1 The LTTng Documentation
2 =======================
3 Philippe Proulx <pproulx@efficios.com>
7 include::../common/copyright.txt[]
10 include::../common/welcome.txt[]
13 include::../common/audience.txt[]
17 === What's in this documentation?
19 The LTTng Documentation is divided into the following sections:
21 * **<<nuts-and-bolts,Nuts and bolts>>** explains the
22 rudiments of software tracing and the rationale behind the
25 You can skip this section if you’re familiar with software tracing and
26 with the LTTng project.
28 * **<<installing-lttng,Installation>>** describes the steps to
29 install the LTTng packages on common Linux distributions and from
32 You can skip this section if you already properly installed LTTng on
35 * **<<getting-started,Quick start>>** is a concise guide to
36 getting started quickly with LTTng kernel and user space tracing.
38 We recommend this section if you're new to LTTng or to software tracing
41 You can skip this section if you're not new to LTTng.
43 * **<<core-concepts,Core concepts>>** explains the concepts at
46 It's a good idea to become familiar with the core concepts
47 before attempting to use the toolkit.
49 * **<<plumbing,Components of LTTng>>** describes the various components
50 of the LTTng machinery, like the daemons, the libraries, and the
51 command-line interface.
52 * **<<instrumenting,Instrumentation>>** shows different ways to
53 instrument user applications and the Linux kernel.
55 Instrumenting source code is essential to provide a meaningful
58 You can skip this section if you do not have a programming background.
60 * **<<controlling-tracing,Tracing control>>** is divided into topics
61 which demonstrate how to use the vast array of features that
62 LTTng{nbsp}{revision} offers.
63 * **<<reference,Reference>>** contains reference tables.
64 * **<<glossary,Glossary>>** is a specialized dictionary of terms related
65 to LTTng or to the field of software tracing.
68 include::../common/convention.txt[]
71 include::../common/acknowledgements.txt[]
75 == What's new in LTTng {revision}?
77 LTTng{nbsp}{revision} bears the name _KeKriek_. From
78 http://brasseriedunham.com/[Brasserie Dunham], the _**KeKriek**_ is a
79 sour mashed golden wheat ale fermented with local sour cherries from
80 Tougas orchards. Fresh sweet cherry notes with some tartness, lively
81 carbonation with a dry finish.
83 New features and changes in LTTng{nbsp}{revision}:
85 * **Tracing control**:
86 ** You can put more than one wildcard special character (`*`), and not
87 only at the end, when you <<enabling-disabling-events,create an event
88 rule>>, in both the instrumentation point name and the literal
90 link:/man/1/lttng-enable-event/v{revision}/#doc-filter-syntax[filter expressions]:
95 # lttng enable-event --kernel 'x86_*_local_timer_*' \
96 --filter='name == "*a*b*c*d*e" && count >= 23'
103 $ lttng enable-event --userspace '*_my_org:*msg*'
107 ** New trigger and notification API for
108 <<liblttng-ctl-lttng,`liblttng-ctl`>>. This new subsystem allows you
109 to register triggers which emit a notification when a given
110 condition is satisfied. As of LTTng{nbsp}{revision}, only
111 <<channel,channel>> buffer usage conditions are available.
112 Documentation is available in the
113 https://github.com/lttng/lttng-tools/tree/stable-{revision}/include/lttng[`liblttng-ctl`
115 <<notif-trigger-api,Get notified when a channel's buffer usage is too
118 ** You can now embed the whole textual LTTng-tools man pages into the
119 executables at build time with the `--enable-embedded-help`
120 configuration option. Thanks to this option, you don't need the
121 http://www.methods.co.nz/asciidoc/[AsciiDoc] and
122 https://directory.fsf.org/wiki/Xmlto[xmlto] tools at build time, and
123 a manual pager at run time, to get access to this documentation.
125 * **User space tracing**:
126 ** New blocking mode: an LTTng-UST tracepoint can now block until
127 <<channel,sub-buffer>> space is available instead of discarding event
128 records in <<channel-overwrite-mode-vs-discard-mode,discard mode>>.
129 With this feature, you can be sure that no event records are
130 discarded during your application's execution at the expense of
133 For example, the following command lines create a user space tracing
134 channel with an infinite blocking timeout and run an application
135 instrumented with LTTng-UST which is explicitly allowed to block:
141 $ lttng enable-channel --userspace --blocking-timeout=-1 blocking-channel
142 $ lttng enable-event --userspace --channel=blocking-channel --all
144 $ LTTNG_UST_ALLOW_BLOCKING=1 my-app
148 See the complete <<blocking-timeout-example,blocking timeout example>>.
150 * **Linux kernel tracing**:
151 ** Linux 4.10, 4.11, and 4.12 support.
152 ** The thread state dump events recorded by LTTng-modules now contain
153 the task's CPU identifier. This improves the precision of the
154 scheduler model for analyses.
155 ** Extended man:socketpair(2) system call tracing data.
161 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
162 generation_ is a modern toolkit for tracing Linux systems and
163 applications. So your first question might be:
170 As the history of software engineering progressed and led to what
171 we now take for granted--complex, numerous and
172 interdependent software applications running in parallel on
173 sophisticated operating systems like Linux--the authors of such
174 components, software developers, began feeling a natural
175 urge to have tools that would ensure the robustness and good performance
176 of their masterpieces.
178 One major achievement in this field is, inarguably, the
179 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
180 an essential tool for developers to find and fix bugs. But even the best
181 debugger won't help make your software run faster, and nowadays, faster
182 software means either more work done by the same hardware, or cheaper
183 hardware for the same work.
185 A _profiler_ is often the tool of choice to identify performance
186 bottlenecks. Profiling is suitable to identify _where_ performance is
187 lost in a given software. The profiler outputs a profile, a statistical
188 summary of observed events, which you may use to discover which
189 functions took the most time to execute. However, a profiler won't
190 report _why_ some identified functions are the bottleneck. Bottlenecks
191 might only occur when specific conditions are met, conditions that are
192 sometimes impossible to capture by a statistical profiler, or impossible
193 to reproduce with an application altered by the overhead of an
194 event-based profiler. For a thorough investigation of software
195 performance issues, a history of execution is essential, with the
196 recorded values of variables and context fields you choose, and
197 with as little influence as possible on the instrumented software. This
198 is where tracing comes in handy.
200 _Tracing_ is a technique used to understand what goes on in a running
201 software system. The software used for tracing is called a _tracer_,
202 which is conceptually similar to a tape recorder. When recording,
203 specific instrumentation points placed in the software source code
204 generate events that are saved on a giant tape: a _trace_ file. You
205 can trace user applications and the operating system at the same time,
206 opening the possibility of resolving a wide range of problems that would
207 otherwise be extremely challenging.
209 Tracing is often compared to _logging_. However, tracers and loggers are
210 two different tools, serving two different purposes. Tracers are
211 designed to record much lower-level events that occur much more
212 frequently than log messages, often in the range of thousands per
213 second, with very little execution overhead. Logging is more appropriate
214 for a very high-level analysis of less frequent events: user accesses,
215 exceptional conditions (errors and warnings, for example), database
216 transactions, instant messaging communications, and such. Simply put,
217 logging is one of the many use cases that can be satisfied with tracing.
219 The list of recorded events inside a trace file can be read manually
220 like a log file for the maximum level of detail, but it is generally
221 much more interesting to perform application-specific analyses to
222 produce reduced statistics and graphs that are useful to resolve a
223 given problem. Trace viewers and analyzers are specialized tools
226 In the end, this is what LTTng is: a powerful, open source set of
227 tools to trace the Linux kernel and user applications at the same time.
228 LTTng is composed of several components actively maintained and
229 developed by its link:/community/#where[community].
232 [[lttng-alternatives]]
233 === Alternatives to noch:{LTTng}
235 Excluding proprietary solutions, a few competing software tracers
238 * https://github.com/dtrace4linux/linux[dtrace4linux] is a port of
239 Sun Microsystems's DTrace to Linux. The cmd:dtrace tool interprets
240 user scripts and is responsible for loading code into the
241 Linux kernel for further execution and collecting the outputted data.
242 * https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF] is a
243 subsystem in the Linux kernel in which a virtual machine can execute
244 programs passed from the user space to the kernel. You can attach
245 such programs to tracepoints and KProbes thanks to a system call, and
246 they can output data to the user space when executed thanks to
247 different mechanisms (pipe, VM register values, and eBPF maps, to name
249 * https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]
250 is the de facto function tracer of the Linux kernel. Its user
251 interface is a set of special files in sysfs.
252 * https://perf.wiki.kernel.org/[perf] is
253 a performance analyzing tool for Linux which supports hardware
254 performance counters, tracepoints, as well as other counters and
255 types of probes. perf's controlling utility is the cmd:perf command
257 * http://linux.die.net/man/1/strace[strace]
258 is a command-line utility which records system calls made by a
259 user process, as well as signal deliveries and changes of process
260 state. strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace]
261 to fulfill its function.
262 * http://www.sysdig.org/[sysdig], like SystemTap, uses scripts to
263 analyze Linux kernel events. You write scripts, or _chisels_ in
264 sysdig's jargon, in Lua and sysdig executes them while the system is
265 being traced or afterwards. sysdig's interface is the cmd:sysdig
266 command-line tool as well as the curses-based cmd:csysdig tool.
267 * https://sourceware.org/systemtap/[SystemTap] is a Linux kernel and
268 user space tracer which uses custom user scripts to produce plain text
269 traces. SystemTap converts the scripts to the C language, and then
270 compiles them as Linux kernel modules which are loaded to produce
271 trace data. SystemTap's primary user interface is the cmd:stap
274 The main distinctive features of LTTng is that it produces correlated
275 kernel and user space traces, as well as doing so with the lowest
276 overhead amongst other solutions. It produces trace files in the
277 http://diamon.org/ctf[CTF] format, a file format optimized
278 for the production and analyses of multi-gigabyte data.
280 LTTng is the result of more than 10 years of active open source
281 development by a community of passionate developers.
282 LTTng{nbsp}{revision} is currently available on major desktop and server
285 The main interface for tracing control is a single command-line tool
286 named cmd:lttng. The latter can create several tracing sessions, enable
287 and disable events on the fly, filter events efficiently with custom
288 user expressions, start and stop tracing, and much more. LTTng can
289 record the traces on the file system or send them over the network, and
290 keep them totally or partially. You can view the traces once tracing
291 becomes inactive or in real-time.
293 <<installing-lttng,Install LTTng now>> and
294 <<getting-started,start tracing>>!
300 **LTTng** is a set of software <<plumbing,components>> which interact to
301 <<instrumenting,instrument>> the Linux kernel and user applications, and
302 to <<controlling-tracing,control tracing>> (start and stop
303 tracing, enable and disable event rules, and the rest). Those
304 components are bundled into the following packages:
306 * **LTTng-tools**: Libraries and command-line interface to
308 * **LTTng-modules**: Linux kernel modules to instrument and
310 * **LTTng-UST**: Libraries and Java/Python packages to instrument and
311 trace user applications.
313 Most distributions mark the LTTng-modules and LTTng-UST packages as
314 optional when installing LTTng-tools (which is always required). In the
315 following sections, we always provide the steps to install all three,
318 * You only need to install LTTng-modules if you intend to trace the
320 * You only need to install LTTng-UST if you intend to trace user
324 .Availability of LTTng{nbsp}{revision} for major Linux distributions as of 25 July 2017.
326 |Distribution |Available in releases |Alternatives
328 |https://www.ubuntu.com/[Ubuntu]
329 |Ubuntu{nbsp}14.04 _Trusty Tahr_ and Ubuntu{nbsp}16.04 _Xenial Xerus_:
330 <<ubuntu-ppa,use the LTTng Stable{nbsp}{revision} PPA>>.
331 |link:/docs/v2.9#doc-ubuntu[LTTng{nbsp}2.9 for Ubuntu{nbsp}17.04 _Zesty Zapus_].
333 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
334 other Ubuntu releases.
336 |https://getfedora.org/[Fedora]
338 |link:/docs/v2.9#doc-fedora[LTTng{nbsp}2.9 for Fedora 26].
340 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
342 |https://www.debian.org/[Debian]
344 |link:/docs/v2.9#doc-debian[LTTng{nbsp}2.9 for Debian "stretch"
345 (stable), Debian "buster" (testing), and Debian "sid" (unstable)].
347 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
349 |https://www.archlinux.org/[Arch Linux]
351 |link:/docs/v2.9#doc-arch-linux[LTTng{nbsp}2.9 in the latest AUR packages].
353 |https://alpinelinux.org/[Alpine Linux]
355 |link:/docs/v2.9#doc-alpine-linux[LTTng{nbsp}2.9 for Alpine Linux "edge"].
357 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
359 |https://www.redhat.com/[RHEL] and https://www.suse.com/[SLES]
360 |See http://packages.efficios.com/[EfficiOS Enterprise Packages].
363 |https://buildroot.org/[Buildroot]
365 |link:/docs/v2.9#doc-buildroot[LTTng{nbsp}2.9 for Buildroot{nbsp}2017.02 and
366 Buildroot{nbsp}2017.05].
368 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
370 |http://www.openembedded.org/wiki/Main_Page[OpenEmbedded] and
371 https://www.yoctoproject.org/[Yocto]
373 |link:/docs/v2.9#doc-oe-yocto[LTTng{nbsp}2.9 for Yocto Project{nbsp}2.3 _Pyro_]
374 (`openembedded-core` layer).
376 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
381 === [[ubuntu-official-repositories]]Ubuntu
384 ==== noch:{LTTng} Stable {revision} PPA
386 The https://launchpad.net/~lttng/+archive/ubuntu/stable-{revision}[LTTng
387 Stable{nbsp}{revision} PPA] offers the latest stable
388 LTTng{nbsp}{revision} packages for:
390 * Ubuntu{nbsp}14.04 _Trusty Tahr_
391 * Ubuntu{nbsp}16.04 _Xenial Xerus_
393 To install LTTng{nbsp}{revision} from the LTTng Stable{nbsp}{revision} PPA:
395 . Add the LTTng Stable{nbsp}{revision} PPA repository and update the
401 # apt-add-repository ppa:lttng/stable-2.10
406 . Install the main LTTng{nbsp}{revision} packages:
411 # apt-get install lttng-tools
412 # apt-get install lttng-modules-dkms
413 # apt-get install liblttng-ust-dev
417 . **If you need to instrument and trace
418 <<java-application,Java applications>>**, install the LTTng-UST
424 # apt-get install liblttng-ust-agent-java
428 . **If you need to instrument and trace
429 <<python-application,Python{nbsp}3 applications>>**, install the
430 LTTng-UST Python agent:
435 # apt-get install python3-lttngust
440 [[enterprise-distributions]]
441 === RHEL, SUSE, and other enterprise distributions
443 To install LTTng on enterprise Linux distributions, such as Red Hat
444 Enterprise Linux (RHEL) and SUSE Linux Enterprise Server (SUSE), please
445 see http://packages.efficios.com/[EfficiOS Enterprise Packages].
448 [[building-from-source]]
449 === Build from source
451 To build and install LTTng{nbsp}{revision} from source:
453 . Using your distribution's package manager, or from source, install
454 the following dependencies of LTTng-tools and LTTng-UST:
457 * https://sourceforge.net/projects/libuuid/[libuuid]
458 * http://directory.fsf.org/wiki/Popt[popt]
459 * http://liburcu.org/[Userspace RCU]
460 * http://www.xmlsoft.org/[libxml2]
463 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
469 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.10.tar.bz2 &&
470 tar -xf lttng-modules-latest-2.10.tar.bz2 &&
471 cd lttng-modules-2.10.* &&
473 sudo make modules_install &&
478 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
484 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.10.tar.bz2 &&
485 tar -xf lttng-ust-latest-2.10.tar.bz2 &&
486 cd lttng-ust-2.10.* &&
496 .Java and Python application tracing
498 If you need to instrument and trace <<java-application,Java
499 applications>>, pass the `--enable-java-agent-jul`,
500 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
501 `configure` script, depending on which Java logging framework you use.
503 If you need to instrument and trace <<python-application,Python
504 applications>>, pass the `--enable-python-agent` option to the
505 `configure` script. You can set the `PYTHON` environment variable to the
506 path to the Python interpreter for which to install the LTTng-UST Python
514 By default, LTTng-UST libraries are installed to
515 dir:{/usr/local/lib}, which is the de facto directory in which to
516 keep self-compiled and third-party libraries.
518 When <<building-tracepoint-providers-and-user-application,linking an
519 instrumented user application with `liblttng-ust`>>:
521 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
523 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
524 man:gcc(1), man:g++(1), or man:clang(1).
528 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
534 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.10.tar.bz2 &&
535 tar -xf lttng-tools-latest-2.10.tar.bz2 &&
536 cd lttng-tools-2.10.* &&
544 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
545 previous steps automatically for a given version of LTTng and confine
546 the installed files in a specific directory. This can be useful to test
547 LTTng without installing it on your system.
553 This is a short guide to get started quickly with LTTng kernel and user
556 Before you follow this guide, make sure to <<installing-lttng,install>>
559 This tutorial walks you through the steps to:
561 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
562 . <<tracing-your-own-user-application,Trace a user application>> written
564 . <<viewing-and-analyzing-your-traces,View and analyze the
568 [[tracing-the-linux-kernel]]
569 === Trace the Linux kernel
571 The following command lines start with the `#` prompt because you need
572 root privileges to trace the Linux kernel. You can also trace the kernel
573 as a regular user if your Unix user is a member of the
574 <<tracing-group,tracing group>>.
576 . Create a <<tracing-session,tracing session>> which writes its traces
577 to dir:{/tmp/my-kernel-trace}:
582 # lttng create my-kernel-session --output=/tmp/my-kernel-trace
586 . List the available kernel tracepoints and system calls:
591 # lttng list --kernel
592 # lttng list --kernel --syscall
596 . Create <<event,event rules>> which match the desired instrumentation
597 point names, for example the `sched_switch` and `sched_process_fork`
598 tracepoints, and the man:open(2) and man:close(2) system calls:
603 # lttng enable-event --kernel sched_switch,sched_process_fork
604 # lttng enable-event --kernel --syscall open,close
608 You can also create an event rule which matches _all_ the Linux kernel
609 tracepoints (this will generate a lot of data when tracing):
614 # lttng enable-event --kernel --all
618 . <<basic-tracing-session-control,Start tracing>>:
627 . Do some operation on your system for a few seconds. For example,
628 load a website, or list the files of a directory.
