Re: [Qemu-devel] [PATCH v2] Add DSDT node for AppleSMC

[Gabriel L. Somlo]

On Fri, Dec 20, 2013 at 10:38:30PM +0100, Igor Mammedov wrote:
> > +    Device (SMC) {
> > +        Name(_HID, EisaId("APP0001"))
> > +        /* _STA will be patched to 0x0B if AppleSMC is present */
> > +        ACPI_EXTRACT_NAME_WORD_CONST CONCAT_SYM(SMC_PFX, smc_sta)
> typically dynamic variables are put in SSDT

When I tried statically building the SMC node as part of the SSDT, it
was not recognized by Mac OS X -- it only seems to work when it's part
of the DSDT.

> > +        Name(_STA, 0xFF00)
> > +        Name(_CRS, ResourceTemplate () {
> > +            IO (Decode16, 0x0300, 0x0300, 0x01, 0x20)
> > +            IRQNoFlags() { 6 }
> > +        })
> > +    }
> Device looks pretty simplistic, Have you considered building it completely
> dynamically?
> See build_ssdt() for examples, you might need to add extra building block for
> ResourceTemplate{} the rest could be done with existing build_* primitives.

If we consider "pasting" the SMC into the DSDT or not, based on
whether it was configured via -device applesmc or not, it comes
down to compiling the following ASL:

    Device (SMC) {
        Name(_HID, EisaId("APP0001"))
        Name(_STA, 0x0B)
        Name(_CRS, ResourceTemplate () {
            IO (Decode16, 0x0300, 0x0300, 0x01, 0x20)
            IRQNoFlags() { 6 }
        })
    }

into AML, and somehow conditionally appending the resulting blob,
unmodified, to the DSDT.

I would explore this approach before I'd consider compiling the blob
dynamically each time during QEMU startup, which would IMHO be harder
to follow.

However, I thought flipping one uint16_t from 0x0B to 0x00 (i.e., the
return value of SMC._STA) is much less intrusive to the DSDT than
pasting an entire blob dynamically during startup, which is why I
thought my approach was a bit cleaner.

Let me know what you think.

Thanks,
--Gabriel