Re: [RFC] ACPI: Add GPIO-signaled event simulator.

[Mika Westerberg <mika.westerberg@linux.intel.com>]

On Tue, Aug 12, 2014 at 07:15:41AM -0700, Alexandre Courbot wrote:
> On Tue, Aug 12, 2014 at 3:01 AM, Mika Westerberg
> <mika.westerberg@linux.intel.com> wrote:
> > On Fri, Aug 08, 2014 at 02:36:02PM +0200, Linus Walleij wrote:
> >> On Thu, Jul 24, 2014 at 5:51 PM, Tomasz Nowicki
> >> <tomasz.nowicki@linaro.org> wrote:
> >>
> >> > GPIO signaled events is quite new thing in Linux kernel.
> >> > AFAIK, there are not many board which can take advantage of it.
> >> > However, GPIO events are very useful feature during work on ACPI
> >> > subsystems.
> >>
> >> Overall this seems like a pretty nice debug feature.
> >>
> >> > This commit emulates GPIO h/w behaviour and consists on read/write
> >> > operation to debugfs file. GPIO device instance is still required in DSDT
> >> > table along with _AEI resources and event methods.
> >> >
> >> > Reading from file provides pin to GPIO device map e.g. :
> >> > $ cat /sys/kernel/debug/acpi/gpio_event
> >> > GPIO device name: /__SB.GPI0
> >> > Available GPIO pin map:
> >> > /__SB.GPI0 <-> pin 0x100
> >> >
> >> > Based on that, user can trigger method corresponding to device pin number:
> >> > $ echo "/__SB.GPI0 0x100" > /sys/kernel/debug/acpi/gpio_event
> >>
> >> I need input from Rafael and Mika as to whether this is a
> >> good interface.
> >
> > Maybe it would make sense to move this into drivers/gpio/gpiolib-acpi.c
> > and hide it behind some Kconfig entry?
> 
> I actually like that this feature is contained in its own file - it
> makes it easier to Kconfig-isolate and keeps the base ACPI code short
> and readable. We spent some effort decoupling the sysfs and legacy
> interfaces from the main gpiolib source file, let's keep that policy
> going. ;)

Well, OK but then it needs to make some functions available through the
private gpiolib.h so that the gpiolib-acpi.c is able to call them. Of
course it depends on whether the idea of exporting debugfs entries from
acpi_gpiochip_request_interrupt() instead is going to be implemented.