Re: [PATCH v3 bpf-next] cpumap: bulk skb using netif_receive_skb_list

On Thu, 22 Apr 2021 12:59:31 +0200
Lorenzo Bianconi [off-list ref] wrote:

quoted

[...]  

quoted

quoted

+   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?  

Ack, right. I think we should keep the struct layout in order to maintain
back-compatibility. I will fix it in v4.
 

quoted

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  

Running an old version of xdp_monitor with a patched kernel, I
verified the xdp sample does not crash but it reports wrong values
(e.g. pps are reported as drops for tracepoint disagliment).
I think we have two possibilities here:
- assuming tracepoints are not a stable ABI, we can just fix xdp
samples available in the kernel tree and provide a patch for
xdp-project

- keep current tracepoint layout and just rename current drop variable
in bpf/cpumap.c in something like skb_alloc_drop.

I like this one the most, rename to skb_alloc_drop, but keep layout.

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