On 06/03/2015 09:40 PM, Stefan Hajnoczi wrote:
> On Wed, Jun 3, 2015 at 10:42 AM, Jason Wang wrote:
>>
>> On 06/03/2015 03:56 PM, Thibaut Collet wrote:
>>> Hi,
>>>
>>> thanks for your point, I did not notice the problem of the lost of ARP
>>> announce by discarding the message.
>>> So I must rew
On Wed, Jun 3, 2015 at 10:42 AM, Jason Wang wrote:
>
>
> On 06/03/2015 03:56 PM, Thibaut Collet wrote:
>> Hi,
>>
>> thanks for your point, I did not notice the problem of the lost of ARP
>> announce by discarding the message.
>> So I must rewrite my patch to define a queue to transmit the ARP
>> a
On Wed, Jun 03, 2015 at 09:56:57AM +0200, Thibaut Collet wrote:
> thanks for your point, I did not notice the problem of the lost of ARP
> announce by discarding the message.
> So I must rewrite my patch to define a queue to transmit the ARP announce
> to the guest. Is it the proper solution?
I'm
On 06/03/2015 03:56 PM, Thibaut Collet wrote:
> Hi,
>
> thanks for your point, I did not notice the problem of the lost of ARP
> announce by discarding the message.
> So I must rewrite my patch to define a queue to transmit the ARP
> announce to the guest. Is it the proper solution?
>
Please ha
On 06/01/2015 06:14 PM, Thibaut Collet wrote:
> Hi,
>
> >- vhost-user set received_disabled to true to prevent this from
> >happening. but looks like qemu_flush_or_purge_queue_packets() change it
> >to false unconditionally. And I don't quite understand this, probably we
> >can just remove this a
Hi,
thanks for your point, I did not notice the problem of the lost of ARP
announce by discarding the message.
So I must rewrite my patch to define a queue to transmit the ARP announce
to the guest. Is it the proper solution?
Thanks.
Best regards.
Thibaut.
On Tue, Jun 2, 2015 at 12:34 PM, Stef
On Fri, May 29, 2015 at 04:28:48PM +0200, Thibaut Collet wrote:
> I agree that virtio-net NIC never enqueues packet to vhost-user
> but qemu_announce_self function (savevm.c file) can do it through
> the qemu_announce_self_iter / qemu_send_packet_raw sequence.
Does vhost-user support live migratio
Hi,
>- vhost-user set received_disabled to true to prevent this from
>happening. but looks like qemu_flush_or_purge_queue_packets() change it
>to false unconditionally. And I don't quite understand this, probably we
>can just remove this and the issue is fixed?
I am afraid that solution can cause
On 05/28/2015 04:03 PM, Thibaut Collet wrote:
> For netdev backend with no receive callback, packets must not be enqueued.
> When
> VM stops the enqueued packets are purged (see fixes ca77d85e1dbf). That
> causes a
> qemu segfault due to a call of a NULL pointer function.
>
> Function to create
Hi,
I agree that virtio-net NIC never enqueues packet to vhost-user
but qemu_announce_self function (savevm.c file) can do it through
the qemu_announce_self_iter / qemu_send_packet_raw sequence.
Regards
Thibaut.
On Fri, May 29, 2015 at 3:12 PM, Stefan Hajnoczi wrote:
> On Thu, May 28, 2015 at
On Thu, May 28, 2015 at 10:03:14AM +0200, Thibaut Collet wrote:
> For netdev backend with no receive callback, packets must not be enqueued.
> When
> VM stops the enqueued packets are purged (see fixes ca77d85e1dbf). That
> causes a
> qemu segfault due to a call of a NULL pointer function.
virti
For netdev backend with no receive callback, packets must not be enqueued. When
VM stops the enqueued packets are purged (see fixes ca77d85e1dbf). That causes a
qemu segfault due to a call of a NULL pointer function.
Function to create net client is modified to allow backend to specify the size
of
12 matches
Mail list logo