On Mon, Aug 29, 2016 at 03:28:34PM +0200, Martin Pieuchot wrote:
> Comments? Oks?
OK bluhm@
> -if_input_process(void *xmq)
> +if_input_process(void *xifidx)
> {
> - struct mbuf_queue *mq = xmq;
> + unsigned int ifidx = (unsigned long)xifidx;
> struct mbuf_list ml;
> - struct m
Unlocking the forwarding path will allow us to gain in latency because
the rest of the kernel will be able to run on another CPU while an
incoming packet is being processed.
However to make use of SMP to improve forwarding performances we will
have to run multiple instances of the incoming path in