629 . <<basic-tracing-session-control,Stop tracing>> and destroy the
640 The man:lttng-destroy(1) command does not destroy the trace data; it
641 only destroys the state of the tracing session.
643 . For the sake of this example, make the recorded trace accessible to
649 # chown -R $(whoami) /tmp/my-kernel-trace
653 See <<viewing-and-analyzing-your-traces,View and analyze the
654 recorded events>> to view the recorded events.
657 [[tracing-your-own-user-application]]
658 === Trace a user application
660 This section steps you through a simple example to trace a
661 _Hello world_ program written in C.
663 To create the traceable user application:
665 . Create the tracepoint provider header file, which defines the
666 tracepoints and the events they can generate:
672 #undef TRACEPOINT_PROVIDER
673 #define TRACEPOINT_PROVIDER hello_world
675 #undef TRACEPOINT_INCLUDE
676 #define TRACEPOINT_INCLUDE "./hello-tp.h"
678 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
681 #include <lttng/tracepoint.h>
691 ctf_string(my_string_field, my_string_arg)
692 ctf_integer(int, my_integer_field, my_integer_arg)
696 #endif /* _HELLO_TP_H */
698 #include <lttng/tracepoint-event.h>
702 . Create the tracepoint provider package source file:
708 #define TRACEPOINT_CREATE_PROBES
709 #define TRACEPOINT_DEFINE
711 #include "hello-tp.h"
715 . Build the tracepoint provider package:
720 $ gcc -c -I. hello-tp.c
724 . Create the _Hello World_ application source file:
731 #include "hello-tp.h"
733 int main(int argc, char *argv[])
737 puts("Hello, World!\nPress Enter to continue...");
740 * The following getchar() call is only placed here for the purpose
741 * of this demonstration, to pause the application in order for
742 * you to have time to list its tracepoints. It is not
748 * A tracepoint() call.
750 * Arguments, as defined in hello-tp.h:
752 * 1. Tracepoint provider name (required)
753 * 2. Tracepoint name (required)
754 * 3. my_integer_arg (first user-defined argument)
755 * 4. my_string_arg (second user-defined argument)
757 * Notice the tracepoint provider and tracepoint names are
758 * NOT strings: they are in fact parts of variables that the
759 * macros in hello-tp.h create.
761 tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
763 for (x = 0; x < argc; ++x) {
764 tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
767 puts("Quitting now!");
768 tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
775 . Build the application:
784 . Link the application with the tracepoint provider package,
785 `liblttng-ust`, and `libdl`:
790 $ gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
794 Here's the whole build process:
797 .User space tracing tutorial's build steps.
798 image::ust-flow.png[]
800 To trace the user application:
802 . Run the application with a few arguments:
807 $ ./hello world and beyond
816 Press Enter to continue...
820 . Start an LTTng <<lttng-sessiond,session daemon>>:
825 $ lttng-sessiond --daemonize
829 Note that a session daemon might already be running, for example as
830 a service that the distribution's service manager started.
832 . List the available user space tracepoints:
837 $ lttng list --userspace
841 You see the `hello_world:my_first_tracepoint` tracepoint listed
842 under the `./hello` process.
844 . Create a <<tracing-session,tracing session>>:
849 $ lttng create my-user-space-session
853 . Create an <<event,event rule>> which matches the
854 `hello_world:my_first_tracepoint` event name:
859 $ lttng enable-event --userspace hello_world:my_first_tracepoint
863 . <<basic-tracing-session-control,Start tracing>>:
872 . Go back to the running `hello` application and press Enter. The
873 program executes all `tracepoint()` instrumentation points and exits.
874 . <<basic-tracing-session-control,Stop tracing>> and destroy the
885 The man:lttng-destroy(1) command does not destroy the trace data; it
886 only destroys the state of the tracing session.
888 By default, LTTng saves the traces in
889 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
890 where +__name__+ is the tracing session name. The
891 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
893 See <<viewing-and-analyzing-your-traces,View and analyze the
894 recorded events>> to view the recorded events.
897 [[viewing-and-analyzing-your-traces]]
898 === View and analyze the recorded events
900 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
901 kernel>> and <<tracing-your-own-user-application,Trace a user
902 application>> tutorials, you can inspect the recorded events.
904 Many tools are available to read LTTng traces:
906 * **cmd:babeltrace** is a command-line utility which converts trace
907 formats; it supports the format that LTTng produces, CTF, as well as a
908 basic text output which can be ++grep++ed. The cmd:babeltrace command
909 is part of the http://diamon.org/babeltrace[Babeltrace] project.
910 * Babeltrace also includes
911 **https://www.python.org/[Python] bindings** so
912 that you can easily open and read an LTTng trace with your own script,
913 benefiting from the power of Python.
914 * http://tracecompass.org/[**Trace Compass**]
915 is a graphical user interface for viewing and analyzing any type of
916 logs or traces, including LTTng's.
917 * https://github.com/lttng/lttng-analyses[**LTTng analyses**] is a
918 project which includes many high-level analyses of LTTng kernel
919 traces, like scheduling statistics, interrupt frequency distribution,
920 top CPU usage, and more.
922 NOTE: This section assumes that the traces recorded during the previous
923 tutorials were saved to their default location, in the
924 dir:{$LTTNG_HOME/lttng-traces} directory. The env:LTTNG_HOME
925 environment variable defaults to `$HOME` if not set.
928 [[viewing-and-analyzing-your-traces-bt]]
929 ==== Use the cmd:babeltrace command-line tool
931 The simplest way to list all the recorded events of a trace is to pass
932 its path to cmd:babeltrace with no options:
936 $ babeltrace ~/lttng-traces/my-user-space-session*
939 cmd:babeltrace finds all traces recursively within the given path and
940 prints all their events, merging them in chronological order.
942 You can pipe the output of cmd:babeltrace into a tool like man:grep(1) for
947 $ babeltrace /tmp/my-kernel-trace | grep _switch
950 You can pipe the output of cmd:babeltrace into a tool like man:wc(1) to
951 count the recorded events:
955 $ babeltrace /tmp/my-kernel-trace | grep _open | wc --lines
959 [[viewing-and-analyzing-your-traces-bt-python]]
960 ==== Use the Babeltrace Python bindings
962 The <<viewing-and-analyzing-your-traces-bt,text output of cmd:babeltrace>>
963 is useful to isolate events by simple matching using man:grep(1) and
964 similar utilities. However, more elaborate filters, such as keeping only
965 event records with a field value falling within a specific range, are
966 not trivial to write using a shell. Moreover, reductions and even the
967 most basic computations involving multiple event records are virtually
968 impossible to implement.
970 Fortunately, Babeltrace ships with Python 3 bindings which makes it easy
971 to read the event records of an LTTng trace sequentially and compute the
974 The following script accepts an LTTng Linux kernel trace path as its
975 first argument and prints the short names of the top 5 running processes
976 on CPU 0 during the whole trace:
981 from collections import Counter
987 if len(sys.argv) != 2:
988 msg = 'Usage: python3 {} TRACEPATH'.format(sys.argv[0])
989 print(msg, file=sys.stderr)
992 # A trace collection contains one or more traces
993 col = babeltrace.TraceCollection()
995 # Add the trace provided by the user (LTTng traces always have
997 if col.add_trace(sys.argv[1], 'ctf') is None:
998 raise RuntimeError('Cannot add trace')
1000 # This counter dict contains execution times:
1002 # task command name -> total execution time (ns)
1003 exec_times = Counter()
1005 # This contains the last `sched_switch` timestamp
1009 for event in col.events:
1010 # Keep only `sched_switch` events
1011 if event.name != 'sched_switch':
1014 # Keep only events which happened on CPU 0
1015 if event['cpu_id'] != 0:
1019 cur_ts = event.timestamp
1025 # Previous task command (short) name
1026 prev_comm = event['prev_comm']
1028 # Initialize entry in our dict if not yet done
1029 if prev_comm not in exec_times:
1030 exec_times[prev_comm] = 0
1032 # Compute previous command execution time
1033 diff = cur_ts - last_ts
1035 # Update execution time of this command
1036 exec_times[prev_comm] += diff
1038 # Update last timestamp
1042 for name, ns in exec_times.most_common(5):
1044 print('{:20}{} s'.format(name, s))
1049 if __name__ == '__main__':
1050 sys.exit(0 if top5proc() else 1)
1057 $ python3 top5proc.py /tmp/my-kernel-trace/kernel
1063 swapper/0 48.607245889 s
1064 chromium 7.192738188 s
1065 pavucontrol 0.709894415 s
1066 Compositor 0.660867933 s
1067 Xorg.bin 0.616753786 s
1070 Note that `swapper/0` is the "idle" process of CPU 0 on Linux; since we
1071 weren't using the CPU that much when tracing, its first position in the
1076 == [[understanding-lttng]]Core concepts
1078 From a user's perspective, the LTTng system is built on a few concepts,
1079 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
1080 operates by sending commands to the <<lttng-sessiond,session daemon>>.
1081 Understanding how those objects relate to eachother is key in mastering
1084 The core concepts are:
1086 * <<tracing-session,Tracing session>>
1087 * <<domain,Tracing domain>>
1088 * <<channel,Channel and ring buffer>>
1089 * <<"event","Instrumentation point, event rule, event, and event record">>
1095 A _tracing session_ is a stateful dialogue between you and
1096 a <<lttng-sessiond,session daemon>>. You can
1097 <<creating-destroying-tracing-sessions,create a new tracing
1098 session>> with the `lttng create` command.
1100 Anything that you do when you control LTTng tracers happens within a
1101 tracing session. In particular, a tracing session:
1104 * Has its own set of trace files.
1105 * Has its own state of activity (started or stopped).
1106 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
1108 * Has its own <<channel,channels>> which have their own
1109 <<event,event rules>>.
1112 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
1113 image::concepts.png[]
1115 Those attributes and objects are completely isolated between different
1118 A tracing session is analogous to a cash machine session:
1119 the operations you do on the banking system through the cash machine do
1120 not alter the data of other users of the same system. In the case of
1121 the cash machine, a session lasts as long as your bank card is inside.
1122 In the case of LTTng, a tracing session lasts from the `lttng create`
1123 command to the `lttng destroy` command.
1126 .Each Unix user has its own set of tracing sessions.
1127 image::many-sessions.png[]
1130 [[tracing-session-mode]]
1131 ==== Tracing session mode
1133 LTTng can send the generated trace data to different locations. The
1134 _tracing session mode_ dictates where to send it. The following modes
1135 are available in LTTng{nbsp}{revision}:
1138 LTTng writes the traces to the file system of the machine being traced
1141 Network streaming mode::
1142 LTTng sends the traces over the network to a
1143 <<lttng-relayd,relay daemon>> running on a remote system.
1146 LTTng does not write the traces by default. Instead, you can request
1147 LTTng to <<taking-a-snapshot,take a snapshot>>, that is, a copy of the
1148 current tracing buffers, and to write it to the target's file system
1149 or to send it over the network to a <<lttng-relayd,relay daemon>>
1150 running on a remote system.
1153 This mode is similar to the network streaming mode, but a live
1154 trace viewer can connect to the distant relay daemon to
1155 <<lttng-live,view event records as LTTng generates them>> by
1162 A _tracing domain_ is a namespace for event sources. A tracing domain
1163 has its own properties and features.
1165 There are currently five available tracing domains:
1169 * `java.util.logging` (JUL)
1173 You must specify a tracing domain when using some commands to avoid
1174 ambiguity. For example, since all the domains support named tracepoints
1175 as event sources (instrumentation points that you manually insert in the
1176 source code), you need to specify a tracing domain when
1177 <<enabling-disabling-events,creating an event rule>> because all the
1178 tracing domains could have tracepoints with the same names.
1180 Some features are reserved to specific tracing domains. Dynamic function
1181 entry and return instrumentation points, for example, are currently only
1182 supported in the Linux kernel tracing domain, but support for other
1183 tracing domains could be added in the future.
1185 You can create <<channel,channels>> in the Linux kernel and user space
1186 tracing domains. The other tracing domains have a single default
1191 === Channel and ring buffer
1193 A _channel_ is an object which is responsible for a set of ring buffers.
1194 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1195 tracer emits an event, it can record it to one or more
1196 sub-buffers. The attributes of a channel determine what to do when
1197 there's no space left for a new event record because all sub-buffers
1198 are full, where to send a full sub-buffer, and other behaviours.
1200 A channel is always associated to a <<domain,tracing domain>>. The
1201 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1202 a default channel which you cannot configure.
1204 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1205 an event, it records it to the sub-buffers of all
1206 the enabled channels with a satisfied event rule, as long as those
1207 channels are part of active <<tracing-session,tracing sessions>>.
1210 [[channel-buffering-schemes]]
1211 ==== Per-user vs. per-process buffering schemes
1213 A channel has at least one ring buffer _per CPU_. LTTng always
1214 records an event to the ring buffer associated to the CPU on which it
1217 Two _buffering schemes_ are available when you
1218 <<enabling-disabling-channels,create a channel>> in the
1219 user space <<domain,tracing domain>>:
1221 Per-user buffering::
1222 Allocate one set of ring buffers--one per CPU--shared by all the
1223 instrumented processes of each Unix user.
1227 .Per-user buffering scheme.
1228 image::per-user-buffering.png[]
1231 Per-process buffering::
1232 Allocate one set of ring buffers--one per CPU--for each
1233 instrumented process.
1237 .Per-process buffering scheme.
1238 image::per-process-buffering.png[]
1241 The per-process buffering scheme tends to consume more memory than the
1242 per-user option because systems generally have more instrumented
1243 processes than Unix users running instrumented processes. However, the
1244 per-process buffering scheme ensures that one process having a high
1245 event throughput won't fill all the shared sub-buffers of the same
1248 The Linux kernel tracing domain has only one available buffering scheme
1249 which is to allocate a single set of ring buffers for the whole system.
1250 This scheme is similar to the per-user option, but with a single, global
1251 user "running" the kernel.
1254 [[channel-overwrite-mode-vs-discard-mode]]
1255 ==== Overwrite vs. discard event loss modes
1257 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1258 arc in the following animation) of a specific channel's ring buffer.
1259 When there's no space left in a sub-buffer, the tracer marks it as
1260 consumable (red) and another, empty sub-buffer starts receiving the
1261 following event records. A <<lttng-consumerd,consumer daemon>>
1262 eventually consumes the marked sub-buffer (returns to white).
1265 [role="docsvg-channel-subbuf-anim"]
1270 In an ideal world, sub-buffers are consumed faster than they are filled,
1271 as is the case in the previous animation. In the real world,
1272 however, all sub-buffers can be full at some point, leaving no space to
1273 record the following events.
1275 By default, LTTng-modules and LTTng-UST are _non-blocking_ tracers: when
1276 no empty sub-buffer is available, it is acceptable to lose event records
1277 when the alternative would be to cause substantial delays in the
1278 instrumented application's execution. LTTng privileges performance over
1279 integrity; it aims at perturbing the traced system as little as possible
1280 in order to make tracing of subtle race conditions and rare interrupt
1283 Starting from LTTng{nbsp}2.10, the LTTng user space tracer, LTTng-UST,
1284 supports a _blocking mode_. See the <<blocking-timeout-example,blocking
1285 timeout example>> to learn how to use the blocking mode.
1287 When it comes to losing event records because no empty sub-buffer is
1288 available, or because the <<opt-blocking-timeout,blocking timeout>> is
1289 reached, the channel's _event loss mode_ determines what to do. The
1290 available event loss modes are:
1293 Drop the newest event records until a the tracer releases a
1296 This is the only available mode when you specify a
1297 <<opt-blocking-timeout,blocking timeout>>.
1300 Clear the sub-buffer containing the oldest event records and start
1301 writing the newest event records there.
1303 This mode is sometimes called _flight recorder mode_ because it's
1305 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1306 always keep a fixed amount of the latest data.
1308 Which mechanism you should choose depends on your context: prioritize
1309 the newest or the oldest event records in the ring buffer?
1311 Beware that, in overwrite mode, the tracer abandons a whole sub-buffer
1312 as soon as a there's no space left for a new event record, whereas in
1313 discard mode, the tracer only discards the event record that doesn't
1316 In discard mode, LTTng increments a count of lost event records when
1317 an event record is lost and saves this count to the trace. In
1318 overwrite mode, LTTng keeps no information when it overwrites a
1319 sub-buffer before consuming it.
1321 There are a few ways to decrease your probability of losing event
1323 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1324 how you can fine-une the sub-buffer count and size of a channel to
1325 virtually stop losing event records, though at the cost of greater
1329 [[channel-subbuf-size-vs-subbuf-count]]
1330 ==== Sub-buffer count and size
1332 When you <<enabling-disabling-channels,create a channel>>, you can
1333 set its number of sub-buffers and their size.
1335 Note that there is noticeable CPU overhead introduced when
1336 switching sub-buffers (marking a full one as consumable and switching
1337 to an empty one for the following events to be recorded). Knowing this,
1338 the following list presents a few practical situations along with how
1339 to configure the sub-buffer count and size for them:
1341 * **High event throughput**: In general, prefer bigger sub-buffers to
1342 lower the risk of losing event records.
1344 Having bigger sub-buffers also ensures a lower
1345 <<channel-switch-timer,sub-buffer switching frequency>>.
1347 The number of sub-buffers is only meaningful if you create the channel
1348 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1349 other sub-buffers are left unaltered.
1351 * **Low event throughput**: In general, prefer smaller sub-buffers
1352 since the risk of losing event records is low.
1354 Because events occur less frequently, the sub-buffer switching frequency
1355 should remain low and thus the tracer's overhead should not be a
1358 * **Low memory system**: If your target system has a low memory
1359 limit, prefer fewer first, then smaller sub-buffers.
