Josh Steadmon <steadmon@xxxxxxxxxx> writes: >> ... understandable if there weren't any suitable mechanism to simply log >> "the control passed at this spot at this time" kind of event in the >> trace2 subsystem, but I do not think it is the case. > > Ack, changed this to a "trace2_printf()" instead. Annoyingly the JSON > Event trace2 target that we use at $DAYJOB doesn't log these events, but > I can add another patch to enable that. Ahh, OK, I was concentrating solely on the producing side, and forgot to consider that the consuming side may not be prepared for non enter/leave pair of events. That's understandable, but if you are updating the consuming side to be able to do so, that would be even better. > Yeah, thanks, this did need to be reworked. I pushed the regions down > into pack_objects() and receive_status(), which look like the only two > places we might spend much time. Sounds good. This is a tangent, but I doubt we have many users without sideband support. In the longer term we may be able to drop the non-sideband codepath, which would automatically simplify the flow quite a bit around here. But that is totally outside of this topic. Thanks.
