Hi Tim,
> Looking for feedback on this quickie patch. This enables netlink to
> deliver link-change events, as reported by drivers via
> netif_carrier_{on,off}.
I've taken a closer look at the approach and like it more than my own
implementation. What you are currently missing is some code to assure
that a device does not have an event queued after removal - look what my
patch does in netdev_unregister(). Patching f.e. the vlan driver to do
some netif_carrier_on/off in the stop method is a good test case. Also,
do not forget to call the network notifier chain beside emitting a
netlink message - the kernel isinterested in the event, too.
IMHO we should merge the patches, taking your one workqueue per device
approach and my usage of IFF_RUNNING as a mirror bit. What do other
netdev people think?
Cheers, Stefan
|