1 The LTTng Documentation
2 =======================
3 Philippe Proulx <pproulx@efficios.com>
7 include::../common/copyright.txt[]
10 include::../common/warning-not-maintained.txt[]
13 include::../common/welcome.txt[]
16 include::../common/audience.txt[]
20 === What's in this documentation?
22 The LTTng Documentation is divided into the following sections:
24 * **<<nuts-and-bolts,Nuts and bolts>>** explains the
25 rudiments of software tracing and the rationale behind the
28 You can skip this section if you’re familiar with software tracing and
29 with the LTTng project.
31 * **<<installing-lttng,Installation>>** describes the steps to
32 install the LTTng packages on common Linux distributions and from
35 You can skip this section if you already properly installed LTTng on
38 * **<<getting-started,Quick start>>** is a concise guide to
39 getting started quickly with LTTng kernel and user space tracing.
41 We recommend this section if you're new to LTTng or to software tracing
44 You can skip this section if you're not new to LTTng.
46 * **<<core-concepts,Core concepts>>** explains the concepts at
49 It's a good idea to become familiar with the core concepts
50 before attempting to use the toolkit.
52 * **<<plumbing,Components of LTTng>>** describes the various components
53 of the LTTng machinery, like the daemons, the libraries, and the
54 command-line interface.
55 * **<<instrumenting,Instrumentation>>** shows different ways to
56 instrument user applications and the Linux kernel.
58 Instrumenting source code is essential to provide a meaningful
61 You can skip this section if you do not have a programming background.
63 * **<<controlling-tracing,Tracing control>>** is divided into topics
64 which demonstrate how to use the vast array of features that
65 LTTng{nbsp}{revision} offers.
66 * **<<reference,Reference>>** contains reference links and tables.
67 * **<<glossary,Glossary>>** is a specialized dictionary of terms related
68 to LTTng or to the field of software tracing.
71 include::../common/convention.txt[]
74 include::../common/acknowledgements.txt[]
78 == What's new in LTTng {revision}?
80 * **Tracing control**:
81 ** Dynamic filter support for <<event,event rules>> in the Linux kernel
82 <<domain,tracing domain>>. For example:
87 lttng enable-event --kernel irq_handler_entry --filter='irq == 28'
91 ** Wildcard support in the instrumentation point name of an event rule
92 in the Linux kernel tracing domain. For example:
97 lttng enable-event --kernel 'sched_*'
101 ** New `lttng track` and `lttng untrack` commands to make
102 <<pid-tracking,PID tracking>> super-fast for both the Linux kernel
103 and the user space tracing domains.
105 When LTTng _tracks_ one or more PIDs, only the processes having those PIDs
106 can emit events for a given tracing session.
108 ** New `--shm-path` option of the `lttng create` command to specify the
109 path where LTTng creates the shared memory holding the ring buffers.
111 This feature is useful when used with persistent memory file systems to
112 extract the latest recorded trace data in the event of a crash requiring
115 The new man:lttng-crash(1) command-line utility can extract trace data
116 from such a file (see <<persistent-memory-file-systems,Record trace data
117 on persistent memory file systems>>).
119 * **User space tracing**:
120 ** New <<python-application,LTTng-UST Python agent>> which makes it easy
121 to trace existing Python applications that are using the standard
122 https://docs.python.org/3/howto/logging.html[`logging` package].
124 This agent is compatible with both the Python 2 and Python 3 languages.
126 ** New <<tracelog,`tracelog()`>> facility to ease the migration from
129 `tracelog()` is similar to <<tracef,`tracef()`>>,
130 but it accepts an additional log level parameter.
132 ** Plugin support in LTTng-UST to provide a custom clock source and to
133 retrieve the current CPU number.
135 This feature exists for very advanced use cases.
138 https://github.com/lttng/lttng-ust/tree/stable-{revision}/doc/examples/clock-override[clock-override]
140 https://github.com/lttng/lttng-ust/tree/stable-{revision}/doc/examples/getcpu-override[getcpu-override]
141 examples for more details.
143 Moreover, LTTng{nbsp}{revision} boasts great stability, benifiting from
144 piles of bug fixes and more-than-welcome internal refactorings.
146 To learn more about the new features of LTTng{nbsp}{revision}, see
147 https://lttng.org/blog/2015/10/14/lttng-2.7-released/[the release announcement].
153 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
154 generation_ is a modern toolkit for tracing Linux systems and
155 applications. So your first question might be:
162 As the history of software engineering progressed and led to what
163 we now take for granted--complex, numerous and
164 interdependent software applications running in parallel on
165 sophisticated operating systems like Linux--the authors of such
166 components, software developers, began feeling a natural
167 urge to have tools that would ensure the robustness and good performance
168 of their masterpieces.
170 One major achievement in this field is, inarguably, the
171 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
172 an essential tool for developers to find and fix bugs. But even the best
173 debugger won't help make your software run faster, and nowadays, faster
174 software means either more work done by the same hardware, or cheaper
175 hardware for the same work.
177 A _profiler_ is often the tool of choice to identify performance
178 bottlenecks. Profiling is suitable to identify _where_ performance is
179 lost in a given software. The profiler outputs a profile, a statistical
180 summary of observed events, which you may use to discover which
181 functions took the most time to execute. However, a profiler won't
182 report _why_ some identified functions are the bottleneck. Bottlenecks
183 might only occur when specific conditions are met, conditions that are
184 sometimes impossible to capture by a statistical profiler, or impossible
185 to reproduce with an application altered by the overhead of an
186 event-based profiler. For a thorough investigation of software
187 performance issues, a history of execution is essential, with the
188 recorded values of variables and context fields you choose, and
189 with as little influence as possible on the instrumented software. This
190 is where tracing comes in handy.
192 _Tracing_ is a technique used to understand what goes on in a running
193 software system. The software used for tracing is called a _tracer_,
194 which is conceptually similar to a tape recorder. When recording,
195 specific instrumentation points placed in the software source code
196 generate events that are saved on a giant tape: a _trace_ file. You
197 can trace user applications and the operating system at the same time,
198 opening the possibility of resolving a wide range of problems that would
199 otherwise be extremely challenging.
201 Tracing is often compared to _logging_. However, tracers and loggers are
202 two different tools, serving two different purposes. Tracers are
203 designed to record much lower-level events that occur much more
204 frequently than log messages, often in the range of thousands per
205 second, with very little execution overhead. Logging is more appropriate
206 for a very high-level analysis of less frequent events: user accesses,
207 exceptional conditions (errors and warnings, for example), database
208 transactions, instant messaging communications, and such. Simply put,
209 logging is one of the many use cases that can be satisfied with tracing.
211 The list of recorded events inside a trace file can be read manually
212 like a log file for the maximum level of detail, but it is generally
213 much more interesting to perform application-specific analyses to
214 produce reduced statistics and graphs that are useful to resolve a
215 given problem. Trace viewers and analyzers are specialized tools
218 In the end, this is what LTTng is: a powerful, open source set of
219 tools to trace the Linux kernel and user applications at the same time.
220 LTTng is composed of several components actively maintained and
221 developed by its link:/community/#where[community].
224 [[lttng-alternatives]]
225 === Alternatives to noch:{LTTng}
227 Excluding proprietary solutions, a few competing software tracers
230 * https://github.com/dtrace4linux/linux[dtrace4linux] is a port of
231 Sun Microsystems's DTrace to Linux. The cmd:dtrace tool interprets
232 user scripts and is responsible for loading code into the
233 Linux kernel for further execution and collecting the outputted data.
234 * https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF] is a
235 subsystem in the Linux kernel in which a virtual machine can execute
236 programs passed from the user space to the kernel. You can attach
237 such programs to tracepoints and KProbes thanks to a system call, and
238 they can output data to the user space when executed thanks to
239 different mechanisms (pipe, VM register values, and eBPF maps, to name
241 * https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]
242 is the de facto function tracer of the Linux kernel. Its user
243 interface is a set of special files in sysfs.
244 * https://perf.wiki.kernel.org/[perf] is
245 a performance analyzing tool for Linux which supports hardware
246 performance counters, tracepoints, as well as other counters and
247 types of probes. perf's controlling utility is the cmd:perf command
249 * http://linux.die.net/man/1/strace[strace]
250 is a command-line utility which records system calls made by a
251 user process, as well as signal deliveries and changes of process
252 state. strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace]
253 to fulfill its function.
254 * http://www.sysdig.org/[sysdig], like SystemTap, uses scripts to
255 analyze Linux kernel events. You write scripts, or _chisels_ in
256 sysdig's jargon, in Lua and sysdig executes them while the system is
257 being traced or afterwards. sysdig's interface is the cmd:sysdig
258 command-line tool as well as the curses-based cmd:csysdig tool.
259 * https://sourceware.org/systemtap/[SystemTap] is a Linux kernel and
260 user space tracer which uses custom user scripts to produce plain text
261 traces. SystemTap converts the scripts to the C language, and then
262 compiles them as Linux kernel modules which are loaded to produce
263 trace data. SystemTap's primary user interface is the cmd:stap
266 The main distinctive features of LTTng is that it produces correlated
267 kernel and user space traces, as well as doing so with the lowest
268 overhead amongst other solutions. It produces trace files in the
269 http://diamon.org/ctf[CTF] format, a file format optimized
270 for the production and analyses of multi-gigabyte data.
272 LTTng is the result of more than 10 years of active open source
273 development by a community of passionate developers.
274 LTTng{nbsp}{revision} is currently available on major desktop and server
277 The main interface for tracing control is a single command-line tool
278 named cmd:lttng. The latter can create several tracing sessions, enable
279 and disable events on the fly, filter events efficiently with custom
280 user expressions, start and stop tracing, and much more. LTTng can
281 record the traces on the file system or send them over the network, and
282 keep them totally or partially. You can view the traces once tracing
283 becomes inactive or in real-time.
285 <<installing-lttng,Install LTTng now>> and
286 <<getting-started,start tracing>>!
292 include::../common/warning-no-installation.txt[]
294 **LTTng** is a set of software <<plumbing,components>> which interact to
295 <<instrumenting,instrument>> the Linux kernel and user applications, and
296 to <<controlling-tracing,control tracing>> (start and stop
297 tracing, enable and disable event rules, and the rest). Those
298 components are bundled into the following packages:
300 * **LTTng-tools**: Libraries and command-line interface to
302 * **LTTng-modules**: Linux kernel modules to instrument and
304 * **LTTng-UST**: Libraries and Java/Python packages to instrument and
305 trace user applications.
307 Most distributions mark the LTTng-modules and LTTng-UST packages as
308 optional when installing LTTng-tools (which is always required). Note
311 * You only need to install LTTng-modules if you intend to trace the
313 * You only need to install LTTng-UST if you intend to trace user
317 [[building-from-source]]
318 === Build from source
320 To build and install LTTng{nbsp}{revision} from source:
322 . Using your distribution's package manager, or from source, install
323 the following dependencies of LTTng-tools and LTTng-UST:
326 * https://sourceforge.net/projects/libuuid/[libuuid]
327 * http://directory.fsf.org/wiki/Popt[popt]
328 * http://liburcu.org/[Userspace RCU]
329 * http://www.xmlsoft.org/[libxml2]
332 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
338 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.7.tar.bz2 &&
339 tar -xf lttng-modules-latest-2.7.tar.bz2 &&
340 cd lttng-modules-2.7.* &&
342 sudo make modules_install &&
347 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
353 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.7.tar.bz2 &&
354 tar -xf lttng-ust-latest-2.7.tar.bz2 &&
355 cd lttng-ust-2.7.* &&
365 .Java and Python application tracing
367 If you need to instrument and trace <<java-application,Java
368 applications>>, pass the `--enable-java-agent-jul`,
369 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
370 `configure` script, depending on which Java logging framework you use.
372 If you need to instrument and trace <<python-application,Python
373 applications>>, pass the `--enable-python-agent` option to the
374 `configure` script. You can set the `PYTHON` environment variable to the
375 path to the Python interpreter for which to install the LTTng-UST Python
383 By default, LTTng-UST libraries are installed to
384 dir:{/usr/local/lib}, which is the de facto directory in which to
385 keep self-compiled and third-party libraries.
387 When <<building-tracepoint-providers-and-user-application,linking an
388 instrumented user application with `liblttng-ust`>>:
390 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
392 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
393 man:gcc(1), man:g++(1), or man:clang(1).
397 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
403 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.7.tar.bz2 &&
404 tar -xf lttng-tools-latest-2.7.tar.bz2 &&
405 cd lttng-tools-2.7.* &&
413 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
414 previous steps automatically for a given version of LTTng and confine
415 the installed files in a specific directory. This can be useful to test
416 LTTng without installing it on your system.
422 This is a short guide to get started quickly with LTTng kernel and user
425 Before you follow this guide, make sure to <<installing-lttng,install>>
428 This tutorial walks you through the steps to:
430 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
431 . <<tracing-your-own-user-application,Trace a user application>> written
433 . <<viewing-and-analyzing-your-traces,View and analyze the
437 [[tracing-the-linux-kernel]]
438 === Trace the Linux kernel
440 The following command lines start with cmd:sudo because you need root
441 privileges to trace the Linux kernel. You can avoid using cmd:sudo if
442 your Unix user is a member of the <<lttng-sessiond,tracing group>>.
444 . Create a <<tracing-session,tracing session>>:
449 sudo lttng create my-kernel-session
453 . List the available kernel tracepoints and system calls:
462 . Create an <<event,event rule>> which matches the desired event names,
463 for example `sched_switch` and `sched_process_fork`:
468 sudo lttng enable-event --kernel sched_switch,sched_process_fork
472 You can also create an event rule which _matches_ all the Linux kernel
473 tracepoints (this will generate a lot of data when tracing):
478 sudo lttng enable-event --kernel --all
491 . Do some operation on your system for a few seconds. For example,
492 load a website, or list the files of a directory.
493 . Stop tracing and destroy the tracing session:
503 The `destroy` command does not destroy the trace data; it only destroys
504 the state of the tracing session.
506 By default, LTTng saves the traces in
507 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
508 where +__name__+ is the tracing session name. Note that the
509 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
511 See <<viewing-and-analyzing-your-traces,View and analyze the
512 recorded events>> to view the recorded events.
515 [[tracing-your-own-user-application]]
516 === Trace a user application
518 This section steps you through a simple example to trace a
519 _Hello world_ program written in C.
521 To create the traceable user application:
523 . Create the tracepoint provider header file, which defines the
524 tracepoints and the events they can generate:
530 #undef TRACEPOINT_PROVIDER
531 #define TRACEPOINT_PROVIDER hello_world
533 #undef TRACEPOINT_INCLUDE
534 #define TRACEPOINT_INCLUDE "./hello-tp.h"
536 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
539 #include <lttng/tracepoint.h>
549 ctf_string(my_string_field, my_string_arg)
550 ctf_integer(int, my_integer_field, my_integer_arg)
554 #endif /* _HELLO_TP_H */
556 #include <lttng/tracepoint-event.h>
560 . Create the tracepoint provider package source file:
566 #define TRACEPOINT_CREATE_PROBES
567 #define TRACEPOINT_DEFINE
569 #include "hello-tp.h"
573 . Build the tracepoint provider package:
578 gcc -c -I. hello-tp.c
582 . Create the _Hello World_ application source file:
589 #include "hello-tp.h"
591 int main(int argc, char *argv[])
595 puts("Hello, World!\nPress Enter to continue...");
598 * The following getchar() call is only placed here for the purpose
599 * of this demonstration, to pause the application in order for
600 * you to have time to list its tracepoints. It is not
606 * A tracepoint() call.
608 * Arguments, as defined in hello-tp.h:
610 * 1. Tracepoint provider name (required)
611 * 2. Tracepoint name (required)
612 * 3. my_integer_arg (first user-defined argument)
613 * 4. my_string_arg (second user-defined argument)
615 * Notice the tracepoint provider and tracepoint names are
616 * NOT strings: they are in fact parts of variables that the
617 * macros in hello-tp.h create.
619 tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
621 for (x = 0; x < argc; ++x) {
622 tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
625 puts("Quitting now!");
626 tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
633 . Build the application:
642 . Link the application with the tracepoint provider package,
643 `liblttng-ust`, and `libdl`:
648 gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
652 Here's the whole build process:
655 .User space tracing tutorial's build steps.
656 image::ust-flow.png[]
658 To trace the user application:
660 . Run the application with a few arguments:
665 ./hello world and beyond
674 Press Enter to continue...
678 . Start an LTTng <<lttng-sessiond,session daemon>>:
683 lttng-sessiond --daemonize
687 Note that a session daemon might already be running, for example as
688 a service that the distribution's service manager started.
690 . List the available user space tracepoints:
695 lttng list --userspace
699 You see the `hello_world:my_first_tracepoint` tracepoint listed
700 under the `./hello` process.
702 . Create a <<tracing-session,tracing session>>:
707 lttng create my-user-space-session
711 . Create an <<event,event rule>> which matches the
712 `hello_world:my_first_tracepoint` event name:
717 lttng enable-event --userspace hello_world:my_first_tracepoint
730 . Go back to the running `hello` application and press Enter. The
731 program executes all `tracepoint()` instrumentation points and exits.
732 . Stop tracing and destroy the tracing session:
742 The `destroy` command does not destroy the trace data; it only destroys
743 the state of the tracing session.
745 By default, LTTng saves the traces in
746 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
747 where +__name__+ is the tracing session name. Note that the
748 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
750 See <<viewing-and-analyzing-your-traces,View and analyze the
751 recorded events>> to view the recorded events.
754 [[viewing-and-analyzing-your-traces]]
755 === View and analyze the recorded events
757 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
758 kernel>> and <<tracing-your-own-user-application,Trace a user
759 application>> tutorials, you can inspect the recorded events.
761 Many tools are available to read LTTng traces:
763 * **cmd:babeltrace** is a command-line utility which converts trace
764 formats; it supports the format that LTTng produces, CTF, as well as a
765 basic text output which can be ++grep++ed. The cmd:babeltrace command
766 is part of the http://diamon.org/babeltrace[Babeltrace] project.
767 * Babeltrace also includes
768 **https://www.python.org/[Python] bindings** so
769 that you can easily open and read an LTTng trace with your own script,
770 benefiting from the power of Python.
771 * http://tracecompass.org/[**Trace Compass**]
772 is a graphical user interface for viewing and analyzing any type of
773 logs or traces, including LTTng's.
