Re: [PATCH] ACPI: Add GPIO-signaled event emulator.

[Tomasz Nowicki <tomasz.nowicki@linaro.org>]

On 26.08.2014 00:39, Rafael J. Wysocki wrote:
> On Thursday, August 21, 2014 04:39:46 PM Tomasz Nowicki wrote:
>> Hi Mika,
>>
>> On 21.08.2014 12:45, Mika Westerberg wrote:
>>> On Wed, Aug 20, 2014 at 04:58:20PM +0200, Tomasz Nowicki wrote:
>
> [cut]
>
>>>> +
>>>> +static int gpio_evt_trigger(void *data, u64 val)
>>>> +{
>>>> +	struct gpio_pin_data *pin_data = (struct gpio_pin_data *)data;
>>>> +	int pin = pin_data->pin;
>>>> +
>>>> +	if (ACPI_FAILURE(acpi_execute_simple_method(pin_data->handle, NULL,
>>>> +						    pin <= 255 ? 0 : pin)))
>>>> +		pr_err(PREFIX "evaluating event method failed\n");
>>>
>>> acpi_execute_simple_method() passes one argument to the method. You
>>> can't use it with _Lxx or _Exx which don't expect any arguments.
>>> Otherwise you get this:
>>>
>>> [  122.258191] ACPI: \_SB_.GPO2._E12: Excess arguments - Caller passed 1, method requires 0 (20140724/nsarguments-263)
>> Right, I will fix it.
>
> OK, so here's my concern.
>
> If AML does any kind of tracking of state in _Exx/_Lxx, you'll likely totally
> confuse it by calling those things at random.
>
> I'm not sure I'm seeing a compelling reason to put this thing into the tree
> for this reason.

Yes you are right, but this emulator is only for debugging/development 
purposes and goes with clear statement "DANGER to use on production 
kernel", like APEI error injection feature.

IMO, this emulator would help with:
1. testing and developing AML methods
2. working on ACPI subsystems that need GPIO-signaled events without 
GPIO h/w

Regards,
Tomasz