Re: [PATCH] gpio / ACPI: Handle ACPI events in accordance with the spec

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Wednesday, April 10, 2013 01:54:26 PM Mika Westerberg wrote:
> On Wed, Apr 10, 2013 at 12:39:21PM +0200, Rafael J. Wysocki wrote:
> > On Wednesday, April 10, 2013 12:17:47 PM Mika Westerberg wrote:
> > > On Wed, Apr 10, 2013 at 11:17:57AM +0300, Mathias Nyman wrote:
> > > > On 04/10/2013 10:53 AM, Mika Westerberg wrote:
> > > > >On Tue, Apr 09, 2013 at 03:57:25PM +0200, Rafael J. Wysocki wrote:
> > > > >>+void acpi_gpiochip_free_interrupts(struct gpio_chip *chip)
> > > > >>+{
> > > > >>+	acpi_handle handle;
> > > > >>+	acpi_status status;
> > > > >>+	struct list_head *evt_pins;
> > > > >>+	struct acpi_gpio_evt_pin *evt_pin, *ep;
> > > > >>+
> > > > >>+	if (!chip->dev || !chip->to_irq)
> > > > >>+		return;
> > > > >>+
> > > > >>+	handle = ACPI_HANDLE(chip->dev);
> > > > >>+	if (!handle)
> > > > >>+		return;
> > > > >>+
> > > > >>+	status = acpi_get_data(handle, acpi_gpio_evt_dh, (void **)&evt_pins);
> > > > >>+	if (ACPI_FAILURE(status))
> > > > >>+		return;
> > > > >>+
> > > > >>+	list_for_each_entry_safe_reverse(evt_pin, ep, evt_pins, node) {
> > > > >>+		devm_free_irq(chip->dev, evt_pin->irq, evt_pin);
> > > > >
> > > > >How about using normal request/free_irq() functions for both _EVT and
> > > > >non-_EVT events? Since we now need to call acpi_gpiochip_free_interrupts()
> > > > >anyway, I don't see the point using devm_* functions here.
> > > > >
> > > > 
> > > > Then we need to create a list of non-_EVT events, or add them to the
> > > > evt_pins list.
> > > 
> > > Good point. Maybe we can add them to the evt_pins list and handle the same
> > > way as _EVT (except that we need to call _Exx and _Lxx methods instead of
> > > _EVT)?
> > 
> > The difference is that the evt_pins data is only needed for _EVT execution,
> > because _EVT takes the pin argument.  _Lxx/_Exx don't take arguments and
> > there's no need to add extra data structures for them, as devm_ does what's
> > needed.
> 
> OK, thanks for the explanation.
> 
> > Of course, plain request/free_irq may be used for the _EVT events only
> > at the expense of a little more complexity in acpi_gpiochip_request_interrupts().
> 
> I'm not sure whether it is a good thing to mix devm_ and plain request/free
> here.

Yes, that was my motivation for doing it the way I did it.

> And more complexity is always bad so I guess we can stay with this
> implementation now.

> Feel free to add my
> 
> Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>

Cool, thanks!

I wonder if Linus or Grant have any objections, though.  Guys?

The patch has more to do with ACPI than with GPIO and I think it would be
good to follow the spec from the outset here, so I'd like to push it for
v3.10.

Thanks,
Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.