On 3/18/25 3:07 AM, Wang Liang wrote:
quoted
quoted
It seems to me that checking master->l3mdev_ops like this is racy.
vrf device leaves the l3mdev ops set; that is probably the better way
to go.
Thanks.
Only l3s ipvlan set the dev->l3mdev_ops to NULL at present, I will delete
'dev->l3mdev_ops = NULL' in ipvlan_l3s_unregister(), is that ok?
I think that avoids the race you saw. vrf has had the ops for 9 years or
so and not seen that problem.