On Mon, 14 Oct 2002, Stefan Rompf wrote:
> Alexey, Jamal,
>
> the first version of this patch just added the feature to distribute
> state changes created by netif_carrier_on()/_off() inside the kernel and
> to netlink via netdev_state_change(). This is still the part most
> important to me - not only in 2.5, but also as a 2.4 backport.
>
> I don't think RFC2863 state keeping makes much sense if we cannot
> forward the results to userspace. So if you are happy with keeping the
> current semantics, but using the more sophisticated implementation of
> forwarding worked out last weekend, I'll be happy to provide a new patch
> (also adressing Jamal's nitpicks). Just agree on one way.
the RFC2863 semantics should stay; the thing we havent agreed on is
how to do it. Maybe for the first phase patch you dont need to expose them
to user space - maybe in the way Alexey suggested as attributes without
loosing any of the flag bits...
Alexey should make that call.
cheers,
jamal
|