From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexey Khoroshilov Subject: Re: [PATCH] USB: input: gtco.c: fix usb_dev leak Date: Mon, 27 Jan 2014 14:29:23 +0400 Message-ID: <52E63503.3060804@ispras.ru> References: <1390087466-9898-1-git-send-email-khoroshilov@ispras.ru> <20140121195957.GB29615@core.coreip.homeip.net> <52E5FD48.7040208@ispras.ru> <20140127065416.GA11945@core.coreip.homeip.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20140127065416.GA11945@core.coreip.homeip.net> Sender: linux-kernel-owner@vger.kernel.org To: Dmitry Torokhov , Greg Kroah-Hartman Cc: linux-input@vger.kernel.org, linux-kernel@vger.kernel.org, ldv-project@linuxtesting.org List-Id: linux-input@vger.kernel.org On 27.01.2014 10:54, Dmitry Torokhov wrote: > Hi Alexey, > > On Mon, Jan 27, 2014 at 10:31:36AM +0400, Alexey Khoroshilov wrote: >> On 21.01.2014 23:59, Dmitry Torokhov wrote: >>> On Sun, Jan 19, 2014 at 03:24:26AM +0400, Alexey Khoroshilov wrote: >>>> There is usb_get_dev() in gtco_probe(), but there is no usb_put_dev() >>>> anywhere in the driver. >>>> >>>> The patch adds usb_get_dev() to failure handling code of gtco_probe() >>>> and to gtco_disconnect((). >>> Hmm, I think gtco should simply not use usb_get_dev() in the first >>> place. >>> >>> Thanks. >> Dear Dmitry, >> >> Could you please clarify why usb_get_dev() not needed here? >> We store reference to usb_dev in gtco structure, so we should refcount it. >> What is wrong in this reasoning? > The lifetime of gtco structure is already directly tied to lifetime of > usb_dev: when destroying usb_dev driver core will call remove() function > of currently bound driver (in our case gtco) which will destroy gtco > memory. > > Taking additional reference is not needed here. > > Hope this helps. Thank you, that helps a lot. By the way, usb_skeleton suggests to use usb_get_dev()/usb_put_dev() nevertheless. Greg, may be it makes sense to fix usb_skeleton as well? -- Alexey