* Grygorii Strashko <grygorii.strashko@xxxxxx> [160408 10:17]: > On 04/08/2016 06:14 PM, Tony Lindgren wrote: > > * Grygorii Strashko <grygorii.strashko@xxxxxx> [160408 03:41]: > >> Hi Tony, > >> > >> On 04/05/2016 01:40 AM, Tony Lindgren wrote: > >>> Hi, > >>> > >>> * Marcin Niestroj <m.niestroj@xxxxxxxxxxxxxxxx> [160404 08:57]: > >>>> Support configuration of ext_wakeup sources. This patch makes it > >>>> possible to enable ext_wakeup and set it's polarity, depending on board > >>>> configuration. AM335x's dedicated PMIC (tps65217) uses ext_wakeup to > >>>> notify about power-button presses. Handling power-button presses enables > >>>> to recover from RTC-only power states correctly. > >>> > >>> I suggest you just set this pin up as a minimal gpiochip. That way > >>> we can use the standard binding :) And if we get lucky, this pin can > >>> also trigger during runtime. > >>> > >> > >> Following my comments on v2 of this patch I propose to rollback to > >> this version of the patch. > >> > >> It seems doesn't fit in gpiochip, it's more likely irqchip, but since > >> rtc can't generate IRQ there are nor reasons for these genetic > >> experiments and RTC's specific bindings looks more suitable, at least for me. > > > > Hmm well gpiochips typically are irqchip too. IMO the generic binding > > here sounds like "gpio-wakeup" as it's an input with polarity and with > > an optional interrupt. > > > > Plain irqchip would work too in this case if there are no other GPIO > > specific features. Some other RTCs may have more GPIO like features. > > > > In any case, setting the ext_wakeup up as an irqchip means that the > > RTC controller can be used as a dedicated wakeirq with Linux :) > > It can't :( It can't generate IRQ when state of ext_wakeup line has > been changed. Hmm to me it certainly seems it should be capable of generating and RTC interrupt as we have EXT_WAKEUP_STATUS register containing four bits, one for each wakeup line. We also have EXT_WAKEUP_DB_EN for debounce, EXT_WAKEUP_POL for polarity, and EXT_WAKEUP_EN to enable in addition to status.. These certainly make it look like a gpiochip or irqchip. > > Are you guys sure there's no wake pin events during runtime? AFAIK > > the RTCs just typically produce an interrupt when programmed to > > do so, they don't know the state of the SoC. > > > > I do not think that It can be fit in gpiochip or irqchip - > in my opinion right way is to proceed with bindings proposed by > Marcin in this patch v1. > > But, probably, it could fit in pinctrl: > - this is pin's configuration > - this is one-time configuration > - or - configuration which can be applied before suspend > and resorted after suspend. > > if you still wanna try some generic framework. Yeah the pin certainly should be configured by pinctrl if it is muxable. However, the functionality of the pin is clearly tied to the RTC driver. My guess the reason it does not generate interrupts is that it's status needs to be cleared each time it triggers? That fits the irq mask/unmask for generic wakeirq like we do for the padconf. Sounds like some experiments and reasoning is needed :) I guess the first test to run is tools/testing/selftests/timers/rtctest.c to make sure the RTC interrupts trigger properly. Regards, Tony -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
