Re: [PATCH v6 5/5] platform/x86: bitland-mifs-wmi: Add per-machine ops table

[Miloš Vlku]

Hi Anton,

Ran your questions on my TM2307 (kernel 7.2.9, for-next plus Chris's
GET-only change).

On battery the driver refuses balanced-performance and performance, like
you said. But if I write the mode straight to QFAN, both 3 and 4 stick on
battery without complaint. So same as your TM2309 - the firmware is fine
with it and these boards don't really need the capability check.

Fn+K only ever goes 1 -> 3 -> 2. I pressed it around twenty times while
watching QFAN and it never landed on 4. So I agree with your mapping:
performance should be Turbo (3), and full speed should be max-power so
nobody gets it by accident from the desktop slider.

You're right about the suspend problem too. My patch didn't handle a
stale 0 at all, so it would have broken suspend the same way it did on
the TM2426. Reading 0 as balanced makes sense. Thanks for catching that.

kb_mode reads back -EIO here on for-next, so your patch should hide it on
this board as well. I haven't built your patch yet though, so that's just
the precondition - happy to test it properly if another report helps.

Command 0x10 looks the same as yours: subcommand 1 reads SOH1 and 2 is
bit 0 of LONL at 0xA4. The only difference is that subcommand 3 compares
ADPW against 100 instead of 140. That looks like the most this board can
take over USB-C: ADPW reads 100 with the stock 100 W charger on either
port, still 100 with a 160 W one, and 0 on battery. The ports report USB
PD 2.0, so no EPR. My limit is off (LONL 0, AFBC 100), SOH1 100.

One more thing, since you use thermald: with int3400_thermal loaded,
thermald --adaptive sets the MMIO PL1 from the GDDV table here too (about
25/50/90 W), but the CPU still drew 27 W because the MSR PL1 stayed at
28 W. Restarting thermald puts it back to 28 W; profile changes don't.
With it opened after thermald starts, sustained load settles around 56 W
at 101 C, and balanced at 45 W. Does your MSR PL1 also read 28 W after
thermald starts?

Happy to do the joint TM2307/TM2309 entry with you once Chris's series is
out - either way round is fine. I'll put my Tested-by on Chris's thread
rather than here.

Thanks,
Milos