1361 Even if the system is limited in memory, you want to keep the
1362 sub-buffers as big as possible to avoid a high sub-buffer switching
1365 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1366 which means event data is very compact. For example, the average
1367 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1368 sub-buffer size of 1{nbsp}MiB is considered big.
1370 The previous situations highlight the major trade-off between a few big
1371 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1372 frequency vs. how much data is lost in overwrite mode. Assuming a
1373 constant event throughput and using the overwrite mode, the two
1374 following configurations have the same ring buffer total size:
1377 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1382 * **2 sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1383 switching frequency, but if a sub-buffer overwrite happens, half of
1384 the event records so far (4{nbsp}MiB) are definitely lost.
1385 * **8 sub-buffers of 1{nbsp}MiB each**: Expect 4{nbsp}times the tracer's
1386 overhead as the previous configuration, but if a sub-buffer
1387 overwrite happens, only the eighth of event records so far are
1390 In discard mode, the sub-buffers count parameter is pointless: use two
1391 sub-buffers and set their size according to the requirements of your
1395 [[channel-switch-timer]]
1396 ==== Switch timer period
1398 The _switch timer period_ is an important configurable attribute of
1399 a channel to ensure periodic sub-buffer flushing.
1401 When the _switch timer_ expires, a sub-buffer switch happens. You can
1402 set the switch timer period attribute when you
1403 <<enabling-disabling-channels,create a channel>> to ensure that event
1404 data is consumed and committed to trace files or to a distant relay
1405 daemon periodically in case of a low event throughput.
1408 [role="docsvg-channel-switch-timer"]
1413 This attribute is also convenient when you use big sub-buffers to cope
1414 with a sporadic high event throughput, even if the throughput is
1418 [[channel-read-timer]]
1419 ==== Read timer period
1421 By default, the LTTng tracers use a notification mechanism to signal a
1422 full sub-buffer so that a consumer daemon can consume it. When such
1423 notifications must be avoided, for example in real-time applications,
1424 you can use the channel's _read timer_ instead. When the read timer
1425 fires, the <<lttng-consumerd,consumer daemon>> checks for full,
1426 consumable sub-buffers.
1429 [[tracefile-rotation]]
1430 ==== Trace file count and size
1432 By default, trace files can grow as large as needed. You can set the
1433 maximum size of each trace file that a channel writes when you
1434 <<enabling-disabling-channels,create a channel>>. When the size of
1435 a trace file reaches the channel's fixed maximum size, LTTng creates
1436 another file to contain the next event records. LTTng appends a file
1437 count to each trace file name in this case.
1439 If you set the trace file size attribute when you create a channel, the
1440 maximum number of trace files that LTTng creates is _unlimited_ by
1441 default. To limit them, you can also set a maximum number of trace
1442 files. When the number of trace files reaches the channel's fixed
1443 maximum count, the oldest trace file is overwritten. This mechanism is
1444 called _trace file rotation_.
1448 === Instrumentation point, event rule, event, and event record
1450 An _event rule_ is a set of conditions which must be **all** satisfied
1451 for LTTng to record an occuring event.
1453 You set the conditions when you <<enabling-disabling-events,create
1456 You always attach an event rule to <<channel,channel>> when you create
1459 When an event passes the conditions of an event rule, LTTng records it
1460 in one of the attached channel's sub-buffers.
1462 The available conditions, as of LTTng{nbsp}{revision}, are:
1464 * The event rule _is enabled_.
1465 * The instrumentation point's type _is{nbsp}T_.
1466 * The instrumentation point's name (sometimes called _event name_)
1467 _matches{nbsp}N_, but _is not{nbsp}E_.
1468 * The instrumentation point's log level _is as severe as{nbsp}L_, or
1469 _is exactly{nbsp}L_.
1470 * The fields of the event's payload _satisfy_ a filter
1471 expression{nbsp}__F__.
1473 As you can see, all the conditions but the dynamic filter are related to
1474 the event rule's status or to the instrumentation point, not to the
1475 occurring events. This is why, without a filter, checking if an event
1476 passes an event rule is not a dynamic task: when you create or modify an
1477 event rule, all the tracers of its tracing domain enable or disable the
1478 instrumentation points themselves once. This is possible because the
1479 attributes of an instrumentation point (type, name, and log level) are
1480 defined statically. In other words, without a dynamic filter, the tracer
1481 _does not evaluate_ the arguments of an instrumentation point unless it
1482 matches an enabled event rule.
1484 Note that, for LTTng to record an event, the <<channel,channel>> to
1485 which a matching event rule is attached must also be enabled, and the
1486 tracing session owning this channel must be active.
1489 .Logical path from an instrumentation point to an event record.
1490 image::event-rule.png[]
1492 .Event, event record, or event rule?
1494 With so many similar terms, it's easy to get confused.
1496 An **event** is the consequence of the execution of an _instrumentation
1497 point_, like a tracepoint that you manually place in some source code,
1498 or a Linux kernel KProbe. An event is said to _occur_ at a specific
1499 time. Different actions can be taken upon the occurrence of an event,
1500 like record the event's payload to a buffer.
1502 An **event record** is the representation of an event in a sub-buffer. A
1503 tracer is responsible for capturing the payload of an event, current
1504 context variables, the event's ID, and the event's timestamp. LTTng
1505 can append this sub-buffer to a trace file.
1507 An **event rule** is a set of conditions which must all be satisfied for
1508 LTTng to record an occuring event. Events still occur without
1509 satisfying event rules, but LTTng does not record them.
1514 == Components of noch:{LTTng}
1516 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1517 to call LTTng a simple _tool_ since it is composed of multiple
1518 interacting components. This section describes those components,
1519 explains their respective roles, and shows how they connect together to
1520 form the LTTng ecosystem.
1522 The following diagram shows how the most important components of LTTng
1523 interact with user applications, the Linux kernel, and you:
1526 .Control and trace data paths between LTTng components.
1527 image::plumbing.png[]
1529 The LTTng project incorporates:
1531 * **LTTng-tools**: Libraries and command-line interface to
1532 control tracing sessions.
1533 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1534 ** <<lttng-consumerd,Consumer daemon>> (man:lttng-consumerd(8)).
1535 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1536 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1537 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1538 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1540 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1541 headers to instrument and trace any native user application.
1542 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1543 *** `liblttng-ust-libc-wrapper`
1544 *** `liblttng-ust-pthread-wrapper`
1545 *** `liblttng-ust-cyg-profile`
1546 *** `liblttng-ust-cyg-profile-fast`
1547 *** `liblttng-ust-dl`
1548 ** User space tracepoint provider source files generator command-line
1549 tool (man:lttng-gen-tp(1)).
1550 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1551 Java applications using `java.util.logging` or
1552 Apache log4j 1.2 logging.
1553 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1554 Python applications using the standard `logging` package.
1555 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1557 ** LTTng kernel tracer module.
1558 ** Tracing ring buffer kernel modules.
1559 ** Probe kernel modules.
1560 ** LTTng logger kernel module.
1564 === Tracing control command-line interface
1567 .The tracing control command-line interface.
1568 image::plumbing-lttng-cli.png[]
1570 The _man:lttng(1) command-line tool_ is the standard user interface to
1571 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1572 is part of LTTng-tools.
1574 The cmd:lttng tool is linked with
1575 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1576 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1578 The cmd:lttng tool has a Git-like interface:
1582 $ lttng <GENERAL OPTIONS> <COMMAND> <COMMAND OPTIONS>
1585 The <<controlling-tracing,Tracing control>> section explores the
1586 available features of LTTng using the cmd:lttng tool.
1589 [[liblttng-ctl-lttng]]
1590 === Tracing control library
1593 .The tracing control library.
1594 image::plumbing-liblttng-ctl.png[]
1596 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1597 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1598 underlying protocol's details. `liblttng-ctl` is part of LTTng-tools.
1600 The <<lttng-cli,cmd:lttng command-line tool>>
1601 is linked with `liblttng-ctl`.
1603 You can use `liblttng-ctl` in C or $$C++$$ source code by including its
1608 #include <lttng/lttng.h>
1611 Some objects are referenced by name (C string), such as tracing
1612 sessions, but most of them require to create a handle first using
1613 `lttng_create_handle()`.
1615 The best available developer documentation for `liblttng-ctl` is, as of
1616 LTTng{nbsp}{revision}, its installed header files. Every function and
1617 structure is thoroughly documented.
1621 === User space tracing library
1624 .The user space tracing library.
1625 image::plumbing-liblttng-ust.png[]
1627 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1628 is the LTTng user space tracer. It receives commands from a
1629 <<lttng-sessiond,session daemon>>, for example to
1630 enable and disable specific instrumentation points, and writes event
1631 records to ring buffers shared with a
1632 <<lttng-consumerd,consumer daemon>>.
1633 `liblttng-ust` is part of LTTng-UST.
1635 Public C header files are installed beside `liblttng-ust` to
1636 instrument any <<c-application,C or $$C++$$ application>>.
1638 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1639 packages, use their own library providing tracepoints which is
1640 linked with `liblttng-ust`.
1642 An application or library does not have to initialize `liblttng-ust`
1643 manually: its constructor does the necessary tasks to properly register
1644 to a session daemon. The initialization phase also enables the
1645 instrumentation points matching the <<event,event rules>> that you
1649 [[lttng-ust-agents]]
1650 === User space tracing agents
1653 .The user space tracing agents.
1654 image::plumbing-lttng-ust-agents.png[]
1656 The _LTTng-UST Java and Python agents_ are regular Java and Python
1657 packages which add LTTng tracing capabilities to the
1658 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1660 In the case of Java, the
1661 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1662 core logging facilities] and
1663 https://logging.apache.org/log4j/1.2/[Apache log4j 1.2] are supported.
1664 Note that Apache Log4{nbsp}2 is not supported.
1666 In the case of Python, the standard
1667 https://docs.python.org/3/library/logging.html[`logging`] package
1668 is supported. Both Python 2 and Python 3 modules can import the
1669 LTTng-UST Python agent package.
1671 The applications using the LTTng-UST agents are in the
1672 `java.util.logging` (JUL),
1673 log4j, and Python <<domain,tracing domains>>.
1675 Both agents use the same mechanism to trace the log statements. When an
1676 agent is initialized, it creates a log handler that attaches to the root
1677 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
1678 When the application executes a log statement, it is passed to the
1679 agent's log handler by the root logger. The agent's log handler calls a
1680 native function in a tracepoint provider package shared library linked
1681 with <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
1682 other fields, like its logger name and its log level. This native
1683 function contains a user space instrumentation point, hence tracing the
1686 The log level condition of an
1687 <<event,event rule>> is considered when tracing
1688 a Java or a Python application, and it's compatible with the standard
1689 JUL, log4j, and Python log levels.
1693 === LTTng kernel modules
1696 .The LTTng kernel modules.
1697 image::plumbing-lttng-modules.png[]
1699 The _LTTng kernel modules_ are a set of Linux kernel modules
1700 which implement the kernel tracer of the LTTng project. The LTTng
1701 kernel modules are part of LTTng-modules.
1703 The LTTng kernel modules include:
1705 * A set of _probe_ modules.
1707 Each module attaches to a specific subsystem
1708 of the Linux kernel using its tracepoint instrument points. There are
1709 also modules to attach to the entry and return points of the Linux
1710 system call functions.
1712 * _Ring buffer_ modules.
1714 A ring buffer implementation is provided as kernel modules. The LTTng
1715 kernel tracer writes to the ring buffer; a
1716 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
1718 * The _LTTng kernel tracer_ module.
1719 * The _LTTng logger_ module.
1721 The LTTng logger module implements the special path:{/proc/lttng-logger}
1722 file so that any executable can generate LTTng events by opening and
1723 writing to this file.
1725 See <<proc-lttng-logger-abi,LTTng logger>>.
1727 Generally, you do not have to load the LTTng kernel modules manually
1728 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
1729 daemon>> loads the necessary modules when starting. If you have extra
1730 probe modules, you can specify to load them to the session daemon on
1733 The LTTng kernel modules are installed in
1734 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
1735 the kernel release (see `uname --kernel-release`).
1742 .The session daemon.
1743 image::plumbing-sessiond.png[]
1745 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
1746 managing tracing sessions and for controlling the various components of
1747 LTTng. The session daemon is part of LTTng-tools.
1749 The session daemon sends control requests to and receives control
1752 * The <<lttng-ust,user space tracing library>>.
1754 Any instance of the user space tracing library first registers to
1755 a session daemon. Then, the session daemon can send requests to
1756 this instance, such as:
1759 ** Get the list of tracepoints.
1760 ** Share an <<event,event rule>> so that the user space tracing library
1761 can enable or disable tracepoints. Amongst the possible conditions
1762 of an event rule is a filter expression which `liblttng-ust` evalutes
1763 when an event occurs.
1764 ** Share <<channel,channel>> attributes and ring buffer locations.
1767 The session daemon and the user space tracing library use a Unix
1768 domain socket for their communication.
1770 * The <<lttng-ust-agents,user space tracing agents>>.
1772 Any instance of a user space tracing agent first registers to
1773 a session daemon. Then, the session daemon can send requests to
1774 this instance, such as:
1777 ** Get the list of loggers.
1778 ** Enable or disable a specific logger.
1781 The session daemon and the user space tracing agent use a TCP connection
1782 for their communication.
1784 * The <<lttng-modules,LTTng kernel tracer>>.
1785 * The <<lttng-consumerd,consumer daemon>>.
1787 The session daemon sends requests to the consumer daemon to instruct
1788 it where to send the trace data streams, amongst other information.
1790 * The <<lttng-relayd,relay daemon>>.
1792 The session daemon receives commands from the
1793 <<liblttng-ctl-lttng,tracing control library>>.
1795 The root session daemon loads the appropriate
1796 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
1797 a <<lttng-consumerd,consumer daemon>> as soon as you create
1798 an <<event,event rule>>.
1800 The session daemon does not send and receive trace data: this is the
1801 role of the <<lttng-consumerd,consumer daemon>> and
1802 <<lttng-relayd,relay daemon>>. It does, however, generate the
1803 http://diamon.org/ctf/[CTF] metadata stream.
1805 Each Unix user can have its own session daemon instance. The
1806 tracing sessions managed by different session daemons are completely
1809 The root user's session daemon is the only one which is
1810 allowed to control the LTTng kernel tracer, and its spawned consumer
1811 daemon is the only one which is allowed to consume trace data from the
1812 LTTng kernel tracer. Note, however, that any Unix user which is a member
1813 of the <<tracing-group,tracing group>> is allowed
1814 to create <<channel,channels>> in the
1815 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
1818 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
1819 session daemon when using its `create` command if none is currently
1820 running. You can also start the session daemon manually.
1827 .The consumer daemon.
1828 image::plumbing-consumerd.png[]
1830 The _consumer daemon_, man:lttng-consumerd(8), is a daemon which shares
1831 ring buffers with user applications or with the LTTng kernel modules to
1832 collect trace data and send it to some location (on disk or to a
1833 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
1834 is part of LTTng-tools.
1836 You do not start a consumer daemon manually: a consumer daemon is always
1837 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
1838 <<event,event rule>>, that is, before you start tracing. When you kill
1839 its owner session daemon, the consumer daemon also exits because it is
1840 the session daemon's child process. Command-line options of
1841 man:lttng-sessiond(8) target the consumer daemon process.
1843 There are up to two running consumer daemons per Unix user, whereas only
1844 one session daemon can run per user. This is because each process can be
1845 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
1846 and 64-bit processes, it is more efficient to have separate
1847 corresponding 32-bit and 64-bit consumer daemons. The root user is an
1848 exception: it can have up to _three_ running consumer daemons: 32-bit
1849 and 64-bit instances for its user applications, and one more
1850 reserved for collecting kernel trace data.
1858 image::plumbing-relayd.png[]
1860 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
1861 between remote session and consumer daemons, local trace files, and a
1862 remote live trace viewer. The relay daemon is part of LTTng-tools.
1864 The main purpose of the relay daemon is to implement a receiver of
1865 <<sending-trace-data-over-the-network,trace data over the network>>.
1866 This is useful when the target system does not have much file system
1867 space to record trace files locally.
1869 The relay daemon is also a server to which a
1870 <<lttng-live,live trace viewer>> can
1871 connect. The live trace viewer sends requests to the relay daemon to
1872 receive trace data as the target system emits events. The
1873 communication protocol is named _LTTng live_; it is used over TCP
1876 Note that you can start the relay daemon on the target system directly.
1877 This is the setup of choice when the use case is to view events as
1878 the target system emits them without the need of a remote system.
1882 == [[using-lttng]]Instrumentation
1884 There are many examples of tracing and monitoring in our everyday life:
1886 * You have access to real-time and historical weather reports and
1887 forecasts thanks to weather stations installed around the country.
1888 * You know your heart is safe thanks to an electrocardiogram.
1889 * You make sure not to drive your car too fast and to have enough fuel
1890 to reach your destination thanks to gauges visible on your dashboard.
1892 All the previous examples have something in common: they rely on
1893 **instruments**. Without the electrodes attached to the surface of your
1894 body's skin, cardiac monitoring is futile.
1896 LTTng, as a tracer, is no different from those real life examples. If
1897 you're about to trace a software system or, in other words, record its
1898 history of execution, you better have **instrumentation points** in the
1899 subject you're tracing, that is, the actual software.
1901 Various ways were developed to instrument a piece of software for LTTng
1902 tracing. The most straightforward one is to manually place
1903 instrumentation points, called _tracepoints_, in the software's source
1904 code. It is also possible to add instrumentation points dynamically in
1905 the Linux kernel <<domain,tracing domain>>.
1907 If you're only interested in tracing the Linux kernel, your
1908 instrumentation needs are probably already covered by LTTng's built-in
1909 <<lttng-modules,Linux kernel tracepoints>>. You may also wish to trace a
1910 user application which is already instrumented for LTTng tracing.
1911 In such cases, you can skip this whole section and read the topics of
1912 the <<controlling-tracing,Tracing control>> section.
1914 Many methods are available to instrument a piece of software for LTTng
1917 * <<c-application,User space instrumentation for C and $$C++$$
1919 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
1920 * <<java-application,User space Java agent>>.