774 * https://github.com/lttng/lttng-analyses[**LTTng analyses**] is a
775 project which includes many high-level analyses of LTTng kernel
776 traces, like scheduling statistics, interrupt frequency distribution,
777 top CPU usage, and more.
779 NOTE: This section assumes that the traces recorded during the previous
780 tutorials were saved to their default location, in the
781 dir:{$LTTNG_HOME/lttng-traces} directory. Note that the env:LTTNG_HOME
782 environment variable defaults to `$HOME` if not set.
785 [[viewing-and-analyzing-your-traces-bt]]
786 ==== Use the cmd:babeltrace command-line tool
788 The simplest way to list all the recorded events of a trace is to pass
789 its path to cmd:babeltrace with no options:
793 babeltrace ~/lttng-traces/my-user-space-session*
796 cmd:babeltrace finds all traces recursively within the given path and
797 prints all their events, merging them in chronological order.
799 You can pipe the output of cmd:babeltrace into a tool like man:grep(1) for
804 babeltrace ~/lttng-traces/my-kernel-session* | grep sys_
807 You can pipe the output of cmd:babeltrace into a tool like man:wc(1) to
808 count the recorded events:
812 babeltrace ~/lttng-traces/my-kernel-session* | grep sys_read | wc --lines
816 [[viewing-and-analyzing-your-traces-bt-python]]
817 ==== Use the Babeltrace Python bindings
819 The <<viewing-and-analyzing-your-traces-bt,text output of cmd:babeltrace>>
820 is useful to isolate events by simple matching using man:grep(1) and
821 similar utilities. However, more elaborate filters, such as keeping only
822 event records with a field value falling within a specific range, are
823 not trivial to write using a shell. Moreover, reductions and even the
824 most basic computations involving multiple event records are virtually
825 impossible to implement.
827 Fortunately, Babeltrace ships with Python 3 bindings which makes it easy
828 to read the event records of an LTTng trace sequentially and compute the
831 The following script accepts an LTTng Linux kernel trace path as its
832 first argument and prints the short names of the top 5 running processes
833 on CPU 0 during the whole trace:
838 from collections import Counter
844 if len(sys.argv) != 2:
845 msg = 'Usage: python3 {} TRACEPATH'.format(sys.argv[0])
846 print(msg, file=sys.stderr)
849 # A trace collection contains one or more traces
850 col = babeltrace.TraceCollection()
852 # Add the trace provided by the user (LTTng traces always have
854 if col.add_trace(sys.argv[1], 'ctf') is None:
855 raise RuntimeError('Cannot add trace')
857 # This counter dict contains execution times:
859 # task command name -> total execution time (ns)
860 exec_times = Counter()
862 # This contains the last `sched_switch` timestamp
866 for event in col.events:
867 # Keep only `sched_switch` events
868 if event.name != 'sched_switch':
871 # Keep only events which happened on CPU 0
872 if event['cpu_id'] != 0:
876 cur_ts = event.timestamp
882 # Previous task command (short) name
883 prev_comm = event['prev_comm']
885 # Initialize entry in our dict if not yet done
886 if prev_comm not in exec_times:
887 exec_times[prev_comm] = 0
889 # Compute previous command execution time
890 diff = cur_ts - last_ts
892 # Update execution time of this command
893 exec_times[prev_comm] += diff
895 # Update last timestamp
899 for name, ns in exec_times.most_common(5):
901 print('{:20}{} s'.format(name, s))
906 if __name__ == '__main__':
907 sys.exit(0 if top5proc() else 1)
914 python3 top5proc.py ~/lttng-traces/my-kernel-session*/kernel
920 swapper/0 48.607245889 s
921 chromium 7.192738188 s
922 pavucontrol 0.709894415 s
923 Compositor 0.660867933 s
924 Xorg.bin 0.616753786 s
927 Note that `swapper/0` is the "idle" process of CPU 0 on Linux; since we
928 weren't using the CPU that much when tracing, its first position in the
933 == [[understanding-lttng]]Core concepts
935 From a user's perspective, the LTTng system is built on a few concepts,
936 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
937 operates by sending commands to the <<lttng-sessiond,session daemon>>.
938 Understanding how those objects relate to eachother is key in mastering
941 The core concepts are:
943 * <<tracing-session,Tracing session>>
944 * <<domain,Tracing domain>>
945 * <<channel,Channel and ring buffer>>
946 * <<"event","Instrumentation point, event rule, event, and event record">>
952 A _tracing session_ is a stateful dialogue between you and
953 a <<lttng-sessiond,session daemon>>. You can
954 <<creating-destroying-tracing-sessions,create a new tracing
955 session>> with the `lttng create` command.
957 Anything that you do when you control LTTng tracers happens within a
958 tracing session. In particular, a tracing session:
961 * Has its own set of trace files.
962 * Has its own state of activity (started or stopped).
963 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
965 * Has its own <<channel,channels>> which have their own
966 <<event,event rules>>.
969 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
970 image::concepts.png[]
972 Those attributes and objects are completely isolated between different
975 A tracing session is analogous to a cash machine session:
976 the operations you do on the banking system through the cash machine do
977 not alter the data of other users of the same system. In the case of
978 the cash machine, a session lasts as long as your bank card is inside.
979 In the case of LTTng, a tracing session lasts from the `lttng create`
980 command to the `lttng destroy` command.
983 .Each Unix user has its own set of tracing sessions.
984 image::many-sessions.png[]
987 [[tracing-session-mode]]
988 ==== Tracing session mode
990 LTTng can send the generated trace data to different locations. The
991 _tracing session mode_ dictates where to send it. The following modes
992 are available in LTTng{nbsp}{revision}:
995 LTTng writes the traces to the file system of the machine being traced
998 Network streaming mode::
999 LTTng sends the traces over the network to a
1000 <<lttng-relayd,relay daemon>> running on a remote system.
1003 LTTng does not write the traces by default. Instead, you can request
1004 LTTng to <<taking-a-snapshot,take a snapshot>>, that is, a copy of the
1005 current tracing buffers, and to write it to the target's file system
1006 or to send it over the network to a <<lttng-relayd,relay daemon>>
1007 running on a remote system.
1010 This mode is similar to the network streaming mode, but a live
1011 trace viewer can connect to the distant relay daemon to
1012 <<lttng-live,view event records as LTTng generates them>> by
1019 A _tracing domain_ is a namespace for event sources. A tracing domain
1020 has its own properties and features.
1022 There are currently five available tracing domains:
1026 * `java.util.logging` (JUL)
1030 You must specify a tracing domain when using some commands to avoid
1031 ambiguity. For example, since all the domains support named tracepoints
1032 as event sources (instrumentation points that you manually insert in the
1033 source code), you need to specify a tracing domain when
1034 <<enabling-disabling-events,creating an event rule>> because all the
1035 tracing domains could have tracepoints with the same names.
1037 Some features are reserved to specific tracing domains. Dynamic function
1038 entry and return instrumentation points, for example, are currently only
1039 supported in the Linux kernel tracing domain, but support for other
1040 tracing domains could be added in the future.
1042 You can create <<channel,channels>> in the Linux kernel and user space
1043 tracing domains. The other tracing domains have a single default
1048 === Channel and ring buffer
1050 A _channel_ is an object which is responsible for a set of ring buffers.
1051 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1052 tracer emits an event, it can record it to one or more
1053 sub-buffers. The attributes of a channel determine what to do when
1054 there's no space left for a new event record because all sub-buffers
1055 are full, where to send a full sub-buffer, and other behaviours.
1057 A channel is always associated to a <<domain,tracing domain>>. The
1058 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1059 a default channel which you cannot configure.
1061 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1062 an event, it records it to the sub-buffers of all
1063 the enabled channels with a satisfied event rule, as long as those
1064 channels are part of active <<tracing-session,tracing sessions>>.
1067 [[channel-buffering-schemes]]
1068 ==== Per-user vs. per-process buffering schemes
1070 A channel has at least one ring buffer _per CPU_. LTTng always
1071 records an event to the ring buffer associated to the CPU on which it
1074 Two _buffering schemes_ are available when you
1075 <<enabling-disabling-channels,create a channel>> in the
1076 user space <<domain,tracing domain>>:
1078 Per-user buffering::
1079 Allocate one set of ring buffers--one per CPU--shared by all the
1080 instrumented processes of each Unix user.
1084 .Per-user buffering scheme.
1085 image::per-user-buffering.png[]
1088 Per-process buffering::
1089 Allocate one set of ring buffers--one per CPU--for each
1090 instrumented process.
1094 .Per-process buffering scheme.
1095 image::per-process-buffering.png[]
1098 The per-process buffering scheme tends to consume more memory than the
1099 per-user option because systems generally have more instrumented
1100 processes than Unix users running instrumented processes. However, the
1101 per-process buffering scheme ensures that one process having a high
1102 event throughput won't fill all the shared sub-buffers of the same
1105 The Linux kernel tracing domain has only one available buffering scheme
1106 which is to allocate a single set of ring buffers for the whole system.
1107 This scheme is similar to the per-user option, but with a single, global
1108 user "running" the kernel.
1111 [[channel-overwrite-mode-vs-discard-mode]]
1112 ==== Overwrite vs. discard event loss modes
1114 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1115 arc in the following animation) of a specific channel's ring buffer.
1116 When there's no space left in a sub-buffer, the tracer marks it as
1117 consumable (red) and another, empty sub-buffer starts receiving the
1118 following event records. A <<lttng-consumerd,consumer daemon>>
1119 eventually consumes the marked sub-buffer (returns to white).
1122 [role="docsvg-channel-subbuf-anim"]
1127 In an ideal world, sub-buffers are consumed faster than they are filled,
1128 as is the case in the previous animation. In the real world,
1129 however, all sub-buffers can be full at some point, leaving no space to
1130 record the following events.
1132 By design, LTTng is a _non-blocking_ tracer: when no empty sub-buffer is
1133 available, it is acceptable to lose event records when the alternative
1134 would be to cause substantial delays in the instrumented application's
1135 execution. LTTng privileges performance over integrity; it aims at
1136 perturbing the traced system as little as possible in order to make
1137 tracing of subtle race conditions and rare interrupt cascades possible.
1139 When it comes to losing event records because no empty sub-buffer is
1140 available, the channel's _event loss mode_ determines what to do. The
1141 available event loss modes are:
1144 Drop the newest event records until a the tracer
1145 releases a sub-buffer.
1148 Clear the sub-buffer containing the oldest event records and start
1149 writing the newest event records there.
1151 This mode is sometimes called _flight recorder mode_ because it's
1153 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1154 always keep a fixed amount of the latest data.
1156 Which mechanism you should choose depends on your context: prioritize
1157 the newest or the oldest event records in the ring buffer?
1159 Beware that, in overwrite mode, the tracer abandons a whole sub-buffer
1160 as soon as a there's no space left for a new event record, whereas in
1161 discard mode, the tracer only discards the event record that doesn't
1164 In discard mode, LTTng increments a count of lost event records when
1165 an event record is lost and saves this count to the trace. In
1166 overwrite mode, LTTng keeps no information when it overwrites a
1167 sub-buffer before consuming it.
1169 There are a few ways to decrease your probability of losing event
1171 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1172 how you can fine-une the sub-buffer count and size of a channel to
1173 virtually stop losing event records, though at the cost of greater
1177 [[channel-subbuf-size-vs-subbuf-count]]
1178 ==== Sub-buffer count and size
1180 When you <<enabling-disabling-channels,create a channel>>, you can
1181 set its number of sub-buffers and their size.
1183 Note that there is noticeable CPU overhead introduced when
1184 switching sub-buffers (marking a full one as consumable and switching
1185 to an empty one for the following events to be recorded). Knowing this,
1186 the following list presents a few practical situations along with how
1187 to configure the sub-buffer count and size for them:
1189 * **High event throughput**: In general, prefer bigger sub-buffers to
1190 lower the risk of losing event records.
1192 Having bigger sub-buffers also ensures a lower sub-buffer switching
1195 The number of sub-buffers is only meaningful if you create the channel
1196 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1197 other sub-buffers are left unaltered.
1199 * **Low event throughput**: In general, prefer smaller sub-buffers
1200 since the risk of losing event records is low.
1202 Because events occur less frequently, the sub-buffer switching frequency
1203 should remain low and thus the tracer's overhead should not be a
1206 * **Low memory system**: If your target system has a low memory
1207 limit, prefer fewer first, then smaller sub-buffers.
1209 Even if the system is limited in memory, you want to keep the
1210 sub-buffers as big as possible to avoid a high sub-buffer switching
1213 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1214 which means event data is very compact. For example, the average
1215 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1216 sub-buffer size of 1{nbsp}MiB is considered big.
1218 The previous situations highlight the major trade-off between a few big
1219 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1220 frequency vs. how much data is lost in overwrite mode. Assuming a
1221 constant event throughput and using the overwrite mode, the two
1222 following configurations have the same ring buffer total size:
1225 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1230 * **2 sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1231 switching frequency, but if a sub-buffer overwrite happens, half of
1232 the event records so far (4{nbsp}MiB) are definitely lost.
1233 * **8 sub-buffers of 1{nbsp}MiB each**: Expect 4{nbsp}times the tracer's
1234 overhead as the previous configuration, but if a sub-buffer
1235 overwrite happens, only the eighth of event records so far are
1238 In discard mode, the sub-buffers count parameter is pointless: use two
1239 sub-buffers and set their size according to the requirements of your
1243 [[channel-switch-timer]]
1244 ==== Switch timer period
1246 The _switch timer period_ is an important configurable attribute of
1247 a channel to ensure periodic sub-buffer flushing.
1249 When the _switch timer_ expires, a sub-buffer switch happens. You can
1250 set the switch timer period attribute when you
1251 <<enabling-disabling-channels,create a channel>> to ensure that event
1252 data is consumed and committed to trace files or to a distant relay
1253 daemon periodically in case of a low event throughput.
1256 [role="docsvg-channel-switch-timer"]
1261 This attribute is also convenient when you use big sub-buffers to cope
1262 with a sporadic high event throughput, even if the throughput is
1266 [[channel-read-timer]]
1267 ==== Read timer period
1269 By default, the LTTng tracers use a notification mechanism to signal a
1270 full sub-buffer so that a consumer daemon can consume it. When such
1271 notifications must be avoided, for example in real-time applications,
1272 you can use the channel's _read timer_ instead. When the read timer
1273 fires, the <<lttng-consumerd,consumer daemon>> checks for full,
1274 consumable sub-buffers.
1277 [[tracefile-rotation]]
1278 ==== Trace file count and size
1280 By default, trace files can grow as large as needed. You can set the
1281 maximum size of each trace file that a channel writes when you
1282 <<enabling-disabling-channels,create a channel>>. When the size of
1283 a trace file reaches the channel's fixed maximum size, LTTng creates
1284 another file to contain the next event records. LTTng appends a file
1285 count to each trace file name in this case.
1287 If you set the trace file size attribute when you create a channel, the
1288 maximum number of trace files that LTTng creates is _unlimited_ by
1289 default. To limit them, you can also set a maximum number of trace
1290 files. When the number of trace files reaches the channel's fixed
1291 maximum count, the oldest trace file is overwritten. This mechanism is
1292 called _trace file rotation_.
1296 === Instrumentation point, event rule, event, and event record
1298 An _event rule_ is a set of conditions which must be **all** satisfied
1299 for LTTng to record an occuring event.
1301 You set the conditions when you <<enabling-disabling-events,create
1304 You always attach an event rule to <<channel,channel>> when you create
1307 When an event passes the conditions of an event rule, LTTng records it
1308 in one of the attached channel's sub-buffers.
1310 The available conditions, as of LTTng{nbsp}{revision}, are:
1312 * The event rule _is enabled_.
1313 * The instrumentation point's type _is{nbsp}T_.
1314 * The instrumentation point's name (sometimes called _event name_)
1315 _matches{nbsp}N_, but _is not{nbsp}E_.
1316 * The instrumentation point's log level _is as severe as{nbsp}L_, or
1317 _is exactly{nbsp}L_.
1318 * The fields of the event's payload _satisfy_ a filter
1319 expression{nbsp}__F__.
1321 As you can see, all the conditions but the dynamic filter are related to
1322 the event rule's status or to the instrumentation point, not to the
1323 occurring events. This is why, without a filter, checking if an event
1324 passes an event rule is not a dynamic task: when you create or modify an
1325 event rule, all the tracers of its tracing domain enable or disable the
1326 instrumentation points themselves once. This is possible because the
1327 attributes of an instrumentation point (type, name, and log level) are
1328 defined statically. In other words, without a dynamic filter, the tracer
1329 _does not evaluate_ the arguments of an instrumentation point unless it
1330 matches an enabled event rule.
1332 Note that, for LTTng to record an event, the <<channel,channel>> to
1333 which a matching event rule is attached must also be enabled, and the
1334 tracing session owning this channel must be active.
1337 .Logical path from an instrumentation point to an event record.
1338 image::event-rule.png[]
1340 .Event, event record, or event rule?
1342 With so many similar terms, it's easy to get confused.
1344 An **event** is the consequence of the execution of an _instrumentation
1345 point_, like a tracepoint that you manually place in some source code,
1346 or a Linux kernel KProbe. An event is said to _occur_ at a specific
1347 time. Different actions can be taken upon the occurrence of an event,
1348 like record the event's payload to a buffer.
1350 An **event record** is the representation of an event in a sub-buffer. A
1351 tracer is responsible for capturing the payload of an event, current
1352 context variables, the event's ID, and the event's timestamp. LTTng
1353 can append this sub-buffer to a trace file.
1355 An **event rule** is a set of conditions which must all be satisfied for
1356 LTTng to record an occuring event. Events still occur without
1357 satisfying event rules, but LTTng does not record them.
1362 == Components of noch:{LTTng}
1364 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1365 to call LTTng a simple _tool_ since it is composed of multiple
1366 interacting components. This section describes those components,
1367 explains their respective roles, and shows how they connect together to
1368 form the LTTng ecosystem.
1370 The following diagram shows how the most important components of LTTng
1371 interact with user applications, the Linux kernel, and you:
1374 .Control and trace data paths between LTTng components.
1375 image::plumbing.png[]
1377 The LTTng project incorporates:
1379 * **LTTng-tools**: Libraries and command-line interface to
1380 control tracing sessions.
1381 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1382 ** <<lttng-consumerd,Consumer daemon>> (cmd:lttng-consumerd).
