[Spice-devel] [RfC PATCH 7/7] async qxl I/O

[Gerd Hoffmann]

   Hi,

>> +/* old spice-protocol */
>> +#ifndef QXL_INTERRUPT_IO_CMD
>> +#define QXL_INTERRUPT_IO_CMD               (1<<  2)
>> +#define QXL_IO_UPDATE_AREA_ASYNC           (QXL_IO_DESTROY_ALL_SURFACES +1)
>> +#define QXL_IO_NOTIFY_OOM_ASYNC            (QXL_IO_DESTROY_ALL_SURFACES +2)
>> +#define QXL_IO_MEMSLOT_ADD_ASYNC           (QXL_IO_DESTROY_ALL_SURFACES +3)
>> +#define QXL_IO_CREATE_PRIMARY_ASYNC        (QXL_IO_DESTROY_ALL_SURFACES +4)
>> +#define QXL_IO_DESTROY_PRIMARY_ASYNC       (QXL_IO_DESTROY_ALL_SURFACES +5)
>> +#define QXL_IO_DESTROY_SURFACE_ASYNC       (QXL_IO_DESTROY_ALL_SURFACES +6)
>> +#define QXL_IO_DESTROY_ALL_SURFACES_ASYNC  (QXL_IO_DESTROY_ALL_SURFACES +7)
>> +#endif
>> +
>
> Can you drop this? I think it's better to just not support the io if spice-protocol
> is too old, or better yet just require a newer spice-protocol for building (and
> bump spice-protocol version).

I'd prefer to not require a new spice-protocol, at least not initially. 
  Of course this stuff can only be merged upstream once the 
spice-protocol bits are final & committed.

> missing for windows driver, which does QXL_IO_MEMSLOT_ADD before CREATE_PRIMARY,
> and with async this turns to QXL_IO_MEMSLOT_ADD_ASYNC followed by QXL_IO_UPDATE_IRQ
>
> +    case QXL_IO_UPDATE_IRQ:

Ok, will fix.

cheers,
   Gerd