On Wed, 26 Feb 2025 at 21:04, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote: > > Top-posting not welcome. ? > On Wed, Feb 26, 2025 at 8:52 PM Antheas Kapenekakis <lkml@xxxxxxxxxxx> wrote: > > > > > > What about adding "quiet" as a "hidden choice" to amd-pmf such that it > > > would allow the test_bit(*bit, handler->choices) check in > > > _store_class_profile() to pass, but it would not cause this "choice" > > > to become visible in the new I/F (or when amd-pmf becomes the only > > > platform-profile driver) and it would be aliased to "low-power" > > > internally? > > > > This is what this patch series essentially does. It makes amd-pmf > > accept all choices but only show its own in its own handler and when > > it is the only option > > No, it does more than this. I would say functionality-wise no. The patch could be minified further. > For instance, it is not necessary to do > anything about PLATFORM_PROFILE_BALANCED_PERFORMANCE in it. I do not see a difference between QUIET and BALANCED_PERFORMANCE, any driver occluding either causes the same issue. Severity is debatably lower on BP though. > The structure of it is questionable either. It really should be two > patches, one modifying the ACPI platform-profile driver and the other > changing amd-pmf on top of this. Ack. I can spin it up as 2 patches. > Moreover, I'm not entirely convinced that the "secondary" driver > concept is needed to address the problem at hand. Any suggestions on that front would be welcome. This is just the way I came up with doing it. Best, Antheas