1921 * <<python-application,User space Python agent>>.
1922 * <<proc-lttng-logger-abi,LTTng logger>>.
1923 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
1927 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
1929 The procedure to instrument a C or $$C++$$ user application with
1930 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
1932 . <<tracepoint-provider,Create the source files of a tracepoint provider
1934 . <<probing-the-application-source-code,Add tracepoints to
1935 the application's source code>>.
1936 . <<building-tracepoint-providers-and-user-application,Build and link
1937 a tracepoint provider package and the user application>>.
1939 If you need quick, man:printf(3)-like instrumentation, you can skip
1940 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
1943 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
1944 instrument a user application with `liblttng-ust`.
1947 [[tracepoint-provider]]
1948 ==== Create the source files of a tracepoint provider package
1950 A _tracepoint provider_ is a set of compiled functions which provide
1951 **tracepoints** to an application, the type of instrumentation point
1952 supported by LTTng-UST. Those functions can emit events with
1953 user-defined fields and serialize those events as event records to one
1954 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
1955 macro, which you <<probing-the-application-source-code,insert in a user
1956 application's source code>>, calls those functions.
1958 A _tracepoint provider package_ is an object file (`.o`) or a shared
1959 library (`.so`) which contains one or more tracepoint providers.
1960 Its source files are:
1962 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
1963 * A <<tpp-source,tracepoint provider package source>> (`.c`).
1965 A tracepoint provider package is dynamically linked with `liblttng-ust`,
1966 the LTTng user space tracer, at run time.
1969 .User application linked with `liblttng-ust` and containing a tracepoint provider.
1970 image::ust-app.png[]
1972 NOTE: If you need quick, man:printf(3)-like instrumentation, you can
1973 skip creating and using a tracepoint provider and use
1974 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
1978 ===== Create a tracepoint provider header file template
1980 A _tracepoint provider header file_ contains the tracepoint
1981 definitions of a tracepoint provider.
1983 To create a tracepoint provider header file:
1985 . Start from this template:
1989 .Tracepoint provider header file template (`.h` file extension).
1991 #undef TRACEPOINT_PROVIDER
1992 #define TRACEPOINT_PROVIDER provider_name
1994 #undef TRACEPOINT_INCLUDE
1995 #define TRACEPOINT_INCLUDE "./tp.h"
1997 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
2000 #include <lttng/tracepoint.h>
2003 * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
2004 * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
2009 #include <lttng/tracepoint-event.h>
2015 * `provider_name` with the name of your tracepoint provider.
2016 * `"tp.h"` with the name of your tracepoint provider header file.
2018 . Below the `#include <lttng/tracepoint.h>` line, put your
2019 <<defining-tracepoints,tracepoint definitions>>.
2021 Your tracepoint provider name must be unique amongst all the possible
2022 tracepoint provider names used on the same target system. We
2023 suggest to include the name of your project or company in the name,
2024 for example, `org_lttng_my_project_tpp`.
2026 TIP: [[lttng-gen-tp]]You can use the man:lttng-gen-tp(1) tool to create
2027 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
2028 write are the <<defining-tracepoints,tracepoint definitions>>.
2031 [[defining-tracepoints]]
2032 ===== Create a tracepoint definition
2034 A _tracepoint definition_ defines, for a given tracepoint:
2036 * Its **input arguments**. They are the macro parameters that the
2037 `tracepoint()` macro accepts for this particular tracepoint
2038 in the user application's source code.
2039 * Its **output event fields**. They are the sources of event fields
2040 that form the payload of any event that the execution of the
2041 `tracepoint()` macro emits for this particular tracepoint.
2043 You can create a tracepoint definition by using the
2044 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
2046 <<tpp-header,tracepoint provider header file template>>.
2048 The syntax of the `TRACEPOINT_EVENT()` macro is:
2051 .`TRACEPOINT_EVENT()` macro syntax.
2054 /* Tracepoint provider name */
2057 /* Tracepoint name */
2060 /* Input arguments */
2065 /* Output event fields */
2074 * `provider_name` with your tracepoint provider name.
2075 * `tracepoint_name` with your tracepoint name.
2076 * `arguments` with the <<tpp-def-input-args,input arguments>>.
2077 * `fields` with the <<tpp-def-output-fields,output event field>>
2080 This tracepoint emits events named `provider_name:tracepoint_name`.
2083 .Event name's length limitation
2085 The concatenation of the tracepoint provider name and the
2086 tracepoint name must not exceed **254 characters**. If it does, the
2087 instrumented application compiles and runs, but LTTng throws multiple
2088 warnings and you could experience serious issues.
2091 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
2094 .`TP_ARGS()` macro syntax.
2103 * `type` with the C type of the argument.
2104 * `arg_name` with the argument name.
2106 You can repeat `type` and `arg_name` up to 10 times to have
2107 more than one argument.
2109 .`TP_ARGS()` usage with three arguments.
2121 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
2122 tracepoint definition with no input arguments.
2124 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
2125 `ctf_*()` macros. Each `ctf_*()` macro defines one event field. See
2126 man:lttng-ust(3) for a complete description of the available `ctf_*()`
2127 macros. A `ctf_*()` macro specifies the type, size, and byte order of
2130 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
2131 C expression that the tracer evalutes at the `tracepoint()` macro site
2132 in the application's source code. This expression provides a field's
2133 source of data. The argument expression can include input argument names
2134 listed in the `TP_ARGS()` macro.
2136 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
2137 must be unique within a given tracepoint definition.
2139 Here's a complete tracepoint definition example:
2141 .Tracepoint definition.
2143 The following tracepoint definition defines a tracepoint which takes
2144 three input arguments and has four output event fields.
2148 #include "my-custom-structure.h"
2154 const struct my_custom_structure*, my_custom_structure,
2159 ctf_string(query_field, query)
2160 ctf_float(double, ratio_field, ratio)
2161 ctf_integer(int, recv_size, my_custom_structure->recv_size)
2162 ctf_integer(int, send_size, my_custom_structure->send_size)
2167 You can refer to this tracepoint definition with the `tracepoint()`
2168 macro in your application's source code like this:
2172 tracepoint(my_provider, my_tracepoint,
2173 my_structure, some_ratio, the_query);
2177 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2178 if they satisfy an enabled <<event,event rule>>.
2181 [[using-tracepoint-classes]]
2182 ===== Use a tracepoint class
2184 A _tracepoint class_ is a class of tracepoints which share the same
2185 output event field definitions. A _tracepoint instance_ is one
2186 instance of such a defined tracepoint class, with its own tracepoint
2189 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2190 shorthand which defines both a tracepoint class and a tracepoint
2191 instance at the same time.
2193 When you build a tracepoint provider package, the C or $$C++$$ compiler
2194 creates one serialization function for each **tracepoint class**. A
2195 serialization function is responsible for serializing the event fields
2196 of a tracepoint to a sub-buffer when tracing.
2198 For various performance reasons, when your situation requires multiple
2199 tracepoint definitions with different names, but with the same event
2200 fields, we recommend that you manually create a tracepoint class
2201 and instantiate as many tracepoint instances as needed. One positive
2202 effect of such a design, amongst other advantages, is that all
2203 tracepoint instances of the same tracepoint class reuse the same
2204 serialization function, thus reducing
2205 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2207 .Use a tracepoint class and tracepoint instances.
2209 Consider the following three tracepoint definitions:
2221 ctf_integer(int, userid, userid)
2222 ctf_integer(size_t, len, len)
2234 ctf_integer(int, userid, userid)
2235 ctf_integer(size_t, len, len)
2247 ctf_integer(int, userid, userid)
2248 ctf_integer(size_t, len, len)
2253 In this case, we create three tracepoint classes, with one implicit
2254 tracepoint instance for each of them: `get_account`, `get_settings`, and
2255 `get_transaction`. However, they all share the same event field names
2256 and types. Hence three identical, yet independent serialization
2257 functions are created when you build the tracepoint provider package.
2259 A better design choice is to define a single tracepoint class and three
2260 tracepoint instances:
2264 /* The tracepoint class */
2265 TRACEPOINT_EVENT_CLASS(
2266 /* Tracepoint provider name */
2269 /* Tracepoint class name */
2272 /* Input arguments */
2278 /* Output event fields */
2280 ctf_integer(int, userid, userid)
2281 ctf_integer(size_t, len, len)
2285 /* The tracepoint instances */
2286 TRACEPOINT_EVENT_INSTANCE(
2287 /* Tracepoint provider name */
2290 /* Tracepoint class name */
2293 /* Tracepoint name */
2296 /* Input arguments */
2302 TRACEPOINT_EVENT_INSTANCE(
2311 TRACEPOINT_EVENT_INSTANCE(
2324 [[assigning-log-levels]]
2325 ===== Assign a log level to a tracepoint definition
2327 You can assign an optional _log level_ to a
2328 <<defining-tracepoints,tracepoint definition>>.
2330 Assigning different levels of severity to tracepoint definitions can
2331 be useful: when you <<enabling-disabling-events,create an event rule>>,
2332 you can target tracepoints having a log level as severe as a specific
2335 The concept of LTTng-UST log levels is similar to the levels found
2336 in typical logging frameworks:
2338 * In a logging framework, the log level is given by the function
2339 or method name you use at the log statement site: `debug()`,
2340 `info()`, `warn()`, `error()`, and so on.
2341 * In LTTng-UST, you statically assign the log level to a tracepoint
2342 definition; any `tracepoint()` macro invocation which refers to
2343 this definition has this log level.
2345 You can assign a log level to a tracepoint definition with the
2346 `TRACEPOINT_LOGLEVEL()` macro. You must use this macro _after_ the
2347 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2348 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2351 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2354 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2356 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2361 * `provider_name` with the tracepoint provider name.
2362 * `tracepoint_name` with the tracepoint name.
2363 * `log_level` with the log level to assign to the tracepoint
2364 definition named `tracepoint_name` in the `provider_name`
2365 tracepoint provider.
2367 See man:lttng-ust(3) for a list of available log level names.
2369 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2373 /* Tracepoint definition */
2382 ctf_integer(int, userid, userid)
2383 ctf_integer(size_t, len, len)
2387 /* Log level assignment */
2388 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2394 ===== Create a tracepoint provider package source file
2396 A _tracepoint provider package source file_ is a C source file which
2397 includes a <<tpp-header,tracepoint provider header file>> to expand its
2398 macros into event serialization and other functions.
2400 You can always use the following tracepoint provider package source
2404 .Tracepoint provider package source file template.
2406 #define TRACEPOINT_CREATE_PROBES
2411 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2412 header file>> name. You may also include more than one tracepoint
2413 provider header file here to create a tracepoint provider package
2414 holding more than one tracepoint providers.
2417 [[probing-the-application-source-code]]
2418 ==== Add tracepoints to an application's source code
2420 Once you <<tpp-header,create a tracepoint provider header file>>, you
2421 can use the `tracepoint()` macro in your application's
2422 source code to insert the tracepoints that this header
2423 <<defining-tracepoints,defines>>.
2425 The `tracepoint()` macro takes at least two parameters: the tracepoint
2426 provider name and the tracepoint name. The corresponding tracepoint
2427 definition defines the other parameters.
2429 .`tracepoint()` usage.
2431 The following <<defining-tracepoints,tracepoint definition>> defines a
2432 tracepoint which takes two input arguments and has two output event
2436 .Tracepoint provider header file.
2438 #include "my-custom-structure.h"
2445 const char*, cmd_name
2448 ctf_string(cmd_name, cmd_name)
2449 ctf_integer(int, number_of_args, argc)
2454 You can refer to this tracepoint definition with the `tracepoint()`
2455 macro in your application's source code like this:
2458 .Application's source file.
2462 int main(int argc, char* argv[])
2464 tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2470 Note how the application's source code includes
2471 the tracepoint provider header file containing the tracepoint
2472 definitions to use, path:{tp.h}.
2475 .`tracepoint()` usage with a complex tracepoint definition.
2477 Consider this complex tracepoint definition, where multiple event
2478 fields refer to the same input arguments in their argument expression
2482 .Tracepoint provider header file.
2484 /* For `struct stat` */
2485 #include <sys/types.h>
2486 #include <sys/stat.h>
2498 ctf_integer(int, my_constant_field, 23 + 17)
2499 ctf_integer(int, my_int_arg_field, my_int_arg)
2500 ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2501 ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2502 my_str_arg[2] + my_str_arg[3])
2503 ctf_string(my_str_arg_field, my_str_arg)
2504 ctf_integer_hex(off_t, size_field, st->st_size)
2505 ctf_float(double, size_dbl_field, (double) st->st_size)
2506 ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2507 size_t, strlen(my_str_arg) / 2)
2512 You can refer to this tracepoint definition with the `tracepoint()`
2513 macro in your application's source code like this:
2516 .Application's source file.
2518 #define TRACEPOINT_DEFINE
2525 stat("/etc/fstab", &s);
2526 tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2532 If you look at the event record that LTTng writes when tracing this
2533 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2534 it should look like this:
2536 .Event record fields
2538 |Field's name |Field's value
2539 |`my_constant_field` |40
2540 |`my_int_arg_field` |23
2541 |`my_int_arg_field2` |529
2543 |`my_str_arg_field` |`Hello, World!`
2544 |`size_field` |0x12d
2545 |`size_dbl_field` |301.0
2546 |`half_my_str_arg_field` |`Hello,`
2550 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2551 compute--they use the call stack, for example. To avoid this
2552 computation when the tracepoint is disabled, you can use the
2553 `tracepoint_enabled()` and `do_tracepoint()` macros.
2555 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2559 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2561 tracepoint_enabled(provider_name, tracepoint_name)
2562 do_tracepoint(provider_name, tracepoint_name, ...)
2567 * `provider_name` with the tracepoint provider name.
2568 * `tracepoint_name` with the tracepoint name.
2570 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2571 `tracepoint_name` from the provider named `provider_name` is enabled
2574 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2575 if the tracepoint is enabled. Using `tracepoint()` with
2576 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2577 the `tracepoint_enabled()` check, thus a race condition is
2578 possible in this situation:
2581 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2583 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2584 stuff = prepare_stuff();
2587 tracepoint(my_provider, my_tracepoint, stuff);
2590 If the tracepoint is enabled after the condition, then `stuff` is not
2591 prepared: the emitted event will either contain wrong data, or the whole
2592 application could crash (segmentation fault, for example).
2594 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2595 `STAP_PROBEV()` call. If you need it, you must emit
2599 [[building-tracepoint-providers-and-user-application]]
2600 ==== Build and link a tracepoint provider package and an application
2602 Once you have one or more <<tpp-header,tracepoint provider header
2603 files>> and a <<tpp-source,tracepoint provider package source file>>,
2604 you can create the tracepoint provider package by compiling its source
2605 file. From here, multiple build and run scenarios are possible. The
2606 following table shows common application and library configurations
2607 along with the required command lines to achieve them.
2609 In the following diagrams, we use the following file names:
2612 Executable application.
2615 Application's object file.
2618 Tracepoint provider package object file.
2621 Tracepoint provider package archive file.
2624 Tracepoint provider package shared object file.
2627 User library object file.
2630 User library shared object file.
2632 We use the following symbols in the diagrams of table below:
2635 .Symbols used in the build scenario diagrams.
2636 image::ust-sit-symbols.png[]
2638 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2639 variable in the following instructions.
2641 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2642 .Common tracepoint provider package scenarios.
2644 |Scenario |Instructions
2647 The instrumented application is statically linked with
2648 the tracepoint provider package object.
2650 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2653 include::../common/ust-sit-step-tp-o.txt[]
2655 To build the instrumented application:
2657 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2662 #define TRACEPOINT_DEFINE
2666 . Compile the application source file:
2675 . Build the application:
2680 $ gcc -o app app.o tpp.o -llttng-ust -ldl
2684 To run the instrumented application:
2686 * Start the application:
2696 The instrumented application is statically linked with the
2697 tracepoint provider package archive file.
2699 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
2702 To create the tracepoint provider package archive file:
2704 . Compile the <<tpp-source,tracepoint provider package source file>>:
2713 . Create the tracepoint provider package archive file:
2718 $ ar rcs tpp.a tpp.o
2722 To build the instrumented application:
2724 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2729 #define TRACEPOINT_DEFINE
2733 . Compile the application source file:
2742 . Build the application:
2747 $ gcc -o app app.o tpp.a -llttng-ust -ldl
2751 To run the instrumented application:
2753 * Start the application:
2763 The instrumented application is linked with the tracepoint provider
2764 package shared object.
2766 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
2769 include::../common/ust-sit-step-tp-so.txt[]
2771 To build the instrumented application:
2773 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2778 #define TRACEPOINT_DEFINE
2782 . Compile the application source file:
2791 . Build the application:
2796 $ gcc -o app app.o -ldl -L. -ltpp
2800 To run the instrumented application:
2802 * Start the application:
2812 The tracepoint provider package shared object is preloaded before the
2813 instrumented application starts.
2815 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
2818 include::../common/ust-sit-step-tp-so.txt[]
2820 To build the instrumented application:
2822 . In path:{app.c}, before including path:{tpp.h}, add the
2828 #define TRACEPOINT_DEFINE
2829 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2833 . Compile the application source file:
2842 . Build the application:
2847 $ gcc -o app app.o -ldl
2851 To run the instrumented application with tracing support:
2853 * Preload the tracepoint provider package shared object and
2854 start the application:
2859 $ LD_PRELOAD=./libtpp.so ./app
2863 To run the instrumented application without tracing support:
2865 * Start the application:
2875 The instrumented application dynamically loads the tracepoint provider
2876 package shared object.
2878 See the <<dlclose-warning,warning about `dlclose()`>>.
2880 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
2883 include::../common/ust-sit-step-tp-so.txt[]
2885 To build the instrumented application:
2887 . In path:{app.c}, before including path:{tpp.h}, add the
2893 #define TRACEPOINT_DEFINE
2894 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2898 . Compile the application source file:
2907 . Build the application:
2912 $ gcc -o app app.o -ldl
2916 To run the instrumented application:
2918 * Start the application:
2928 The application is linked with the instrumented user library.
2930 The instrumented user library is statically linked with the tracepoint
2931 provider package object file.