1383 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1384 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1385 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1386 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1388 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1389 headers to instrument and trace any native user application.
1390 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1391 *** `liblttng-ust-libc-wrapper`
1392 *** `liblttng-ust-pthread-wrapper`
1393 *** `liblttng-ust-cyg-profile`
1394 *** `liblttng-ust-cyg-profile-fast`
1395 *** `liblttng-ust-dl`
1396 ** User space tracepoint provider source files generator command-line
1397 tool (man:lttng-gen-tp(1)).
1398 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1399 Java applications using `java.util.logging` or
1400 Apache log4j 1.2 logging.
1401 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1402 Python applications using the standard `logging` package.
1403 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1405 ** LTTng kernel tracer module.
1406 ** Tracing ring buffer kernel modules.
1407 ** Probe kernel modules.
1408 ** LTTng logger kernel module.
1412 === Tracing control command-line interface
1415 .The tracing control command-line interface.
1416 image::plumbing-lttng-cli.png[]
1418 The _man:lttng(1) command-line tool_ is the standard user interface to
1419 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1420 is part of LTTng-tools.
1422 The cmd:lttng tool is linked with
1423 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1424 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1426 The cmd:lttng tool has a Git-like interface:
1430 lttng <general options> <command> <command options>
1433 The <<controlling-tracing,Tracing control>> section explores the
1434 available features of LTTng using the cmd:lttng tool.
1437 [[liblttng-ctl-lttng]]
1438 === Tracing control library
1441 .The tracing control library.
1442 image::plumbing-liblttng-ctl.png[]
1444 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1445 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1446 underlying protocol's details. `liblttng-ctl` is part of LTTng-tools.
1448 The <<lttng-cli,cmd:lttng command-line tool>>
1449 is linked with `liblttng-ctl`.
1451 You can use `liblttng-ctl` in C or $$C++$$ source code by including its
1456 #include <lttng/lttng.h>
1459 Some objects are referenced by name (C string), such as tracing
1460 sessions, but most of them require to create a handle first using
1461 `lttng_create_handle()`.
1463 The best available developer documentation for `liblttng-ctl` is, as of
1464 LTTng{nbsp}{revision}, its installed header files. Every function and
1465 structure is thoroughly documented.
1469 === User space tracing library
1472 .The user space tracing library.
1473 image::plumbing-liblttng-ust.png[]
1475 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1476 is the LTTng user space tracer. It receives commands from a
1477 <<lttng-sessiond,session daemon>>, for example to
1478 enable and disable specific instrumentation points, and writes event
1479 records to ring buffers shared with a
1480 <<lttng-consumerd,consumer daemon>>.
1481 `liblttng-ust` is part of LTTng-UST.
1483 Public C header files are installed beside `liblttng-ust` to
1484 instrument any <<c-application,C or $$C++$$ application>>.
1486 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1487 packages, use their own library providing tracepoints which is
1488 linked with `liblttng-ust`.
1490 An application or library does not have to initialize `liblttng-ust`
1491 manually: its constructor does the necessary tasks to properly register
1492 to a session daemon. The initialization phase also enables the
1493 instrumentation points matching the <<event,event rules>> that you
1497 [[lttng-ust-agents]]
1498 === User space tracing agents
1501 .The user space tracing agents.
1502 image::plumbing-lttng-ust-agents.png[]
1504 The _LTTng-UST Java and Python agents_ are regular Java and Python
1505 packages which add LTTng tracing capabilities to the
1506 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1508 In the case of Java, the
1509 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1510 core logging facilities] and
1511 https://logging.apache.org/log4j/1.2/[Apache log4j 1.2] are supported.
1512 Note that Apache Log4{nbsp}2 is not supported.
1514 In the case of Python, the standard
1515 https://docs.python.org/3/library/logging.html[`logging`] package
1516 is supported. Both Python 2 and Python 3 modules can import the
1517 LTTng-UST Python agent package.
1519 The applications using the LTTng-UST agents are in the
1520 `java.util.logging` (JUL),
1521 log4j, and Python <<domain,tracing domains>>.
1523 Both agents use the same mechanism to trace the log statements. When an
1524 agent is initialized, it creates a log handler that attaches to the root
1525 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
1526 When the application executes a log statement, it is passed to the
1527 agent's log handler by the root logger. The agent's log handler calls a
1528 native function in a tracepoint provider package shared library linked
1529 with <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
1530 other fields, like its logger name and its log level. This native
1531 function contains a user space instrumentation point, hence tracing the
1534 The log level condition of an
1535 <<event,event rule>> is considered when tracing
1536 a Java or a Python application, and it's compatible with the standard
1537 JUL, log4j, and Python log levels.
1541 === LTTng kernel modules
1544 .The LTTng kernel modules.
1545 image::plumbing-lttng-modules.png[]
1547 The _LTTng kernel modules_ are a set of Linux kernel modules
1548 which implement the kernel tracer of the LTTng project. The LTTng
1549 kernel modules are part of LTTng-modules.
1551 The LTTng kernel modules include:
1553 * A set of _probe_ modules.
1555 Each module attaches to a specific subsystem
1556 of the Linux kernel using its tracepoint instrument points. There are
1557 also modules to attach to the entry and return points of the Linux
1558 system call functions.
1560 * _Ring buffer_ modules.
1562 A ring buffer implementation is provided as kernel modules. The LTTng
1563 kernel tracer writes to the ring buffer; a
1564 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
1566 * The _LTTng kernel tracer_ module.
1567 * The _LTTng logger_ module.
1569 The LTTng logger module implements the special path:{/proc/lttng-logger}
1570 file so that any executable can generate LTTng events by opening and
1571 writing to this file.
1573 See <<proc-lttng-logger-abi,LTTng logger>>.
1575 Generally, you do not have to load the LTTng kernel modules manually
1576 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
1577 daemon>> loads the necessary modules when starting. If you have extra
1578 probe modules, you can specify to load them to the session daemon on
1581 The LTTng kernel modules are installed in
1582 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
1583 the kernel release (see `uname --kernel-release`).
1590 .The session daemon.
1591 image::plumbing-sessiond.png[]
1593 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
1594 managing tracing sessions and for controlling the various components of
1595 LTTng. The session daemon is part of LTTng-tools.
1597 The session daemon sends control requests to and receives control
1600 * The <<lttng-ust,user space tracing library>>.
1602 Any instance of the user space tracing library first registers to
1603 a session daemon. Then, the session daemon can send requests to
1604 this instance, such as:
1607 ** Get the list of tracepoints.
1608 ** Share an <<event,event rule>> so that the user space tracing library
1609 can enable or disable tracepoints. Amongst the possible conditions
1610 of an event rule is a filter expression which `liblttng-ust` evalutes
1611 when an event occurs.
1612 ** Share <<channel,channel>> attributes and ring buffer locations.
1615 The session daemon and the user space tracing library use a Unix
1616 domain socket for their communication.
1618 * The <<lttng-ust-agents,user space tracing agents>>.
1620 Any instance of a user space tracing agent first registers to
1621 a session daemon. Then, the session daemon can send requests to
1622 this instance, such as:
1625 ** Get the list of loggers.
1626 ** Enable or disable a specific logger.
1629 The session daemon and the user space tracing agent use a TCP connection
1630 for their communication.
1632 * The <<lttng-modules,LTTng kernel tracer>>.
1633 * The <<lttng-consumerd,consumer daemon>>.
1635 The session daemon sends requests to the consumer daemon to instruct
1636 it where to send the trace data streams, amongst other information.
1638 * The <<lttng-relayd,relay daemon>>.
1640 The session daemon receives commands from the
1641 <<liblttng-ctl-lttng,tracing control library>>.
1643 The root session daemon loads the appropriate
1644 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
1645 a <<lttng-consumerd,consumer daemon>> as soon as you create
1646 an <<event,event rule>>.
1648 The session daemon does not send and receive trace data: this is the
1649 role of the <<lttng-consumerd,consumer daemon>> and
1650 <<lttng-relayd,relay daemon>>. It does, however, generate the
1651 http://diamon.org/ctf/[CTF] metadata stream.
1653 Each Unix user can have its own session daemon instance. The
1654 tracing sessions managed by different session daemons are completely
1657 The root user's session daemon is the only one which is
1658 allowed to control the LTTng kernel tracer, and its spawned consumer
1659 daemon is the only one which is allowed to consume trace data from the
1660 LTTng kernel tracer. Note, however, that any Unix user which is a member
1661 of the <<tracing-group,tracing group>> is allowed
1662 to create <<channel,channels>> in the
1663 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
1666 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
1667 session daemon when using its `create` command if none is currently
1668 running. You can also start the session daemon manually.
1675 .The consumer daemon.
1676 image::plumbing-consumerd.png[]
1678 The _consumer daemon_, cmd:lttng-consumerd, is a daemon which shares
1679 ring buffers with user applications or with the LTTng kernel modules to
1680 collect trace data and send it to some location (on disk or to a
1681 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
1682 is part of LTTng-tools.
1684 You do not start a consumer daemon manually: a consumer daemon is always
1685 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
1686 <<event,event rule>>, that is, before you start tracing. When you kill
1687 its owner session daemon, the consumer daemon also exits because it is
1688 the session daemon's child process. Command-line options of
1689 man:lttng-sessiond(8) target the consumer daemon process.
1691 There are up to two running consumer daemons per Unix user, whereas only
1692 one session daemon can run per user. This is because each process can be
1693 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
1694 and 64-bit processes, it is more efficient to have separate
1695 corresponding 32-bit and 64-bit consumer daemons. The root user is an
1696 exception: it can have up to _three_ running consumer daemons: 32-bit
1697 and 64-bit instances for its user applications, and one more
1698 reserved for collecting kernel trace data.
1706 image::plumbing-relayd.png[]
1708 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
1709 between remote session and consumer daemons, local trace files, and a
1710 remote live trace viewer. The relay daemon is part of LTTng-tools.
1712 The main purpose of the relay daemon is to implement a receiver of
1713 <<sending-trace-data-over-the-network,trace data over the network>>.
1714 This is useful when the target system does not have much file system
1715 space to record trace files locally.
1717 The relay daemon is also a server to which a
1718 <<lttng-live,live trace viewer>> can
1719 connect. The live trace viewer sends requests to the relay daemon to
1720 receive trace data as the target system emits events. The
1721 communication protocol is named _LTTng live_; it is used over TCP
1724 Note that you can start the relay daemon on the target system directly.
1725 This is the setup of choice when the use case is to view events as
1726 the target system emits them without the need of a remote system.
1730 == [[using-lttng]]Instrumentation
1732 There are many examples of tracing and monitoring in our everyday life:
1734 * You have access to real-time and historical weather reports and
1735 forecasts thanks to weather stations installed around the country.
1736 * You know your heart is safe thanks to an electrocardiogram.
1737 * You make sure not to drive your car too fast and to have enough fuel
1738 to reach your destination thanks to gauges visible on your dashboard.
1740 All the previous examples have something in common: they rely on
1741 **instruments**. Without the electrodes attached to the surface of your
1742 body's skin, cardiac monitoring is futile.
1744 LTTng, as a tracer, is no different from those real life examples. If
1745 you're about to trace a software system or, in other words, record its
1746 history of execution, you better have **instrumentation points** in the
1747 subject you're tracing, that is, the actual software.
1749 Various ways were developed to instrument a piece of software for LTTng
1750 tracing. The most straightforward one is to manually place
1751 instrumentation points, called _tracepoints_, in the software's source
1752 code. It is also possible to add instrumentation points dynamically in
1753 the Linux kernel <<domain,tracing domain>>.
1755 If you're only interested in tracing the Linux kernel, your
1756 instrumentation needs are probably already covered by LTTng's built-in
1757 <<lttng-modules,Linux kernel tracepoints>>. You may also wish to trace a
1758 user application which is already instrumented for LTTng tracing.
1759 In such cases, you can skip this whole section and read the topics of
1760 the <<controlling-tracing,Tracing control>> section.
1762 Many methods are available to instrument a piece of software for LTTng
1765 * <<c-application,User space instrumentation for C and $$C++$$
1767 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
1768 * <<java-application,User space Java agent>>.
1769 * <<python-application,User space Python agent>>.
1770 * <<proc-lttng-logger-abi,LTTng logger>>.
1771 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
1775 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
1777 The procedure to instrument a C or $$C++$$ user application with
1778 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
1780 . <<tracepoint-provider,Create the source files of a tracepoint provider
1782 . <<probing-the-application-source-code,Add tracepoints to
1783 the application's source code>>.
1784 . <<building-tracepoint-providers-and-user-application,Build and link
1785 a tracepoint provider package and the user application>>.
1787 If you need quick, man:printf(3)-like instrumentation, you can skip
1788 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
1791 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
1792 instrument a user application with `liblttng-ust`.
1795 [[tracepoint-provider]]
1796 ==== Create the source files of a tracepoint provider package
1798 A _tracepoint provider_ is a set of compiled functions which provide
1799 **tracepoints** to an application, the type of instrumentation point
1800 supported by LTTng-UST. Those functions can emit events with
1801 user-defined fields and serialize those events as event records to one
1802 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
1803 macro, which you <<probing-the-application-source-code,insert in a user
1804 application's source code>>, calls those functions.
1806 A _tracepoint provider package_ is an object file (`.o`) or a shared
1807 library (`.so`) which contains one or more tracepoint providers.
1808 Its source files are:
1810 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
1811 * A <<tpp-source,tracepoint provider package source>> (`.c`).
1813 A tracepoint provider package is dynamically linked with `liblttng-ust`,
1814 the LTTng user space tracer, at run time.
1817 .User application linked with `liblttng-ust` and containing a tracepoint provider.
1818 image::ust-app.png[]
1820 NOTE: If you need quick, man:printf(3)-like instrumentation, you can
1821 skip creating and using a tracepoint provider and use
1822 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
1826 ===== Create a tracepoint provider header file template
1828 A _tracepoint provider header file_ contains the tracepoint
1829 definitions of a tracepoint provider.
1831 To create a tracepoint provider header file:
1833 . Start from this template:
1837 .Tracepoint provider header file template (`.h` file extension).
1839 #undef TRACEPOINT_PROVIDER
1840 #define TRACEPOINT_PROVIDER provider_name
1842 #undef TRACEPOINT_INCLUDE
1843 #define TRACEPOINT_INCLUDE "./tp.h"
1845 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
1848 #include <lttng/tracepoint.h>
1851 * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
1852 * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
1857 #include <lttng/tracepoint-event.h>
1863 * `provider_name` with the name of your tracepoint provider.
1864 * `"tp.h"` with the name of your tracepoint provider header file.
1866 . Below the `#include <lttng/tracepoint.h>` line, put your
1867 <<defining-tracepoints,tracepoint definitions>>.
1869 Your tracepoint provider name must be unique amongst all the possible
1870 tracepoint provider names used on the same target system. We
1871 suggest to include the name of your project or company in the name,
1872 for example, `org_lttng_my_project_tpp`.
1874 TIP: [[lttng-gen-tp]]You can use the man:lttng-gen-tp(1) tool to create
1875 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
1876 write are the <<defining-tracepoints,tracepoint definitions>>.
1879 [[defining-tracepoints]]
1880 ===== Create a tracepoint definition
1882 A _tracepoint definition_ defines, for a given tracepoint:
1884 * Its **input arguments**. They are the macro parameters that the
1885 `tracepoint()` macro accepts for this particular tracepoint
1886 in the user application's source code.
1887 * Its **output event fields**. They are the sources of event fields
1888 that form the payload of any event that the execution of the
1889 `tracepoint()` macro emits for this particular tracepoint.
1891 You can create a tracepoint definition by using the
1892 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
1894 <<tpp-header,tracepoint provider header file template>>.
1896 The syntax of the `TRACEPOINT_EVENT()` macro is:
1899 .`TRACEPOINT_EVENT()` macro syntax.
1902 /* Tracepoint provider name */
1905 /* Tracepoint name */
1908 /* Input arguments */
1913 /* Output event fields */
1922 * `provider_name` with your tracepoint provider name.
1923 * `tracepoint_name` with your tracepoint name.
1924 * `arguments` with the <<tpp-def-input-args,input arguments>>.
1925 * `fields` with the <<tpp-def-output-fields,output event field>>
1928 This tracepoint emits events named `provider_name:tracepoint_name`.
1931 .Event name's length limitation
1933 The concatenation of the tracepoint provider name and the
1934 tracepoint name must not exceed **254 characters**. If it does, the
1935 instrumented application compiles and runs, but LTTng throws multiple
1936 warnings and you could experience serious issues.
1939 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
1942 .`TP_ARGS()` macro syntax.
1951 * `type` with the C type of the argument.
1952 * `arg_name` with the argument name.
1954 You can repeat `type` and `arg_name` up to 10 times to have
1955 more than one argument.
1957 .`TP_ARGS()` usage with three arguments.
1969 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
1970 tracepoint definition with no input arguments.
1972 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
1973 `ctf_*()` macros. Each `ctf_*()` macro defines one event field.
1974 See <<liblttng-ust-tp-fields,Tracepoint fields macros>> for a
1975 complete description of the available `ctf_*()` macros.
1976 A `ctf_*()` macro specifies the type, size, and byte order of
1979 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
1980 C expression that the tracer evalutes at the `tracepoint()` macro site
1981 in the application's source code. This expression provides a field's
1982 source of data. The argument expression can include input argument names
1983 listed in the `TP_ARGS()` macro.
1985 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
1986 must be unique within a given tracepoint definition.
1988 Here's a complete tracepoint definition example:
1990 .Tracepoint definition.
1992 The following tracepoint definition defines a tracepoint which takes
1993 three input arguments and has four output event fields.
1997 #include "my-custom-structure.h"
2003 const struct my_custom_structure*, my_custom_structure,
2008 ctf_string(query_field, query)
2009 ctf_float(double, ratio_field, ratio)
2010 ctf_integer(int, recv_size, my_custom_structure->recv_size)
2011 ctf_integer(int, send_size, my_custom_structure->send_size)
2016 You can refer to this tracepoint definition with the `tracepoint()`
2017 macro in your application's source code like this:
2021 tracepoint(my_provider, my_tracepoint,
2022 my_structure, some_ratio, the_query);
2026 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2027 if they satisfy an enabled <<event,event rule>>.
2030 [[using-tracepoint-classes]]
2031 ===== Use a tracepoint class
2033 A _tracepoint class_ is a class of tracepoints which share the same
2034 output event field definitions. A _tracepoint instance_ is one
2035 instance of such a defined tracepoint class, with its own tracepoint
2038 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2039 shorthand which defines both a tracepoint class and a tracepoint
2040 instance at the same time.
