[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9ac8402e-fd27-519c-0c1c-d1b306b3441b@lab.ntt.co.jp>
Date: Fri, 24 May 2019 10:33:47 +0900
From: Toshiaki Makita <makita.toshiaki@....ntt.co.jp>
To: Jesper Dangaard Brouer <brouer@...hat.com>
Cc: Alexei Starovoitov <ast@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>,
"David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <jakub.kicinski@...ronome.com>,
Jesper Dangaard Brouer <hawk@...nel.org>,
John Fastabend <john.fastabend@...il.com>,
netdev@...r.kernel.org, xdp-newbies@...r.kernel.org,
bpf@...r.kernel.org
Subject: Re: [PATCH bpf-next 2/3] xdp: Add tracepoint for bulk XDP_TX
On 2019/05/23 22:12, Jesper Dangaard Brouer wrote:
> On Thu, 23 May 2019 19:56:47 +0900
> Toshiaki Makita <makita.toshiaki@....ntt.co.jp> wrote:
>
>> This is introduced for admins to check what is happening on XDP_TX when
>> bulk XDP_TX is in use.
>>
>> Signed-off-by: Toshiaki Makita <makita.toshiaki@....ntt.co.jp>
>> ---
>> include/trace/events/xdp.h | 25 +++++++++++++++++++++++++
>> kernel/bpf/core.c | 1 +
>> 2 files changed, 26 insertions(+)
>>
>> diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h
>> index e95cb86..e06ea65 100644
>> --- a/include/trace/events/xdp.h
>> +++ b/include/trace/events/xdp.h
>> @@ -50,6 +50,31 @@
>> __entry->ifindex)
>> );
>>
>> +TRACE_EVENT(xdp_bulk_tx,
>> +
>
> You are using this tracepoint like/instead of trace_xdp_devmap_xmit if
> I understand correctly? Or maybe the trace_xdp_redirect tracepoint.
Yes, I have trace_xdp_devmap_xmit in mind, which is for XDP_REDIRECT.
> The point is that is will be good if the tracepoints can share the
> TP_STRUCT layout beginning, as it allows for attaching and reusing eBPF
> code that is only interested in the top part of the struct.
It's good, but this tracepoint does not have map concept so differs from
xdp_devmap_xmit.
> I would also want to see some identifier, that trace programs can use
> to group and corrolate these events, you do have ifindex, but most
> other XDP tracepoints also have "prog_id".
I have considered that too. The problem is that we cannot pass a
reliable prog_id since bulk xmit happens after RCU critical section of
XDP_TX.
xdp_devmap_xmit does not have prog_id and I guess there is a similar
reason for it?
--
Toshiaki Makita
Powered by blists - more mailing lists