2933 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
2936 include::../common/ust-sit-step-tp-o-fpic.txt[]
2938 To build the instrumented user library:
2940 . In path:{emon.c}, before including path:{tpp.h}, add the
2946 #define TRACEPOINT_DEFINE
2950 . Compile the user library source file:
2955 $ gcc -I. -fpic -c emon.c
2959 . Build the user library shared object:
2964 $ gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
2968 To build the application:
2970 . Compile the application source file:
2979 . Build the application:
2984 $ gcc -o app app.o -L. -lemon
2988 To run the application:
2990 * Start the application:
3000 The application is linked with the instrumented user library.
3002 The instrumented user library is linked with the tracepoint provider
3003 package shared object.
3005 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3008 include::../common/ust-sit-step-tp-so.txt[]
3010 To build the instrumented user library:
3012 . In path:{emon.c}, before including path:{tpp.h}, add the
3018 #define TRACEPOINT_DEFINE
3022 . Compile the user library source file:
3027 $ gcc -I. -fpic -c emon.c
3031 . Build the user library shared object:
3036 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3040 To build the application:
3042 . Compile the application source file:
3051 . Build the application:
3056 $ gcc -o app app.o -L. -lemon
3060 To run the application:
3062 * Start the application:
3072 The tracepoint provider package shared object is preloaded before the
3075 The application is linked with the instrumented user library.
3077 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
3080 include::../common/ust-sit-step-tp-so.txt[]
3082 To build the instrumented user library:
3084 . In path:{emon.c}, before including path:{tpp.h}, add the
3090 #define TRACEPOINT_DEFINE
3091 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3095 . Compile the user library source file:
3100 $ gcc -I. -fpic -c emon.c
3104 . Build the user library shared object:
3109 $ gcc -shared -o libemon.so emon.o -ldl
3113 To build the application:
3115 . Compile the application source file:
3124 . Build the application:
3129 $ gcc -o app app.o -L. -lemon
3133 To run the application with tracing support:
3135 * Preload the tracepoint provider package shared object and
3136 start the application:
3141 $ LD_PRELOAD=./libtpp.so ./app
3145 To run the application without tracing support:
3147 * Start the application:
3157 The application is linked with the instrumented user library.
3159 The instrumented user library dynamically loads the tracepoint provider
3160 package shared object.
3162 See the <<dlclose-warning,warning about `dlclose()`>>.
3164 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3167 include::../common/ust-sit-step-tp-so.txt[]
3169 To build the instrumented user library:
3171 . In path:{emon.c}, before including path:{tpp.h}, add the
3177 #define TRACEPOINT_DEFINE
3178 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3182 . Compile the user library source file:
3187 $ gcc -I. -fpic -c emon.c
3191 . Build the user library shared object:
3196 $ gcc -shared -o libemon.so emon.o -ldl
3200 To build the application:
3202 . Compile the application source file:
3211 . Build the application:
3216 $ gcc -o app app.o -L. -lemon
3220 To run the application:
3222 * Start the application:
3232 The application dynamically loads the instrumented user library.
3234 The instrumented user library is linked with the tracepoint provider
3235 package shared object.
3237 See the <<dlclose-warning,warning about `dlclose()`>>.
3239 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3242 include::../common/ust-sit-step-tp-so.txt[]
3244 To build the instrumented user library:
3246 . In path:{emon.c}, before including path:{tpp.h}, add the
3252 #define TRACEPOINT_DEFINE
3256 . Compile the user library source file:
3261 $ gcc -I. -fpic -c emon.c
3265 . Build the user library shared object:
3270 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3274 To build the application:
3276 . Compile the application source file:
3285 . Build the application:
3290 $ gcc -o app app.o -ldl -L. -lemon
3294 To run the application:
3296 * Start the application:
3306 The application dynamically loads the instrumented user library.
3308 The instrumented user library dynamically loads the tracepoint provider
3309 package shared object.
3311 See the <<dlclose-warning,warning about `dlclose()`>>.
3313 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3316 include::../common/ust-sit-step-tp-so.txt[]
3318 To build the instrumented user library:
3320 . In path:{emon.c}, before including path:{tpp.h}, add the
3326 #define TRACEPOINT_DEFINE
3327 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3331 . Compile the user library source file:
3336 $ gcc -I. -fpic -c emon.c
3340 . Build the user library shared object:
3345 $ gcc -shared -o libemon.so emon.o -ldl
3349 To build the application:
3351 . Compile the application source file:
3360 . Build the application:
3365 $ gcc -o app app.o -ldl -L. -lemon
3369 To run the application:
3371 * Start the application:
3381 The tracepoint provider package shared object is preloaded before the
3384 The application dynamically loads the instrumented user library.
3386 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3389 include::../common/ust-sit-step-tp-so.txt[]
3391 To build the instrumented user library:
3393 . In path:{emon.c}, before including path:{tpp.h}, add the
3399 #define TRACEPOINT_DEFINE
3400 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3404 . Compile the user library source file:
3409 $ gcc -I. -fpic -c emon.c
3413 . Build the user library shared object:
3418 $ gcc -shared -o libemon.so emon.o -ldl
3422 To build the application:
3424 . Compile the application source file:
3433 . Build the application:
3438 $ gcc -o app app.o -L. -lemon
3442 To run the application with tracing support:
3444 * Preload the tracepoint provider package shared object and
3445 start the application:
3450 $ LD_PRELOAD=./libtpp.so ./app
3454 To run the application without tracing support:
3456 * Start the application:
3466 The application is statically linked with the tracepoint provider
3467 package object file.
3469 The application is linked with the instrumented user library.
3471 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3474 include::../common/ust-sit-step-tp-o.txt[]
3476 To build the instrumented user library:
3478 . In path:{emon.c}, before including path:{tpp.h}, add the
3484 #define TRACEPOINT_DEFINE
3488 . Compile the user library source file:
3493 $ gcc -I. -fpic -c emon.c
3497 . Build the user library shared object:
3502 $ gcc -shared -o libemon.so emon.o
3506 To build the application:
3508 . Compile the application source file:
3517 . Build the application:
3522 $ gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3526 To run the instrumented application:
3528 * Start the application:
3538 The application is statically linked with the tracepoint provider
3539 package object file.
3541 The application dynamically loads the instrumented user library.
3543 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3546 include::../common/ust-sit-step-tp-o.txt[]
3548 To build the application:
3550 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3555 #define TRACEPOINT_DEFINE
3559 . Compile the application source file:
3568 . Build the application:
3573 $ gcc -Wl,--export-dynamic -o app app.o tpp.o \
3578 The `--export-dynamic` option passed to the linker is necessary for the
3579 dynamically loaded library to ``see'' the tracepoint symbols defined in
3582 To build the instrumented user library:
3584 . Compile the user library source file:
3589 $ gcc -I. -fpic -c emon.c
3593 . Build the user library shared object:
3598 $ gcc -shared -o libemon.so emon.o
3602 To run the application:
3604 * Start the application:
3616 .Do not use man:dlclose(3) on a tracepoint provider package
3618 Never use man:dlclose(3) on any shared object which:
3620 * Is linked with, statically or dynamically, a tracepoint provider
3622 * Calls man:dlopen(3) itself to dynamically open a tracepoint provider
3623 package shared object.
3625 This is currently considered **unsafe** due to a lack of reference
3626 counting from LTTng-UST to the shared object.
3628 A known workaround (available since glibc 2.2) is to use the
3629 `RTLD_NODELETE` flag when calling man:dlopen(3) initially. This has the
3630 effect of not unloading the loaded shared object, even if man:dlclose(3)
3633 You can also preload the tracepoint provider package shared object with
3634 the env:LD_PRELOAD environment variable to overcome this limitation.
3638 [[using-lttng-ust-with-daemons]]
3639 ===== Use noch:{LTTng-UST} with daemons
3641 If your instrumented application calls man:fork(2), man:clone(2),
3642 or BSD's man:rfork(2), without a following man:exec(3)-family
3643 system call, you must preload the path:{liblttng-ust-fork.so} shared
3644 object when you start the application.
3648 $ LD_PRELOAD=liblttng-ust-fork.so ./my-app
3651 If your tracepoint provider package is
3652 a shared library which you also preload, you must put both
3653 shared objects in env:LD_PRELOAD:
3657 $ LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3663 ===== Use noch:{LTTng-UST} with applications which close file descriptors that don't belong to them
3665 If your instrumented application closes one or more file descriptors
3666 which it did not open itself, you must preload the
3667 path:{liblttng-ust-fd.so} shared object when you start the application:
3671 $ LD_PRELOAD=liblttng-ust-fd.so ./my-app
3674 Typical use cases include closing all the file descriptors after
3675 man:fork(2) or man:rfork(2) and buggy applications doing
3679 [[lttng-ust-pkg-config]]
3680 ===== Use noch:{pkg-config}
3682 On some distributions, LTTng-UST ships with a
3683 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3684 metadata file. If this is your case, then you can use cmd:pkg-config to
3685 build an application on the command line:
3689 $ gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3693 [[instrumenting-32-bit-app-on-64-bit-system]]
3694 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3696 In order to trace a 32-bit application running on a 64-bit system,
3697 LTTng must use a dedicated 32-bit
3698 <<lttng-consumerd,consumer daemon>>.
3700 The following steps show how to build and install a 32-bit consumer
3701 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3702 build and install the 32-bit LTTng-UST libraries, and how to build and
3703 link an instrumented 32-bit application in that context.
3705 To build a 32-bit instrumented application for a 64-bit target system,
3706 assuming you have a fresh target system with no installed Userspace RCU
3709 . Download, build, and install a 32-bit version of Userspace RCU:
3714 $ cd $(mktemp -d) &&
3715 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
3716 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
3717 cd userspace-rcu-0.9.* &&
3718 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
3720 sudo make install &&
3725 . Using your distribution's package manager, or from source, install
3726 the following 32-bit versions of the following dependencies of
3727 LTTng-tools and LTTng-UST:
3730 * https://sourceforge.net/projects/libuuid/[libuuid]
3731 * http://directory.fsf.org/wiki/Popt[popt]
3732 * http://www.xmlsoft.org/[libxml2]
3735 . Download, build, and install a 32-bit version of the latest
3736 LTTng-UST{nbsp}{revision}:
3741 $ cd $(mktemp -d) &&
3742 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.10.tar.bz2 &&
3743 tar -xf lttng-ust-latest-2.10.tar.bz2 &&
3744 cd lttng-ust-2.10.* &&
3745 ./configure --libdir=/usr/local/lib32 \
3746 CFLAGS=-m32 CXXFLAGS=-m32 \
3747 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3749 sudo make install &&
3756 Depending on your distribution,
3757 32-bit libraries could be installed at a different location than
3758 `/usr/lib32`. For example, Debian is known to install
3759 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
3761 In this case, make sure to set `LDFLAGS` to all the
3762 relevant 32-bit library paths, for example:
3766 $ LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
3770 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
3771 the 32-bit consumer daemon:
3776 $ cd $(mktemp -d) &&
3777 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.10.tar.bz2 &&
3778 tar -xf lttng-tools-latest-2.10.tar.bz2 &&
3779 cd lttng-tools-2.10.* &&
3780 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
3781 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' \
3782 --disable-bin-lttng --disable-bin-lttng-crash \
3783 --disable-bin-lttng-relayd --disable-bin-lttng-sessiond &&
3785 cd src/bin/lttng-consumerd &&
3786 sudo make install &&
3791 . From your distribution or from source,
3792 <<installing-lttng,install>> the 64-bit versions of
3793 LTTng-UST and Userspace RCU.
3794 . Download, build, and install the 64-bit version of the
3795 latest LTTng-tools{nbsp}{revision}:
3800 $ cd $(mktemp -d) &&
3801 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.10.tar.bz2 &&
3802 tar -xf lttng-tools-latest-2.10.tar.bz2 &&
3803 cd lttng-tools-2.10.* &&
3804 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
3805 --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
3807 sudo make install &&
3812 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
3813 when linking your 32-bit application:
3816 -m32 -L/usr/lib32 -L/usr/local/lib32 \
3817 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
3820 For example, let's rebuild the quick start example in
3821 <<tracing-your-own-user-application,Trace a user application>> as an
3822 instrumented 32-bit application:
3827 $ gcc -m32 -c -I. hello-tp.c
3828 $ gcc -m32 -c hello.c
3829 $ gcc -m32 -o hello hello.o hello-tp.o \
3830 -L/usr/lib32 -L/usr/local/lib32 \
3831 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
3836 No special action is required to execute the 32-bit application and
3837 to trace it: use the command-line man:lttng(1) tool as usual.
3844 man:tracef(3) is a small LTTng-UST API designed for quick,
3845 man:printf(3)-like instrumentation without the burden of
3846 <<tracepoint-provider,creating>> and
3847 <<building-tracepoint-providers-and-user-application,building>>
3848 a tracepoint provider package.
3850 To use `tracef()` in your application:
3852 . In the C or C++ source files where you need to use `tracef()`,
3853 include `<lttng/tracef.h>`:
3858 #include <lttng/tracef.h>
3862 . In the application's source code, use `tracef()` like you would use
3870 tracef("my message: %d (%s)", my_integer, my_string);
3876 . Link your application with `liblttng-ust`:
3881 $ gcc -o app app.c -llttng-ust
3885 To trace the events that `tracef()` calls emit:
3887 * <<enabling-disabling-events,Create an event rule>> which matches the
3888 `lttng_ust_tracef:*` event name:
3893 $ lttng enable-event --userspace 'lttng_ust_tracef:*'
3898 .Limitations of `tracef()`
3900 The `tracef()` utility function was developed to make user space tracing
3901 super simple, albeit with notable disadvantages compared to
3902 <<defining-tracepoints,user-defined tracepoints>>:
3904 * All the emitted events have the same tracepoint provider and
3905 tracepoint names, respectively `lttng_ust_tracef` and `event`.
3906 * There is no static type checking.
3907 * The only event record field you actually get, named `msg`, is a string
3908 potentially containing the values you passed to `tracef()`
3909 using your own format string. This also means that you cannot filter
3910 events with a custom expression at run time because there are no
3912 * Since `tracef()` uses the C standard library's man:vasprintf(3)
3913 function behind the scenes to format the strings at run time, its
3914 expected performance is lower than with user-defined tracepoints,
3915 which do not require a conversion to a string.
3917 Taking this into consideration, `tracef()` is useful for some quick
3918 prototyping and debugging, but you should not consider it for any
3919 permanent and serious applicative instrumentation.
3925 ==== Use `tracelog()`
3927 The man:tracelog(3) API is very similar to <<tracef,`tracef()`>>, with
3928 the difference that it accepts an additional log level parameter.
3930 The goal of `tracelog()` is to ease the migration from logging to
3933 To use `tracelog()` in your application:
3935 . In the C or C++ source files where you need to use `tracelog()`,
3936 include `<lttng/tracelog.h>`:
3941 #include <lttng/tracelog.h>
3945 . In the application's source code, use `tracelog()` like you would use
3946 man:printf(3), except for the first parameter which is the log
3954 tracelog(TRACE_WARNING, "my message: %d (%s)",
3955 my_integer, my_string);
3961 See man:lttng-ust(3) for a list of available log level names.
3963 . Link your application with `liblttng-ust`:
3968 $ gcc -o app app.c -llttng-ust
3972 To trace the events that `tracelog()` calls emit with a log level
3973 _as severe as_ a specific log level:
3975 * <<enabling-disabling-events,Create an event rule>> which matches the
3976 `lttng_ust_tracelog:*` event name and a minimum level
3982 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
3983 --loglevel=TRACE_WARNING
3987 To trace the events that `tracelog()` calls emit with a
3988 _specific log level_:
3990 * Create an event rule which matches the `lttng_ust_tracelog:*`
3991 event name and a specific log level:
3996 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
3997 --loglevel-only=TRACE_INFO
4002 [[prebuilt-ust-helpers]]
4003 === Prebuilt user space tracing helpers
4005 The LTTng-UST package provides a few helpers in the form or preloadable
4006 shared objects which automatically instrument system functions and
4009 The helper shared objects are normally found in dir:{/usr/lib}. If you
4010 built LTTng-UST <<building-from-source,from source>>, they are probably
4011 located in dir:{/usr/local/lib}.
4013 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
4016 path:{liblttng-ust-libc-wrapper.so}::
4017 path:{liblttng-ust-pthread-wrapper.so}::
4018 <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
4019 memory and POSIX threads function tracing>>.
4021 path:{liblttng-ust-cyg-profile.so}::
4022 path:{liblttng-ust-cyg-profile-fast.so}::
4023 <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
4025 path:{liblttng-ust-dl.so}::
4026 <<liblttng-ust-dl,Dynamic linker tracing>>.
4028 To use a user space tracing helper with any user application:
4030 * Preload the helper shared object when you start the application:
4035 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
4039 You can preload more than one helper:
4044 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
4050 [[liblttng-ust-libc-pthread-wrapper]]
4051 ==== Instrument C standard library memory and POSIX threads functions
4053 The path:{liblttng-ust-libc-wrapper.so} and
4054 path:{liblttng-ust-pthread-wrapper.so} helpers
4055 add instrumentation to some C standard library and POSIX
4059 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
4061 |TP provider name |TP name |Instrumented function
4063 .6+|`lttng_ust_libc` |`malloc` |man:malloc(3)
4064 |`calloc` |man:calloc(3)
4065 |`realloc` |man:realloc(3)
4066 |`free` |man:free(3)
4067 |`memalign` |man:memalign(3)
4068 |`posix_memalign` |man:posix_memalign(3)
4072 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
4074 |TP provider name |TP name |Instrumented function
4076 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
4077 |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
4078 |`pthread_mutex_trylock` |man:pthread_mutex_trylock(3p)
4079 |`pthread_mutex_unlock` |man:pthread_mutex_unlock(3p)
4082 When you preload the shared object, it replaces the functions listed
4083 in the previous tables by wrappers which contain tracepoints and call
4084 the replaced functions.
