On 12 December 2017 at 09:16, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> Hi Ulf,
>
> On Mon, Dec 11, 2017 at 9:59 PM, Ulf Hansson <ulf.hansson@xxxxxxxxxx> wrote:
>> That together with an option of allowing "consumed resource-devices"
>> (irqchip) to be included in the wakeup path. I am thinking, perhaps
>> another driver PM flag (DPM_FLAG_WAKEUP_PATH), that the PM core looks
>> at and sets ->power.wakeup_path flag for the device and its parents.
>
> This is complicated by the fact that currently the device and irq
> subsystems don't really share data. E.g. {en,dis}able_irq_wake() and
> irq_set_irq_wake() don't take a device parameter, only an interrupt number,
> and conversion from interrupt numbers to devices is non-trivial.
> That's why I handled it in the irqchip drivers in the series referenced
> below.
Yes, that seems like a good way forward for now.
We can always look into how to extend the irq subsystem to deal with
this internally, but I rather leave that to be done as a separate
task.
[...]
Kind regards
Uffe