2042 When you build a tracepoint provider package, the C or $$C++$$ compiler
2043 creates one serialization function for each **tracepoint class**. A
2044 serialization function is responsible for serializing the event fields
2045 of a tracepoint to a sub-buffer when tracing.
2047 For various performance reasons, when your situation requires multiple
2048 tracepoint definitions with different names, but with the same event
2049 fields, we recommend that you manually create a tracepoint class
2050 and instantiate as many tracepoint instances as needed. One positive
2051 effect of such a design, amongst other advantages, is that all
2052 tracepoint instances of the same tracepoint class reuse the same
2053 serialization function, thus reducing
2054 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2056 .Use a tracepoint class and tracepoint instances.
2058 Consider the following three tracepoint definitions:
2070 ctf_integer(int, userid, userid)
2071 ctf_integer(size_t, len, len)
2083 ctf_integer(int, userid, userid)
2084 ctf_integer(size_t, len, len)
2096 ctf_integer(int, userid, userid)
2097 ctf_integer(size_t, len, len)
2102 In this case, we create three tracepoint classes, with one implicit
2103 tracepoint instance for each of them: `get_account`, `get_settings`, and
2104 `get_transaction`. However, they all share the same event field names
2105 and types. Hence three identical, yet independent serialization
2106 functions are created when you build the tracepoint provider package.
2108 A better design choice is to define a single tracepoint class and three
2109 tracepoint instances:
2113 /* The tracepoint class */
2114 TRACEPOINT_EVENT_CLASS(
2115 /* Tracepoint provider name */
2118 /* Tracepoint class name */
2121 /* Input arguments */
2127 /* Output event fields */
2129 ctf_integer(int, userid, userid)
2130 ctf_integer(size_t, len, len)
2134 /* The tracepoint instances */
2135 TRACEPOINT_EVENT_INSTANCE(
2136 /* Tracepoint provider name */
2139 /* Tracepoint class name */
2142 /* Tracepoint name */
2145 /* Input arguments */
2151 TRACEPOINT_EVENT_INSTANCE(
2160 TRACEPOINT_EVENT_INSTANCE(
2173 [[assigning-log-levels]]
2174 ===== Assign a log level to a tracepoint definition
2176 You can assign an optional _log level_ to a
2177 <<defining-tracepoints,tracepoint definition>>.
2179 Assigning different levels of severity to tracepoint definitions can
2180 be useful: when you <<enabling-disabling-events,create an event rule>>,
2181 you can target tracepoints having a log level as severe as a specific
2184 The concept of LTTng-UST log levels is similar to the levels found
2185 in typical logging frameworks:
2187 * In a logging framework, the log level is given by the function
2188 or method name you use at the log statement site: `debug()`,
2189 `info()`, `warn()`, `error()`, and so on.
2190 * In LTTng-UST, you statically assign the log level to a tracepoint
2191 definition; any `tracepoint()` macro invocation which refers to
2192 this definition has this log level.
2194 You can assign a log level to a tracepoint definition with the
2195 `TRACEPOINT_LOGLEVEL()` macro. You must use this macro _after_ the
2196 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2197 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2200 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2203 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2205 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2210 * `provider_name` with the tracepoint provider name.
2211 * `tracepoint_name` with the tracepoint name.
2212 * `log_level` with the log level to assign to the tracepoint
2213 definition named `tracepoint_name` in the `provider_name`
2214 tracepoint provider.
2216 See <<liblttng-ust-tracepoint-loglevel,Tracepoint log levels>> for
2217 a list of available log level names.
2219 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2223 /* Tracepoint definition */
2232 ctf_integer(int, userid, userid)
2233 ctf_integer(size_t, len, len)
2237 /* Log level assignment */
2238 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2244 ===== Create a tracepoint provider package source file
2246 A _tracepoint provider package source file_ is a C source file which
2247 includes a <<tpp-header,tracepoint provider header file>> to expand its
2248 macros into event serialization and other functions.
2250 You can always use the following tracepoint provider package source
2254 .Tracepoint provider package source file template.
2256 #define TRACEPOINT_CREATE_PROBES
2261 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2262 header file>> name. You may also include more than one tracepoint
2263 provider header file here to create a tracepoint provider package
2264 holding more than one tracepoint providers.
2267 [[probing-the-application-source-code]]
2268 ==== Add tracepoints to an application's source code
2270 Once you <<tpp-header,create a tracepoint provider header file>>, you
2271 can use the `tracepoint()` macro in your application's
2272 source code to insert the tracepoints that this header
2273 <<defining-tracepoints,defines>>.
2275 The `tracepoint()` macro takes at least two parameters: the tracepoint
2276 provider name and the tracepoint name. The corresponding tracepoint
2277 definition defines the other parameters.
2279 .`tracepoint()` usage.
2281 The following <<defining-tracepoints,tracepoint definition>> defines a
2282 tracepoint which takes two input arguments and has two output event
2286 .Tracepoint provider header file.
2288 #include "my-custom-structure.h"
2295 const char*, cmd_name
2298 ctf_string(cmd_name, cmd_name)
2299 ctf_integer(int, number_of_args, argc)
2304 You can refer to this tracepoint definition with the `tracepoint()`
2305 macro in your application's source code like this:
2308 .Application's source file.
2312 int main(int argc, char* argv[])
2314 tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2320 Note how the application's source code includes
2321 the tracepoint provider header file containing the tracepoint
2322 definitions to use, path:{tp.h}.
2325 .`tracepoint()` usage with a complex tracepoint definition.
2327 Consider this complex tracepoint definition, where multiple event
2328 fields refer to the same input arguments in their argument expression
2332 .Tracepoint provider header file.
2334 /* For `struct stat` */
2335 #include <sys/types.h>
2336 #include <sys/stat.h>
2348 ctf_integer(int, my_constant_field, 23 + 17)
2349 ctf_integer(int, my_int_arg_field, my_int_arg)
2350 ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2351 ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2352 my_str_arg[2] + my_str_arg[3])
2353 ctf_string(my_str_arg_field, my_str_arg)
2354 ctf_integer_hex(off_t, size_field, st->st_size)
2355 ctf_float(double, size_dbl_field, (double) st->st_size)
2356 ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2357 size_t, strlen(my_str_arg) / 2)
2362 You can refer to this tracepoint definition with the `tracepoint()`
2363 macro in your application's source code like this:
2366 .Application's source file.
2368 #define TRACEPOINT_DEFINE
2375 stat("/etc/fstab", &s);
2376 tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2382 If you look at the event record that LTTng writes when tracing this
2383 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2384 it should look like this:
2386 .Event record fields
2388 |Field's name |Field's value
2389 |`my_constant_field` |40
2390 |`my_int_arg_field` |23
2391 |`my_int_arg_field2` |529
2393 |`my_str_arg_field` |`Hello, World!`
2394 |`size_field` |0x12d
2395 |`size_dbl_field` |301.0
2396 |`half_my_str_arg_field` |`Hello,`
2400 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2401 compute--they use the call stack, for example. To avoid this
2402 computation when the tracepoint is disabled, you can use the
2403 `tracepoint_enabled()` and `do_tracepoint()` macros.
2405 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2409 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2411 tracepoint_enabled(provider_name, tracepoint_name)
2412 do_tracepoint(provider_name, tracepoint_name, ...)
2417 * `provider_name` with the tracepoint provider name.
2418 * `tracepoint_name` with the tracepoint name.
2420 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2421 `tracepoint_name` from the provider named `provider_name` is enabled
2424 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2425 if the tracepoint is enabled. Using `tracepoint()` with
2426 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2427 the `tracepoint_enabled()` check, thus a race condition is
2428 possible in this situation:
2431 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2433 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2434 stuff = prepare_stuff();
2437 tracepoint(my_provider, my_tracepoint, stuff);
2440 If the tracepoint is enabled after the condition, then `stuff` is not
2441 prepared: the emitted event will either contain wrong data, or the whole
2442 application could crash (segmentation fault, for example).
2444 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2445 `STAP_PROBEV()` call. If you need it, you must emit
2449 [[building-tracepoint-providers-and-user-application]]
2450 ==== Build and link a tracepoint provider package and an application
2452 Once you have one or more <<tpp-header,tracepoint provider header
2453 files>> and a <<tpp-source,tracepoint provider package source file>>,
2454 you can create the tracepoint provider package by compiling its source
2455 file. From here, multiple build and run scenarios are possible. The
2456 following table shows common application and library configurations
2457 along with the required command lines to achieve them.
2459 In the following diagrams, we use the following file names:
2462 Executable application.
2465 Application's object file.
2468 Tracepoint provider package object file.
2471 Tracepoint provider package archive file.
2474 Tracepoint provider package shared object file.
2477 User library object file.
2480 User library shared object file.
2482 We use the following symbols in the diagrams of table below:
2485 .Symbols used in the build scenario diagrams.
2486 image::ust-sit-symbols.png[]
2488 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2489 variable in the following instructions.
2491 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2492 .Common tracepoint provider package scenarios.
2494 |Scenario |Instructions
2497 The instrumented application is statically linked with
2498 the tracepoint provider package object.
2500 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2503 include::../common/ust-sit-step-tp-o.txt[]
2505 To build the instrumented application:
2507 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2512 #define TRACEPOINT_DEFINE
2516 . Compile the application source file:
2525 . Build the application:
2530 gcc -o app app.o tpp.o -llttng-ust -ldl
2534 To run the instrumented application:
2536 * Start the application:
2546 The instrumented application is statically linked with the
2547 tracepoint provider package archive file.
2549 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
2552 To create the tracepoint provider package archive file:
2554 . Compile the <<tpp-source,tracepoint provider package source file>>:
2563 . Create the tracepoint provider package archive file:
2572 To build the instrumented application:
2574 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2579 #define TRACEPOINT_DEFINE
2583 . Compile the application source file:
2592 . Build the application:
2597 gcc -o app app.o tpp.a -llttng-ust -ldl
2601 To run the instrumented application:
2603 * Start the application:
2613 The instrumented application is linked with the tracepoint provider
2614 package shared object.
2616 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
2619 include::../common/ust-sit-step-tp-so.txt[]
2621 To build the instrumented application:
2623 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2628 #define TRACEPOINT_DEFINE
2632 . Compile the application source file:
2641 . Build the application:
2646 gcc -o app app.o -ldl -L. -ltpp
2650 To run the instrumented application:
2652 * Start the application:
2662 The tracepoint provider package shared object is preloaded before the
2663 instrumented application starts.
2665 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
2668 include::../common/ust-sit-step-tp-so.txt[]
2670 To build the instrumented application:
2672 . In path:{app.c}, before including path:{tpp.h}, add the
2678 #define TRACEPOINT_DEFINE
2679 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2683 . Compile the application source file:
2692 . Build the application:
2697 gcc -o app app.o -ldl
2701 To run the instrumented application with tracing support:
2703 * Preload the tracepoint provider package shared object and
2704 start the application:
2709 LD_PRELOAD=./libtpp.so ./app
2713 To run the instrumented application without tracing support:
2715 * Start the application:
2725 The instrumented application dynamically loads the tracepoint provider
2726 package shared object.
2728 See the <<dlclose-warning,warning about `dlclose()`>>.
2730 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
2733 include::../common/ust-sit-step-tp-so.txt[]
2735 To build the instrumented application:
2737 . In path:{app.c}, before including path:{tpp.h}, add the
2743 #define TRACEPOINT_DEFINE
2744 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2748 . Compile the application source file:
2757 . Build the application:
2762 gcc -o app app.o -ldl
2766 To run the instrumented application:
2768 * Start the application:
2778 The application is linked with the instrumented user library.
2780 The instrumented user library is statically linked with the tracepoint
2781 provider package object file.
2783 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
2786 include::../common/ust-sit-step-tp-o-fpic.txt[]
2788 To build the instrumented user library:
2790 . In path:{emon.c}, before including path:{tpp.h}, add the
2796 #define TRACEPOINT_DEFINE
2800 . Compile the user library source file:
2805 gcc -I. -fpic -c emon.c
2809 . Build the user library shared object:
2814 gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
2818 To build the application:
2820 . Compile the application source file:
2829 . Build the application:
2834 gcc -o app app.o -L. -lemon
2838 To run the application:
2840 * Start the application:
2850 The application is linked with the instrumented user library.
2852 The instrumented user library is linked with the tracepoint provider
2853 package shared object.
2855 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
2858 include::../common/ust-sit-step-tp-so.txt[]
2860 To build the instrumented user library:
2862 . In path:{emon.c}, before including path:{tpp.h}, add the
2868 #define TRACEPOINT_DEFINE
2872 . Compile the user library source file:
2877 gcc -I. -fpic -c emon.c
2881 . Build the user library shared object:
2886 gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
2890 To build the application:
2892 . Compile the application source file:
2901 . Build the application:
2906 gcc -o app app.o -L. -lemon
2910 To run the application:
2912 * Start the application:
2922 The tracepoint provider package shared object is preloaded before the
2925 The application is linked with the instrumented user library.
2927 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
2930 include::../common/ust-sit-step-tp-so.txt[]
2932 To build the instrumented user library:
2934 . In path:{emon.c}, before including path:{tpp.h}, add the
2940 #define TRACEPOINT_DEFINE
2941 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2945 . Compile the user library source file:
2950 gcc -I. -fpic -c emon.c
2954 . Build the user library shared object:
2959 gcc -shared -o libemon.so emon.o -ldl
2963 To build the application:
2965 . Compile the application source file:
2974 . Build the application:
2979 gcc -o app app.o -L. -lemon
2983 To run the application with tracing support:
2985 * Preload the tracepoint provider package shared object and
2986 start the application:
2991 LD_PRELOAD=./libtpp.so ./app
2995 To run the application without tracing support:
2997 * Start the application:
3007 The application is linked with the instrumented user library.
3009 The instrumented user library dynamically loads the tracepoint provider
3010 package shared object.
3012 See the <<dlclose-warning,warning about `dlclose()`>>.
3014 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3017 include::../common/ust-sit-step-tp-so.txt[]
3019 To build the instrumented user library:
3021 . In path:{emon.c}, before including path:{tpp.h}, add the
3027 #define TRACEPOINT_DEFINE
3028 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3032 . Compile the user library source file:
3037 gcc -I. -fpic -c emon.c
3041 . Build the user library shared object:
3046 gcc -shared -o libemon.so emon.o -ldl
3050 To build the application:
3052 . Compile the application source file:
3061 . Build the application:
3066 gcc -o app app.o -L. -lemon
3070 To run the application:
3072 * Start the application:
3082 The application dynamically loads the instrumented user library.
3084 The instrumented user library is linked with the tracepoint provider
3085 package shared object.
3087 See the <<dlclose-warning,warning about `dlclose()`>>.
3089 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3092 include::../common/ust-sit-step-tp-so.txt[]
3094 To build the instrumented user library:
3096 . In path:{emon.c}, before including path:{tpp.h}, add the
3102 #define TRACEPOINT_DEFINE
3106 . Compile the user library source file:
3111 gcc -I. -fpic -c emon.c
3115 . Build the user library shared object:
3120 gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3124 To build the application:
3126 . Compile the application source file:
3135 . Build the application:
3140 gcc -o app app.o -ldl -L. -lemon
3144 To run the application:
3146 * Start the application:
3156 The application dynamically loads the instrumented user library.
3158 The instrumented user library dynamically loads the tracepoint provider
3159 package shared object.
3161 See the <<dlclose-warning,warning about `dlclose()`>>.
3163 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3166 include::../common/ust-sit-step-tp-so.txt[]
3168 To build the instrumented user library:
3170 . In path:{emon.c}, before including path:{tpp.h}, add the
3176 #define TRACEPOINT_DEFINE
3177 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3181 . Compile the user library source file:
3186 gcc -I. -fpic -c emon.c
3190 . Build the user library shared object:
3195 gcc -shared -o libemon.so emon.o -ldl
3199 To build the application:
3201 . Compile the application source file:
3210 . Build the application:
3215 gcc -o app app.o -ldl -L. -lemon
3219 To run the application:
3221 * Start the application:
3231 The tracepoint provider package shared object is preloaded before the
3234 The application dynamically loads the instrumented user library.
3236 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3239 include::../common/ust-sit-step-tp-so.txt[]
3241 To build the instrumented user library:
3243 . In path:{emon.c}, before including path:{tpp.h}, add the
3249 #define TRACEPOINT_DEFINE
3250 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3254 . Compile the user library source file:
3259 gcc -I. -fpic -c emon.c
3263 . Build the user library shared object:
3268 gcc -shared -o libemon.so emon.o -ldl
3272 To build the application:
3274 . Compile the application source file:
3283 . Build the application:
3288 gcc -o app app.o -L. -lemon
3292 To run the application with tracing support:
3294 * Preload the tracepoint provider package shared object and
3295 start the application:
3300 LD_PRELOAD=./libtpp.so ./app
3304 To run the application without tracing support:
3306 * Start the application:
3316 The application is statically linked with the tracepoint provider
3317 package object file.
3319 The application is linked with the instrumented user library.
3321 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3324 include::../common/ust-sit-step-tp-o.txt[]
3326 To build the instrumented user library:
3328 . In path:{emon.c}, before including path:{tpp.h}, add the
3334 #define TRACEPOINT_DEFINE
3338 . Compile the user library source file:
3343 gcc -I. -fpic -c emon.c
3347 . Build the user library shared object:
3352 gcc -shared -o libemon.so emon.o
3356 To build the application:
3358 . Compile the application source file:
3367 . Build the application:
3372 gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3376 To run the instrumented application:
3378 * Start the application:
3388 The application is statically linked with the tracepoint provider
3389 package object file.
3391 The application dynamically loads the instrumented user library.
3393 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3396 include::../common/ust-sit-step-tp-o.txt[]
3398 To build the application:
3400 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3405 #define TRACEPOINT_DEFINE
3409 . Compile the application source file:
3418 . Build the application:
3423 gcc -Wl,--export-dynamic -o app app.o tpp.o \
3428 The `--export-dynamic` option passed to the linker is necessary for the
3429 dynamically loaded library to ``see'' the tracepoint symbols defined in
3432 To build the instrumented user library:
3434 . Compile the user library source file:
3439 gcc -I. -fpic -c emon.c
3443 . Build the user library shared object:
3448 gcc -shared -o libemon.so emon.o
3452 To run the application:
3454 * Start the application:
3466 .Do not use man:dlclose(3) on a tracepoint provider package
3468 Never use man:dlclose(3) on any shared object which:
3470 * Is linked with, statically or dynamically, a tracepoint provider
3472 * Calls man:dlopen(3) itself to dynamically open a tracepoint provider
3473 package shared object.
