[dpdk-dev] [PATCH v4 00/44] net/virtio: Virtio PMD rework

[Maxime Coquelin]

On 2/1/21 2:42 PM, Ilya Maximets wrote:
> On 2/1/21 2:16 PM, Maxime Coquelin wrote:
>> Hi Ilya,
>>
>> On 2/1/21 2:03 PM, Ilya Maximets wrote:
>>> CC: ovs-dev
>>>
>>> On 2/1/21 2:00 PM, Ilya Maximets wrote:
>>>> On 1/26/21 11:15 AM, Maxime Coquelin wrote:
>>>>>
>>>>> Only functionnal change in this second part is making the
>>>>> Vhost-user server mode blocking at init time, as long as
>>>>> a client is not connected. The goal of this change is to
>>>>> make the Vhost-user support much more robust, as without
>>>>> blocking, the driver has to assume features that are going
>>>>> to be supported by the client, which is very fragile and
>>>>> error prone. As a side-effect, it also simplifies the
>>>>> logic nin several place of the virtio-user PMD.
>>>>
>>>> Hi, Maxime.
>>>>
>>>> I didn't actually look at the code, but I have a question.
>>>> Does above text mean that with this change OVS will hang inside
>>>> driver_register() or similar function until client is connected
>>>> to dpdkvhostuser (server mode) port?
>>>>
>>>> If so, I think, we will not be able to support server mode
>>>> in OVS anymore and will have to actually remove it.  It's
>>>> deprecated for a long time now, but I think it might still be in
>>>> use by some people, especially for virtio-user usecase.
>>
>> No, there is no impact for OVS. My explanation was maybe a bit
>> confusion, sorry about that.
>>
>> I meant that Virtio-user PMD, acting as Vhost-user master in server
>> mode, will be blocking waiting for client (OVS dpdkvhostuserclient in
>> your case) connection. This makes the Virtio-user PMD in server mode to
>> behave the same as QEMU Vhost-user port in server mode.
> 
> Oh.  OK.  So the only affected party is 'net_virtio_user' PMD and vhost
> library is not affected, right?

Right!

> And in this case the following command will block OVS:
> 
>   ovs-vsctl add-port br0 vu -- set Interface vu type=dpdk \
>       options:dpdk-devargs="net_virtio_user,path=vu.sock,server=1"
> 
> Is that correct?
> 
> (I'm not sure if this works right now, though.  I didn't test it.)

This is correct, even though I don't know a use-case for it (chained
OVS-DPDK?).

Thanks,
Maxime
> 
>>
>> Only noticeable effect on OVS side should be more reliability, as
>> without this change, Virtio-user PMD could assume OVS Vhost-user backend
>> would support features before the connection happens.
>>
>>>> Best regards, Ilya Maximets.
>>>>
>>>
>>
>> Regards,
>> Maxime
>>
>