On Wed, 2016-08-24 at 17:24 +0200, bigeasy@xxxxxxxxxxxxx wrote: > On 2016-08-24 15:16:08 [+0000], Shevchenko, Andriy wrote: > > > > On Wed, 2016-08-24 at 17:03 +0200, bigeasy@xxxxxxxxxxxxx wrote: > > > > > > On 2016-08-24 14:56:12 [+0000], Shevchenko, Andriy wrote: > > > > > > > > > > > > > Today I discovered that runtime PM rather broken in 8250 _by > > > > design_. > > > > > > It > > > > definitely requires irq_safe flag to be set in runtime PM. But > > > > that > > > > flag > > > > demolishes an effort since it keeps parent (and thus all other > > > > devices > > > > which parent needs to be on) always on. > > > > ^^^^^ (1) > > > > > > > > > The question is is somebody is going to fix this and when? > > > > > > Why is it broken? > > > > I meant that without irq_safe flag set there are too many places > > where > > "sleep while in atomic" happens. On the other hand the irq_safe flag > > prevents parent to go to suspend. > > This must be new then. I don't remember _any_ "sleeping while atomic" > warnings back then. Maybe Tony knows something. Maybe it does not > occur on TI's PM code. I would have to reanimate my beagle bone to > check that. If you remove pm_runtime_irq_safe() you would be lucky to see them. Any pm_runtime_get_sync() call will issue one if run in atomic context. I found a lot of places in serial_core.c where it may happen. -- Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> Intel Finland Oy -- To unsubscribe from this list: send the line "unsubscribe linux-serial" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
