Re: [Qemu-devel] [PATCH 1/8] spapr-iommu: add a bus for spapr-iommu devices

[Alexey Kardashevskiy]

On 04/15/2014 01:16 AM, Andreas Färber wrote:
> Am 10.04.2014 17:18, schrieb Alexey Kardashevskiy:
>> On 04/11/2014 12:52 AM, Andreas Färber wrote:
>>> Am 10.04.2014 16:40, schrieb Alexey Kardashevskiy:
>>>> On 04/10/2014 10:40 PM, Alexander Graf wrote:
>>>>>
>>>>> Juan, is a different command line device order supposed to work with
>>>>> migration?
>>>>
>>>>
>>>> We discussed this on IRC with Paolo and the conclusion is that yes, the
>>>> order should not matter.
>>>
>>> Huh?! If you ever tried changing the order of PCI devices such as
>>> virtio-blk-pci on the command line (or changing between if=virtio and
>>> -device virtio-blk-pci) then surely it does change the order of the
>>> /dev/vdX the guest sees and will be migration-incompatible. The order of
>>> -drive and -device however, for instance, does not seem to matter. But
>>> in this generality, your statement does not seem to reflect current reality.
>>
>> As far as I can tell libvirt is trying to be precise and specifies PCI bus
>> name and PCI address for PCI devices.
> 
> Now that is very different from a generic "the device order should not
> matter"! :)

"should" != "does" ;)

For PCI it is possible to get same guest adding PCI device in any order
(sure, with parameters essential for that such as bus address). And if the
user or libvirt wants to change the order, he/she/it should follow these
rules. For VIO it is not possible.



-- 
Alexey