Re: Kernel 4.19 network performance - forwarding/routing normal users traffic

[David Ahern <dsahern@gmail.com>]

On 11/9/18 9:21 AM, David Ahern wrote:
>> Is there possible to add only counters from xdp for vlans ?
>> This will help me in testing.
> I will take a look today at adding counters that you can dump using
> bpftool. It will be a temporary solution for this xdp program only.
> 

Same tree, kernel-tables-wip-02 branch. Compile kernel and install.
Compile samples as before.

If you give the userspace program a -t arg, it loop showing stats.
Ctrl-C to break. The xdp programs are not detached on exit.

Example:

./xdp_fwd -t 5 eth1 eth2 eth3 eth4

15:59:32:       rx         tx  dropped  skipped    l3_dev    fib_dev
index  3:   901158     901158        0       18         0          0
index  4:   901159     901158        0       20         0     901139
index 10:        0          0        0        0        19         19
index 11:        0          0        0        0    901139     901139
index 15:        0          0        0        0        19         19
index 16:        0          0        0        0    901139          0

Rx and Tx counters are for the physical port.

VLANs show up as l3_dev (ingress) and fib_dev (egress).

dropped is anytime the xdp program returns XDP_DROP (e.g., invalid packet)

skipped is anytime the program returns XDP_PASS (e.g., not ipv4 or ipv6,
local traffic, or needs full stack assist).