> > > + TP_PROTO(struct ib_mad_send_wr_private *wr, struct > > rdma_mad_trace_addr *addr), > > > + TP_ARGS(wr, addr), > > > + > > > + TP_STRUCT__entry( > > > + __field(u8, base_version) > > > + __field(u8, mgmt_class) > > > + __field(u8, class_version) > > > + __field(u8, port_num) > > > + __field(u32, qp_num) > > > + __field(u8, method) > > > + __field(u8, sl) > > > + __field(u16, attr_id) > > > + __field(u32, attr_mod) > > > + __field(u64, wrtid) > > > + __field(u64, tid) > > > + __field(u16, status) > > > + __field(u16, class_specific) > > > + __field(u32, length) > > > + __field(u32, dlid) > > > + __field(u32, rqpn) > > > + __string(dev_name, wr->mad_agent_priv->agent.device- > > >name) > > > + __field(void *, agent_priv) > > > + __field(unsigned long, timeout) > > > + __field(int, retries_left) > > > + __field(int, max_retries) > > > + __field(int, retry) > > > + __field(u32, rqkey) > > > + __field(u16, pkey) > > > + ), > > > > Does it really make sense to extract all this data? Can we just trace the bulk > > packet header? I have no idea what convention is in tracing.. > A bulk header would preclude filtering or triggering on the fields in the bulk portion. That is one of the most useful aspects of the static tracing. Mike
