On 2022-10-06 09:16:25 [+0200], Florian Paul Schmidt wrote:
> On 10/6/2022 8:27 AM, Sebastian Andrzej Siewior wrote:
> > that looks fine to me.
> > The runtime_resume path points to intel_dmc_load_program(). The
> > preempt-disable section there was introduced by commit
> > dff457d74e7ea ("drm/i915: Speed up DMC firmware loading")
> >
> > and the other thing seems to be intel_irq_reset() not sure where but it
> > is the runtime_suspend path.
> >
> > So disabling runtime-suspend should help.
>
> OK, so it was a user error :) Thanks for teaching me about rtla in the
> process! This will turn out to be helpful in the future!
So that got introduced in the v5.17 cycle then independent of RT itself?
If so I'm not sure what to do about it. Probably adding it to the i915
pile…
> I passed acpi=off as kernel parameter and that seems to have helped! The
> closest to runtime-suspend I could find was:
>
> /sys/devices/system/cpu/power/runtime_enabled
You should have one for the i915 somewhere under
/sys/devices/pci*/*/drm/*/power
there is powertop which should also have a disable button (it is usually
used to enable it).
Disabling acpi could be a big hammer in case there are also useful
things coming out it.
> which is "disabled" now.
>
> > Sebastian
>
> Thanks again and kind regards,
> FPS
Sebastian