3475 This is currently considered **unsafe** due to a lack of reference
3476 counting from LTTng-UST to the shared object.
3478 A known workaround (available since glibc 2.2) is to use the
3479 `RTLD_NODELETE` flag when calling man:dlopen(3) initially. This has the
3480 effect of not unloading the loaded shared object, even if man:dlclose(3)
3483 You can also preload the tracepoint provider package shared object with
3484 the env:LD_PRELOAD environment variable to overcome this limitation.
3488 [[using-lttng-ust-with-daemons]]
3489 ===== Use noch:{LTTng-UST} with daemons
3491 If your instrumented application calls man:fork(2), man:clone(2),
3492 or BSD's man:rfork(2), without a following man:exec(3)-family
3493 system call, you must preload the path:{liblttng-ust-fork.so} shared
3494 object when starting the application.
3498 LD_PRELOAD=liblttng-ust-fork.so ./my-app
3501 If your tracepoint provider package is
3502 a shared library which you also preload, you must put both
3503 shared objects in env:LD_PRELOAD:
3507 LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3511 [[lttng-ust-pkg-config]]
3512 ===== Use noch:{pkg-config}
3514 On some distributions, LTTng-UST ships with a
3515 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3516 metadata file. If this is your case, then you can use cmd:pkg-config to
3517 build an application on the command line:
3521 gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3525 [[instrumenting-32-bit-app-on-64-bit-system]]
3526 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3528 In order to trace a 32-bit application running on a 64-bit system,
3529 LTTng must use a dedicated 32-bit
3530 <<lttng-consumerd,consumer daemon>>.
3532 The following steps show how to build and install a 32-bit consumer
3533 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3534 build and install the 32-bit LTTng-UST libraries, and how to build and
3535 link an instrumented 32-bit application in that context.
3537 To build a 32-bit instrumented application for a 64-bit target system,
3538 assuming you have a fresh target system with no installed Userspace RCU
3541 . Download, build, and install a 32-bit version of Userspace RCU:
3547 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
3548 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
3549 cd userspace-rcu-0.9.* &&
3550 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
3552 sudo make install &&
3557 . Using your distribution's package manager, or from source, install
3558 the following 32-bit versions of the following dependencies of
3559 LTTng-tools and LTTng-UST:
3562 * https://sourceforge.net/projects/libuuid/[libuuid]
3563 * http://directory.fsf.org/wiki/Popt[popt]
3564 * http://www.xmlsoft.org/[libxml2]
3567 . Download, build, and install a 32-bit version of the latest
3568 LTTng-UST{nbsp}{revision}:
3574 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.7.tar.bz2 &&
3575 tar -xf lttng-ust-latest-2.7.tar.bz2 &&
3576 cd lttng-ust-2.7.* &&
3577 ./configure --libdir=/usr/local/lib32 \
3578 CFLAGS=-m32 CXXFLAGS=-m32 \
3579 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3581 sudo make install &&
3588 Depending on your distribution,
3589 32-bit libraries could be installed at a different location than
3590 `/usr/lib32`. For example, Debian is known to install
3591 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
3593 In this case, make sure to set `LDFLAGS` to all the
3594 relevant 32-bit library paths, for example:
3598 LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
3602 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
3603 the 32-bit consumer daemon:
3609 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.7.tar.bz2 &&
3610 tar -xf lttng-tools-latest-2.7.tar.bz2 &&
3611 cd lttng-tools-2.7.* &&
3612 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
3613 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3615 cd src/bin/lttng-consumerd &&
3616 sudo make install &&
3621 . From your distribution or from source,
3622 <<installing-lttng,install>> the 64-bit versions of
3623 LTTng-UST and Userspace RCU.
3624 . Download, build, and install the 64-bit version of the
3625 latest LTTng-tools{nbsp}{revision}:
3631 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.7.tar.bz2 &&
3632 tar -xf lttng-tools-latest-2.7.tar.bz2 &&
3633 cd lttng-tools-2.7.* &&
3634 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
3635 --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
3637 sudo make install &&
3642 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
3643 when linking your 32-bit application:
3646 -m32 -L/usr/lib32 -L/usr/local/lib32 \
3647 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
3650 For example, let's rebuild the quick start example in
3651 <<tracing-your-own-user-application,Trace a user application>> as an
3652 instrumented 32-bit application:
3657 gcc -m32 -c -I. hello-tp.c
3659 gcc -m32 -o hello hello.o hello-tp.o \
3660 -L/usr/lib32 -L/usr/local/lib32 \
3661 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
3666 No special action is required to execute the 32-bit application and
3667 to trace it: use the command-line man:lttng(1) tool as usual.
3674 `tracef()` is a small LTTng-UST API designed for quick,
3675 man:printf(3)-like instrumentation without the burden of
3676 <<tracepoint-provider,creating>> and
3677 <<building-tracepoint-providers-and-user-application,building>>
3678 a tracepoint provider package.
3680 To use `tracef()` in your application:
3682 . In the C or C++ source files where you need to use `tracef()`,
3683 include `<lttng/tracef.h>`:
3688 #include <lttng/tracef.h>
3692 . In the application's source code, use `tracef()` like you would use
3700 tracef("my message: %d (%s)", my_integer, my_string);
3706 . Link your application with `liblttng-ust`:
3711 gcc -o app app.c -llttng-ust
3715 To trace the events that `tracef()` calls emit:
3717 * <<enabling-disabling-events,Create an event rule>> which matches the
3718 `lttng_ust_tracef:*` event name:
3723 lttng enable-event --userspace 'lttng_ust_tracef:*'
3728 .Limitations of `tracef()`
3730 The `tracef()` utility function was developed to make user space tracing
3731 super simple, albeit with notable disadvantages compared to
3732 <<defining-tracepoints,user-defined tracepoints>>:
3734 * All the emitted events have the same tracepoint provider and
3735 tracepoint names, respectively `lttng_ust_tracef` and `event`.
3736 * There is no static type checking.
3737 * The only event record field you actually get, named `msg`, is a string
3738 potentially containing the values you passed to `tracef()`
3739 using your own format string. This also means that you cannot filter
3740 events with a custom expression at run time because there are no
3742 * Since `tracef()` uses the C standard library's man:vasprintf(3)
3743 function behind the scenes to format the strings at run time, its
3744 expected performance is lower than with user-defined tracepoints,
3745 which do not require a conversion to a string.
3747 Taking this into consideration, `tracef()` is useful for some quick
3748 prototyping and debugging, but you should not consider it for any
3749 permanent and serious applicative instrumentation.
3755 ==== Use `tracelog()`
3757 The `tracelog()` API is very similar to <<tracef,`tracef()`>>, with
3758 the difference that it accepts an additional log level parameter.
3760 The goal of `tracelog()` is to ease the migration from logging to
3763 To use `tracelog()` in your application:
3765 . In the C or C++ source files where you need to use `tracelog()`,
3766 include `<lttng/tracelog.h>`:
3771 #include <lttng/tracelog.h>
3775 . In the application's source code, use `tracelog()` like you would use
3776 man:printf(3), except for the first parameter which is the log
3784 tracelog(TRACE_WARNING, "my message: %d (%s)",
3785 my_integer, my_string);
3791 See <<liblttng-ust-tracepoint-loglevel,Tracepoint log levels>> for
3792 a list of available log level names.
3794 . Link your application with `liblttng-ust`:
3799 gcc -o app app.c -llttng-ust
3803 To trace the events that `tracelog()` calls emit with a log level
3804 _as severe as_ a specific log level:
3806 * <<enabling-disabling-events,Create an event rule>> which matches the
3807 `lttng_ust_tracelog:*` event name and a minimum level
3813 lttng enable-event --userspace 'lttng_ust_tracelog:*'
3814 --loglevel=TRACE_WARNING
3818 To trace the events that `tracelog()` calls emit with a
3819 _specific log level_:
3821 * Create an event rule which matches the `lttng_ust_tracelog:*`
3822 event name and a specific log level:
3827 lttng enable-event --userspace 'lttng_ust_tracelog:*'
3828 --loglevel-only=TRACE_INFO
3833 [[prebuilt-ust-helpers]]
3834 === Prebuilt user space tracing helpers
3836 The LTTng-UST package provides a few helpers in the form or preloadable
3837 shared objects which automatically instrument system functions and
3840 The helper shared objects are normally found in dir:{/usr/lib}. If you
3841 built LTTng-UST <<building-from-source,from source>>, they are probably
3842 located in dir:{/usr/local/lib}.
3844 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
3847 path:{liblttng-ust-libc-wrapper.so}::
3848 path:{liblttng-ust-pthread-wrapper.so}::
3849 <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
3850 memory and POSIX threads function tracing>>.
3852 path:{liblttng-ust-cyg-profile.so}::
3853 path:{liblttng-ust-cyg-profile-fast.so}::
3854 <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
3856 path:{liblttng-ust-dl.so}::
3857 <<liblttng-ust-dl,Dynamic linker tracing>>.
3859 To use a user space tracing helper with any user application:
3861 * Preload the helper shared object when you start the application:
3866 LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
3870 You can preload more than one helper:
3875 LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
3881 [[liblttng-ust-libc-pthread-wrapper]]
3882 ==== Instrument C standard library memory and POSIX threads functions
3884 The path:{liblttng-ust-libc-wrapper.so} and
3885 path:{liblttng-ust-pthread-wrapper.so} helpers
3886 add instrumentation to some C standard library and POSIX
3890 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
3892 |TP provider name |TP name |Instrumented function
3894 .6+|`lttng_ust_libc` |`malloc` |man:malloc(3)
3895 |`calloc` |man:calloc(3)
3896 |`realloc` |man:realloc(3)
3897 |`free` |man:free(3)
3898 |`memalign` |man:memalign(3)
3899 |`posix_memalign` |man:posix_memalign(3)
3903 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
3905 |TP provider name |TP name |Instrumented function
3907 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
3908 |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
3909 |`pthread_mutex_trylock` |man:pthread_mutex_trylock(3p)
3910 |`pthread_mutex_unlock` |man:pthread_mutex_unlock(3p)
3913 When you preload the shared object, it replaces the functions listed
3914 in the previous tables by wrappers which contain tracepoints and call
3915 the replaced functions.
3918 [[liblttng-ust-cyg-profile]]
3919 ==== Instrument function entry and exit
3921 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
3922 to the entry and exit points of functions.
3924 man:gcc(1) and man:clang(1) have an option named
3925 https://gcc.gnu.org/onlinedocs/gcc/Code-Gen-Options.html[`-finstrument-functions`]
3926 which generates instrumentation calls for entry and exit to functions.
3927 The LTTng-UST function tracing helpers,
3928 path:{liblttng-ust-cyg-profile.so} and
3929 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
3930 to add tracepoints to the two generated functions (which contain
3931 `cyg_profile` in their names, hence the helper's name).
3933 To use the LTTng-UST function tracing helper, the source files to
3934 instrument must be built using the `-finstrument-functions` compiler
3937 There are two versions of the LTTng-UST function tracing helper:
3939 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
3940 that you should only use when it can be _guaranteed_ that the
3941 complete event stream is recorded without any lost event record.
3942 Any kind of duplicate information is left out.
3944 This version contains the following tracepoints:
3948 .Points instrumented by preloading path:{liblttng-ust-cyg-profile-fast.so}.
3950 |TP provider name |TP name |Instrumented points
3952 .2+|`lttng_ust_cyg_profile_fast`
3958 Address of called function.
3965 Assuming no event record is lost, having only the function addresses on
3966 entry is enough to create a call graph, since an event record always
3967 contains the ID of the CPU that generated it.
3969 You can use a tool like man:addr2line(1) to convert function addresses
3970 back to source file names and line numbers.
3972 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
3973 which also works in use cases where event records might get discarded or
3974 not recorded from application startup.
3975 In these cases, the trace analyzer needs more information to be
3976 able to reconstruct the program flow.
3978 This version contains the following tracepoints:
3982 .Points instrumented by preloading path:{liblttng-ust-cyg-profile.so}.
3984 |TP provider name |TP name |Instrumented point
3986 .2+|`lttng_ust_cyg_profile`
3992 Address of called function.
4001 Address of called function.
4008 TIP: It's sometimes a good idea to limit the number of source files that
4009 you compile with the `-finstrument-functions` option to prevent LTTng
4010 from writing an excessive amount of trace data at run time. When using
4011 man:gcc(1), you can use the
4012 `-finstrument-functions-exclude-function-list` option to avoid
4013 instrument entries and exits of specific function names.
4015 All the tracepoints that this helper contains have the
4016 <<liblttng-ust-tracepoint-loglevel,log level>> `TRACE_DEBUG_FUNCTION`.
4021 ==== Instrument the dynamic linker
4023 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4024 man:dlopen(3) and man:dlclose(3) function calls.
4027 .Functions instrumented by preloading path:{liblttng-ust-dl.so}.
4029 |TP provider name |TP name |Instrumented function
4037 Memory base address (where the dynamic linker placed the shared
4041 File system path to the loaded shared object.
4044 File size of the the loaded shared object.
4047 Last modification time (seconds since Epoch time) of the loaded shared
4054 Memory base address (where the dynamic linker placed the shared
4060 [[java-application]]
4061 === User space Java agent
4063 You can instrument a Java application which uses one of the following
4066 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4067 (JUL) core logging facilities.
4068 * http://logging.apache.org/log4j/1.2/[**Apache log4j 1.2**], since
4069 LTTng 2.6. Note that Apache Log4j{nbsp}2 is not supported.
4071 Each log statement emits an LTTng event once the
4072 application initializes the <<lttng-ust-agents,LTTng-UST Java agent>>
4076 .LTTng-UST Java agent imported by a Java application.
4077 image::java-app.png[]
4079 NOTE: We use http://openjdk.java.net/[OpenJDK] 7 for development and
4080 https://ci.lttng.org/[continuous integration], thus this version is
4081 directly supported. However, the LTTng-UST Java agent is also
4082 tested with OpenJDK 6.
4084 To use the LTTng-UST Java agent:
4086 . In the Java application's source code, import the LTTng-UST Java
4092 import org.lttng.ust.agent.LTTngAgent;
4096 . As soon as possible after the entry point of the application,
4097 initialize the LTTng-UST Java agent:
4102 LTTngAgent lttngAgent = LTTngAgent.getLTTngAgent();
4106 Any log statement that the application executes before this
4107 initialization does not emit an LTTng event.
4109 . Use `java.util.logging` and/or log4j log statements and configuration
4110 as usual. Since the LTTng-UST Java agent adds a handler to the _root_
4111 loggers, you can trace any log statement from any logger.
4113 . Before exiting the application, dispose the LTTng-UST Java agent:
4118 lttngAgent.dispose();
4122 This is not strictly necessary, but it is recommended for a clean
4123 disposal of the agent's resources.
4125 Any log statement that the application executes after this disposal does
4126 not emit an LTTng event.
4128 . Include the LTTng-UST Java agent's JAR file, path:{liblttng-ust-agent.jar},
4130 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class path]
4131 when building the Java application.
4133 path:{liblttng-ust-agent.jar} is typically located in
4134 dir:{/usr/share/java}.
4136 IMPORTANT: The LTTng-UST Java agent must be
4137 <<installing-lttng,installed>> for the logging framework your
4140 .[[jul]]Use the LTTng-UST Java agent with `java.util.logging`.
4145 import java.util.logging.Logger;
4146 import org.lttng.ust.agent.LTTngAgent;
4150 private static final int answer = 42;
4152 public static void main(String[] argv) throws Exception
4155 Logger logger = Logger.getLogger("jello");
4157 // Call this as soon as possible (before logging)
4158 LTTngAgent lttngAgent = LTTngAgent.getLTTngAgent();
4161 logger.info("some info");
4162 logger.warning("some warning");
4164 logger.finer("finer information; the answer is " + answer);
4166 logger.severe("error!");
4168 // Not mandatory, but cleaner
4169 lttngAgent.dispose();
4174 You can build this example like this:
4178 javac -cp /usr/share/java/liblttng-ust-agent.jar Test.java
4181 You can run the compiled class like this:
4185 java -cp /usr/share/java/liblttng-ust-agent.jar:. Test
4189 .[[log4j]]Use the LTTng-UST Java agent with Apache log4j 1.2.
4194 import org.apache.log4j.Logger;
4195 import org.apache.log4j.BasicConfigurator;
4196 import org.lttng.ust.agent.LTTngAgent;
4200 private static final int answer = 42;
4202 public static void main(String[] argv) throws Exception
4204 // Create and configure a logger
4205 Logger logger = Logger.getLogger(Test.class);
4206 BasicConfigurator.configure();
4208 // Call this as soon as possible (before logging)
4209 LTTngAgent lttngAgent = LTTngAgent.getLTTngAgent();
4212 logger.info("some info");
4213 logger.warn("some warning");
4215 logger.debug("debug information; the answer is " + answer);
4217 logger.error("error!");
4218 logger.fatal("fatal error!");
4220 // Not mandatory, but cleaner
4221 lttngAgent.dispose();
4226 You can build this example like this:
4230 javac -cp /usr/share/java/liblttng-ust-agent.jar:$LOG4JCP Test.java
4233 where `$LOG4JCP` is the path to log4j's JAR file.
4235 You can run the compiled class like this:
4239 java -cp /usr/share/java/liblttng-ust-agent.jar:$LOG4JCP:. Test
4243 When you <<enabling-disabling-events,create an event rule>>, use the
4244 `--jul` (`java.util.logging`) or `--log4j` (log4j) option to target
4246 <<domain,tracing domain>>. You can also use the `--loglevel` or
4247 `--loglevel-only` option to target a range of JUL/log4j log levels or a
4248 specific JUL/log4j log level.
4252 [[python-application]]
4253 === User space Python agent
4255 You can instrument a Python 2 or Python 3 application which uses the
4256 standard https://docs.python.org/3/library/logging.html[`logging`]
4259 Each log statement emits an LTTng event once the
4260 application module imports the
4261 <<lttng-ust-agents,LTTng-UST Python agent>> package.
4264 .A Python application importing the LTTng-UST Python agent.
