Re: [PATCHv2 bpf-next] samples/bpf: add xdp program on egress for xdp_redirect_map

On Thu, Nov 26, 2020 at 11:51:19AM +0100, Jesper Dangaard Brouer wrote:

On Thu, 26 Nov 2020 16:43:25 +0800
Hangbin Liu [off-list ref] wrote:

quoted

Current sample test xdp_redirect_map only count pkts on ingress. But we
can't know whether the pkts are redirected or dropped. So add a counter
on egress interface so we could know how many pkts are redirect in fact.

This is not true.

The 2nd devmap XDP-prog will run in the same RX-context, so it doesn't
tell us if the redirect was successful.  I looked up the code, and the
2nd XDP-prog is even allowed to run when the egress driver doesn't
support the NDO to xmit (dev->netdev_ops->ndo_xdp_xmit), which is very
misleading, if you place a output counter here.

-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer


static inline int __xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
			       struct net_device *dev_rx)
{
	struct xdp_frame *xdpf;
	int err;

	if (!dev->netdev_ops->ndo_xdp_xmit)
		return -EOPNOTSUPP;

Err... You are right. I didn't read this part carefully.

Based on Toke's suggestion, maybe edit the src mac address to the egress
iface's mac address is a better example. WDYT?

And does it make any sense to run perf test for 2nd xdp program
for xdp_redirect_map?

Thanks
Hangbin

	err = xdp_ok_fwd_dev(dev, xdp->data_end - xdp->data);
	if (unlikely(err))
		return err;

	xdpf = xdp_convert_buff_to_frame(xdp);
	if (unlikely(!xdpf))
		return -EOVERFLOW;

	bq_enqueue(dev, xdpf, dev_rx);
	return 0;
}


int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
		    struct net_device *dev_rx)
{
	struct net_device *dev = dst->dev;

	if (dst->xdp_prog) {
		xdp = dev_map_run_prog(dev, xdp, dst->xdp_prog);
		if (!xdp)
			return 0;
	}
	return __xdp_enqueue(dev, xdp, dev_rx);
}