Re: [PATCH] gpio: omap: fix irq triggering in smart-idle wakeup mode

[Tony Lindgren]

* Grygorii Strashko <grygorii.strashko@xxxxxx> [160418 08:59]:
> On 04/15/2016 09:54 PM, Tony Lindgren wrote:
> > * santosh shilimkar <santosh.shilimkar@xxxxxxxxxx> [160415 08:22]:
> >> On 4/15/2016 2:26 AM, Grygorii Strashko wrote:
> >>>
> >>> Santosh, Tony, do you want me to perform any additional actions regarding this patch?
> >>>
> >> This patch should be run across family of SOCs to make
> >> sure wakeup works on all of those if not done already
> > 
> > Also, I'm not sure if we can just drop this code in question.
> > 
> > After this patch, what function updates the GPIO wkup_en registers
> > depending on enable_irq_wake()/disable_irq_wake()?
> > 
> 
> The main purpose of this patch is to *not* modify GPIO wkup_en registers
> depending of enable_irq_wake()/disable_irq_wake() :), instead all
> non wake up IRQs should be masked during suspend.

OK that makes sense.

> The GPIO wkup_en registers should be always in sync with GPIO irq_en when
> GPIO IP is in smart-idle wakeup mode. And this is done now from 
> omap_gpio_unmask_irq/omap_gpio_mask_irq(). See also [1].
> 
> In general, it is more or less similar to GIC + wakeupgen:
> - during normal work (including cpuidle) GIC irq_en and Wakeupgen wkup_en
> should be in sync always
> - during suspend - only IRQs, marked as wake up sources, should be left
> unmasked.
> 
> Also, I've found old thread [2] where Santosh proposed to use IRQCHIP_MASK_ON_SUSPEND.
> And it was not possible, at that time, but now IRQCHIP_MASK_ON_SUSPEND can be used :),
> because OMAP GPIO driver was switched to use generic irq handler instead of chained, so
> now OMAP GPIO irqs are properly handled by IRQ PM core.  
> [chained irqs (and chained irq handles) are not disabled during suspend/resume and they are
>  not maintained by IRQ PM core as result they can trigger way too early on resume when
>  OMAP GPIO is not ready/powered.]

OK. For my tests this patch does not change anything. I noticed however
that we still have some additional bug somewhere where GPIO wake up
events work fine for omap3 PM runtime, but are flakey for suspend.

> I've tested it on: am57x-evm, am437x-idk-evm, omap4-panda

OK thanks! Based on my tests and the above:

Acked-by: Tony Lindgren <tony@xxxxxxxxxxx>

Regards,

Tony

> [1] https://lkml.org/lkml/2016/4/12/676
> [2] https://lkml.org/lkml/2012/8/26/1
>     https://groups.google.com/forum/#!msg/linux.kernel/iXJ5Y568B3Q/hZ39bSlcs0kJ