Re: [Qemu-devel] Re: [PATCH 11/11] test-vmstate: add test case to verify we don't change VMState

[Anthony Liguori <anthony@codemonkey.ws>]

On 03/23/2011 11:45 AM, Peter Maydell wrote:
> On 23 March 2011 16:27, Anthony Liguori<anthony@codemonkey.ws>  wrote:
>> Migration uses the VMStateDescription name as a section identifier.  The
>> section identifiers MUST be unique for a given device.  Otherwise, if both
>> devices are present, migration fails miserably.
> So how does it work if you have two devices of the same type
> in the system? I'd assumed that the unique identifier would
> be one based on the actually instantiated devices.

It's a combination of the section name + a device instance ID that has 
some magic behind making it unique and stable.  But instance ID is not 
unique on it's own.

>>   It also means that if the
>> wrong devices are created on the destination, instead of predictable
>> failure, you get unpredictable guest corruption.
>>
>> The Right Way to support what you're describing above is to have a single
>> VMStateField array and two VMStateDescriptions.  IOW:
> This doesn't make sense because it's decoupling the information
> about minimum version ID etc (in the VMStateDescription) from
> the information about which fields are in which version (which
> is in the VMStateField array when it's initialised via
> VMSTATE_*_V() macros). So although you get to avoid duplicating
> all the fields in the arrangement you suggest you still end
> up with version_id, minimum_version_id etc being duplicated
> and needing to stay in sync.

So maybe we just need to remove name from VMStateDescription and force 
it to come from DeviceInfo.

Regards,

Anthony Liguori

> -- PMM
>