On Tue, May 28, 2024 at 03:27:34PM +0300, Laurent Pinchart wrote:
> Hi Linus,
>
> On Tue, May 28, 2024 at 01:54:29PM +0200, Linus Walleij wrote:
> > On Mon, May 20, 2024 at 9:59 PM Laurent Pinchart wrote:
> >
> > > From: Haibo Chen <haibo.chen@xxxxxxx>
> > >
> > > The ADP5585 is a 10/11 input/output port expander with a built in keypad
> > > matrix decoder, programmable logic, reset generator, and PWM generator.
> > > This driver supports the GPIO function using the platform device
> > > registered by the core MFD driver.
> > >
> > > The driver is derived from an initial implementation from NXP, available
> > > in commit 451f61b46b76 ("MLK-25917-2 gpio: adp5585-gpio: add
> > > adp5585-gpio support") in their BSP kernel tree. It has been extensively
> > > rewritten.
> > >
> > > Signed-off-by: Haibo Chen <haibo.chen@xxxxxxx>
> > > Signed-off-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> >
> > (...)
> >
> > > +static int adp5585_gpio_direction_input(struct gpio_chip *chip, unsigned int off)
> > > +{
> > > + struct adp5585_gpio_dev *adp5585_gpio = gpiochip_get_data(chip);
> > > + unsigned int bank = ADP5585_BANK(off);
> > > + unsigned int bit = ADP5585_BIT(off);
> > > +
> > > + guard(mutex)(&adp5585_gpio->lock);
> > > +
> > > + return regmap_update_bits(adp5585_gpio->regmap,
> > > + ADP5585_GPIO_DIRECTION_A + bank,
> > > + bit, 0);
> >
> > First, I love the guarded mutex!
>
> Yes, it's neat :-)
>
> > But doesn't regmap already contain a mutex? Or is this one of those
> > cases where regmap has been instantiated without a lock?
>
> regmap indeed includes a lock, but it will lock each register access
> independently. In adp5585_gpio_get_value() we need to read two
> registers atomically, so we need to cover them by a single lock.
>
> I could drop the lock from regmap, but I would then likely need to
> introduce a lock in the parent mfd device that both the PWM and GPIO
> children would use to protect bus access. That may make sense in the
> future, but is a bit overkill for now I think.
Actually, I think I can drop the lock. Concurrent access to the
registers from different GPIOs are protected by the regmap lock, and
concurrent access to groups of registers for the same GPIO isn't a valid
use case as callers shouldn't do that.
> > > + gc = &adp5585_gpio->gpio_chip;
> > > + gc->parent = dev;
> > > + gc->direction_input = adp5585_gpio_direction_input;
> > > + gc->direction_output = adp5585_gpio_direction_output;
> >
> > And chance to implemen ->get_direction()?
>
> Sure, I'll add that to v2.
>
> > Other than this I think the driver is ready for merge, so with the
> > comments fixed or addressed:
> > Reviewed-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
>
> Thank you.
--
Regards,
Laurent Pinchart
