On 5/3/2020 2:06 PM, Florian Fainelli wrote:

Le 2020-05-01 à 10:58, Allen a écrit :

quoted

quoted

quoted

  It maps to "eth0". Please let me know if you need further details.

I suppose I should have been clearer, what network device driver created
eth0?

 This was seen on a VM.
eth0 [52:54:00:c1:cd:65]: virtio_net (up)

I have reproduced it here with virtio_net and am now looking into this,
at first glance it does not look like we are properly holding the device
reference count for the case where DSA was probed via platform device
configuration.

There is a DSA master reference counting issue, but with dsa-loop, the
DSA master is already properly reference counted thanks to the
dev_get_by_name() call, I will keep digging.
-- 
Florian