4265 image::python-app.png[]
4267 To use the LTTng-UST Python agent:
4269 . In the Python application's source code, import the LTTng-UST Python
4279 The LTTng-UST Python agent automatically adds its logging handler to the
4280 root logger at import time.
4282 Any log statement that the application executes before this import does
4283 not emit an LTTng event.
4285 IMPORTANT: The LTTng-UST Python agent must be
4286 <<installing-lttng,installed>>.
4288 . Use log statements and logging configuration as usual.
4289 Since the LTTng-UST Python agent adds a handler to the _root_
4290 logger, you can trace any log statement from any logger.
4292 .Use the LTTng-UST Python agent.
4302 logging.basicConfig()
4303 logger = logging.getLogger('my-logger')
4306 logger.debug('debug message')
4307 logger.info('info message')
4308 logger.warn('warn message')
4309 logger.error('error message')
4310 logger.critical('critical message')
4314 if __name__ == '__main__':
4318 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
4319 logging handler which prints to the standard error stream, is not
4320 strictly required for LTTng-UST tracing to work, but in versions of
4321 Python preceding 3.2, you could see a warning message which indicates
4322 that no handler exists for the logger `my-logger`.
4325 When you <<enabling-disabling-events,create an event rule>>, use the
4326 `--python` option to target the Python
4327 <<domain,tracing domain>>. You can also use
4328 the `--loglevel` or `--loglevel-only` option to target a range of
4329 Python log levels or a specific Python log level.
4331 When an application imports the LTTng-UST Python agent, the agent tries
4332 to register to a <<lttng-sessiond,session daemon>>. Note that you must
4333 start the session daemon _before_ you start the Python application.
4334 If a session daemon is found, the agent tries to register to it
4335 during 5{nbsp}seconds, after which the application continues without
4336 LTTng tracing support. You can override this timeout value with the
4337 env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
4340 If the session daemon stops while a Python application with an imported
4341 LTTng-UST Python agent runs, the agent retries to connect and to
4342 register to a session daemon every 3{nbsp}seconds. You can override this
4343 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
4348 [[proc-lttng-logger-abi]]
4351 The `lttng-tracer` Linux kernel module, part of
4352 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger file
4353 path:{/proc/lttng-logger} when it's loaded. Any application can write
4354 text data to this file to emit an LTTng event.
4357 .An application writes to the LTTng logger file to emit an LTTng event.
4358 image::lttng-logger.png[]
4360 The LTTng logger is the quickest method--not the most efficient,
4361 however--to add instrumentation to an application. It is designed
4362 mostly to instrument shell scripts:
4366 echo "Some message, some $variable" > /proc/lttng-logger
4369 Any event that the LTTng logger emits is named `lttng_logger` and
4370 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
4371 other instrumentation points in the kernel tracing domain, **any Unix
4372 user** can <<enabling-disabling-events,create an event rule>> which
4373 matches its event name, not only the root user or users in the tracing
4376 To use the LTTng logger:
4378 * From any application, write text data to the path:{/proc/lttng-logger}
4381 The `msg` field of `lttng_logger` event records contains the
4384 NOTE: The maximum message length of an LTTng logger event is
4385 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
4386 than one event to contain the remaining data.
4388 You should not use the LTTng logger to trace a user application which
4389 can be instrumented in a more efficient way, namely:
4391 * <<c-application,C and $$C++$$ applications>>.
4392 * <<java-application,Java applications>>.
4393 * <<python-application,Python applications>>.
4396 [[instrumenting-linux-kernel]]
4397 === LTTng kernel tracepoints
4399 NOTE: This section shows how to _add_ instrumentation points to the
4400 Linux kernel. The kernel's subsystems are already thoroughly
4401 instrumented at strategic places for LTTng when you
4402 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
4406 There are two methods to instrument the Linux kernel:
4408 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
4409 tracepoint which uses the `TRACE_EVENT()` API.
4411 Choose this if you want to instrumentation a Linux kernel tree with an
4412 instrumentation point compatible with ftrace, perf, and SystemTap.
4414 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
4415 instrument an out-of-tree kernel module.
4417 Choose this if you don't need ftrace, perf, or SystemTap support.
4421 [[linux-add-lttng-layer]]
4422 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
4424 This section shows how to add an LTTng layer to existing ftrace
4425 instrumentation using the `TRACE_EVENT()` API.
4427 This section does not document the `TRACE_EVENT()` macro. You can
4428 read the following articles to learn more about this API:
4430 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part 1)]
4431 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part 2)]
4432 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part 3)]
4434 The following procedure assumes that your ftrace tracepoints are
4435 correctly defined in their own header and that they are created in
4436 one source file using the `CREATE_TRACE_POINTS` definition.
4438 To add an LTTng layer over an existing ftrace tracepoint:
4440 . Make sure the following kernel configuration options are
4446 * `CONFIG_HIGH_RES_TIMERS`
4447 * `CONFIG_TRACEPOINTS`
4450 . Build the Linux source tree with your custom ftrace tracepoints.
4451 . Boot the resulting Linux image on your target system.
4453 Confirm that the tracepoints exist by looking for their names in the
4454 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
4455 is your subsystem's name.
4457 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
4463 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
4464 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
4465 cd lttng-modules-2.8.*
4469 . In dir:{instrumentation/events/lttng-module}, relative to the root
4470 of the LTTng-modules source tree, create a header file named
4471 +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
4472 LTTng-modules tracepoint definitions using the LTTng-modules
4475 Start with this template:
4479 .path:{instrumentation/events/lttng-module/my_subsys.h}
4482 #define TRACE_SYSTEM my_subsys
4484 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
4485 #define _LTTNG_MY_SUBSYS_H
4487 #include "../../../probes/lttng-tracepoint-event.h"
4488 #include <linux/tracepoint.h>
4490 LTTNG_TRACEPOINT_EVENT(
4492 * Format is identical to TRACE_EVENT()'s version for the three
4493 * following macro parameters:
4496 TP_PROTO(int my_int, const char *my_string),
4497 TP_ARGS(my_int, my_string),
4499 /* LTTng-modules specific macros */
4501 ctf_integer(int, my_int_field, my_int)
4502 ctf_string(my_bar_field, my_bar)
4506 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
4508 #include "../../../probes/define_trace.h"
4512 The entries in the `TP_FIELDS()` section are the list of fields for the
4513 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
4514 ftrace's `TRACE_EVENT()` macro.
4516 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
4517 complete description of the available `ctf_*()` macros.
4519 . Create the LTTng-modules probe's kernel module C source file,
4520 +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
4525 .path:{probes/lttng-probe-my-subsys.c}
4527 #include <linux/module.h>
4528 #include "../lttng-tracer.h"
4531 * Build-time verification of mismatch between mainline
4532 * TRACE_EVENT() arguments and the LTTng-modules adaptation
4533 * layer LTTNG_TRACEPOINT_EVENT() arguments.
4535 #include <trace/events/my_subsys.h>
4537 /* Create LTTng tracepoint probes */
4538 #define LTTNG_PACKAGE_BUILD
4539 #define CREATE_TRACE_POINTS
4540 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
4542 #include "../instrumentation/events/lttng-module/my_subsys.h"
4544 MODULE_LICENSE("GPL and additional rights");
4545 MODULE_AUTHOR("Your name <your-email>");
4546 MODULE_DESCRIPTION("LTTng my_subsys probes");
4547 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
4548 __stringify(LTTNG_MODULES_MINOR_VERSION) "."
4549 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
4550 LTTNG_MODULES_EXTRAVERSION);
4554 . Edit path:{probes/KBuild} and add your new kernel module object
4555 next to the existing ones:
4559 .path:{probes/KBuild}
4563 obj-m += lttng-probe-module.o
4564 obj-m += lttng-probe-power.o
4566 obj-m += lttng-probe-my-subsys.o
4572 . Build and install the LTTng kernel modules:
4577 make KERNELDIR=/path/to/linux
4578 sudo make modules_install
4582 Replace `/path/to/linux` with the path to the Linux source tree where
4583 you defined and used tracepoints with ftrace's `TRACE_EVENT()` macro.
4585 Note that you can also use the
4586 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
4587 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
4588 C code that need to be executed before the event fields are recorded.
4590 The best way to learn how to use the previous LTTng-modules macros is to
4591 inspect the existing LTTng-modules tracepoint definitions in the
4592 dir:{instrumentation/events/lttng-module} header files. Compare them
4593 with the Linux kernel mainline versions in the
4594 dir:{include/trace/events} directory of the Linux source tree.
4598 [[lttng-tracepoint-event-code]]
4599 ===== Use custom C code to access the data for tracepoint fields
4601 Although we recommended to always use the
4602 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
4603 the arguments and fields of an LTTng-modules tracepoint when possible,
4604 sometimes you need a more complex process to access the data that the
4605 tracer records as event record fields. In other words, you need local
4606 variables and multiple C{nbsp}statements instead of simple
4607 argument-based expressions that you pass to the
4608 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
4610 You can use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
4611 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
4612 a block of C{nbsp}code to be executed before LTTng records the fields.
4613 The structure of this macro is:
4616 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
4618 LTTNG_TRACEPOINT_EVENT_CODE(
4620 * Format identical to the LTTNG_TRACEPOINT_EVENT()
4621 * version for the following three macro parameters:
4624 TP_PROTO(int my_int, const char *my_string),
4625 TP_ARGS(my_int, my_string),
4627 /* Declarations of custom local variables */
4630 unsigned long b = 0;
4631 const char *name = "(undefined)";
4632 struct my_struct *my_struct;
4636 * Custom code which uses both tracepoint arguments
4637 * (in TP_ARGS()) and local variables (in TP_locvar()).
4639 * Local variables are actually members of a structure pointed
4640 * to by the special variable tp_locvar.
4644 tp_locvar->a = my_int + 17;
4645 tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
4646 tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
4647 tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
4648 put_my_struct(tp_locvar->my_struct);
4657 * Format identical to the LTTNG_TRACEPOINT_EVENT()
4658 * version for this, except that tp_locvar members can be
4659 * used in the argument expression parameters of
4660 * the ctf_*() macros.
4663 ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
4664 ctf_integer(int, my_struct_a, tp_locvar->a)
4665 ctf_string(my_string_field, my_string)
4666 ctf_string(my_struct_name, tp_locvar->name)
4671 IMPORTANT: The C code defined in `TP_code()` must not have any side
4672 effects when executed. In particular, the code must not allocate
4673 memory or get resources without deallocating this memory or putting
4674 those resources afterwards.
4677 [[instrumenting-linux-kernel-tracing]]
4678 ==== Load and unload a custom probe kernel module
4680 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
4681 kernel module>> in the kernel before it can emit LTTng events.
4683 To load the default probe kernel modules and a custom probe kernel
4686 * Use the `--extra-kmod-probes` option to give extra probe modules
4687 to load when starting a root <<lttng-sessiond,session daemon>>:
4690 .Load the `my_subsys`, `usb`, and the default probe modules.
4694 sudo lttng-sessiond --extra-kmod-probes=my_subsys,usb
4699 You only need to pass the subsystem name, not the whole kernel module
4702 To load _only_ a given custom probe kernel module:
4704 * Use the `--kmod-probes` option to give the probe modules
4705 to load when starting a root session daemon:
4708 .Load only the `my_subsys` and `usb` probe modules.
4712 sudo lttng-sessiond --kmod-probes=my_subsys,usb
4717 To confirm that a probe module is loaded:
4724 lsmod | grep lttng_probe_usb
4728 To unload the loaded probe modules:
4730 * Kill the session daemon with `SIGTERM`:
4735 sudo pkill lttng-sessiond
4739 You can also use man:modprobe(8)'s `--remove` option if the session
4740 daemon terminates abnormally.
4743 [[controlling-tracing]]
4746 Once an application or a Linux kernel is
4747 <<instrumenting,instrumented>> for LTTng tracing,
4750 This section is divided in topics on how to use the various
4751 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
4752 command-line tool>>, to _control_ the LTTng daemons and tracers.
4754 Note that the <<online-lttng-manpages,Online LTTng man pages>> are
4755 more comprehensive than the guides of this section. Refer to them if
4756 your use case is not included in this section.
4760 === Start a session daemon
4762 In some situations, you need to run a <<lttng-sessiond,session daemon>>
4763 _before_ you can use the cmd:lttng command-line tool.
4765 You will see the following error when you run a command while no session
4769 Error: No session daemon is available
4772 The only command that automatically runs a session daemon is `create`,
4773 which you use to <<creating-destroying-tracing-sessions,create a tracing
4774 session>>. While this is most of the time the first operation that you
4775 do, sometimes it's not. Some examples are:
4777 * <<list-instrumentation-points,List the available instrumentation points>>.
4778 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
4780 [[tracing-group]] Each Unix user must have its own running session
4781 daemon to trace user applications. The session daemon that the root user
4782 starts is the only one allowed to control the LTTng kernel tracer. Users
4783 that are part of the _tracing group_ can control the root session
4784 daemon. The default tracing group name is `tracing`; you can set it to
4785 something else with the `--group` option when you start the root session
4788 To start a user session daemon:
4790 * Run cmd:lttng-sessiond:
4795 lttng-sessiond --daemonize
4799 To start the root session daemon:
4801 * Run cmd:lttng-sessiond as the root user:
4806 sudo lttng-sessiond --daemonize
4810 In both cases, remove the `--daemonize` option to start the session
4811 daemon in foreground.
4813 To stop a session daemon, use cmd:kill on its process ID (standard
4816 Note that some Linux distributions could manage the LTTng session daemon
4817 as a service. In this case, you should use the service manager to
4818 start, restart, and stop session daemons.
4821 [[creating-destroying-tracing-sessions]]
4822 === Create and destroy a tracing session
4824 Almost all the LTTng control operations happen in the scope of
4825 a <<tracing-session,tracing session>>, which is the dialogue between the
4826 <<lttng-sessiond,session daemon>> and you.
4828 To create a tracing session with a generated name:
4830 * Use the `create` command:
4839 The created tracing session's name is `auto` followed by the
4842 To create a tracing session with a specific name:
4844 * Use the optional argument of the `create` command:
4849 lttng create my-session
4853 Replace `my-session` with the specific tracing session name.
4855 LTTng appends the creation date to the created tracing session's name.
4857 LTTng writes the traces of a tracing session in
4858 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
4859 name of the tracing session. Note that the env:LTTNG_HOME environment
4860 variable defaults to `$HOME` if not set.
4862 To output LTTng traces to a non-default location:
4864 * Use the `--output` option of the `create` command:
4869 lttng create my-session --output=/tmp/some-directory
4873 You may create as many tracing sessions as you wish.
4875 To list all the existing tracing sessions for your Unix user:
4877 * Use the `list` command:
4886 When you create a tracing session, it is set as the _current tracing
4887 session_. The following man:lttng(1) commands operate on the current
4888 tracing session when you don't specify one:
4890 [role="list-3-cols"]
4906 To change the current tracing session:
4908 * Use the `set-session` command:
4913 lttng set-session new-session
4917 Replace `new-session` by the name of the new current tracing session.
4919 When you are done tracing in a given tracing session, you can destroy
4920 it. This operation frees the resources taken by the tracing session
4921 to destroy; it does not destroy the trace data that LTTng wrote for
4922 this tracing session.
4924 To destroy the current tracing session:
4926 * Use the `destroy` command:
4936 [[list-instrumentation-points]]
4937 === List the available instrumentation points
4939 The <<lttng-sessiond,session daemon>> can query the running instrumented
4940 user applications and the Linux kernel to get a list of available
4941 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
4942 they are tracepoints and system calls. For the user space tracing
4943 domain, they are tracepoints. For the other tracing domains, they are
4946 To list the available instrumentation points:
4948 * Use the `list` command with the requested tracing domain's option
4952 * `--kernel`: Linux kernel tracepoints (your Unix user must be a root
4953 user, or it must be a member of the tracing group).
4954 * `--kernel --syscall`: Linux kernel system calls (your Unix user must
4955 be a root user, or it must be a member of the tracing group).
4956 * `--userspace`: user space tracepoints.
4957 * `--jul`: `java.util.logging` loggers.
4958 * `--log4j`: Apache log4j loggers.
4959 * `--python`: Python loggers.
4962 .List the available user space tracepoints.
4966 lttng list --userspace
4970 .List the available Linux kernel system call tracepoints.
4974 lttng list --kernel --syscall
4979 [[enabling-disabling-events]]
4980 === Create and enable an event rule
4982 Once you <<creating-destroying-tracing-sessions,create a tracing
4983 session>>, you can create <<event,event rules>> with the
4984 `enable-event` command.
4986 You specify each condition with a command-line option. The available
4987 condition options are shown in the following table.
4989 [role="growable",cols="asciidoc,asciidoc,default"]
4990 .Condition command-line options for the `enable-event` command.
4992 |Option |Description |Applicable tracing domains
4998 . +--probe=__ADDR__+
4999 . +--function=__ADDR__+
5002 Instead of using the default _tracepoint_ instrumentation type, use:
5004 . A Linux system call.
5005 . A Linux https://lwn.net/Articles/132196/[KProbe] (symbol or address).
5006 . The entry and return points of a Linux function (symbol or address).
5010 |First positional argument.
5013 Tracepoint or system call name. In the case of a Linux KProbe or
5014 function, this is a custom name given to the event rule. With the
5015 JUL, log4j, and Python domains, this is a logger name.
5017 With a tracepoint, logger, or system call name, the last character
5018 can be `*` to match anything that remains.
5025 . +--loglevel=__LEVEL__+
5026 . +--loglevel-only=__LEVEL__+
5029 . Match only tracepoints or log statements with a logging level at
5030 least as severe as +__LEVEL__+.
5031 . Match only tracepoints or log statements with a logging level
5032 equal to +__LEVEL__+.
5034 You can get the list of available logging level names with
5035 `lttng enable-event --help`.
5037 |User space, JUL, log4j, and Python.
5039 |+--exclude=__EXCLUSIONS__+
5042 When you use a `*` character at the end of the tracepoint or logger
5043 name (first positional argument), exclude the specific names in the
5044 comma-delimited list +__EXCLUSIONS__+.
5047 User space, JUL, log4j, and Python.
5049 |+--filter=__EXPR__+
5052 Match only events which satisfy the expression +__EXPR__+.
5054 +__EXPR__+ is a C-like logical expression where identifiers are event
5055 fields (preceded with `$ctx.` for context fields). Nested expressions
5056 with `(` and `)`, and all the logical and comparison operators of the C
5057 language are supported. The precedence rules of those operators are the
5058 same as in the C language.
