Tracing Summit 2017

Here is the content of my presentation at the Tracing Summit 2017.

Aside from the presentation, you will find an example demonstrating an example of a Babeltrace 2 component implemented in Python. The component is imaginatively named sink.demo.stacktrace.

Using sink.demo.stacktrace

Building the software

In order to retrieve the function names from the addresses recorded in the LTTng trace, the binary must contain DWARF debug information and the compiler must instrument the various functions' entry and exit points.

Make sure to build your binary with the following CFLAGS:

CFLAGS="-finstrument-functions -g"

While the component will work even if the binary has no DWARF information, it will only be able to display function names which are exported (public symbols).

Tracing the software

This component depends on two events produced by LTTng-ust:

• lttng_ust_cyg_profile_fast:func_entry
• lttng_ust_cyg_profile_fast:func_exit

LTTng provides the liblttng-ust-cygprofile-fast.so shared object utility in order to trace these events on the functions' entry and exit. For more information, see the LTTng official documentation which has a dedicated section on this mechanism.

Setup a tracing session

$ lttng create my_tracing_session
$ lttng enable-event --userspace --all
$ lttng add-context --userspace --type ip --type vpid
$ lttng start

Launch the program you want to trace

We will launch the ls utility, built with the debug information enabled.

LD_PRELOAD="liblttng-ust-cyg-profile-fast.so" ls

Stop and destroy the tracing session

$ lttng stop
$ lttng destroy

Displaying a callstack

In order for babeltrace to see our new component, we need to make it aware of the path in which it is saved. The --plugin-path option is used to add a path to babeltrace's search path.

Let's list the plugins found by babeltrace and make sure we see our demo component.

$ babeltrace list-plugins --plugin-path ~/EfficiOS/confs/Tracing.Summit.2017/

Now we simply need to instruct babeltrace to

• Look for our plug-in in the appropriate path
• Insert a filter.lttng-utils.debug-info filter in the graph so that the function name, source filename and line number are added to the trace
• Use our new component, sink.demo.stacktrace as a sink

$ babeltrace --debug-info --plugin-path ~/EfficiOS/confs/Tracing.Summit.2017/ \
  --component sink.demo.stacktrace ~/lttng-traces/stacktrace-20171025-170540/