Re: [PATCH 13/13] migration: Use vmstate_register_any() for vmware_vga

[Juan Quintela]

Stefan Berger <stefanb@linux.ibm.com> wrote:
> On 10/19/23 15:08, Juan Quintela wrote:
>> I have no idea if we can have more than one vmware_vga device, so play
>> it safe.
>>
>> Signed-off-by: Juan Quintela <quintela@redhat.com>
>
> Reviewed-by: Stefan Berger <stefanb@linux.ibm.com>
>
>> ---
>>   hw/display/vmware_vga.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/hw/display/vmware_vga.c b/hw/display/vmware_vga.c
>> index 09591fbd39..7490d43881 100644
>> --- a/hw/display/vmware_vga.c
>> +++ b/hw/display/vmware_vga.c
>> @@ -1264,7 +1264,7 @@ static void vmsvga_init(DeviceState *dev, struct 
>> vmsvga_state_s *s,
>>
>>       vga_common_init(&s->vga, OBJECT(dev), &error_fatal);
>>       vga_init(&s->vga, OBJECT(dev), address_space, io, true);
>> -    vmstate_register(NULL, 0, &vmstate_vga_common, &s->vga);
>> +    vmstate_register_any(NULL, &vmstate_vga_common, &s->vga);
>
> And the first one registered with 'any' will again have instance_id =
> 0 assigned. So there's no side effect to be expected with any of these
> device, I suppose.

I will really change all the remaining registrations with 0 to any.

* If there is only a registration for that device: nothing changes
* If there is more than one registration for that device: It *could*
  work (from the migration point of view).

But then there are devices that *clearly* will not be able to have more
than one instance, so 0 is the best option there.  On top of my head:

* cpu-timers
* replay
* migration global_state

And the rest that I put on the cover letter are basically devices that
are used only once on the board initilization routine, so I feel safe
leaving them with zero.

Thanks for the review, Juan.