Re: [PATCHv17 bpf-next 1/6] bpf: run devmap xdp_prog on flush instead of bulk enqueue
From: Jesper Dangaard Brouer <hidden>
Date: 2021-01-27 16:01:38
Also in:
bpf
On Mon, 25 Jan 2021 20:45:11 +0800 Hangbin Liu [off-list ref] wrote:
By using xdp_redirect_map in sample/bpf and send pkts via pktgen cmd: ./pktgen_sample03_burst_single_flow.sh -i eno1 -d $dst_ip -m $dst_mac -t 10 -s 64 There are about +/- 0.1M deviation for native testing, the performance improved for the base-case, but some drop back with xdp devmap prog attached. Version | Test | Generic | Native | Native + 2nd xdp_prog 5.10 rc6 | xdp_redirect_map i40e->i40e | 2.0M | 9.1M | 8.0M 5.10 rc6 | xdp_redirect_map i40e->veth | 1.7M | 11.0M | 9.7M 5.10 rc6 + patch | xdp_redirect_map i40e->i40e | 2.0M | 9.5M | 7.5M 5.10 rc6 + patch | xdp_redirect_map i40e->veth | 1.7M | 11.6M | 9.1M
I want to highlight the improvement 9.1M -> 9.5M. This is the native (40e->i40e) case where the isn't any "2nd xdp_prog". This means that when we introduced the "2nd xdp_prog", we lost a little performance without noticing (death-by-a-1000-paper-cuts), for the baseline case where this feature is not used/activated. This patch regains that performance for our baseline. That in itself make this patch worth it. -- Best regards, Jesper Dangaard Brouer MSc.CS, Principal Kernel Engineer at Red Hat LinkedIn: http://www.linkedin.com/in/brouer