Re: [lttng-dev] Can't receive a lot of functions that i registerd as trace-points

[Kienan Stewart via lttng-dev <lttng-dev@lists.lttng.org>]

Hi Lin,

Some of the event rules you have listed are for items that aren't 
instrumented. Others have the wrong name.

I've put some more specific indications below. You can check which 
events are available in the kernel domain with

`lttng list --kernel`

And for syscalls:

`lttng list --kernel --syscall`

There is also documentation available on adding LTTng probes for 
existing ftrace tracepoints at 
https://lttng.org/docs/v2.13/#doc-linux-add-lttng-layer (for 2.13) or 
https://lttng.org/docs/v2.12/#doc-linux-add-lttng-layer


> Hello,
> 
> I have a simple udp server/client and I want to trace these functions:
> 
>      Event rules:
>        netif_rx_ni_entry (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        netif_rx_entry (loglevel: TRACE_EMERG (0)) (type: tracepoint) 

In LTTng, these events start with "net_if_", not "netif_"

> [enabled]
>        napi_gro_receive_entry (loglevel: TRACE_EMERG (0)) (type: 
> tracepoint) [enabled]

Try "net_napi_gro_receive_entry" instead.

>        softirq_raise (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        irq_handler_entry (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        napi_poll (loglevel: TRACE_EMERG (0)) (type: tracepoint) [enabled]
>        softirq_entry (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        e1000_xmit_frame (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        ndo_start_xmit (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]

lttng-modules doesn't provide any specific instrumentation for the e1000 
and ndo drivers. Are you using additional kernel modules that add the 
instrumentation?

If you are, please ensure that the module is loaded, or it can be added 
to the list of probes loaded by the sessiond using `--extra-kmod-probes` 
(c.f. 
https://lttng.org/man/8/lttng-sessiond/v2.13/#doc-opt--extra-kmod-probes)

If not, you will need to implement the probes as mentioned earlier. If 
you need probes within a specific timeline, either patches and/or
financial support are welcome to create and maintain them. If you're
interested in funding this, please feel free to reach out to EfficiOS.

>        net_dev_start_xmit (loglevel: TRACE_EMERG (0)) (type: tracepoint) > [enabled]

This might be "net_dev_xmit" instead of "net_dev_start_xmit"

>        dev_queue_xmit (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        net_dev_queue (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]
>        netif_receive_skb_list_entry (loglevel: TRACE_EMERG (0)) (type: 
> tracepoint) [enabled]
>        netif_receive_skb (loglevel: TRACE_EMERG (0)) (type: tracepoint) 
> [enabled]

These are also prefixed with "net_if_" instead of "netif_".

>        inet_sock_set_state (loglevel: TRACE_EMERG (0)) (type: 
> tracepoint) [enabled]

I don't think this is instrumented in lttng-modules at this time.

> 
> I start tracing, and when it is finished, I will open my trace directory 
> into TraceCompass, But I can see just some traces from these functions:
> 
> - irq_handler_entry
> - napi_poll
> - net_dev_queue
> 
> I'm using  2.12.14.
> 
> Do you have any idea?
> 
> Regards,
> 

Hope this helps.

thanks,
kienan

> _______________________________________________
> lttng-dev mailing list
> lttng-dev@lists.lttng.org
> https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev