question about traffic control in for input (ingress) traffic

question about traffic control in for input (ingress) traffic

[Kevin Wilson]

Hello,
I know that traffic control is usually used for output.
I want to learn  traffic control kernel implementation for input
(ingress) traffic
I made several tries with the tc userspace tool to add
traffic control entries, but I did not succeed.
What I mean is that in the kernel, in the ingress filter method
ing_filter(), I always reach with noop_qdisc, no matter what I try
with the  tc tool:

Any ideas?  can I register the route scheduler for the Rx path ?

In http://lxr.free-electrons.com/source/net/core/dev.c
 static int ing_filter(struct sk_buff *skb, struct netdev_queue *rxq)
 {
  ...
  struct Qdisc *q;


         q = rxq->qdisc;
         if (q != &noop_qdisc) {
         }

         return result;
 }

regards,
Kevin

Re: question about traffic control in for input (ingress) traffic

[Andy Furniss]

Kevin Wilson wrote:
> Hello,
> I know that traffic control is usually used for output.
> I want to learn  traffic control kernel implementation for input
> (ingress) traffic
> I made several tries with the tc userspace tool to add
> traffic control entries, but I did not succeed.

You can only add filters/policers to ingress, if you want to shape you 
have to use the ifb device and redirect traffic to it with a filter.

eg.

modprobe ifb
tc qdisc add dev eth0 ingress
ip link set ifb0 up

tc filter add dev eth0 parent ffff: \
protocol ip prio 10 u32 match u32 0 0 flowid 1:0 \
action mirred egress redirect dev ifb0

Then you add normal qdiscs etc to ifb0 just like you would for output.