Re: [PATCH] backlight: gpio-backlight: Fix warning when the GPIO is on a I2C chip

[Jingoo Han <jg1.han@samsung.com>]

On Friday, May 09, 2014 12:09 PM, Tony Lindgren wrote:
> On Friday, May 09, 2014 11:25 AM, Jingoo Han wrote:
> > On Friday, May 09, 2014 10:25 AM, Lee Jones wrote:
> > >
> > > If the GPIO for the backlight is on an I2C chip, we currently
> > > get nasty warnings like this during the boot:
> > >
> > > WARNING: CPU: 0 PID: 6 at drivers/gpio/gpiolib.c:2364 gpiod_set_raw_value+0x40/0x4c()
> > > Modules linked in:
> > > CPU: 0 PID: 6 Comm: kworker/u2:0 Not tainted 3.15.0-rc4-12393-gcde9f4e #400
> > > Workqueue: deferwq deferred_probe_work_func
> > > [<c0014cbc>] (unwind_backtrace) from [<c001191c>] (show_stack+0x10/0x14)
> > > [<c001191c>] (show_stack) from [<c0566ae0>] (dump_stack+0x80/0x9c)
> > > [<c0566ae0>] (dump_stack) from [<c003f61c>] (warn_slowpath_common+0x68/0x8c)
> > > [<c003f61c>] (warn_slowpath_common) from [<c003f65c>] (warn_slowpath_null+0x1c/0x24)
> > > [<c003f65c>] (warn_slowpath_null) from [<c02f7e10>] (gpiod_set_raw_value+0x40/0x4c)
> > > [<c02f7e10>] (gpiod_set_raw_value) from [<c0308fbc>] (gpio_backlight_update_status+0x4c/0x74)
> > > [<c0308fbc>] (gpio_backlight_update_status) from [<c030914c>] (gpio_backlight_probe+0x168/0x254)
> > > [<c030914c>] (gpio_backlight_probe) from [<c0378fa8>] (platform_drv_probe+0x18/0x48)
> > > [<c0378fa8>] (platform_drv_probe) from [<c0377c40>] (driver_probe_device+0x10c/0x238)
> > > [<c0377c40>] (driver_probe_device) from [<c0376330>] (bus_for_each_drv+0x44/0x8c)
> > > [<c0376330>] (bus_for_each_drv) from [<c0377afc>] (device_attach+0x74/0x8c)
> > > [<c0377afc>] (device_attach) from [<c03771c4>] (bus_probe_device+0x88/0xb0)
> > > [<c03771c4>] (bus_probe_device) from [<c03775c8>] (deferred_probe_work_func+0x64/0x94)
> > > [<c03775c8>] (deferred_probe_work_func) from [<c00572e8>] (process_one_work+0x1b4/0x4bc)
> > > [<c00572e8>] (process_one_work) from [<c00579d0>] (worker_thread+0x11c/0x398)
> > > [<c00579d0>] (worker_thread) from [<c005dfd8>] (kthread+0xc8/0xe4)
> > > [<c005dfd8>] (kthread) from [<c000e768>] (ret_from_fork+0x14/0x2c)
> > >
> > > Fix this by using gpio_set_value_cansleep() as suggested in
> > > drivers/gpio/gpiolib.c:2364. This is what the other backlight drivers
> > > are also doing.
> >
> > OK, I see.
> > However, gpio_backlight drive can be used by a lot of gpio drivers.
> > In some cases, 'can_sleep' is 'false' and gpio_set_value_cansleep()
> > is unnecessary.
> >
> > In my opinion, gpio_set_value_cansleep() or gpio_set_value() can be
> > called selectively by 'can_sleep' value.
> >
> > How about the following?
> >
> > -       gpio_set_value(gbl->gpio, brightness ? gbl->active : !gbl->active);
> > +       if (gpio_cansleep(gbl->gpio))
> > +               gpio_set_value_cansleep(gbl->gpio,
> > +                                       brightness ? gbl->active : !gbl->active);
> > +       else
> > +               gpio_set_value(gbl->gpio, brightness ? gbl->active : !gbl->active);
> 
> It should be always fine to use gpio_set_value_cansleep(), see your
> old thread from few years ago related to another backlight driver:
> 
> https://lkml.org/lkml/2012/12/5/343

(+cc Linus Walleij, Alexandre Courbot, Russell King)

OK, I see.

gpio_set_value_cansleep() calls gpiod_set_raw_value_cansleep(),
and gpio_set_value() calls gpiod_set_raw_value() as below.

./drivers/gpio/gpiolib.c
void gpiod_set_raw_value_cansleep(struct gpio_desc *desc, int value)
{
        might_sleep_if(extra_checks);
        if (!desc)
                return;
        _gpiod_set_raw_value(desc, value);
}
EXPORT_SYMBOL_GPL(gpiod_set_raw_value_cansleep);

./drivers/gpio/gpiolib.c
void gpiod_set_raw_value(struct gpio_desc *desc, int value)
{
        if (!desc)
                return;
        /* Should be using gpio_set_value_cansleep() */
        WARN_ON(desc->chip->can_sleep);
        _gpiod_set_raw_value(desc, value);
}
EXPORT_SYMBOL_GPL(gpiod_set_raw_value);

Then, the difference between gpio_set_value_cansleep() or
gpio_set_value() is whether might_sleep_if(extra_checks) is
called or not.

So, you said that "It should be always fine to use 
gpio_set_value_cansleep()", right?

Linus Walleij,
Is there any reason to keep these two functions such as
gpiod_set_raw_value_cansleep() and gpiod_set_raw_value()?

Best regards,
Jingoo Han