_dev_pm_detach() even when the device "->remove()" function fails. > > But it only sets clientdata to NULL if it succeeds. > > > > That looks a bit odd. > > > > I didn't try to fix it, though. I just thought I'd point out the oddity. > > Well, given that the driver core doesn't even check the return value of > dev->bus->remove(dev), I think doing the unconditional acpi_dev_pm_detach() > is actually correct and clientdata should be cleared unconditionally too. Actually, can we remove the whole clientdata setting there? Commit 0998d0631001 (device-core: Ensure drvdata = NULL when no driver is bound) modified the driver core to always clear out that field. Same seems to apply if driver probe fails.