[MPTCP] Re: [MPTCP][PATCH mptcp-next 2/3] mptcp: add tracepoint for get_mapping_status

Hello, On Mon, 2021-02-22 at 15:30 +0800, Geliang Tang wrote: > This patch added the tracepoint for the mapping status function > get_mapping_status. > > Signed-off-by: Geliang Tang <geliangtang(a)gmail.com&gt; > --- > include/trace/events/mptcp.h | 51 ++++++++++++++++++++++++++++++++++++ > net/mptcp/subflow.c | 13 +++------ > 2 files changed, 54 insertions(+), 10 deletions(-) > > diff --git a/include/trace/events/mptcp.h b/include/trace/events/mptcp.h > index b36b308f48e2..f06a7d33b237 100644 > --- a/include/trace/events/mptcp.h > +++ b/include/trace/events/mptcp.h > @@ -39,6 +39,57 @@ TRACE_EVENT(mptcp_subflow_get_send, > __entry->ssk_1, __entry->ratio_1) > ); > > +TRACE_EVENT(get_mapping_status, > + > + TP_PROTO(struct mptcp_ext *mpext), > + > + TP_ARGS(mpext), > + > + TP_STRUCT__entry( > + __array(char, str, 128) > + ), > + > + TP_fast_assign( > + memset(__entry->str, 0, 128); > + > + if (!mpext) { > + sprintf(__entry->str, "mpext is NULL"); > + } else { > + if (!mpext->data_len) { > + sprintf(__entry->str, "Infinite mapping not handled"); > + } else { > + if (mpext->data_fin == 1) { > + if (mpext->data_len == 1) { > + sprintf(__entry->str, > + "data_fin: with no payload seq=%llu", > + mpext->data_seq); > + } else { > + sprintf(__entry->str, > + "data_fin: with mapping seq=%llu dsn64=%d", > + mpext->data_seq + mpext->data_len - 1, > + mpext->dsn64); > + } > + } > + > + if (mpext->use_map) { > + sprintf(__entry->str + strlen(__entry->str), > + "use_map: dsn64=%d data_seq=%llu subflow_seq=%u data_len=%u", > + mpext->dsn64, mpext->data_seq, > + mpext->subflow_seq, mpext->data_len); > + } > + > + if (mpext->use_ack) { > + sprintf(__entry->str + strlen(__entry->str), > + "use_ack: ack64=%d data_ack=%llu", > + mpext->ack64, mpext->data_ack); > + } > + } > + } > + ), I'm sorry for the very late feedback. I think the tracepoint location is good, but the tracepoint hooks contains a bit too-much mptcp-related logic. What dumping every mpext field? Side note: we possibly would need another trace point in ack_update_msk() to track the incoming data_ack and window/snd_una updates