5060 When a comparison includes a non-existent event field, the whole filter
5061 expression evaluates to false.
5063 C integer and floating point number constants are supported, as well as
5064 literal strings between double quotes (`"`). Literal strings can
5065 contain a wildcard character (`*`) at the end to match anything that
5066 remains. This wildcard can be escaped using `\*`.
5068 Note that, although it is possible to use this option with the JUL,
5069 log4j, and Python tracing domains, the tracer evalutes the expression
5070 against the equivalent user space event.
5077 for more details about those command-line options.
5079 You attach an event rule to a <<channel,channel>> on creation. If you
5080 do not specify the channel with the `--channel` option, and if the event
5081 rule to create is the first in its <<domain,tracing domain>> for a given
5082 tracing session, then LTTng creates a _default channel_ for you. This
5083 default channel is reused in subsequent invocations of the
5084 `enable-event` command for the same tracing domain.
5086 An event rule is always enabled at creation time.
5088 The following examples show how you can combine the previous
5089 command-line options to create simple to more complex event rules.
5091 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5095 lttng enable-event --kernel sched_switch
5099 .Create an event rule matching four Linux kernel system calls (default channel).
5103 lttng enable-event --kernel --syscall open,write,read,close
5107 .Create an event rule matching a Linux kernel tracepoint with a filter expression (default channel).
5111 lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
5114 IMPORTANT: Make sure to always quote the filter string when you
5115 use man:lttng(1) from a shell.
5118 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
5122 lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
5125 IMPORTANT: Make sure to always quote the wildcard character when you
5126 use man:lttng(1) from a shell.
5129 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
5133 lttng enable-event --python my-app.'*' \
5134 --exclude='my-app.module,my-app.hello'
5138 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
5142 lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
5146 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
5150 lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
5154 The event rules of a given channel form a whitelist: as soon as an
5155 emitted event passes one of them, LTTng can record the event. For
5156 example, an event named `my_app:my_tracepoint` emitted from a user space
5157 tracepoint with a `TRACE_ERROR` log level passes both of the following
5162 lttng enable-event --userspace my_app:my_tracepoint
5163 lttng enable-event --userspace my_app:my_tracepoint \
5164 --loglevel=TRACE_INFO
5167 The second event rule is redundant: the first one includes
5171 [[disable-event-rule]]
5172 === Disable an event rule
5174 To disable an event rule that you <<enabling-disabling-events,created>>
5175 previously, use the `disable-event` command. This command disables _all_
5176 the event rules (of a given tracing domain and channel) which match an
5177 instrumentation point. The other conditions are not supported as of
5178 LTTng{nbsp}{revision}.
5180 The LTTng tracer does not record an emitted event which passes
5181 a _disabled_ event rule.
5183 .Disable an event rule matching a Python logger (default channel).
5187 lttng disable-event --python my-logger
5191 .Disable an event rule matching all `java.util.logging` loggers (default channel).
5195 lttng disable-event --jul '*'
5199 .Disable _all_ the event rules of the default channel.
5201 The `--all-events` option is not, like the `--all` option of
5202 `enable-event`, the equivalent of the event name `*` (wildcard): it
5203 disables _all_ the event rules of a given channel.
5207 lttng disable-event --jul --all-events
5211 NOTE: You cannot delete an event rule once you create it.
5215 === Get the status of a tracing session
5217 To get the status of a tracing session, that is, its channels, event
5218 rules, and their attributes:
5220 * Use the `list` command with the tracing session's name:
5225 lttng list my-session
5229 Replace `my-session` with your tracing session's name.
5232 [[basic-tracing-session-control]]
5233 === Start and stop a tracing session
5235 Once you <<creating-destroying-tracing-sessions,create a tracing
5237 <<enabling-disabling-events,create one or more event rules>>,
5238 you can start and stop the tracers for this tracing session.
5240 To start tracing in the current tracing session:
5242 * Use the `start` command:
5251 To stop tracing in the current tracing session:
5253 * Use the `stop` command:
5262 LTTng is very flexible: you can launch user applications before
5263 or after the you start the tracers. The tracers only record the events
5264 if they pass enabled event rules and if they occur while the tracers are
5268 [[enabling-disabling-channels]]
5269 === Create a channel
5271 Once you create a tracing session, you can create a <<channel,channel>>
5272 with the `enable-channel` command.
5274 Note that LTTng automatically creates a default channel when, for a
5275 given <<domain,tracing domain>>, no channels exist and you
5276 <<enabling-disabling-events,create>> the first event rule. This default
5277 channel is named `channel0` and its attributes are set to reasonable
5278 values. Therefore, you only need to create a channel when you need
5279 non-default attributes.
5281 You specify each non-default channel attribute with a command-line
5282 option when you use the `enable-channel` command. The available
5283 command-line options are:
5285 [role="growable",cols="asciidoc,asciidoc"]
5286 .Command-line options for the `enable-channel` command.
5288 |Option |Description
5294 <<channel-overwrite-mode-vs-discard-mode,event loss mode>> instead of
5295 the default _discard_ mode.
5297 |`--buffers-pid` (user space tracing domain only)
5300 Use the per-process <<channel-buffering-schemes,buffering scheme>>
5301 instead of the default per-user buffering scheme.
5303 |+--subbuf-size=__SIZE__+
5306 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
5307 either for each Unix user (default), or for each instrumented process.
5309 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5311 |+--num-subbuf=__COUNT__+
5314 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
5315 for each Unix user (default), or for each instrumented process.
5317 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5319 |+--tracefile-size=__SIZE__+
5322 Set the maximum size of each trace file that this channel writes within
5323 a stream to +__SIZE__+ bytes instead of no maximum.
5325 See <<tracefile-rotation,Trace file count and size>>.
5327 |+--tracefile-count=__COUNT__+
5330 Limit the number of trace files that this channel creates to
5331 +__COUNT__+ channels instead of no limit.
5333 See <<tracefile-rotation,Trace file count and size>>.
5335 |+--switch-timer=__PERIODUS__+
5338 Set the <<channel-switch-timer,switch timer period>>
5339 to +__PERIODUS__+{nbsp}µs.
5341 |+--read-timer=__PERIODUS__+
5344 Set the <<channel-read-timer,read timer period>>
5345 to +__PERIODUS__+{nbsp}µs.
5347 |+--output=__TYPE__+ (Linux kernel tracing domain only)
5350 Set the channel's output type to +__TYPE__+, either `mmap` or `splice`.
5355 for more details about those command-line options.
5357 You can only create a channel in the Linux kernel and user space
5358 <<domain,tracing domains>>: other tracing domains have their own
5359 channel created on the fly when
5360 <<enabling-disabling-events,creating event rules>>.
5364 Because of a current LTTng limitation, you must create all channels
5365 _before_ you <<basic-tracing-session-control,start tracing>> in a given
5366 tracing session, that is, before the first time you run `lttng start`.
5368 Since LTTng automatically creates a default channel when you use the
5369 `enable-event` command with a specific tracing domain, you cannot, for
5370 example, create a Linux kernel event rule, start tracing, and then
5371 create a user space event rule, because no user space channel exists yet
5372 and it's too late to create one.
5374 For this reason, make sure to configure your channels properly
5375 before starting the tracers for the first time!
5378 The following examples show how you can combine the previous
5379 command-line options to create simple to more complex channels.
5381 .Create a Linux kernel channel with default attributes.
5385 lttng enable-channel --kernel my-channel
5389 .Create a user space channel with 4 sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
5393 lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
5394 --buffers-pid my-channel
5398 .Create a Linux kernel channel which rotates 8 trace files of 4{nbsp}MiB each for each stream
5402 lttng enable-channel --kernel --tracefile-count=8 \
5403 --tracefile-size=4194304 my-channel
5407 .Create a user space channel in overwrite (or _flight recorder_) mode.
5411 lttng enable-channel --userspace --overwrite my-channel
5415 You can <<enabling-disabling-events,create>> the same event rule in
5416 two different channels:
5420 lttng enable-event --userspace --channel=my-channel app:tp
5421 lttng enable-event --userspace --channel=other-channel app:tp
5424 If both channels are enabled, when a tracepoint named `app:tp` is
5425 reached, LTTng records two events, one for each channel.
5429 === Disable a channel
5431 To disable a specific channel that you <<enabling-disabling-channels,created>>
5432 previously, use the `disable-channel` command.
5434 .Disable a specific Linux kernel channel.
5438 lttng disable-channel --kernel my-channel
5442 The state of a channel precedes the individual states of event rules
5443 attached to it: event rules which belong to a disabled channel, even if
5444 they are enabled, are also considered disabled.
5448 === Add context fields to a channel
5450 Event record fields in trace files provide important information about
5451 events that occured previously, but sometimes some external context may
5452 help you solve a problem faster. Examples of context fields are:
5454 * The **process ID**, **thread ID**, **process name**, and
5455 **process priority** of the thread in which the event occurs.
5456 * The **hostname** of the system on which the event occurs.
5457 * The current values of many possible **performance counters** using
5459 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
5461 ** Branch instructions, misses, and loads.
5464 To get the full list of available context fields, see
5465 `lttng add-context --help`. Some context fields are reserved for a
5466 specific <<domain,tracing domain>> (Linux kernel or user space).
5468 You add context fields to <<channel,channels>>. All the events
5469 that a channel with added context fields records contain those fields.
5471 To add context fields to one or all the channels of a given tracing
5472 session, use the `add-context` command.
5474 .Add context fields to all the channels of the current tracing session.
5476 The following command line adds the virtual process identifier and
5477 the per-thread CPU cycles count fields to all the user space channels
5478 of the current tracing session.
5482 lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
5486 .Add a context field to a specific channel.
5488 The following command line adds the thread identifier context field
5489 to the Linux kernel channel named `my-channel` in the current
5494 lttng add-context --kernel --channel=my-channel --type=tid
5498 NOTE: You cannot remove context fields from a channel once you add it.
5503 === Track process IDs
5505 It's often useful to allow only specific process IDs (PIDs) to emit
5506 events. For example, you may wish to record all the system calls made by
5507 a given process (Ă la http://linux.die.net/man/1/strace[strace]).
5509 The `track` and `untrack` commands serve this purpose. Both commands
5510 operate on a whitelist of process IDs. You _add_ entries to this
5511 whitelist with the `track` command and remove entries with the `untrack`
5512 command. Any process which has one of the PIDs in the whitelist is
5513 allowed to emit LTTng events which pass an enabled <<event,event rule>>.
5515 NOTE: The PID tracker tracks the _numeric process IDs_. Should a
5516 process with a given tracked ID exit and another process be given this
5517 ID, then the latter would also be allowed to emit events.
5519 .Track and untrack process IDs.
5521 For the sake of the following example, assume the target system has 16
5525 <<creating-destroying-tracing-sessions,create a tracing session>>,
5526 the whitelist contains all the possible PIDs:
5529 .All PIDs are tracked.
5530 image::track-all.png[]
5532 When the whitelist is full and you use the `track` command to specify
5533 some PIDs to track, LTTng first clears the whitelist, then it tracks
5534 the specific PIDs. After:
5538 lttng track --pid=3,4,7,10,13
5544 .PIDs 3, 4, 7, 10, and 13 are tracked.
5545 image::track-3-4-7-10-13.png[]
5547 You can add more PIDs to the whitelist afterwards:
5551 lttng track --pid=1,15,16
5557 .PIDs 1, 15, and 16 are added to the whitelist.
5558 image::track-1-3-4-7-10-13-15-16.png[]
5560 The `untrack` command removes entries from the PID tracker's whitelist.
5561 Given the previous example, the following command:
5565 lttng untrack --pid=3,7,10,13
5568 leads to this whitelist:
5571 .PIDs 3, 7, 10, and 13 are removed from the whitelist.
5572 image::track-1-4-15-16.png[]
5574 LTTng can track all possible PIDs again using the `--all` option:
5578 lttng track --pid --all
5581 The result is, again:
5584 .All PIDs are tracked.
5585 image::track-all.png[]
5588 .Track only specific PIDs
5590 A very typical use case with PID tracking is to start with an empty
5591 whitelist, then <<basic-tracing-session-control,start the tracers>>,
5592 and then add PIDs manually while tracers are active. You can accomplish
5593 this by using the `--all` option of the `untrack` command to clear the
5594 whitelist after you create a tracing session:
5598 lttng untrack --pid --all
5604 .No PIDs are tracked.
5605 image::untrack-all.png[]
5607 If you trace with this whitelist configuration, the tracer records no
5608 events for this <<domain,tracing domain>> because no processes are
5609 tracked. You can use the `track` command as usual to track specific
5614 lttng track --pid=6,11
5620 .PIDs 6 and 11 are tracked.
5621 image::track-6-11.png[]
5626 [[saving-loading-tracing-session]]
5627 === Save and load tracing session configurations
5629 Configuring a <<tracing-session,tracing session>> can be long. Some of
5630 the tasks involved are:
5632 * <<enabling-disabling-channels,Create channels>> with
5633 specific attributes.
5634 * <<adding-context,Add context fields>> to specific channels.
5635 * <<enabling-disabling-events,Create event rules>> with specific log
5636 level and filter conditions.
5638 If you use LTTng to solve real world problems, chances are you have to
5639 record events using the same tracing session setup over and over,
5640 modifying a few variables each time in your instrumented program
5641 or environment. To avoid constant tracing session reconfiguration,
5642 the cmd:lttng command-line tool can save and load tracing session
5643 configurations to/from XML files.
5645 To save a given tracing session configuration:
5647 * Use the `save` command:
5652 lttng save my-session
5656 Replace `my-session` with the name of the tracing session to save.
5658 LTTng saves tracing session configurations to
5659 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
5660 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
5661 the `--output-path` option to change this destination directory.
5663 LTTng saves all configuration parameters, for example:
5665 * The tracing session name.
5666 * The trace data output path.
5667 * The channels with their state and all their attributes.
5668 * The context fields you added to channels.
5669 * The event rules with their state, log level and filter conditions.
5671 To load a tracing session:
5673 * Use the `load` command:
5678 lttng load my-session
5682 Replace `my-session` with the name of the tracing session to load.
5684 When LTTng loads a configuration, it restores your saved tracing session
5685 as if you just configured it manually.
5687 See man:lttng(1) for the complete list of command-line options. You
5688 can also save and load all many sessions at a time, and decide in which
5689 directory to output the XML files.
5692 [[sending-trace-data-over-the-network]]
5693 === Send trace data over the network
5695 LTTng can send the recorded trace data to a remote system over the
5696 network instead of writing it to the local file system.
5698 To send the trace data over the network:
5700 . On the _remote_ system (which can also be the target system),
5701 start an LTTng <<lttng-relayd,relay daemon>>:
5710 . On the _target_ system, create a tracing session configured to
5711 send trace data over the network:
5716 lttng create my-session --set-url=net://remote-system
5720 Replace `remote-system` by the host name or IP address of the
5721 remote system. See `lttng create --help` for the exact URL format.
5723 . On the target system, use the cmd:lttng command-line tool as usual.
5724 When tracing is active, the target's consumer daemon sends sub-buffers
5725 to the relay daemon running on the remote system intead of flushing
5726 them to the local file system. The relay daemon writes the received
5727 packets to the local file system.
5729 The relay daemon writes trace files to
5730 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
5731 +__hostname__+ is the host name of the target system and +__session__+
5732 is the tracing session name. Note that the env:LTTNG_HOME environment
5733 variable defaults to `$HOME` if not set. Use the `--output` option of
5734 cmd:lttng-relayd to write trace files to another base directory.
5739 === View events as LTTng emits them (noch:{LTTng} live)
5741 LTTng live is a network protocol implemented by the
5742 <<lttng-relayd,relay daemon>> to allow compatible trace viewers to
5743 display events as LTTng emits them on the target system while tracing
5746 The relay daemon creates a _tee_: it forwards the trace data to both
5747 the local file system and to connected live viewers:
5750 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
5755 . On the _target system_, create a <<tracing-session,tracing session>>
5761 lttng create my-session --live
5765 This spawns a local relay daemon.
5767 . Start the live viewer and configure it to connect to the relay
5768 daemon. For example, with http://diamon.org/babeltrace[Babeltrace]:
5773 babeltrace --input-format=lttng-live net://localhost/host/hostname/my-session
5780 * `hostname` with the host name of the target system.
5781 * `my-session` with the name of the tracing session to view.
5784 . Configure the tracing session as usual with the cmd:lttng
5785 command-line tool, and <<basic-tracing-session-control,start tracing>>.
5787 You can list the available live tracing sessions with Babeltrace:
5791 babeltrace --input-format=lttng-live net://localhost
5794 You can start the relay daemon on another system. In this case, you need
5795 to specify the relay daemon's URL when you create the tracing session
5796 with the `--set-url` option. You also need to replace `localhost`
5797 in the procedure above with the host name of the system on which the
5798 relay daemon is running.
5800 See man:lttng(1) and man:lttng-relayd(8) for the complete list of
5801 command-line options.
5805 [[taking-a-snapshot]]
5806 === Take a snapshot of the current sub-buffers of a tracing session
5808 The normal behavior of LTTng is to append full sub-buffers to growing
5809 trace data files. This is ideal to keep a full history of the events
5810 that occurred on the target system, but it can
5811 represent too much data in some situations. For example, you may wish
5812 to trace your application continuously until some critical situation
5813 happens, in which case you only need the latest few recorded
5814 events to perform the desired analysis, not multi-gigabyte trace files.
5816 With the `snapshot` command, you can take a snapshot of the current
5817 sub-buffers of a given <<tracing-session,tracing session>>. LTTng can
5818 write the snapshot to the local file system or send it over the network.
5822 . Create a tracing session in _snapshot mode_:
5827 lttng create my-session --snapshot
5831 The <<channel-overwrite-mode-vs-discard-mode,event loss mode>> of
5832 <<channel,channels>> created in this mode is automatically set to
5833 _overwrite_ (flight recorder mode).
5835 . Configure the tracing session as usual with the cmd:lttng
5836 command-line tool, and <<basic-tracing-session-control,start tracing>>.
5838 . **Optional**: When you need to take a snapshot, stop tracing.
5840 You can take a snapshot when the tracers are active, but if you stop
5841 them first, you are sure that the data in the sub-buffers does not
5842 change before you actually take the snapshot.
