Re: [PATCH] hw/acpi: Set memory regions to native endian as a work around

[BALATON Zoltan]

On Mon, 1 May 2023, BALATON Zoltan wrote:
> - unaligned access is not handled correctly in acpi_pm_cnt_write() which now 
> I think is the place where this should be fixed

And also when running the test script on QEMU:

acpi_pm_cnt_read: 0 2 -> 0x80
acpi_pm_cnt_read: 1 2 -> 0x80

where the second should return 0 so these ops are either wrong or the 
memory layer does not do what's intended for:

static const MemoryRegionOps acpi_pm_cnt_ops = {
    .read = acpi_pm_cnt_read,
    .write = acpi_pm_cnt_write,
    .impl.min_access_size = 2,
    .valid.min_access_size = 1,
    .valid.max_access_size = 2,
    .endianness = DEVICE_LITTLE_ENDIAN,
};

I could patch this up in acpi_pm_cnt_read() abd acpi_pm_cnt_write() but 
how this is supposed to work? Do these functions have to handle unaligned 
access or the impl/valid settings should take care of that?

Regards,
BALATON Zoltan