Re: [RFC PATCH 2/3] platform/x86: Add Uniwill laptop driver

[Armin Wolf]

Am 18.06.25 um 13:15 schrieb Werner Sembach:

> Hi,
>
> Am 17.06.25 um 23:50 schrieb Armin Wolf:
> > Am 16.06.25 um 14:32 schrieb Werner Sembach:
> >
> > > Hi,
> > >
> > > Am 15.06.25 um 19:59 schrieb Armin Wolf:
> [...]
> > > > +The EC behind the ``GetSetULong`` method is used by the OEM 
> > > > software supplied by the manufacturer.
> > > > +Reverse-engineering of this software is difficult since it uses an 
> > > > obfuscator, however some parts
> > > > +are not obfuscated.
> > > Iirc Christoffer said there is also another non wmi, but acpi, 
> > > get/set cmd that can write only specific regions but is faster. Just 
> > > wanted to leave this here for potential future optimizations if you 
> > > are interested.
> >
> > What interface are you referring to exactly?
>
> I think he was referring to these 2 methods in the dsdt:
>
>     Scope (_SB)
>     {
>         Device (INOU)
>         {
>             [...]
>
>             Method (ECRR, 1, NotSerialized)
>             {
>                 Local0 = (0xFE410000 + Arg0)
>                 Local1 = MMRW (Local0, Zero, Zero, Zero)
>                 Return (Local1)
>             }
>
>             Method (ECRW, 2, NotSerialized)
>             {
>                 Local0 = (0xFE410000 + Arg0)
>                 MMRW (Local0, One, Zero, Arg1)
>             }
>
>             [...]
>         }
>     }
>
> They can read/write the 0x04xx and 0x07xx addresses but not the 0x18xx 
> ones iirc, don't know about the 0x0fxx ones.
>
> I think 0x18xx was never intended to be written probably, but we use 
> it to implement a custom fan curve option in tuxedo-drivers on older 
> devices that officially don't support it.
>
> Best regards,
>
> Werner
I see, i too noticed the INOU0000 device during reverse engineering. I will update the documentation to inform future developers about this potentially
feaster interface. However for now i suggest that we stick to the WMI interface.

Thanks,
Armin Wolf