Re: [bug report] gpio: protect the pointer to gpio_chip in gpio_device with SRCU

[Bartosz Golaszewski <brgl@xxxxxxxx>]

On Wed, Feb 14, 2024 at 8:36 AM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>
> Hello Bartosz Golaszewski,
>
> The patch d83cee3d2bb1: "gpio: protect the pointer to gpio_chip in
> gpio_device with SRCU" from Jan 23, 2024 (linux-next), leads to the
> following Smatch static checker warning:
>
>         drivers/gpio/gpiolib-sysfs.c:808 gpiochip_sysfs_unregister()
>         error: we previously assumed 'chip' could be null (see line 804)
>
> drivers/gpio/gpiolib-sysfs.c
>     786 void gpiochip_sysfs_unregister(struct gpio_device *gdev)
>     787 {
>     788         struct gpio_desc *desc;
>     789         struct gpio_chip *chip;
>     790
>     791         scoped_guard(mutex, &sysfs_lock) {
>     792                 if (!gdev->mockdev)
>     793                         return;
>     794
>     795                 device_unregister(gdev->mockdev);
>     796
>     797                 /* prevent further gpiod exports */
>     798                 gdev->mockdev = NULL;
>     799         }
>     800
>     801         guard(srcu)(&gdev->srcu);
>     802
>     803         chip = rcu_dereference(gdev->chip);
>     804         if (chip)
>                     ^^^^
> Is this test reversed?
>
>     805                 return;
>     806
>     807         /* unregister gpiod class devices owned by sysfs */
> --> 808         for_each_gpio_desc_with_flag(chip, desc, FLAG_SYSFS) {
>                                              ^^^^
> NULL dereference
>
>     809                 gpiod_unexport(desc);
>     810                 gpiod_free(desc);
>     811         }
>     812 }
>
> regards,
> dan carpenter
>

Thanks for the report Dan, I sent out a fix.

Bart