Call gpio_cansleep only after gpio_request succeeded avoiding an
oops.
Signed-off-by: NUwe Kleine-König <Uwe.Kleine-Koenig@digi.com>
Acked-by: NDavid Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: NRichard Purdie <rpurdie@rpsys.net>

When gpio_direction_output() is called, led_dat->active_low is used
as default value. This means that the led will always be off by
default. cdev.brightness should really have been set to LED_OFF
unconditionally to reflect this behavior.
Signed-off-by: NRaphael Assenat <raph@8d.com>
Signed-off-by: NRichard Purdie <rpurdie@rpsys.net>

Three bugfixes to the leds-gpio driver, plus minor whitespace tweaks:

 - Do the INIT_WORK() before registering each LED, so if its trigger
   becomes immediately active it can schedule work without oopsing..

 - Use normal registration, not platform_driver_probe(), so that
   devices appearing "late" (hotplug type) can still be bound.

 - Mark the driver remove code as "__devexit", preventing oopses
   when the underlying device is removed.

These issues came up when using this driver with some GPIO expanders
living on serial busses, which act unlike "normal" platform devices:
they can appear and vanish along with the serial bus driver.
Signed-off-by: NDavid Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: NRichard Purdie <rpurdie@rpsys.net>

Teach the new leds-gpio driver that some GPIOs can't be accessed from
timer callbacks ... which is how all today's standard triggers use them.
Signed-off-by: NDavid Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: NRichard Purdie <rpurdie@rpsys.net>

This patch adds support for GPIO connected leds via the new GPIO framework.

Information about leds (gpio, polarity, name, default trigger) is passed
to the driver via platform_data.
Signed-off-by: NRaphael Assenat <raph@8d.com>
Signed-off-by: NRichard Purdie <rpurdie@rpsys.net>