4087 [[liblttng-ust-cyg-profile]]
4088 ==== Instrument function entry and exit
4090 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
4091 to the entry and exit points of functions.
4093 man:gcc(1) and man:clang(1) have an option named
4094 https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html[`-finstrument-functions`]
4095 which generates instrumentation calls for entry and exit to functions.
4096 The LTTng-UST function tracing helpers,
4097 path:{liblttng-ust-cyg-profile.so} and
4098 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
4099 to add tracepoints to the two generated functions (which contain
4100 `cyg_profile` in their names, hence the helper's name).
4102 To use the LTTng-UST function tracing helper, the source files to
4103 instrument must be built using the `-finstrument-functions` compiler
4106 There are two versions of the LTTng-UST function tracing helper:
4108 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
4109 that you should only use when it can be _guaranteed_ that the
4110 complete event stream is recorded without any lost event record.
4111 Any kind of duplicate information is left out.
4113 Assuming no event record is lost, having only the function addresses on
4114 entry is enough to create a call graph, since an event record always
4115 contains the ID of the CPU that generated it.
4117 You can use a tool like man:addr2line(1) to convert function addresses
4118 back to source file names and line numbers.
4120 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
4121 which also works in use cases where event records might get discarded or
4122 not recorded from application startup.
4123 In these cases, the trace analyzer needs more information to be
4124 able to reconstruct the program flow.
4126 See man:lttng-ust-cyg-profile(3) to learn more about the instrumentation
4127 points of this helper.
4129 All the tracepoints that this helper provides have the
4130 log level `TRACE_DEBUG_FUNCTION` (see man:lttng-ust(3)).
4132 TIP: It's sometimes a good idea to limit the number of source files that
4133 you compile with the `-finstrument-functions` option to prevent LTTng
4134 from writing an excessive amount of trace data at run time. When using
4135 man:gcc(1), you can use the
4136 `-finstrument-functions-exclude-function-list` option to avoid
4137 instrument entries and exits of specific function names.
4142 ==== Instrument the dynamic linker
4144 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4145 man:dlopen(3) and man:dlclose(3) function calls.
4147 See man:lttng-ust-dl(3) to learn more about the instrumentation points
4152 [[java-application]]
4153 === User space Java agent
4155 You can instrument any Java application which uses one of the following
4158 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4159 (JUL) core logging facilities.
4160 * http://logging.apache.org/log4j/1.2/[**Apache log4j 1.2**], since
4161 LTTng 2.6. Note that Apache Log4j{nbsp}2 is not supported.
4164 .LTTng-UST Java agent imported by a Java application.
4165 image::java-app.png[]
4167 Note that the methods described below are new in LTTng{nbsp}{revision}.
4168 Previous LTTng versions use another technique.
4170 NOTE: We use http://openjdk.java.net/[OpenJDK]{nbsp}8 for development
4171 and https://ci.lttng.org/[continuous integration], thus this version is
4172 directly supported. However, the LTTng-UST Java agent is also tested
4173 with OpenJDK{nbsp}7.
4178 ==== Use the LTTng-UST Java agent for `java.util.logging`
4180 To use the LTTng-UST Java agent in a Java application which uses
4181 `java.util.logging` (JUL):
4183 . In the Java application's source code, import the LTTng-UST
4184 log handler package for `java.util.logging`:
4189 import org.lttng.ust.agent.jul.LttngLogHandler;
4193 . Create an LTTng-UST JUL log handler:
4198 Handler lttngUstLogHandler = new LttngLogHandler();
4202 . Add this handler to the JUL loggers which should emit LTTng events:
4207 Logger myLogger = Logger.getLogger("some-logger");
4209 myLogger.addHandler(lttngUstLogHandler);
4213 . Use `java.util.logging` log statements and configuration as usual.
4214 The loggers with an attached LTTng-UST log handler can emit
4217 . Before exiting the application, remove the LTTng-UST log handler from
4218 the loggers attached to it and call its `close()` method:
4223 myLogger.removeHandler(lttngUstLogHandler);
4224 lttngUstLogHandler.close();
4228 This is not strictly necessary, but it is recommended for a clean
4229 disposal of the handler's resources.
4231 . Include the LTTng-UST Java agent's common and JUL-specific JAR files,
4232 path:{lttng-ust-agent-common.jar} and path:{lttng-ust-agent-jul.jar},
4234 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4235 path] when you build the Java application.
4237 The JAR files are typically located in dir:{/usr/share/java}.
4239 IMPORTANT: The LTTng-UST Java agent must be
4240 <<installing-lttng,installed>> for the logging framework your
4243 .Use the LTTng-UST Java agent for `java.util.logging`.
4248 import java.io.IOException;
4249 import java.util.logging.Handler;
4250 import java.util.logging.Logger;
4251 import org.lttng.ust.agent.jul.LttngLogHandler;
4255 private static final int answer = 42;
4257 public static void main(String[] argv) throws Exception
4260 Logger logger = Logger.getLogger("jello");
4262 // Create an LTTng-UST log handler
4263 Handler lttngUstLogHandler = new LttngLogHandler();
4265 // Add the LTTng-UST log handler to our logger
4266 logger.addHandler(lttngUstLogHandler);
4269 logger.info("some info");
4270 logger.warning("some warning");
4272 logger.finer("finer information; the answer is " + answer);
4274 logger.severe("error!");
4276 // Not mandatory, but cleaner
4277 logger.removeHandler(lttngUstLogHandler);
4278 lttngUstLogHandler.close();
4287 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4290 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4291 <<enabling-disabling-events,create an event rule>> matching the
4292 `jello` JUL logger, and <<basic-tracing-session-control,start tracing>>:
4297 $ lttng enable-event --jul jello
4301 Run the compiled class:
4305 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4308 <<basic-tracing-session-control,Stop tracing>> and inspect the
4318 In the resulting trace, an <<event,event record>> generated by a Java
4319 application using `java.util.logging` is named `lttng_jul:event` and
4320 has the following fields:
4323 Log record's message.
4329 Name of the class in which the log statement was executed.
4332 Name of the method in which the log statement was executed.
4335 Logging time (timestamp in milliseconds).
4338 Log level integer value.
4341 ID of the thread in which the log statement was executed.
4343 You can use the opt:lttng-enable-event(1):--loglevel or
4344 opt:lttng-enable-event(1):--loglevel-only option of the
4345 man:lttng-enable-event(1) command to target a range of JUL log levels
4346 or a specific JUL log level.
4351 ==== Use the LTTng-UST Java agent for Apache log4j
4353 To use the LTTng-UST Java agent in a Java application which uses
4356 . In the Java application's source code, import the LTTng-UST
4357 log appender package for Apache log4j:
4362 import org.lttng.ust.agent.log4j.LttngLogAppender;
4366 . Create an LTTng-UST log4j log appender:
4371 Appender lttngUstLogAppender = new LttngLogAppender();
4375 . Add this appender to the log4j loggers which should emit LTTng events:
4380 Logger myLogger = Logger.getLogger("some-logger");
4382 myLogger.addAppender(lttngUstLogAppender);
4386 . Use Apache log4j log statements and configuration as usual. The
4387 loggers with an attached LTTng-UST log appender can emit LTTng events.
4389 . Before exiting the application, remove the LTTng-UST log appender from
4390 the loggers attached to it and call its `close()` method:
4395 myLogger.removeAppender(lttngUstLogAppender);
4396 lttngUstLogAppender.close();
4400 This is not strictly necessary, but it is recommended for a clean
4401 disposal of the appender's resources.
4403 . Include the LTTng-UST Java agent's common and log4j-specific JAR
4404 files, path:{lttng-ust-agent-common.jar} and
4405 path:{lttng-ust-agent-log4j.jar}, in the
4406 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4407 path] when you build the Java application.
4409 The JAR files are typically located in dir:{/usr/share/java}.
4411 IMPORTANT: The LTTng-UST Java agent must be
4412 <<installing-lttng,installed>> for the logging framework your
4415 .Use the LTTng-UST Java agent for Apache log4j.
4420 import org.apache.log4j.Appender;
4421 import org.apache.log4j.Logger;
4422 import org.lttng.ust.agent.log4j.LttngLogAppender;
4426 private static final int answer = 42;
4428 public static void main(String[] argv) throws Exception
4431 Logger logger = Logger.getLogger("jello");
4433 // Create an LTTng-UST log appender
4434 Appender lttngUstLogAppender = new LttngLogAppender();
4436 // Add the LTTng-UST log appender to our logger
4437 logger.addAppender(lttngUstLogAppender);
4440 logger.info("some info");
4441 logger.warn("some warning");
4443 logger.debug("debug information; the answer is " + answer);
4445 logger.fatal("error!");
4447 // Not mandatory, but cleaner
4448 logger.removeAppender(lttngUstLogAppender);
4449 lttngUstLogAppender.close();
4455 Build this example (`$LOG4JPATH` is the path to the Apache log4j JAR
4460 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH Test.java
4463 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4464 <<enabling-disabling-events,create an event rule>> matching the
4465 `jello` log4j logger, and <<basic-tracing-session-control,start tracing>>:
4470 $ lttng enable-event --log4j jello
4474 Run the compiled class:
4478 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH:. Test
4481 <<basic-tracing-session-control,Stop tracing>> and inspect the
4491 In the resulting trace, an <<event,event record>> generated by a Java
4492 application using log4j is named `lttng_log4j:event` and
4493 has the following fields:
4496 Log record's message.
4502 Name of the class in which the log statement was executed.
4505 Name of the method in which the log statement was executed.
4508 Name of the file in which the executed log statement is located.
4511 Line number at which the log statement was executed.
4517 Log level integer value.
4520 Name of the Java thread in which the log statement was executed.
4522 You can use the opt:lttng-enable-event(1):--loglevel or
4523 opt:lttng-enable-event(1):--loglevel-only option of the
4524 man:lttng-enable-event(1) command to target a range of Apache log4j log levels
4525 or a specific log4j log level.
4529 [[java-application-context]]
4530 ==== Provide application-specific context fields in a Java application
4532 A Java application-specific context field is a piece of state provided
4533 by the application which <<adding-context,you can add>>, using the
4534 man:lttng-add-context(1) command, to each <<event,event record>>
4535 produced by the log statements of this application.
4537 For example, a given object might have a current request ID variable.
4538 You can create a context information retriever for this object and
4539 assign a name to this current request ID. You can then, using the
4540 man:lttng-add-context(1) command, add this context field by name to
4541 the JUL or log4j <<channel,channel>>.
4543 To provide application-specific context fields in a Java application:
4545 . In the Java application's source code, import the LTTng-UST
4546 Java agent context classes and interfaces:
4551 import org.lttng.ust.agent.context.ContextInfoManager;
4552 import org.lttng.ust.agent.context.IContextInfoRetriever;
4556 . Create a context information retriever class, that is, a class which
4557 implements the `IContextInfoRetriever` interface:
4562 class MyContextInfoRetriever implements IContextInfoRetriever
4565 public Object retrieveContextInfo(String key)
4567 if (key.equals("intCtx")) {
4569 } else if (key.equals("strContext")) {
4570 return "context value!";
4579 This `retrieveContextInfo()` method is the only member of the
4580 `IContextInfoRetriever` interface. Its role is to return the current
4581 value of a state by name to create a context field. The names of the
4582 context fields and which state variables they return depends on your
4585 All primitive types and objects are supported as context fields.
4586 When `retrieveContextInfo()` returns an object, the context field
4587 serializer calls its `toString()` method to add a string field to
4588 event records. The method can also return `null`, which means that
4589 no context field is available for the required name.
4591 . Register an instance of your context information retriever class to
4592 the context information manager singleton:
4597 IContextInfoRetriever cir = new MyContextInfoRetriever();
4598 ContextInfoManager cim = ContextInfoManager.getInstance();
4599 cim.registerContextInfoRetriever("retrieverName", cir);
4603 . Before exiting the application, remove your context information
4604 retriever from the context information manager singleton:
4609 ContextInfoManager cim = ContextInfoManager.getInstance();
4610 cim.unregisterContextInfoRetriever("retrieverName");
4614 This is not strictly necessary, but it is recommended for a clean
4615 disposal of some manager's resources.
4617 . Build your Java application with LTTng-UST Java agent support as
4618 usual, following the procedure for either the <<jul,JUL>> or
4619 <<log4j,Apache log4j>> framework.
4622 .Provide application-specific context fields in a Java application.
4627 import java.util.logging.Handler;
4628 import java.util.logging.Logger;
4629 import org.lttng.ust.agent.jul.LttngLogHandler;
4630 import org.lttng.ust.agent.context.ContextInfoManager;
4631 import org.lttng.ust.agent.context.IContextInfoRetriever;
4635 // Our context information retriever class
4636 private static class MyContextInfoRetriever
4637 implements IContextInfoRetriever
4640 public Object retrieveContextInfo(String key) {
4641 if (key.equals("intCtx")) {
4643 } else if (key.equals("strContext")) {
4644 return "context value!";
4651 private static final int answer = 42;
4653 public static void main(String args[]) throws Exception
4655 // Get the context information manager instance
4656 ContextInfoManager cim = ContextInfoManager.getInstance();
4658 // Create and register our context information retriever
4659 IContextInfoRetriever cir = new MyContextInfoRetriever();
4660 cim.registerContextInfoRetriever("myRetriever", cir);
4663 Logger logger = Logger.getLogger("jello");
4665 // Create an LTTng-UST log handler
4666 Handler lttngUstLogHandler = new LttngLogHandler();
4668 // Add the LTTng-UST log handler to our logger
4669 logger.addHandler(lttngUstLogHandler);
4672 logger.info("some info");
4673 logger.warning("some warning");
4675 logger.finer("finer information; the answer is " + answer);
4677 logger.severe("error!");
4679 // Not mandatory, but cleaner
4680 logger.removeHandler(lttngUstLogHandler);
4681 lttngUstLogHandler.close();
4682 cim.unregisterContextInfoRetriever("myRetriever");
4691 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4694 <<creating-destroying-tracing-sessions,Create a tracing session>>
4695 and <<enabling-disabling-events,create an event rule>> matching the
4701 $ lttng enable-event --jul jello
4704 <<adding-context,Add the application-specific context fields>> to the
4709 $ lttng add-context --jul --type='$app.myRetriever:intCtx'
4710 $ lttng add-context --jul --type='$app.myRetriever:strContext'
4713 <<basic-tracing-session-control,Start tracing>>:
4720 Run the compiled class:
4724 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4727 <<basic-tracing-session-control,Stop tracing>> and inspect the
4739 [[python-application]]
4740 === User space Python agent
4742 You can instrument a Python 2 or Python 3 application which uses the
4743 standard https://docs.python.org/3/library/logging.html[`logging`]
4746 Each log statement emits an LTTng event once the
4747 application module imports the
4748 <<lttng-ust-agents,LTTng-UST Python agent>> package.
4751 .A Python application importing the LTTng-UST Python agent.
4752 image::python-app.png[]
4754 To use the LTTng-UST Python agent:
4756 . In the Python application's source code, import the LTTng-UST Python
4766 The LTTng-UST Python agent automatically adds its logging handler to the
4767 root logger at import time.
4769 Any log statement that the application executes before this import does
4770 not emit an LTTng event.
4772 IMPORTANT: The LTTng-UST Python agent must be
4773 <<installing-lttng,installed>>.
4775 . Use log statements and logging configuration as usual.
4776 Since the LTTng-UST Python agent adds a handler to the _root_
4777 logger, you can trace any log statement from any logger.
4779 .Use the LTTng-UST Python agent.
4790 logging.basicConfig()
4791 logger = logging.getLogger('my-logger')
4794 logger.debug('debug message')
4795 logger.info('info message')
4796 logger.warn('warn message')
4797 logger.error('error message')
4798 logger.critical('critical message')
4802 if __name__ == '__main__':
4806 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
4807 logging handler which prints to the standard error stream, is not
4808 strictly required for LTTng-UST tracing to work, but in versions of
4809 Python preceding 3.2, you could see a warning message which indicates
4810 that no handler exists for the logger `my-logger`.
4812 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4813 <<enabling-disabling-events,create an event rule>> matching the
4814 `my-logger` Python logger, and <<basic-tracing-session-control,start
4820 $ lttng enable-event --python my-logger
4824 Run the Python script:
4831 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
4841 In the resulting trace, an <<event,event record>> generated by a Python
4842 application is named `lttng_python:event` and has the following fields:
4845 Logging time (string).
4848 Log record's message.
4854 Name of the function in which the log statement was executed.
4857 Line number at which the log statement was executed.
4860 Log level integer value.
4863 ID of the Python thread in which the log statement was executed.
4866 Name of the Python thread in which the log statement was executed.
4868 You can use the opt:lttng-enable-event(1):--loglevel or
4869 opt:lttng-enable-event(1):--loglevel-only option of the
4870 man:lttng-enable-event(1) command to target a range of Python log levels
4871 or a specific Python log level.
4873 When an application imports the LTTng-UST Python agent, the agent tries
4874 to register to a <<lttng-sessiond,session daemon>>. Note that you must
4875 <<start-sessiond,start the session daemon>> _before_ you run the Python
4876 application. If a session daemon is found, the agent tries to register
4877 to it during 5{nbsp}seconds, after which the application continues
4878 without LTTng tracing support. You can override this timeout value with
4879 the env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
4882 If the session daemon stops while a Python application with an imported
4883 LTTng-UST Python agent runs, the agent retries to connect and to
4884 register to a session daemon every 3{nbsp}seconds. You can override this
4885 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
4890 [[proc-lttng-logger-abi]]
4893 The `lttng-tracer` Linux kernel module, part of
4894 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger file
4895 path:{/proc/lttng-logger} when it's loaded. Any application can write
4896 text data to this file to emit an LTTng event.
4899 .An application writes to the LTTng logger file to emit an LTTng event.
4900 image::lttng-logger.png[]
4902 The LTTng logger is the quickest method--not the most efficient,
4903 however--to add instrumentation to an application. It is designed
4904 mostly to instrument shell scripts:
4908 $ echo "Some message, some $variable" > /proc/lttng-logger
4911 Any event that the LTTng logger emits is named `lttng_logger` and
4912 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
4913 other instrumentation points in the kernel tracing domain, **any Unix
4914 user** can <<enabling-disabling-events,create an event rule>> which
4915 matches its event name, not only the root user or users in the
4916 <<tracing-group,tracing group>>.
