Re: [PATCH] virtio-spec: Define virtio-mmio registers as LE

[Rusty Russell]

Pawel Moll <pawel.moll@xxxxxxx> writes:
> On Tue, 2013-03-05 at 00:11 +0000, Rusty Russell wrote:
>> Pawel Moll <pawel.moll@xxxxxxx> writes:
>> > On Fri, 2013-03-01 at 11:21 +0000, Marc Zyngier wrote:
>> >> > Having said that, Rusty was contemplating enforcing LE config space in
>> >> > the new PCI layout...
>> >> 
>> >> I wouldn't complain about that, and would like to see a similar thing on
>> >> MMIO.
>> >
>> > Wherever PCI goes, MMIO follows :-)
>> 
>> Yes, but if you switch from 'guest-endian' to 'little-endian' how will
>> you tell?  For PCI, we'd detect it by using the new layout.
>
> The version register/value. At some point of time there will be a
> new(ish) MMIO layout anyway to deal with 64-bit addresses, replacing the
> ring page number with two 32-bit hi/lo physical address registers. This
> was discussed not long after the driver got merged...

As long as you have a plan for older guests...

>> I'd rather you specify MMIO as little endian, and we fix the kernel
>> config accessors to be endian aware (ie. 8, 16, 32, 64-bit accessors).
>> Since noone BE is using MMIO right now, it's safe...
>
> That's absolutely fine with me, however I don't see anything I could do
> in the virtio_mmio driver and spec - the virtio_config_ops specifies
> get/set as void * operations and I simply do byte-by-byte copy. Have I
> missed some config/endianess/PCI related discussion?

Yes, that's exactly what I mean, they'd have to be split into
8/16/32/64 accessors.  Which would do byte-by-byte for PCI.

The spec can simply be updated.

Cheers,
Rusty.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/