Hi, * Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> [200702 14:50]: > On Sun, May 24, 2020 at 7:11 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > > any special reason for this change ? It is not really explained in the > > commit description. > > Indeed. Why this change? For a kernel console, we want it to work for important oopses etc without trying to enable DMA or power on regulators for example. And we want to get rid of pm_runtime_irq_safe() usage as that takes a permanent usage count on the parent device. This causes issue where uart can keep an interconnect instance from idling with let's say genpd :) Also, we cannot easily make the serial driver PM runtime generic without removing the pm_runtime_irq_safe() dependency as it would currently wrongly impose the same nasty dependency to all the serial drivers. So when a kernel console is attached, we want to keep it from idling. For PM related testing, just detaching kernel console and configuring it's autosuspend timeout works nicely with dmesg -w. Or actually in that case using pstore console is even better. Andy has a series pending for generic serial PM runtime support, I guess he's going to be posting it as soon as the pending regressions are dealt with. Regards, Tony
