On Tue, 20 Apr 2021 16:05:14 +0200 Lorenzo Bianconi <lorenzo@xxxxxxxxxx> wrote: > Rely on netif_receive_skb_list routine to send skbs converted from > xdp_frames in cpu_map_kthread_run in order to improve i-cache usage. > The proposed patch has been tested running xdp_redirect_cpu bpf sample > available in the kernel tree that is used to redirect UDP frames from > ixgbe driver to a cpumap entry and then to the networking stack. > UDP frames are generated using pkt_gen. Packets are discarded by the > UDP layer. > > $xdp_redirect_cpu --cpu <cpu> --progname xdp_cpu_map0 --dev <eth> > > bpf-next: ~2.35Mpps > bpf-next + cpumap skb-list: ~2.72Mpps > > Since netif_receive_skb_list does not return number of discarded packets, > remove drop counter from xdp_cpumap_kthread tracepoint and update related > xdp samples. > > Signed-off-by: Lorenzo Bianconi <lorenzo@xxxxxxxxxx> > --- > Changes since v2: > - remove drop counter and update related xdp samples > - rebased on top of bpf-next > > Changes since v1: > - fixed comment > - rebased on top of bpf-next tree > --- > include/trace/events/xdp.h | 14 +++++--------- > kernel/bpf/cpumap.c | 16 +++++++--------- > samples/bpf/xdp_monitor_kern.c | 6 ++---- > samples/bpf/xdp_monitor_user.c | 14 ++++++-------- > samples/bpf/xdp_redirect_cpu_kern.c | 12 +++++------- > samples/bpf/xdp_redirect_cpu_user.c | 10 ++++------ > 6 files changed, 29 insertions(+), 43 deletions(-) > > diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h > index fcad3645a70b..52ecfe9c7e25 100644 > --- a/include/trace/events/xdp.h > +++ b/include/trace/events/xdp.h > @@ -184,16 +184,15 @@ DEFINE_EVENT(xdp_redirect_template, xdp_redirect_map_err, > > TRACE_EVENT(xdp_cpumap_kthread, > > - TP_PROTO(int map_id, unsigned int processed, unsigned int drops, > - int sched, struct xdp_cpumap_stats *xdp_stats), > + TP_PROTO(int map_id, unsigned int processed, int sched, > + struct xdp_cpumap_stats *xdp_stats), > > - TP_ARGS(map_id, processed, drops, sched, xdp_stats), > + TP_ARGS(map_id, processed, sched, xdp_stats), > > TP_STRUCT__entry( > __field(int, map_id) > __field(u32, act) > __field(int, cpu) > - __field(unsigned int, drops) > __field(unsigned int, processed) So, struct member @processed will takeover the room for @drops. Can you please test how an old xdp_monitor program will react to this? Will it fail, or extract and show wrong values? The xdp_mointor tool is in several external git repos: https://github.com/netoptimizer/prototype-kernel/blob/master/kernel/samples/bpf/xdp_monitor_kern.c https://github.com/xdp-project/xdp-tutorial/tree/master/tracing02-xdp-monitor Do you have any plans for fixing those tools? > __field(int, sched) > __field(unsigned int, xdp_pass) > @@ -205,7 +204,6 @@ TRACE_EVENT(xdp_cpumap_kthread, > __entry->map_id = map_id; > __entry->act = XDP_REDIRECT; > __entry->cpu = smp_processor_id(); > - __entry->drops = drops; > __entry->processed = processed; > __entry->sched = sched; > __entry->xdp_pass = xdp_stats->pass; > @@ -215,13 +213,11 @@ TRACE_EVENT(xdp_cpumap_kthread, > > TP_printk("kthread" > " cpu=%d map_id=%d action=%s" > - " processed=%u drops=%u" > - " sched=%d" > + " processed=%u sched=%u" > " xdp_pass=%u xdp_drop=%u xdp_redirect=%u", > __entry->cpu, __entry->map_id, > __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB), > - __entry->processed, __entry->drops, > - __entry->sched, > + __entry->processed, __entry->sched, > __entry->xdp_pass, __entry->xdp_drop, __entry->xdp_redirect) > ); -- Best regards, Jesper Dangaard Brouer MSc.CS, Principal Kernel Engineer at Red Hat LinkedIn: http://www.linkedin.com/in/brouer
