Re: [PATCH v2] gpio: pca953x: fix a lockdep warning
From: Bartosz Golaszewski
Date: Wed Sep 07 2016 - 08:09:04 EST
2016-08-29 10:33 GMT+02:00 Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>:
> If an I2C GPIO multiplexer is driven by a GPIO provided by an expander
> when there's a second expander using the same device driver on one of
> the I2C bus segments, lockdep prints a deadlock warning when trying to
> set the direction or the value of the GPIOs provided by the second
> expander.
>
> The below diagram presents the setup:
>
> - - - - -
> ------- --------- Bus segment 1 | |
> | | | |--------------- Devices
> | | SCL/SDA | | | |
> | Linux |-----------| I2C MUX | - - - - -
> | | | | | Bus segment 2
> | | | | |-------------------
> ------- | --------- |
> | | - - - - -
> ------------ | MUX GPIO | |
> | | | Devices
> | GPIO | | | |
> | Expander 1 |---- - - - - -
> | | |
> ------------ | SCL/SDA
> |
> ------------
> | |
> | GPIO |
> | Expander 2 |
> | |
> ------------
>
> The reason for lockdep warning is that we take the chip->i2c_lock in
> pca953x_gpio_set_value() or pca953x_gpio_direction_output() and then
> come right back to pca953x_gpio_set_value() when the GPIO mux kicks
> in. The locks actually protect different expanders, but lockdep
> doesn't see this and says:
>
> Possible unsafe locking scenario:
>
> CPU0
> ----
> lock(&chip->i2c_lock);
> lock(&chip->i2c_lock);
>
> *** DEADLOCK ***
>
> May be due to missing lock nesting notation
>
> To shut lockdep up, use mutex_lock_nested() and use the GPIO base
> number as the subclass argument (it has the same type).
>
> NOTE: this only fixes a specific issue we're experiencing with our
> setup. The problem would probably occur as well with other I2C
> expanders under similar circumstances. A proper fix would probably be
> to implement an I2C-GPIO expander framework that would unduplicate
> common code for all drivers.
>
> Signed-off-by: Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>
> ---
> v1 -> v2:
> - tweaked the commit message
> - expanded the comment
>
> drivers/gpio/gpio-pca953x.c | 10 +++++++++-
> 1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpio/gpio-pca953x.c b/drivers/gpio/gpio-pca953x.c
> index 02f2a56..3387bdd 100644
> --- a/drivers/gpio/gpio-pca953x.c
> +++ b/drivers/gpio/gpio-pca953x.c
> @@ -329,7 +329,15 @@ static void pca953x_gpio_set_value(struct gpio_chip *gc, unsigned off, int val)
> u8 reg_val;
> int ret, offset = 0;
>
> - mutex_lock(&chip->i2c_lock);
> + /*
> + * We're using mutex_lock_nested() here to avoid a lockdep warning
> + * when there are two pca953x expanders, of which one is used to
> + * control an i2c gpio mux.
> + *
> + * We're using the GPIO base number to distinguish the lock
> + * subclasses.
> + */
> + mutex_lock_nested(&chip->i2c_lock, chip->gpio_start);
> if (val)
> reg_val = chip->reg_output[off / BANK_SZ]
> | (1u << (off % BANK_SZ));
> --
> 2.7.4
>
I didn't notice it before, but this patch triggers a different lockdep
warning due to exceeding the max allowed subclass number. I'm working
on converting the driver to regmap, which should fix the issue.
Best regards,
Bartosz Golaszewski