Re: [PATCH v1] s390/virtio_ccw: don't allocate/assign airqs for non-existing queues

[David Hildenbrand <david@redhat.com>]

On 07.04.25 10:54, Michael S. Tsirkin wrote:
> On Mon, Apr 07, 2025 at 09:18:21AM +0200, David Hildenbrand wrote:
>>> Now I am beginning to think we should leave the spec alone
>>> and fix the drivers ... Ugh ....
>>
>> We could always say that starting with feature X, queue indexes are fixed
>> again. E.g., VIRTIO_BALLOON_F_X would have it's virtqueue fixed at index 5,
>> independent of the other (older) features where the virtqueue indexes are
>> determined like today.
>>
>> Won't make the implementation easier, though, I'm afraid.
>>
>> (I also thought about a way to query the virtqueue index for a feature, but
>> that's probably overengineering)
> 
> The best contract we have is the spec. Sometimes it is hopelessly broken
> and we have to fix it, but not in this case.
> 
> Let's do a theoretical excercise, assuming we want to fix the drivers,
> but we also want to have workarounds in place in qemu and in
> drivers to support existing ones. How would we go about it?

QEMU could likely be changed to always offer 
VIRTIO_BALLOON_F_FREE_PAGE_HINT, but not actually use it unless enabled 
for QEMU. That should work, because all action is initiated by the device.

That way, all virtqueue indexes would always be according to the spec.

We'll likely need compat machine handling ....


Regarding Linux, I'll have to think about it further ...

> Maybe we want a feature bit BALLOON_FIXED and ask everyone
> to negotiate it?  But if we go this way, we really need to fix
> the 48 bit limitation too.

I was thinking about the same, but it's all a mess ...

-- 
Cheers,

David / dhildenb