4918 To use the LTTng logger:
4920 * From any application, write text data to the path:{/proc/lttng-logger}
4923 The `msg` field of `lttng_logger` event records contains the
4926 NOTE: The maximum message length of an LTTng logger event is
4927 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
4928 than one event to contain the remaining data.
4930 You should not use the LTTng logger to trace a user application which
4931 can be instrumented in a more efficient way, namely:
4933 * <<c-application,C and $$C++$$ applications>>.
4934 * <<java-application,Java applications>>.
4935 * <<python-application,Python applications>>.
4937 .Use the LTTng logger.
4942 echo 'Hello, World!' > /proc/lttng-logger
4944 df --human-readable --print-type / > /proc/lttng-logger
4947 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4948 <<enabling-disabling-events,create an event rule>> matching the
4949 `lttng_logger` Linux kernel tracepoint, and
4950 <<basic-tracing-session-control,start tracing>>:
4955 $ lttng enable-event --kernel lttng_logger
4959 Run the Bash script:
4966 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
4977 [[instrumenting-linux-kernel]]
4978 === LTTng kernel tracepoints
4980 NOTE: This section shows how to _add_ instrumentation points to the
4981 Linux kernel. The kernel's subsystems are already thoroughly
4982 instrumented at strategic places for LTTng when you
4983 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
4987 There are two methods to instrument the Linux kernel:
4989 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
4990 tracepoint which uses the `TRACE_EVENT()` API.
4992 Choose this if you want to instrumentation a Linux kernel tree with an
4993 instrumentation point compatible with ftrace, perf, and SystemTap.
4995 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
4996 instrument an out-of-tree kernel module.
4998 Choose this if you don't need ftrace, perf, or SystemTap support.
5002 [[linux-add-lttng-layer]]
5003 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
5005 This section shows how to add an LTTng layer to existing ftrace
5006 instrumentation using the `TRACE_EVENT()` API.
5008 This section does not document the `TRACE_EVENT()` macro. You can
5009 read the following articles to learn more about this API:
5011 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part 1)]
5012 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part 2)]
5013 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part 3)]
5015 The following procedure assumes that your ftrace tracepoints are
5016 correctly defined in their own header and that they are created in
5017 one source file using the `CREATE_TRACE_POINTS` definition.
5019 To add an LTTng layer over an existing ftrace tracepoint:
5021 . Make sure the following kernel configuration options are
5027 * `CONFIG_HIGH_RES_TIMERS`
5028 * `CONFIG_TRACEPOINTS`
5031 . Build the Linux source tree with your custom ftrace tracepoints.
5032 . Boot the resulting Linux image on your target system.
5034 Confirm that the tracepoints exist by looking for their names in the
5035 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
5036 is your subsystem's name.
5038 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
5043 $ cd $(mktemp -d) &&
5044 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.10.tar.bz2 &&
5045 tar -xf lttng-modules-latest-2.10.tar.bz2 &&
5046 cd lttng-modules-2.10.*
5050 . In dir:{instrumentation/events/lttng-module}, relative to the root
5051 of the LTTng-modules source tree, create a header file named
5052 +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
5053 LTTng-modules tracepoint definitions using the LTTng-modules
5056 Start with this template:
5060 .path:{instrumentation/events/lttng-module/my_subsys.h}
5063 #define TRACE_SYSTEM my_subsys
5065 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
5066 #define _LTTNG_MY_SUBSYS_H
5068 #include "../../../probes/lttng-tracepoint-event.h"
5069 #include <linux/tracepoint.h>
5071 LTTNG_TRACEPOINT_EVENT(
5073 * Format is identical to TRACE_EVENT()'s version for the three
5074 * following macro parameters:
5077 TP_PROTO(int my_int, const char *my_string),
5078 TP_ARGS(my_int, my_string),
5080 /* LTTng-modules specific macros */
5082 ctf_integer(int, my_int_field, my_int)
5083 ctf_string(my_bar_field, my_bar)
5087 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
5089 #include "../../../probes/define_trace.h"
5093 The entries in the `TP_FIELDS()` section are the list of fields for the
5094 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
5095 ftrace's `TRACE_EVENT()` macro.
5097 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
5098 complete description of the available `ctf_*()` macros.
5100 . Create the LTTng-modules probe's kernel module C source file,
5101 +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
5106 .path:{probes/lttng-probe-my-subsys.c}
5108 #include <linux/module.h>
5109 #include "../lttng-tracer.h"
5112 * Build-time verification of mismatch between mainline
5113 * TRACE_EVENT() arguments and the LTTng-modules adaptation
5114 * layer LTTNG_TRACEPOINT_EVENT() arguments.
5116 #include <trace/events/my_subsys.h>
5118 /* Create LTTng tracepoint probes */
5119 #define LTTNG_PACKAGE_BUILD
5120 #define CREATE_TRACE_POINTS
5121 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
5123 #include "../instrumentation/events/lttng-module/my_subsys.h"
5125 MODULE_LICENSE("GPL and additional rights");
5126 MODULE_AUTHOR("Your name <your-email>");
5127 MODULE_DESCRIPTION("LTTng my_subsys probes");
5128 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
5129 __stringify(LTTNG_MODULES_MINOR_VERSION) "."
5130 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
5131 LTTNG_MODULES_EXTRAVERSION);
5135 . Edit path:{probes/KBuild} and add your new kernel module object
5136 next to the existing ones:
5140 .path:{probes/KBuild}
5144 obj-m += lttng-probe-module.o
5145 obj-m += lttng-probe-power.o
5147 obj-m += lttng-probe-my-subsys.o
5153 . Build and install the LTTng kernel modules:
5158 $ make KERNELDIR=/path/to/linux
5159 # make modules_install && depmod -a
5163 Replace `/path/to/linux` with the path to the Linux source tree where
5164 you defined and used tracepoints with ftrace's `TRACE_EVENT()` macro.
5166 Note that you can also use the
5167 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
5168 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
5169 C code that need to be executed before the event fields are recorded.
5171 The best way to learn how to use the previous LTTng-modules macros is to
5172 inspect the existing LTTng-modules tracepoint definitions in the
5173 dir:{instrumentation/events/lttng-module} header files. Compare them
5174 with the Linux kernel mainline versions in the
5175 dir:{include/trace/events} directory of the Linux source tree.
5179 [[lttng-tracepoint-event-code]]
5180 ===== Use custom C code to access the data for tracepoint fields
5182 Although we recommended to always use the
5183 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
5184 the arguments and fields of an LTTng-modules tracepoint when possible,
5185 sometimes you need a more complex process to access the data that the
5186 tracer records as event record fields. In other words, you need local
5187 variables and multiple C{nbsp}statements instead of simple
5188 argument-based expressions that you pass to the
5189 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
5191 You can use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
5192 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
5193 a block of C{nbsp}code to be executed before LTTng records the fields.
5194 The structure of this macro is:
5197 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
5199 LTTNG_TRACEPOINT_EVENT_CODE(
5201 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5202 * version for the following three macro parameters:
5205 TP_PROTO(int my_int, const char *my_string),
5206 TP_ARGS(my_int, my_string),
5208 /* Declarations of custom local variables */
5211 unsigned long b = 0;
5212 const char *name = "(undefined)";
5213 struct my_struct *my_struct;
5217 * Custom code which uses both tracepoint arguments
5218 * (in TP_ARGS()) and local variables (in TP_locvar()).
5220 * Local variables are actually members of a structure pointed
5221 * to by the special variable tp_locvar.
5225 tp_locvar->a = my_int + 17;
5226 tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
5227 tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
5228 tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
5229 put_my_struct(tp_locvar->my_struct);
5238 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5239 * version for this, except that tp_locvar members can be
5240 * used in the argument expression parameters of
5241 * the ctf_*() macros.
5244 ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
5245 ctf_integer(int, my_struct_a, tp_locvar->a)
5246 ctf_string(my_string_field, my_string)
5247 ctf_string(my_struct_name, tp_locvar->name)
5252 IMPORTANT: The C code defined in `TP_code()` must not have any side
5253 effects when executed. In particular, the code must not allocate
5254 memory or get resources without deallocating this memory or putting
5255 those resources afterwards.
5258 [[instrumenting-linux-kernel-tracing]]
5259 ==== Load and unload a custom probe kernel module
5261 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
5262 kernel module>> in the kernel before it can emit LTTng events.
5264 To load the default probe kernel modules and a custom probe kernel
5267 * Use the opt:lttng-sessiond(8):--extra-kmod-probes option to give extra
5268 probe modules to load when starting a root <<lttng-sessiond,session
5272 .Load the `my_subsys`, `usb`, and the default probe modules.
5276 # lttng-sessiond --extra-kmod-probes=my_subsys,usb
5281 You only need to pass the subsystem name, not the whole kernel module
5284 To load _only_ a given custom probe kernel module:
5286 * Use the opt:lttng-sessiond(8):--kmod-probes option to give the probe
5287 modules to load when starting a root session daemon:
5290 .Load only the `my_subsys` and `usb` probe modules.
5294 # lttng-sessiond --kmod-probes=my_subsys,usb
5299 To confirm that a probe module is loaded:
5306 $ lsmod | grep lttng_probe_usb
5310 To unload the loaded probe modules:
5312 * Kill the session daemon with `SIGTERM`:
5317 # pkill lttng-sessiond
5321 You can also use man:modprobe(8)'s `--remove` option if the session
5322 daemon terminates abnormally.
5325 [[controlling-tracing]]
5328 Once an application or a Linux kernel is
5329 <<instrumenting,instrumented>> for LTTng tracing,
5332 This section is divided in topics on how to use the various
5333 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
5334 command-line tool>>, to _control_ the LTTng daemons and tracers.
5336 NOTE: In the following subsections, we refer to an man:lttng(1) command
5337 using its man page name. For example, instead of _Run the `create`
5338 command to..._, we use _Run the man:lttng-create(1) command to..._.
5342 === Start a session daemon
5344 In some situations, you need to run a <<lttng-sessiond,session daemon>>
5345 (man:lttng-sessiond(8)) _before_ you can use the man:lttng(1)
5348 You will see the following error when you run a command while no session
5352 Error: No session daemon is available
5355 The only command that automatically runs a session daemon is
5356 man:lttng-create(1), which you use to
5357 <<creating-destroying-tracing-sessions,create a tracing session>>. While
5358 this is most of the time the first operation that you do, sometimes it's
5359 not. Some examples are:
5361 * <<list-instrumentation-points,List the available instrumentation points>>.
5362 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
5364 [[tracing-group]] Each Unix user must have its own running session
5365 daemon to trace user applications. The session daemon that the root user
5366 starts is the only one allowed to control the LTTng kernel tracer. Users
5367 that are part of the _tracing group_ can control the root session
5368 daemon. The default tracing group name is `tracing`; you can set it to
5369 something else with the opt:lttng-sessiond(8):--group option when you
5370 start the root session daemon.
5372 To start a user session daemon:
5374 * Run man:lttng-sessiond(8):
5379 $ lttng-sessiond --daemonize
5383 To start the root session daemon:
5385 * Run man:lttng-sessiond(8) as the root user:
5390 # lttng-sessiond --daemonize
5394 In both cases, remove the opt:lttng-sessiond(8):--daemonize option to
5395 start the session daemon in foreground.
5397 To stop a session daemon, use man:kill(1) on its process ID (standard
5400 Note that some Linux distributions could manage the LTTng session daemon
5401 as a service. In this case, you should use the service manager to
5402 start, restart, and stop session daemons.
5405 [[creating-destroying-tracing-sessions]]
5406 === Create and destroy a tracing session
5408 Almost all the LTTng control operations happen in the scope of
5409 a <<tracing-session,tracing session>>, which is the dialogue between the
5410 <<lttng-sessiond,session daemon>> and you.
5412 To create a tracing session with a generated name:
5414 * Use the man:lttng-create(1) command:
5423 The created tracing session's name is `auto` followed by the
5426 To create a tracing session with a specific name:
5428 * Use the optional argument of the man:lttng-create(1) command:
5433 $ lttng create my-session
5437 Replace `my-session` with the specific tracing session name.
5439 LTTng appends the creation date to the created tracing session's name.
5441 LTTng writes the traces of a tracing session in
5442 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
5443 name of the tracing session. Note that the env:LTTNG_HOME environment
5444 variable defaults to `$HOME` if not set.
5446 To output LTTng traces to a non-default location:
5448 * Use the opt:lttng-create(1):--output option of the man:lttng-create(1) command:
5453 $ lttng create my-session --output=/tmp/some-directory
5457 You may create as many tracing sessions as you wish.
5459 To list all the existing tracing sessions for your Unix user:
5461 * Use the man:lttng-list(1) command:
5470 When you create a tracing session, it is set as the _current tracing
5471 session_. The following man:lttng(1) commands operate on the current
5472 tracing session when you don't specify one:
5474 [role="list-3-cols"]
5491 To change the current tracing session:
5493 * Use the man:lttng-set-session(1) command:
5498 $ lttng set-session new-session
5502 Replace `new-session` by the name of the new current tracing session.
5504 When you are done tracing in a given tracing session, you can destroy
5505 it. This operation frees the resources taken by the tracing session
5506 to destroy; it does not destroy the trace data that LTTng wrote for
5507 this tracing session.
5509 To destroy the current tracing session:
5511 * Use the man:lttng-destroy(1) command:
5521 [[list-instrumentation-points]]
5522 === List the available instrumentation points
5524 The <<lttng-sessiond,session daemon>> can query the running instrumented
5525 user applications and the Linux kernel to get a list of available
5526 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
5527 they are tracepoints and system calls. For the user space tracing
5528 domain, they are tracepoints. For the other tracing domains, they are
5531 To list the available instrumentation points:
5533 * Use the man:lttng-list(1) command with the requested tracing domain's
5537 * opt:lttng-list(1):--kernel: Linux kernel tracepoints (your Unix user
5538 must be a root user, or it must be a member of the
5539 <<tracing-group,tracing group>>).
5540 * opt:lttng-list(1):--kernel with opt:lttng-list(1):--syscall: Linux
5541 kernel system calls (your Unix user must be a root user, or it must be
5542 a member of the tracing group).
5543 * opt:lttng-list(1):--userspace: user space tracepoints.
5544 * opt:lttng-list(1):--jul: `java.util.logging` loggers.
5545 * opt:lttng-list(1):--log4j: Apache log4j loggers.
5546 * opt:lttng-list(1):--python: Python loggers.
5549 .List the available user space tracepoints.
5553 $ lttng list --userspace
5557 .List the available Linux kernel system call tracepoints.
5561 $ lttng list --kernel --syscall
5566 [[enabling-disabling-events]]
5567 === Create and enable an event rule
5569 Once you <<creating-destroying-tracing-sessions,create a tracing
5570 session>>, you can create <<event,event rules>> with the
5571 man:lttng-enable-event(1) command.
5573 You specify each condition with a command-line option. The available
5574 condition options are shown in the following table.
5576 [role="growable",cols="asciidoc,asciidoc,default"]
5577 .Condition command-line options for the man:lttng-enable-event(1) command.
5579 |Option |Description |Applicable tracing domains
5585 . +--probe=__ADDR__+
5586 . +--function=__ADDR__+
5589 Instead of using the default _tracepoint_ instrumentation type, use:
5591 . A Linux system call.
5592 . A Linux https://lwn.net/Articles/132196/[KProbe] (symbol or address).
5593 . The entry and return points of a Linux function (symbol or address).
5597 |First positional argument.
5600 Tracepoint or system call name. In the case of a Linux KProbe or
5601 function, this is a custom name given to the event rule. With the
5602 JUL, log4j, and Python domains, this is a logger name.
5604 With a tracepoint, logger, or system call name, the last character
5605 can be `*` to match anything that remains.
5612 . +--loglevel=__LEVEL__+
5613 . +--loglevel-only=__LEVEL__+
5616 . Match only tracepoints or log statements with a logging level at
5617 least as severe as +__LEVEL__+.
5618 . Match only tracepoints or log statements with a logging level
5619 equal to +__LEVEL__+.
5621 See man:lttng-enable-event(1) for the list of available logging level
5624 |User space, JUL, log4j, and Python.
5626 |+--exclude=__EXCLUSIONS__+
5629 When you use a `*` character at the end of the tracepoint or logger
5630 name (first positional argument), exclude the specific names in the
5631 comma-delimited list +__EXCLUSIONS__+.
5634 User space, JUL, log4j, and Python.
5636 |+--filter=__EXPR__+
5639 Match only events which satisfy the expression +__EXPR__+.
5641 See man:lttng-enable-event(1) to learn more about the syntax of a
5648 You attach an event rule to a <<channel,channel>> on creation. If you do
5649 not specify the channel with the opt:lttng-enable-event(1):--channel
5650 option, and if the event rule to create is the first in its
5651 <<domain,tracing domain>> for a given tracing session, then LTTng
5652 creates a _default channel_ for you. This default channel is reused in
5653 subsequent invocations of the man:lttng-enable-event(1) command for the
5654 same tracing domain.
5656 An event rule is always enabled at creation time.
5658 The following examples show how you can combine the previous
5659 command-line options to create simple to more complex event rules.
5661 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5665 $ lttng enable-event --kernel sched_switch
5669 .Create an event rule matching four Linux kernel system calls (default channel).
5673 $ lttng enable-event --kernel --syscall open,write,read,close
5677 .Create event rules matching tracepoints with filter expressions (default channel).
5681 $ lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
5686 $ lttng enable-event --kernel --all \
5687 --filter='$ctx.tid == 1988 || $ctx.tid == 1534'
5692 $ lttng enable-event --jul my_logger \
5693 --filter='$app.retriever:cur_msg_id > 3'
5696 IMPORTANT: Make sure to always quote the filter string when you
5697 use man:lttng(1) from a shell.
5700 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
5704 $ lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
5707 IMPORTANT: Make sure to always quote the wildcard character when you
5708 use man:lttng(1) from a shell.
