* Sudeep Holla [off-list ref] [151204 08:16]:

Hi Tony,

On 04/12/15 15:40, Tony Lindgren wrote:

quoted

* Tony Lindgren [off-list ref] [151203 13:41]:

quoted

* Sudeep Holla [off-list ref] [151203 11:00]:

quoted

I have added irq_set_irq_wake(pcs_soc->irq, state) in pcs_irq_set_wake
which ensures it's marked for wakeup.

Hmm well see the error I pasted in this thread, maybe that provides
more clues.

The irq_set_irq_wake(pcs_soc->irq, state) in pcs_irq_set_wake does not
look right to me as pcs_irq_set_wake toggles the irq_wake for each pin
separately, not for the whole controller.

After thinking more about it we need some way to tell IRQ core that
pcs_soc->irq is wakeup capable. Is that going to happen automatically
via dev_pm_set_dedicated_wake_irq as you mentioned earlier ?

quoted

I think all that can be left out with the snipped from Grygorii, and maybe
also the lock_class_key changes.

If we not calling irq_set_irq_wake(pcs_soc->irq) in pcs_irq_set_wake, do
you see possibility of lockdep recursion in any other paths.

Otherwise we don't need this if we remove irq_set_irq_wake(pcs_soc->irq)
from pcs_irq_set_wake

I think Grygorii is right here and this is correct as it's a counter
once the other issues are sorted out and we have figured out what all
needs to be patched together.

Regards,

Tony