From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jean Delvare Subject: Re: i2c: cdev lock_kernel() pushdown Date: Tue, 15 Jul 2008 18:35:52 +0200 Message-ID: <20080715183552.7e1b797a@hyperion.delvare> References: <20080715174155.2d52e6f6@hyperion.delvare> <20080715101406.6b4517e3@bike.lwn.net> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20080715101406.6b4517e3-vw3g6Xz/EtPk1uMJSBkQmQ@public.gmane.org> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: i2c-bounces-GZX6beZjE8VD60Wz+7aTrA@public.gmane.org Errors-To: i2c-bounces-GZX6beZjE8VD60Wz+7aTrA@public.gmane.org To: Jonathan Corbet Cc: Linux I2C List-Id: linux-i2c@vger.kernel.org On Tue, 15 Jul 2008 10:14:06 -0600, Jonathan Corbet wrote: > Hi, Jean, > > > I am looking at this patch of yours: > > http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=3db633ee352bfe20d4a2b0c3c8a46ce31a6c7149 > > > > I believe that no locking is needed in i2cdev_open(). Do you have any > > reason to think it does? If not, can I simply revert this patch? > > Before now, i2cdev_open() has always had the protection of the BKL. > When I pushed that locking down into the individual open() functions, I > really had to take a pretty conservative approach and assume that the > BKL was needed unless that was really obviously not the case. In > i2cdev_open(), for example, there's: > > i2c_dev = i2c_dev_get_by_minor(minor); > > and I really don't know what keeps *i2c_dev from going away during the > rest of the call. I'm *not* saying there is a problem here; I just > don't know. So the only thing I could really do is to push the BKL > down and let the maintainers sort it out. > > ...all of which is my long-winded way of saying that, if you're > convinced that i2cdev_open() is safe in the absence of the BKL, feel > free to take it out. Good point. i2c_dev is guaranteed to stay thanks to the call to i2c_get_adapter(), however it happens _after_ the call to i2c_dev_get_by_minor(), so without additional locking, this is racy. That being said, I'm not sure how lock_kernel() is supposed to help. Is the BKL held when i2cdev_detach_adapter() is called? If not, then I suspect that the current code is already racy. I'll look into this, thanks for the hint. -- Jean Delvare _______________________________________________ i2c mailing list i2c-GZX6beZjE8VD60Wz+7aTrA@public.gmane.org http://lists.lm-sensors.org/mailman/listinfo/i2c