On Fr, 2013-10-04 at 15:20 +0200, Igor Mammedov wrote: > On Wed, 21 Aug 2013 13:01:32 +0200 > Gerd Hoffmann <kra...@redhat.com> wrote: > > > Hi, > > > > > +#define ICH9_PROC_BASE 0xaf00 > > > +#define ICH9_PROC_LEN 32 > > > > No, please don't. It makes it impossible to assign the 0xa000 -> 0xafff > > I/O port window to a PCI bridge. Please lets stop occupy random io > > ports above 0x1000 and burn I/O address space that way. > I'm curios why 0x1000 is any better than 0xa000, it still would be random > port occupation. Is there any guideline which ports could be used and which > shouldn't?
0x1000 doesn't by us anything. *below* 0x1000 gives us one more pci io window which we can assign to a pci bridge then. > Could we safely move PIIX CPU hotplug 0xaf00-0xaf20 range below 0x1000? I'd love to, but unfortunaly it isn't that easy as the address is hard-coded in the DSDT. Once mst acpi generation patches are in it would be a bit simpler, but we would still break compatibility with seabios versions not fetching the acpi tables from qemu. cheers, Gerd