5849 lttng snapshot record --name=my-first-snapshot
5853 LTTng writes the current sub-buffers of all the current tracing
5854 session's channels to trace files on the local file system. Those trace
5855 files have `my-first-snapshot` in their name.
5857 There is no difference between the format of a normal trace file and the
5858 format of a snapshot: viewers of LTTng traces also support LTTng
5861 By default, LTTng writes snapshot files to the path shown by
5862 `lttng snapshot list-output`. You can change this path or decide to send
5863 snapshots over the network using either:
5865 . An output path or URL that you specify when you create the
5867 . An snapshot output path or URL that you add using
5868 `lttng snapshot add-output`
5869 . An output path or URL that you provide directly to the
5870 `lttng snapshot record` command.
5872 Method 3 overrides method 2, which overrides method 1. When you
5873 specify a URL, a relay daemon must listen on a remote system (see
5874 <<sending-trace-data-over-the-network,Send trace data over the network>>).
5879 === Use the machine interface
5881 With any command of the cmd:lttng command-line tool, you can use the
5882 `--mi=xml` argument (before the command name) to get an XML machine
5883 interface output, for example:
5887 lttng --mi=xml enable-event --kernel --syscall open
5890 A schema definition (XSD) is
5891 https://github.com/lttng/lttng-tools/blob/stable-{revision}/src/common/mi_lttng.xsd[available]
5892 to ease the integration with external tools as much as possible.
5896 [[persistent-memory-file-systems]]
5897 === Record trace data on persistent memory file systems
5899 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
5900 (NVRAM) is random-access memory that retains its information when power
5901 is turned off (non-volatile). Systems with such memory can store data
5902 structures in RAM and retrieve them after a reboot, without flushing
5903 to typical _storage_.
5905 Linux supports NVRAM file systems thanks to either
5906 http://pramfs.sourceforge.net/[PRAMFS] or
5907 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
5908 (requires Linux 4.1+).
5910 This section does not describe how to operate such file systems;
5911 we assume that you have a working persistent memory file system.
5913 When you create a <<tracing-session,tracing session>>, you can specify
5914 the path of the shared memory holding the sub-buffers. If you specify a
5915 location on an NVRAM file system, then you can retrieve the latest
5916 recorded trace data when the system reboots after a crash.
5918 To record trace data on a persistent memory file system and retrieve the
5919 trace data after a system crash:
5921 . Create a tracing session with a sub-buffer shared memory path located
5922 on an NVRAM file system:
5927 lttng create my-session -shm-path=/path/to/shm
5931 . Configure the tracing session as usual with the cmd:lttng
5932 command-line tool, and <<basic-tracing-session-control,start tracing>>.
5934 . After a system crash, use the cmd:lttng-crash command-line tool to
5935 view the trace data recorded on the NVRAM file system:
5940 lttng-crash /path/to/shm
5944 The binary layout of the ring buffer files is not exactly the same as
5945 the trace files layout. This is why you need to use the cmd:lttng-crash
5946 utility instead of your preferred trace viewer directly.
5948 To convert the ring buffer files to LTTng trace files:
5950 * Use the `--extract` option of cmd:lttng-crash:
5955 lttng-crash --extract=/path/to/trace /path/to/shm
5959 See man:lttng-crash(1) for the complete list of command-line options.
5965 This section presents various references for LTTng packages such as
5966 links to online manpages, tables that the rest of the text needs,
5967 descriptions of library functions, and more.
5970 [[online-lttng-manpages]]
5971 === Online noch:{LTTng} manpages
5973 LTTng packages currently install the following link:/man[man pages],
5974 available online using the links below:
5978 ** man:lttng-crash(1)
5979 ** man:lttng-sessiond(8)
5980 ** man:lttng-relayd(8)
5982 ** man:lttng-gen-tp(1)
5984 ** man:lttng-ust-cyg-profile(3)
5985 ** man:lttng-ust-dl(3)
5989 === noch:{LTTng-UST}
5991 This section presents references of the LTTng-UST package.
5995 ==== noch:{LTTng-UST} library (+liblttng‑ust+)
5997 The LTTng-UST library, or `liblttng-ust`, is the main shared object
5998 against which user applications are linked to make LTTng user space
6001 The <<c-application,C application>> guide shows the complete
6002 process to instrument, build and run a C/$$C++$$ application using
6003 LTTng-UST, while this section contains a few important tables.
6006 [[liblttng-ust-tp-fields]]
6007 ===== Tracepoint fields macros (for `TP_FIELDS()`)
6009 The available macros to define tracepoint fields, which you must use
6010 within `TP_FIELDS()` in `TRACEPOINT_EVENT()`, are:
6012 [role="func-desc growable",cols="asciidoc,asciidoc"]
6013 .Available macros to define LTTng-UST tracepoint fields
6015 |Macro |Description and parameters
6018 +ctf_integer(__t__, __n__, __e__)+
6020 +ctf_integer_nowrite(__t__, __n__, __e__)+
6022 Standard integer, displayed in base 10.
6025 Integer C type (`int`, `long`, `size_t`, ...).
6031 Argument expression.
6033 |+ctf_integer_hex(__t__, __n__, __e__)+
6035 Standard integer, displayed in base 16.
6044 Argument expression.
6046 |+ctf_integer_network(__t__, __n__, __e__)+
6048 Integer in network byte order (big-endian), displayed in base 10.
6057 Argument expression.
6059 |+ctf_integer_network_hex(__t__, __n__, __e__)+
6061 Integer in network byte order, displayed in base 16.
6070 Argument expression.
6073 +ctf_float(__t__, __n__, __e__)+
6075 +ctf_float_nowrite(__t__, __n__, __e__)+
6077 Floating point number.
6080 Floating point number C type (`float` or `double`).
6086 Argument expression.
6089 +ctf_string(__n__, __e__)+
6091 +ctf_string_nowrite(__n__, __e__)+
6093 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
6099 Argument expression.
6102 +ctf_array(__t__, __n__, __e__, __s__)+
6104 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
6106 Statically-sized array of integers
6109 Array element C type.
6115 Argument expression.
6121 +ctf_array_text(__t__, __n__, __e__, __s__)+
6123 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
6125 Statically-sized array, printed as text.
6127 The string does not need to be null-terminated.
6130 Array element C type (always `char`).
6136 Argument expression.
6142 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
6144 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
6146 Dynamically-sized array of integers.
6148 The type of +__E__+ must be unsigned.
6151 Array element C type.
6157 Argument expression.
6160 Length expression C type.
6166 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
6168 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
6170 Dynamically-sized array, displayed as text.
6172 The string does not need to be null-terminated.
6174 The type of +__E__+ must be unsigned.
6176 The behaviour is undefined if +__e__+ is `NULL`.
6179 Sequence element C type (always `char`).
6185 Argument expression.
6188 Length expression C type.
6194 The `_nowrite` versions omit themselves from the session trace, but are
6195 otherwise identical. This means the tracer does not write the `_nowrite`
6196 fields to the trace. Their primary purpose is to make some of the event
6197 context available to the <<enabling-disabling-events,event filters>>
6198 without having to commit the data to sub-buffers.
6201 [[liblttng-ust-tracepoint-loglevel]]
6202 ===== Tracepoint log levels (for `TRACEPOINT_LOGLEVEL()`)
6204 The following table shows the available log level values for the
6205 `TRACEPOINT_LOGLEVEL()` macro:
6211 Action must be taken immediately.
6214 Critical conditions.
6223 Normal, but significant, condition.
6226 Informational message.
6228 `TRACE_DEBUG_SYSTEM`::
6229 Debug information with system-level scope (set of programs).
6231 `TRACE_DEBUG_PROGRAM`::
6232 Debug information with program-level scope (set of processes).
6234 `TRACE_DEBUG_PROCESS`::
6235 Debug information with process-level scope (set of modules).
6237 `TRACE_DEBUG_MODULE`::
6238 Debug information with module (executable/library) scope (set of units).
6240 `TRACE_DEBUG_UNIT`::
6241 Debug information with compilation unit scope (set of functions).
6243 `TRACE_DEBUG_FUNCTION`::
6244 Debug information with function-level scope.
6246 `TRACE_DEBUG_LINE`::
6247 Debug information with line-level scope (TRACEPOINT_EVENT default).
6250 Debug-level message.
6252 Log levels `TRACE_EMERG` through `TRACE_INFO` and `TRACE_DEBUG` match
6253 http://man7.org/linux/man-pages/man3/syslog.3.html[syslog]
6254 level semantics. Log levels `TRACE_DEBUG_SYSTEM` through `TRACE_DEBUG`
6255 offer more fine-grained selection of debug information.
6258 [[lttng-modules-ref]]
6259 === noch:{LTTng-modules}
6261 This section presents references of the LTTng-modules package.
6265 [[lttng-modules-tp-fields]]
6266 ==== Tracepoint fields macros (for `TP_FIELDS()`)
6268 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
6269 tracepoint fields, which must be listed within `TP_FIELDS()` in
6270 `LTTNG_TRACEPOINT_EVENT()`, are:
6272 [role="func-desc growable",cols="asciidoc,asciidoc"]
6273 .Available macros to define LTTng-modules tracepoint fields
6275 |Macro |Description and parameters
6278 +ctf_integer(__t__, __n__, __e__)+
6280 +ctf_integer_nowrite(__t__, __n__, __e__)+
6282 +ctf_user_integer(__t__, __n__, __e__)+
6284 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
6286 Standard integer, displayed in base 10.
6289 Integer C type (`int`, `long`, `size_t`, ...).
6295 Argument expression.
6298 +ctf_integer_hex(__t__, __n__, __e__)+
6300 +ctf_user_integer_hex(__t__, __n__, __e__)+
6302 Standard integer, displayed in base 16.
6311 Argument expression.
6313 |+ctf_integer_oct(__t__, __n__, __e__)+
6315 Standard integer, displayed in base 8.
6324 Argument expression.
6327 +ctf_integer_network(__t__, __n__, __e__)+
6329 +ctf_user_integer_network(__t__, __n__, __e__)+
6331 Integer in network byte order (big-endian), displayed in base 10.
6340 Argument expression.
6343 +ctf_integer_network_hex(__t__, __n__, __e__)+
6345 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
6347 Integer in network byte order, displayed in base 16.
6356 Argument expression.
6359 +ctf_string(__n__, __e__)+
6361 +ctf_string_nowrite(__n__, __e__)+
6363 +ctf_user_string(__n__, __e__)+
6365 +ctf_user_string_nowrite(__n__, __e__)+
6367 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
6373 Argument expression.
6376 +ctf_array(__t__, __n__, __e__, __s__)+
6378 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
6380 +ctf_user_array(__t__, __n__, __e__, __s__)+
6382 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
6384 Statically-sized array of integers
6387 Array element C type.
6393 Argument expression.
6399 +ctf_array_text(__t__, __n__, __e__, __s__)+
6401 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
6403 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
6405 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
6407 Statically-sized array, printed as text.
6409 The string does not need to be null-terminated.
6412 Array element C type (always `char`).
6418 Argument expression.
6424 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
6426 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
6428 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
6430 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
6432 Dynamically-sized array of integers.
6434 The type of +__E__+ must be unsigned.
6437 Array element C type.
6443 Argument expression.
6446 Length expression C type.
6451 |+ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
6453 Dynamically-sized array of integers, displayed in base 16.
6455 The type of +__E__+ must be unsigned.
6458 Array element C type.
6464 Argument expression.
6467 Length expression C type.
6472 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
6474 Dynamically-sized array of integers in network byte order (big-endian),
6475 displayed in base 10.
6477 The type of +__E__+ must be unsigned.
6480 Array element C type.
6486 Argument expression.
6489 Length expression C type.
6495 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
6497 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
6499 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
6501 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
6503 Dynamically-sized array, displayed as text.
6505 The string does not need to be null-terminated.
6507 The type of +__E__+ must be unsigned.
6509 The behaviour is undefined if +__e__+ is `NULL`.
6512 Sequence element C type (always `char`).
6518 Argument expression.
6521 Length expression C type.
6527 Use the `_user` versions when the argument expression, `e`, is
6528 a user space address. In the cases of `ctf_user_integer*()` and
6529 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
6532 The `_nowrite` versions omit themselves from the session trace, but are
6533 otherwise identical. This means the `_nowrite` fields won't be written
6534 in the recorded trace. Their primary purpose is to make some
6535 of the event context available to the
6536 <<enabling-disabling-events,event filters>> without having to
6537 commit the data to sub-buffers.
6543 Terms related to LTTng and to tracing in general:
6546 The http://diamon.org/babeltrace[Babeltrace] project, which includes
6547 the cmd:babeltrace command, some libraries, and Python bindings.
6549 <<channel-buffering-schemes,buffering scheme>>::
6550 A layout of sub-buffers applied to a given channel.
6552 <<channel,channel>>::
6553 An entity which is responsible for a set of ring buffers.
6555 <<event,Event rules>> are always attached to a specific channel.
6558 A reference of time for a tracer.
6560 <<lttng-consumerd,consumer daemon>>::
6561 A process which is responsible for consuming the full sub-buffers
6562 and write them to a file system or send them over the network.
6564 <<channel-overwrite-mode-vs-discard-mode,discard mode>>:: The event loss
6565 mode in which the tracer _discards_ new event records when there's no
6566 sub-buffer space left to store them.
6569 The consequence of the execution of an instrumentation
6570 point, like a tracepoint that you manually place in some source code,
6571 or a Linux kernel KProbe.
6573 An event is said to _occur_ at a specific time. Different actions can
6574 be taken upon the occurrence of an event, like record the event's payload
6577 <<channel-overwrite-mode-vs-discard-mode,event loss mode>>::
6578 The mechanism by which event records of a given channel are lost
6579 (not recorded) when there is no sub-buffer space left to store them.
6581 [[def-event-name]]event name::
6582 The name of an event, which is also the name of the event record.
6583 This is also called the _instrumentation point name_.
6586 A record, in a trace, of the payload of an event which occured.
6588 <<event,event rule>>::
6589 Set of conditions which must be satisfied for one or more occuring
6590 events to be recorded.
6592 `java.util.logging`::
6594 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities].
6596 <<instrumenting,instrumentation>>::
6597 The use of LTTng probes to make a piece of software traceable.
6599 instrumentation point::
6600 A point in the execution path of a piece of software that, when
6601 reached by this execution, can emit an event.
6603 instrumentation point name::
6604 See _<<def-event-name,event name>>_.
6607 A http://logging.apache.org/log4j/1.2/[logging library] for Java
6608 developed by the Apache Software Foundation.
6611 Level of severity of a log statement or user space
6612 instrumentation point.
6615 The _Linux Trace Toolkit: next generation_ project.
6617 <<lttng-cli,cmd:lttng>>::
6618 A command-line tool provided by the LTTng-tools project which you
6619 can use to send and receive control messages to and from a
6623 The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
6624 which is a set of analyzing programs that are used to obtain a
6625 higher level view of an LTTng trace.
6627 cmd:lttng-consumerd::
6628 The name of the consumer daemon program.
6631 A utility provided by the LTTng-tools project which can convert
6632 ring buffer files (usually
6633 <<persistent-memory-file-systems,saved on a persistent memory file system>>)
6636 LTTng Documentation::
6639 <<lttng-live,LTTng live>>::
6640 A communication protocol between the relay daemon and live viewers
6641 which makes it possible to see events "live", as they are received by
6644 <<lttng-modules,LTTng-modules>>::
6645 The https://github.com/lttng/lttng-modules[LTTng-modules] project,
6646 which contains the Linux kernel modules to make the Linux kernel
6647 instrumentation points available for LTTng tracing.
6650 The name of the relay daemon program.
6652 cmd:lttng-sessiond::
6653 The name of the session daemon program.
6656 The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
6657 contains the various programs and libraries used to
6658 <<controlling-tracing,control tracing>>.
6660 <<lttng-ust,LTTng-UST>>::
6661 The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
6662 contains libraries to instrument user applications.
6664 <<lttng-ust-agents,LTTng-UST Java agent>>::
6665 A Java package provided by the LTTng-UST project to allow the
6666 LTTng instrumentation of `java.util.logging` and Apache log4j 1.2
6669 <<lttng-ust-agents,LTTng-UST Python agent>>::
6670 A Python package provided by the LTTng-UST project to allow the
6671 LTTng instrumentation of Python logging statements.
6673 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
6674 The event loss mode in which new event records overwrite older
6675 event records when there's no sub-buffer space left to store them.
6677 <<channel-buffering-schemes,per-process buffering>>::
6678 A buffering scheme in which each instrumented process has its own
6679 sub-buffers for a given user space channel.
6681 <<channel-buffering-schemes,per-user buffering>>::
6682 A buffering scheme in which all the processes of a Unix user share the
6683 same sub-buffer for a given user space channel.
6685 <<lttng-relayd,relay daemon>>::
6686 A process which is responsible for receiving the trace data sent by
6687 a distant consumer daemon.
6690 A set of sub-buffers.
6692 <<lttng-sessiond,session daemon>>::
6693 A process which receives control commands from you and orchestrates
6694 the tracers and various LTTng daemons.
6696 <<taking-a-snapshot,snapshot>>::
6697 A copy of the current data of all the sub-buffers of a given tracing
6698 session, saved as trace files.
6701 One part of an LTTng ring buffer which contains event records.
6704 The time information attached to an event when it is emitted.
6707 A set of files which are the concatenations of one or more
6708 flushed sub-buffers.
6711 The action of recording the events emitted by an application
6712 or by a system, or to initiate such recording by controlling
6716 The http://tracecompass.org[Trace Compass] project and application.
6719 An instrumentation point using the tracepoint mechanism of the Linux
6720 kernel or of LTTng-UST.
6722 tracepoint definition::
6723 The definition of a single tracepoint.
6726 The name of a tracepoint.
6728 tracepoint provider::
6729 A set of functions providing tracepoints to an instrumented user
6732 Not to be confused with a _tracepoint provider package_: many tracepoint
6733 providers can exist within a tracepoint provider package.
6735 tracepoint provider package::
6736 One or more tracepoint providers compiled as an object file or as
6740 A software which records emitted events.
6742 <<domain,tracing domain>>::
6743 A namespace for event sources.
6746 The Unix group in which a Unix user can be to be allowed to trace the
6749 <<tracing-session,tracing session>>::
6750 A stateful dialogue between you and a <<lttng-sessiond,session
6754 An application running in user space, as opposed to a Linux kernel
6755 module, for example.