[SeaBIOS] Graphics card pass-through working with two pass pci-initialization

[Alex Williamson]

On Wed, 2011-06-01 at 16:31 +0200, Jan Kiszka wrote:
> On 2011-06-01 16:20, Isaku Yamahata wrote:
> > On Wed, Jun 01, 2011 at 09:30:12AM +0200, Gerd Hoffmann wrote:
> >>   Hi,
> >>
> >>> 0xE0000000 is hard-coded in the DSDT for both piix and q35 as below.
> >>> If the range is determined dynamically, the area also needs to be
> >>> updated somehow dynamically.
> >>>
> >>> ...
> >>>              Name (_CRS, ResourceTemplate ()
> >>> ...
> >>>                  DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed, NonCacheable, ReadWrite,
> >>>                      0x00000000,         // Address Space Granularity
> >>>                      0xE0000000,         // Address Range Minimum
> >>>                      0xFEBFFFFF,         // Address Range Maximum
> >>>                      0x00000000,         // Address Translation Offset
> >>>                      0x1EC00000,         // Address Length
> >>>                      ,, , AddressRangeMemory, TypeStatic)
> >>
> >> Uhm, indeed.  I know next to nothing about ACPI though.  Ideas anyone  
> >> how this could be done?
> > 
> > Right now what I can think of is.
> > It would be possible to know the offset in AmlCode[] by
> > compiling dsl with -l option. The we can get the mix of source and
> > resulted hex with offset like
> > 
> > iasl -l q35-acpi-dsdt.dsl.i
> > => q35-acpi-dsdt.dsl.lst
> > 
> >      192....                DWordMemory (ResourceProducer, PosDecode, MinFixed, MaxFixed, NonCacheable, ReadWrite,
> >      193....                    0x00000000,
> >      194....                    0xE0000000,
> >      195....                    0xFEBFFFFF,
> >      196....                    0x00000000,
> >      197....                    0x1EC00000,
> >      198....                    ,, , AddressRangeMemory, TypeStatic)
> > 
> > 00000F74....87 17 00 00 0C 01 00 00     "........"
> > 00000F7C....00 00 00 00 00 E0 FF FF     "........"
> > 00000F84....BF FE 00 00 00 00 00 00     "........"
> > 00000F8C....C0 1E ..................    ".."
> > 00000F8E....79 00 ..................    "y."
> > 
> > But this is very iasl specific and fragile...
> 
> Why make this particular thing dynamic? Physical PCs need to define a
> certain fixed PCI memory region as well and then live with it. IMHO, the
> current setting is just too small for normal use.

I don't think that's true.  It's possible a BIOS could walk the PCI bus,
figure out how much MMIO space it needs to reserve, then configure the
chipset to re-route the necessary memory and build the DSDT.  Hardly
anything is actually fixed within the chipset AIUI.  Thanks,

Alex