From mboxrd@z Thu Jan 1 00:00:00 1970 From: Zhang Rui Subject: Re: [PATCHv5 01/10] Thermal: Do kfree in _unregister functions Date: Mon, 03 Mar 2014 21:44:16 +0800 Message-ID: <1393854256.2193.2.camel@rzhang1-mobl4> References: <1389916587-2541-1-git-send-email-durgadoss.r@intel.com> <1389916587-2541-2-git-send-email-durgadoss.r@intel.com> <20140121194624.GA29615@core.coreip.homeip.net> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Return-path: Received: from mga01.intel.com ([192.55.52.88]:61045 "EHLO mga01.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753881AbaCCNoo (ORCPT ); Mon, 3 Mar 2014 08:44:44 -0500 In-Reply-To: <20140121194624.GA29615@core.coreip.homeip.net> Sender: linux-pm-owner@vger.kernel.org List-Id: linux-pm@vger.kernel.org To: Dmitry Torokhov Cc: Durgadoss R , eduardo.valentin@ti.com, linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org, hongbo.zhang@freescale.com, wni@nvidia.com On Tue, 2014-01-21 at 11:46 -0800, Dmitry Torokhov wrote: > Hi Durgadoss, > > On Fri, Jan 17, 2014 at 05:26:18AM +0530, Durgadoss R wrote: > > Currently the thermal_release function does > > kfree for all devices when they call device_unregister. > > This makes code scattering i.e whenever we add new > > devices (to thermal class) we need to add its > > corresponding kfree in thermal_release function. > > And the if-else is also growing lengthy. > > > > Instead, do all kfree() of devices in their own > > _unregister functions. This makes the code look > > clean and easy to maintain. > > > > Signed-off-by: Durgadoss R > > --- > > drivers/thermal/thermal_core.c | 17 +++++------------ > > 1 file changed, 5 insertions(+), 12 deletions(-) > > > > diff --git a/drivers/thermal/thermal_core.c b/drivers/thermal/thermal_core.c > > index 338a88b..165afc6 100644 > > --- a/drivers/thermal/thermal_core.c > > +++ b/drivers/thermal/thermal_core.c > > @@ -1042,18 +1042,7 @@ EXPORT_SYMBOL_GPL(thermal_zone_unbind_cooling_device); > > > > static void thermal_release(struct device *dev) > > { > > - struct thermal_zone_device *tz; > > - struct thermal_cooling_device *cdev; > > - > > - if (!strncmp(dev_name(dev), "thermal_zone", > > - sizeof("thermal_zone") - 1)) { > > - tz = to_thermal_zone(dev); > > - kfree(tz); > > - } else if(!strncmp(dev_name(dev), "cooling_device", > > - sizeof("cooling_device") - 1)){ > > - cdev = to_cooling_device(dev); > > - kfree(cdev); > > - } > > + /* No-op since kfree(dev) is done in _unregister functions */ > > } > > > > static struct class thermal_class = { > > @@ -1146,6 +1135,7 @@ __thermal_cooling_device_register(struct device_node *np, > > unregister: > > release_idr(&thermal_cdev_idr, &thermal_idr_lock, cdev->id); > > device_unregister(&cdev->device); > > + kfree(cdev); > > This is broken, you can not assume that you are holding the last > reference. Please observe help text of DEBUG_KOBJECT_RELEASE config > option, activate said option and try unregistering thermal device. > agreed. If you think it is too complicated to support the different thermal components in thermal_release(), we can introduce per device .release() callback as well. thanks, rui > Thanks. >