Re: [PATCH 1/4] virtio_net: notify MAC address change on device initialization

[Eli Cohen <elic@nvidia.com>]

On 23/01/2023 11:52, Laurent Vivier wrote:
> On 1/22/23 14:47, Eli Cohen wrote:
>>
>> On 22/01/2023 12:05, Laurent Vivier wrote:
>>> In virtnet_probe(), if the device doesn't provide a MAC address the
>>> driver assigns a random one.
>>> As we modify the MAC address we need to notify the device to allow it
>>> to update all the related information.
>>>
>>> The problem can be seen with vDPA and mlx5_vdpa driver as it doesn't
>>> assign a MAC address by default. The virtio_net device uses a random
>>> MAC address (we can see it with "ip link"), but we can't ping a net
>>> namespace from another one using the virtio-vdpa device because the
>>> new MAC address has not been provided to the hardware.
>>>
>>> Signed-off-by: Laurent Vivier <lvivier@redhat.com>
>>> ---
>>>   drivers/net/virtio_net.c | 14 ++++++++++++++
>>>   1 file changed, 14 insertions(+)
>>>
>>> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
>>> index 7723b2a49d8e..25511a86590e 100644
>>> --- a/drivers/net/virtio_net.c
>>> +++ b/drivers/net/virtio_net.c
>>> @@ -3800,6 +3800,8 @@ static int virtnet_probe(struct virtio_device 
>>> *vdev)
>>>           eth_hw_addr_set(dev, addr);
>>>       } else {
>>>           eth_hw_addr_random(dev);
>>> +        dev_info(&vdev->dev, "Assigned random MAC address %pM\n",
>>> +             dev->dev_addr);
>>>       }
>>>       /* Set up our device-specific information */
>>> @@ -3956,6 +3958,18 @@ static int virtnet_probe(struct virtio_device 
>>> *vdev)
>>>       pr_debug("virtnet: registered device %s with %d RX and TX 
>>> vq's\n",
>>>            dev->name, max_queue_pairs);
>>> +    /* a random MAC address has been assigned, notify the device */
>>> +    if (dev->addr_assign_type == NET_ADDR_RANDOM &&
>> Maybe it's better to not count on addr_assign_type and use a local 
>> variable to indicate that virtnet_probe assigned random MAC. The 
>> reason is that the hardware driver might have done that as well and 
>> does not need notification.
>
> eth_hw_addr_random() sets explicitly NET_ADDR_RANDOM, while 
> eth_hw_addr_set() doesn't change addr_assign_type so it doesn't seem 
> this value is set by the hardware driver.
>
> So I guess it's the default value (NET_ADDR_PERM) in this case (even 
> if it's a random address from the point of view of the hardware).
>
> If you prefer I can replace it by "!virtio_has_feature(vdev, 
> VIRTIO_NET_F_MAC)"?
>
My point is this. What if the hardware driver decides to choose a random 
address by calling eth_hw_addr_random(). In this case 
dev->addr_assign_type = NET_ADDR_RANDOM is executed unconditionally. But 
now you will needlessly execute the control command because the hardware 
address already knows about this address.


> Thanks,
> Laurent
>