On Fri, 2 Oct 2026, MiloĻ Vlku wrote:

> Follow-up to my own mail: I have now run those values rather than only
> reading them out of the tables.
> 
> Built current pdx86 for-next with Chris's GET-only function-id change
> and balanced 1 / quiet 2 / performance 3 / full speed 4 on top. All four
> profiles apply and read back correctly, checked against the EC register
> each time:
> 
>   low-power            -> QFAN 2
>   balanced             -> QFAN 1
>   balanced-performance -> QFAN 3
>   performance          -> QFAN 4
> 
> All three power-profiles-daemon profiles work as well, so the desktop
> slider works end to end on TM2307.
> 
> Two things fell out of that:
> 
> - Without Chris's change every one of those writes fails with -ENOMSG
>   here, exactly as you described, and with it all of them succeed. So
>   that is a Tested-by from a second board if it is useful.
> 
> - No machine-specific capability check is needed on this one.
>   WMI_FN_SYSTEM_AC_TYPE is unimplemented, Armin's work reports that as
>   -EOPNOTSUPP, and bitland_pp_check_capability() already treats that as
>   "skip the check". The stock one is fine as-is.
> 
> So the TM2307 entry comes out as just a mode map plus a DMI match.
> Attached, against the v9 ops table and checkpatch-clean, in case it is
> useful to have in hand while you write the TM2309 one. I will resend it
> properly once that series settles -- posting it now mainly so the values
> sit somewhere concrete, and Mingyou can fold it straight into v10 if
> that is easier than carrying a separate patch.

Hi Milos,

The profile map approach used by Mingyou's patch 5 is not generic enough 
to handle all variation out there so please don't base on top of that.

We're currently trying to build a generalized approach for the known 
profile variations:

https://lore.kernel.org/all/20260928153902.139956-1-chris@miget.com/
https://lore.kernel.org/all/20260928064918.80526-1-matias.civadda2342001@gmail.com/

AFAIK, Chris has not yet sent the updated series which would have the 
latest version of the generalized profile approach. But once that is out, 
you should base your model's profiles on top of it.

-- 
 i.