On 01/06/2019 17:49, Greg Kurz wrote:
> On Fri, 31 May 2019 16:36:33 -0300
> Eduardo Habkost <ehabk...@redhat.com> wrote:
>
>> On Tue, May 28, 2019 at 10:48:09AM +0800, Yongji Xie wrote:
>>> On Tue, 28 May 2019 at 02:54, Michael S. Tsirkin <m...@redhat.com> wrote:
>>>>
>>>> On Mon, May 27, 2019 at 12:44:46PM +0200, Greg Kurz wrote:
>>>>> On Fri, 24 May 2019 19:56:06 +0800
>>>>> Yongji Xie <elohi...@gmail.com> wrote:
>>>>>
>>>>>> On Fri, 24 May 2019 at 18:20, Greg Kurz <gr...@kaod.org> wrote:
>>>>>>>
>>>>>>> On Mon, 20 May 2019 19:10:35 -0400
>>>>>>> "Michael S. Tsirkin" <m...@redhat.com> wrote:
>>>>>>>
>>>>>>>> From: Xie Yongji <xieyon...@baidu.com>
>> [...]
>>>>>>>> @@ -1770,6 +1796,13 @@ static bool virtio_broken_needed(void *opaque)
>>>>>>>> return vdev->broken;
>>>>>>>> }
>>>>>>>>
>>>>>>>> +static bool virtio_started_needed(void *opaque)
>>>>>>>> +{
>>>>>>>> + VirtIODevice *vdev = opaque;
>>>>>>>> +
>>>>>>>> + return vdev->started;
>>>>>>>
>>>>>>> Existing machine types don't know about the "virtio/started"
>>>>>>> subsection. This
>>>>>>> breaks migration to older QEMUs if the driver has started the device,
>>>>>>> ie. most
>>>>>>> probably always when it comes to live migration.
>>>>>>>
>>>>>>> My understanding is that we do try to support backward migration
>>>>>>> though. It
>>>>>>> is a regular practice in datacenters to migrate workloads without
>>>>>>> having to
>>>>>>> take care of the QEMU version. FWIW I had to fix similar issues
>>>>>>> downstream
>>>>>>> many times in the past because customers had filed bugs.
>>>>>>>
>>>>>>
>>>>>> If we do need to support backward migration, for this patch, what I
>>>>>> can think of is to only migrate the flag in the case that guest kicks
>>>>>> but not set DRIVER_OK. This could fix backward migration in most case.
>>>>>
>>>>> You mean something like that ?
>>>>>
>>>>> static bool virtio_started_needed(void *opaque)
>>>>> {
>>>>> VirtIODevice *vdev = opaque;
>>>>>
>>>>> return vdev->started && !(vdev->status & VIRTIO_CONFIG_S_DRIVER_OK);
>>>>> }
>>>>>
>>>>>> Not sure if there is a more general approach...
>>>>>>
>>>>>
>>>>> Another approach would be to only implement the started flag for
>>>>> machine version > 4.0. This can be achieved by adding a "use-started"
>>>>> property to the base virtio device, true by default and set to
>>>>> false by hw_compat_4_0.
>>>>
>>>> I think this is best frankly.
>>>>
>>>
>>> Only implement the started flag for machine version > 4.0 might not be
>>> good because vhost-user-blk now need to use this flag. How about only
>>> migrating this flag for machine version > 4.0 instead?
>>
>> Was this implemented? Is migration from QEMU 4.1 to QEMU 4.0
>> currently broken?
>>
>
> Answer is yes for both questions.
>
Is there a fix?
The problem is really easy to reproduce: start a guest with virtio-blk
and migrate once the driver has started.
Thanks,
Laurent
