[lttng-dev] Understand userspace events in babeltrace output

[Zvi Vered]

Hi Jonathan,

Thank you very much !

Best regards,
Zvika

On Fri, Jul 19, 2019 at 11:48 PM Jonathan Rajotte-Julien <
jonathan.rajotte-julien at efficios.com> wrote:

> Hi Zvi,
>
> On Fri, Jul 19, 2019 at 05:28:48PM +0300, Zvi Vered wrote:
> > Hello,
> >
> > Using babeltrace I created a text trace.
> > How can I understand it ?
>
> Here lies the heavy work!
>
> >
> > For example:
> > I logged userspace events. One of my threads is called: RSPOn3
>
> As far as I can tell sched_* are not userspace events they are
> "kernelspace"
> events.
>
> > I can see the following lines:
> >
> > sched_switch: { cpu_id = 0 }, { prev_comm = "RSPOn3", prev_tid = 27307,
> > prev_prio = -58, prev_state = 1, next_comm = "swapper/0", next_tid = 0,
> > next_prio = 20 }
> >
> > sched_wakeup: { cpu_id = 0 }, { comm = "RSPOn3", tid = 27307, prio = -58,
> > success = 1, target_cpu = 0 }
> >
> > sched_switch: { cpu_id = 0 }, { prev_comm = "swapper/0", prev_tid = 0,
> > prev_prio = 20, prev_state = 0, next_comm = "RSPOn3", next_tid = 27307,
> > next_prio = -58 }
> >
> > sched_switch: { cpu_id = 0 }, { prev_comm = "RSPOn3", prev_tid = 27307,
> > prev_prio = -58, prev_state = 1, next_comm = "swapper/0", next_tid = 0,
> > next_prio = 20 }
> >
> > How can I know the meaning of each field in each line ?
>
> You will need to understand the kernel state transition of the scheduler.
> There
> is multiple ways to go about it.
>
> You could use TraceCompass[1][2] or lttng-analyses[3] to ease/complement
> your
> learning experience a bit.
>
> You could dive in the kernel source code and understand the control flow
> surrounding each events and understand the significance of the events
> payload.
>
> Another interesting place to look is the kernel trace documentation [4].
> Some
> subsystems have clear explanation regarding each tracepoint [5].
>
> > 1. sched_switch, sched_wakeup
> > 2. prev_comm
> > 3. prev_state
> > 4. next_comm
> >
> > If a task is blocked (e.g waiting for TCP) and then runs for some time
> and
> > then waits again, how this will be logged ?
>
> Babeltrace is a trace converter, it does not perform any analysis.
> TraceCompass
> and lttng-analyses might have analysis fitting your needs.
>
> Cheers
>
> [1] https://www.eclipse.org/tracecompass/
> [2]
> https://archive.eclipse.org/tracecompass/doc/stable/org.eclipse.tracecompass.doc.user/LTTng-Kernel-Analysis.html#LTTng_Kernel_Analysis
> [3] https://github.com/lttng/lttng-analyses
> [4] https://www.kernel.org/doc/Documentation/trace/
> [5] https://www.kernel.org/doc/Documentation/trace/events-kmem.txt
>
> --
> Jonathan Rajotte-Julien
> EfficiOS
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.lttng.org/pipermail/lttng-dev/attachments/20190720/09e49185/attachment.html>