matt_hsu wrote: > This patch is a a band-aid solution to fix the following issue: > > https://docs.openmoko.org/trac/ticket/1884
Hah, now I see what bug this one was about and that level-triggered interrupts have been suggested before :-) Holger and I discussed this one quite extensively on IRC earlier today. The conclusion is that, indeed, switching to a level-triggered PMU interrupt should solve this issue. Reading back the GPIO in the last moment and aborting the suspend by software would still leave a small race condition. We should consider using that only in the unlikely case that the 2442 still goes to suspend if a wakeup condition is already pending. The other work-arounds are likely to introduce more issues, like adding that parallel line to a CPU GPIO did in HXD8. (CPU could pull nONKEY down during a shutdown, causing the PMU to power up. We solved that somehow in hardware, but it wasn't nice.) Holger is investigating trying the level-triggered interrupt now. - Werner
