On 9/4/24 21:50, Cristian Marussi wrote:
On Wed, Sep 04, 2024 at 05:12:29PM +0100, Sudeep Holla wrote:
On Wed, Sep 04, 2024 at 08:43:24AM +0530, Sibi Sankar wrote:
Ensure that the bad duplicates reported by the platform firmware doesn't
get added to the opp-tables.
I am really interested to know if the platform firmware is presenting
duplicates intentionally for some unknown reasons and we are just speculating
it to be broken firmware or is it really broken firmware.
For me, it is very hard to digest something like OPP tables which is there
for a very long time now is not very well understood by firmware authors.
How many duplicates are we seeing on this platform really ? If it is
just one I can understand. More than one is hard to miss from the OPP
tables in the firmware.
While I am not opposing to make the driver handle these duplicates,
I am just worried if they are put there intentionally for reasons we
don't understand yet or not published.
The number of duplicates reported in logs makes me suspect the same...seems
like intentional/by_design .... but at first I stick to the general issue
of handling bad fw replies and how to survive kernel side at first...but I
indeed share your same concerns...
Hey Cristian/Sudeep,
The number of opps being duplicated is limited to the max
sustainable frequency before we see the turbo frequency. This
was pretty much the case in older non scmi perf qc cpufreq
drivers. They just filter it there, but I've gotten word that
this will get fixed in firmware for this SoC and any future ones
planning to use scmi-perf for cpufreq.
-Sibi
Thanks,
Cristian
