From mboxrd@z Thu Jan 1 00:00:00 1970 From: Vasiliy Kulikov Date: Sun, 19 Sep 2010 14:39:50 +0000 Subject: Re: [PATCH 10/14] scsi: osd: fix device_register() error handling Message-Id: <20100919143948.GA4866@albatros> List-Id: References: <1284900907-24621-1-git-send-email-segooon@gmail.com> <20100919142653.GF6236@bicker> In-Reply-To: <20100919142653.GF6236@bicker> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Dan Carpenter , kernel-janitors@vger.kernel.org, Boaz Harrosh , Benny Halevy , "James E.J. Bottomley" , Tejun Heo , Arnd Bergmann , osd-dev@open-osd.org, linux-scsi@vger.kernel.org, cornelia.huck@de.ibm.com, linux-kernel@vger.kernel.org On Sun, Sep 19, 2010 at 16:26 +0200, Dan Carpenter wrote: > On Sun, Sep 19, 2010 at 04:55:07PM +0400, Vasiliy Kulikov wrote: > > If device_register() fails then call put_device(). > > See comment to device_register. > > > > Signed-off-by: Vasiliy Kulikov > > --- ... > > Hm... So if device_register() fails then we should always call > device_put()? It seems like a lot of existing code does that but I > hadn't realized until now that that is how it works. Yes, almost ALL code using device_register() is buggy :-( > Why can't the device_put() just be added inside the device_register() so > the unwinding works automatically? Because some code already calls device_put(). Also it is documented like not putting the device. However, I'm in doubt why it is written this way. > Also if someone add some more stuff to the end of this function, will > the device_unregister() followed by a device_put() cause problems if we > unwind like this? Yes, device_register() gets one reference, you should put in in both cases - when device_register() failed and when it succeeded, but only one time. device_unregister() puts it, so it is "double putting". > +err_free_something: > + kfree(foo); > + device_unregister(&oud->class_dev); > > +err_put_device: > > + put_device(&oud->class_dev); > > err_put_cdev: > > cdev_del(&oud->cdev); > > err_put_disk: > > If that's the case then the put_device() should be called infront of the > goto. As it is the last call that may fail, it is redundant. Or you mean for future, if someone adds more code after device_register()? Thanks, -- Vasiliy