does the kernel patch set you sent to kernel-team@ yesterday help the testcase you put in comment #22?
Someone just pointed out he got this
[54545.873460] unregister_netdevice: waiting for lo to become free. Usage count = 1
in a 3.8 kernel when using macvlan in a container.
It seems to me that the testcase in comment #22, plus one or two more (starting veth and macvlan containers and checking dmesg), at this point belong in a kernel regression testsuite, bc it seems to keep coming back. I don't know which testsuite would be best for this.
@arges
does the kernel patch set you sent to kernel-team@ yesterday help the testcase you put in comment #22?
Someone just pointed out he got this
[54545.873460] unregister_ netdevice: waiting for lo to become free. Usage count = 1
in a 3.8 kernel when using macvlan in a container.
It seems to me that the testcase in comment #22, plus one or two more (starting veth and macvlan containers and checking dmesg), at this point belong in a kernel regression testsuite, bc it seems to keep coming back. I don't know which testsuite would be best for this.