Re: [Qemu-devel] [PATCH 06/10] acpi-piix4: do not use old_portio-style callbacks

[Hervé Poussineau]

Andreas Färber a écrit :
> Am 04.01.2013 22:29, schrieb Hervé Poussineau:
> >  static const MemoryRegionOps piix4_pci_ops = {
> > -    .old_portio = (MemoryRegionPortio[]) {
> > -        {
> > -            .offset = PCI_UP_BASE - PCI_HOTPLUG_ADDR,   .len = 4, .size = 4,
> > -            .read = pci_up_read,
> > -        },{
> > -            .offset = PCI_DOWN_BASE - PCI_HOTPLUG_ADDR, .len = 4, .size = 4,
> > -            .read = pci_down_read,
> > -        },{
> > -            .offset = PCI_EJ_BASE - PCI_HOTPLUG_ADDR,   .len = 4, .size = 4,
> > -            .read = pci_features_read,
> > -            .write = pciej_write,
> > -        },{
> > -            .offset = PCI_RMV_BASE - PCI_HOTPLUG_ADDR,  .len = 4, .size = 4,
> > -            .read = pcirmv_read,
> > -        },
> > -        PORTIO_END_OF_LIST()
> > +    .read = pci_read,
> > +    .write = pci_write,
> > +    .endianness = DEVICE_NATIVE_ENDIAN,
> > +    .valid = {
> > +        .min_access_size = 4,
> > +        .max_access_size = 4,
> >      },
> > -    .endianness = DEVICE_LITTLE_ENDIAN,
>
> Ugh, question: Why is this changing Little Endian to native?

An error.
However, I was unable to find in the spec if this field is native on 
little endian. Not that it changes anything for now, as x86 is little 
endian...
I will respin patches, except those already in your memory-ioport branch.

Hervé