5711 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
5715 $ lttng enable-event --python my-app.'*' \
5716 --exclude='my-app.module,my-app.hello'
5720 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
5724 $ lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
5728 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
5732 $ lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
5736 The event rules of a given channel form a whitelist: as soon as an
5737 emitted event passes one of them, LTTng can record the event. For
5738 example, an event named `my_app:my_tracepoint` emitted from a user space
5739 tracepoint with a `TRACE_ERROR` log level passes both of the following
5744 $ lttng enable-event --userspace my_app:my_tracepoint
5745 $ lttng enable-event --userspace my_app:my_tracepoint \
5746 --loglevel=TRACE_INFO
5749 The second event rule is redundant: the first one includes
5753 [[disable-event-rule]]
5754 === Disable an event rule
5756 To disable an event rule that you <<enabling-disabling-events,created>>
5757 previously, use the man:lttng-disable-event(1) command. This command
5758 disables _all_ the event rules (of a given tracing domain and channel)
5759 which match an instrumentation point. The other conditions are not
5760 supported as of LTTng{nbsp}{revision}.
5762 The LTTng tracer does not record an emitted event which passes
5763 a _disabled_ event rule.
5765 .Disable an event rule matching a Python logger (default channel).
5769 $ lttng disable-event --python my-logger
5773 .Disable an event rule matching all `java.util.logging` loggers (default channel).
5777 $ lttng disable-event --jul '*'
5781 .Disable _all_ the event rules of the default channel.
5783 The opt:lttng-disable-event(1):--all-events option is not, like the
5784 opt:lttng-enable-event(1):--all option of man:lttng-enable-event(1), the
5785 equivalent of the event name `*` (wildcard): it disables _all_ the event
5786 rules of a given channel.
5790 $ lttng disable-event --jul --all-events
5794 NOTE: You cannot delete an event rule once you create it.
5798 === Get the status of a tracing session
5800 To get the status of the current tracing session, that is, its
5801 parameters, its channels, event rules, and their attributes:
5803 * Use the man:lttng-status(1) command:
5813 To get the status of any tracing session:
5815 * Use the man:lttng-list(1) command with the tracing session's name:
5820 $ lttng list my-session
5824 Replace `my-session` with the desired tracing session's name.
5827 [[basic-tracing-session-control]]
5828 === Start and stop a tracing session
5830 Once you <<creating-destroying-tracing-sessions,create a tracing
5832 <<enabling-disabling-events,create one or more event rules>>,
5833 you can start and stop the tracers for this tracing session.
5835 To start tracing in the current tracing session:
5837 * Use the man:lttng-start(1) command:
5846 LTTng is very flexible: you can launch user applications before
5847 or after the you start the tracers. The tracers only record the events
5848 if they pass enabled event rules and if they occur while the tracers are
5851 To stop tracing in the current tracing session:
5853 * Use the man:lttng-stop(1) command:
5862 If there were <<channel-overwrite-mode-vs-discard-mode,lost event
5863 records>> or lost sub-buffers since the last time you ran
5864 man:lttng-start(1), warnings are printed when you run the
5865 man:lttng-stop(1) command.
5868 [[enabling-disabling-channels]]
5869 === Create a channel
5871 Once you create a tracing session, you can create a <<channel,channel>>
5872 with the man:lttng-enable-channel(1) command.
5874 Note that LTTng automatically creates a default channel when, for a
5875 given <<domain,tracing domain>>, no channels exist and you
5876 <<enabling-disabling-events,create>> the first event rule. This default
5877 channel is named `channel0` and its attributes are set to reasonable
5878 values. Therefore, you only need to create a channel when you need
5879 non-default attributes.
5881 You specify each non-default channel attribute with a command-line
5882 option when you use the man:lttng-enable-channel(1) command. The
5883 available command-line options are:
5885 [role="growable",cols="asciidoc,asciidoc"]
5886 .Command-line options for the man:lttng-enable-channel(1) command.
5888 |Option |Description
5894 <<channel-overwrite-mode-vs-discard-mode,event loss mode>> instead of
5895 the default _discard_ mode.
5897 |`--buffers-pid` (user space tracing domain only)
5900 Use the per-process <<channel-buffering-schemes,buffering scheme>>
5901 instead of the default per-user buffering scheme.
5903 |+--subbuf-size=__SIZE__+
5906 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
5907 either for each Unix user (default), or for each instrumented process.
5909 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5911 |+--num-subbuf=__COUNT__+
5914 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
5915 for each Unix user (default), or for each instrumented process.
5917 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5919 |+--tracefile-size=__SIZE__+
5922 Set the maximum size of each trace file that this channel writes within
5923 a stream to +__SIZE__+ bytes instead of no maximum.
5925 See <<tracefile-rotation,Trace file count and size>>.
5927 |+--tracefile-count=__COUNT__+
5930 Limit the number of trace files that this channel creates to
5931 +__COUNT__+ channels instead of no limit.
5933 See <<tracefile-rotation,Trace file count and size>>.
5935 |+--switch-timer=__PERIODUS__+
5938 Set the <<channel-switch-timer,switch timer period>>
5939 to +__PERIODUS__+{nbsp}µs.
5941 |+--read-timer=__PERIODUS__+
5944 Set the <<channel-read-timer,read timer period>>
5945 to +__PERIODUS__+{nbsp}µs.
5947 |[[opt-blocking-timeout]]+--blocking-timeout=__TIMEOUTUS__+
5950 Set the timeout of user space applications which load LTTng-UST
5951 in blocking mode to +__TIMEOUTUS__+:
5954 Never block (non-blocking mode).
5957 Block forever until space is available in a sub-buffer to record
5960 __n__, a positive value::
5961 Wait for at most __n__ µs when trying to write into a sub-buffer.
5963 Note that, for this option to have any effect on an instrumented
5964 user space application, you need to run the application with a set
5965 env:LTTNG_UST_ALLOW_BLOCKING environment variable.
5967 |+--output=__TYPE__+ (Linux kernel tracing domain only)
5970 Set the channel's output type to +__TYPE__+, either `mmap` or `splice`.
5974 You can only create a channel in the Linux kernel and user space
5975 <<domain,tracing domains>>: other tracing domains have their own channel
5976 created on the fly when <<enabling-disabling-events,creating event
5981 Because of a current LTTng limitation, you must create all channels
5982 _before_ you <<basic-tracing-session-control,start tracing>> in a given
5983 tracing session, that is, before the first time you run
5986 Since LTTng automatically creates a default channel when you use the
5987 man:lttng-enable-event(1) command with a specific tracing domain, you
5988 cannot, for example, create a Linux kernel event rule, start tracing,
5989 and then create a user space event rule, because no user space channel
5990 exists yet and it's too late to create one.
5992 For this reason, make sure to configure your channels properly
5993 before starting the tracers for the first time!
5996 The following examples show how you can combine the previous
5997 command-line options to create simple to more complex channels.
5999 .Create a Linux kernel channel with default attributes.
6003 $ lttng enable-channel --kernel my-channel
6007 .Create a user space channel with 4 sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
6011 $ lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
6012 --buffers-pid my-channel
6016 .[[blocking-timeout-example]]Create a default user space channel with an infinite blocking timeout.
6018 <<creating-destroying-tracing-sessions,Create a tracing-session>>,
6019 create the channel, <<enabling-disabling-events,create an event rule>>,
6020 and <<basic-tracing-session-control,start tracing>>:
6025 $ lttng enable-channel --userspace --blocking-timeout=-1 blocking-channel
6026 $ lttng enable-event --userspace --channel=blocking-channel --all
6030 Run an application instrumented with LTTng-UST and allow it to block:
6034 $ LTTNG_UST_ALLOW_BLOCKING=1 my-app
6038 .Create a Linux kernel channel which rotates 8 trace files of 4{nbsp}MiB each for each stream
6042 $ lttng enable-channel --kernel --tracefile-count=8 \
6043 --tracefile-size=4194304 my-channel
6047 .Create a user space channel in overwrite (or _flight recorder_) mode.
6051 $ lttng enable-channel --userspace --overwrite my-channel
6055 You can <<enabling-disabling-events,create>> the same event rule in
6056 two different channels:
6060 $ lttng enable-event --userspace --channel=my-channel app:tp
6061 $ lttng enable-event --userspace --channel=other-channel app:tp
6064 If both channels are enabled, when a tracepoint named `app:tp` is
6065 reached, LTTng records two events, one for each channel.
6069 === Disable a channel
6071 To disable a specific channel that you <<enabling-disabling-channels,created>>
6072 previously, use the man:lttng-disable-channel(1) command.
6074 .Disable a specific Linux kernel channel.
6078 $ lttng disable-channel --kernel my-channel
6082 The state of a channel precedes the individual states of event rules
6083 attached to it: event rules which belong to a disabled channel, even if
6084 they are enabled, are also considered disabled.
6088 === Add context fields to a channel
6090 Event record fields in trace files provide important information about
6091 events that occured previously, but sometimes some external context may
6092 help you solve a problem faster. Examples of context fields are:
6094 * The **process ID**, **thread ID**, **process name**, and
6095 **process priority** of the thread in which the event occurs.
6096 * The **hostname** of the system on which the event occurs.
6097 * The current values of many possible **performance counters** using
6099 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
6101 ** Branch instructions, misses, and loads.
6103 * Any context defined at the application level (supported for the
6104 JUL and log4j <<domain,tracing domains>>).
6106 To get the full list of available context fields, see
6107 `lttng add-context --list`. Some context fields are reserved for a
6108 specific <<domain,tracing domain>> (Linux kernel or user space).
6110 You add context fields to <<channel,channels>>. All the events
6111 that a channel with added context fields records contain those fields.
6113 To add context fields to one or all the channels of a given tracing
6116 * Use the man:lttng-add-context(1) command.
6118 .Add context fields to all the channels of the current tracing session.
6120 The following command line adds the virtual process identifier and
6121 the per-thread CPU cycles count fields to all the user space channels
6122 of the current tracing session.
6126 $ lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
6130 .Add performance counter context fields by raw ID
6132 See man:lttng-add-context(1) for the exact format of the context field
6133 type, which is partly compatible with the format used in
6138 $ lttng add-context --userspace --type=perf:thread:raw:r0110:test
6139 $ lttng add-context --kernel --type=perf:cpu:raw:r0013c:x86unhalted
6143 .Add a context field to a specific channel.
6145 The following command line adds the thread identifier context field
6146 to the Linux kernel channel named `my-channel` in the current
6151 $ lttng add-context --kernel --channel=my-channel --type=tid
6155 .Add an application-specific context field to a specific channel.
6157 The following command line adds the `cur_msg_id` context field of the
6158 `retriever` context retriever for all the instrumented
6159 <<java-application,Java applications>> recording <<event,event records>>
6160 in the channel named `my-channel`:
6164 $ lttng add-context --kernel --channel=my-channel \
6165 --type='$app:retriever:cur_msg_id'
6168 IMPORTANT: Make sure to always quote the `$` character when you
6169 use man:lttng-add-context(1) from a shell.
6172 NOTE: You cannot remove context fields from a channel once you add it.
6177 === Track process IDs
6179 It's often useful to allow only specific process IDs (PIDs) to emit
6180 events. For example, you may wish to record all the system calls made by
6181 a given process (Ă la http://linux.die.net/man/1/strace[strace]).
6183 The man:lttng-track(1) and man:lttng-untrack(1) commands serve this
6184 purpose. Both commands operate on a whitelist of process IDs. You _add_
6185 entries to this whitelist with the man:lttng-track(1) command and remove
6186 entries with the man:lttng-untrack(1) command. Any process which has one
6187 of the PIDs in the whitelist is allowed to emit LTTng events which pass
6188 an enabled <<event,event rule>>.
6190 NOTE: The PID tracker tracks the _numeric process IDs_. Should a
6191 process with a given tracked ID exit and another process be given this
6192 ID, then the latter would also be allowed to emit events.
6194 .Track and untrack process IDs.
6196 For the sake of the following example, assume the target system has 16
6200 <<creating-destroying-tracing-sessions,create a tracing session>>,
6201 the whitelist contains all the possible PIDs:
6204 .All PIDs are tracked.
6205 image::track-all.png[]
6207 When the whitelist is full and you use the man:lttng-track(1) command to
6208 specify some PIDs to track, LTTng first clears the whitelist, then it
6209 tracks the specific PIDs. After:
6213 $ lttng track --pid=3,4,7,10,13
6219 .PIDs 3, 4, 7, 10, and 13 are tracked.
6220 image::track-3-4-7-10-13.png[]
6222 You can add more PIDs to the whitelist afterwards:
6226 $ lttng track --pid=1,15,16
6232 .PIDs 1, 15, and 16 are added to the whitelist.
6233 image::track-1-3-4-7-10-13-15-16.png[]
6235 The man:lttng-untrack(1) command removes entries from the PID tracker's
6236 whitelist. Given the previous example, the following command:
6240 $ lttng untrack --pid=3,7,10,13
6243 leads to this whitelist:
6246 .PIDs 3, 7, 10, and 13 are removed from the whitelist.
6247 image::track-1-4-15-16.png[]
6249 LTTng can track all possible PIDs again using the opt:track(1):--all
6254 $ lttng track --pid --all
6257 The result is, again:
6260 .All PIDs are tracked.
6261 image::track-all.png[]
6264 .Track only specific PIDs
6266 A very typical use case with PID tracking is to start with an empty
6267 whitelist, then <<basic-tracing-session-control,start the tracers>>, and
6268 then add PIDs manually while tracers are active. You can accomplish this
6269 by using the opt:lttng-untrack(1):--all option of the
6270 man:lttng-untrack(1) command to clear the whitelist after you
6271 <<creating-destroying-tracing-sessions,create a tracing session>>:
6275 $ lttng untrack --pid --all
6281 .No PIDs are tracked.
6282 image::untrack-all.png[]
6284 If you trace with this whitelist configuration, the tracer records no
6285 events for this <<domain,tracing domain>> because no processes are
6286 tracked. You can use the man:lttng-track(1) command as usual to track
6287 specific PIDs, for example:
6291 $ lttng track --pid=6,11
6297 .PIDs 6 and 11 are tracked.
6298 image::track-6-11.png[]
6303 [[saving-loading-tracing-session]]
6304 === Save and load tracing session configurations
6306 Configuring a <<tracing-session,tracing session>> can be long. Some of
6307 the tasks involved are:
6309 * <<enabling-disabling-channels,Create channels>> with
6310 specific attributes.
6311 * <<adding-context,Add context fields>> to specific channels.
6312 * <<enabling-disabling-events,Create event rules>> with specific log
6313 level and filter conditions.
6315 If you use LTTng to solve real world problems, chances are you have to
6316 record events using the same tracing session setup over and over,
6317 modifying a few variables each time in your instrumented program
6318 or environment. To avoid constant tracing session reconfiguration,
6319 the man:lttng(1) command-line tool can save and load tracing session
6320 configurations to/from XML files.
6322 To save a given tracing session configuration:
6324 * Use the man:lttng-save(1) command:
6329 $ lttng save my-session
6333 Replace `my-session` with the name of the tracing session to save.
6335 LTTng saves tracing session configurations to
6336 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
6337 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
6338 the opt:lttng-save(1):--output-path option to change this destination
6341 LTTng saves all configuration parameters, for example:
6343 * The tracing session name.
6344 * The trace data output path.
6345 * The channels with their state and all their attributes.
6346 * The context fields you added to channels.
6347 * The event rules with their state, log level and filter conditions.
6349 To load a tracing session:
6351 * Use the man:lttng-load(1) command:
6356 $ lttng load my-session
6360 Replace `my-session` with the name of the tracing session to load.
6362 When LTTng loads a configuration, it restores your saved tracing session
6363 as if you just configured it manually.
6365 See man:lttng(1) for the complete list of command-line options. You
6366 can also save and load all many sessions at a time, and decide in which
6367 directory to output the XML files.
6370 [[sending-trace-data-over-the-network]]
6371 === Send trace data over the network
6373 LTTng can send the recorded trace data to a remote system over the
6374 network instead of writing it to the local file system.
6376 To send the trace data over the network:
6378 . On the _remote_ system (which can also be the target system),
6379 start an LTTng <<lttng-relayd,relay daemon>> (man:lttng-relayd(8)):
6388 . On the _target_ system, create a tracing session configured to
6389 send trace data over the network:
6394 $ lttng create my-session --set-url=net://remote-system
6398 Replace `remote-system` by the host name or IP address of the
6399 remote system. See man:lttng-create(1) for the exact URL format.
6401 . On the target system, use the man:lttng(1) command-line tool as usual.
6402 When tracing is active, the target's consumer daemon sends sub-buffers
6403 to the relay daemon running on the remote system instead of flushing
6404 them to the local file system. The relay daemon writes the received
6405 packets to the local file system.
6407 The relay daemon writes trace files to
6408 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
6409 +__hostname__+ is the host name of the target system and +__session__+
6410 is the tracing session name. Note that the env:LTTNG_HOME environment
6411 variable defaults to `$HOME` if not set. Use the
6412 opt:lttng-relayd(8):--output option of man:lttng-relayd(8) to write
6413 trace files to another base directory.
6418 === View events as LTTng emits them (noch:{LTTng} live)
6420 LTTng live is a network protocol implemented by the <<lttng-relayd,relay
6421 daemon>> (man:lttng-relayd(8)) to allow compatible trace viewers to
6422 display events as LTTng emits them on the target system while tracing is
6425 The relay daemon creates a _tee_: it forwards the trace data to both
6426 the local file system and to connected live viewers:
6429 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
6434 . On the _target system_, create a <<tracing-session,tracing session>>
6440 $ lttng create my-session --live
6444 This spawns a local relay daemon.
6446 . Start the live viewer and configure it to connect to the relay
6447 daemon. For example, with http://diamon.org/babeltrace[Babeltrace]:
6452 $ babeltrace --input-format=lttng-live \
6453 net://localhost/host/hostname/my-session
6460 * `hostname` with the host name of the target system.
6461 * `my-session` with the name of the tracing session to view.
6464 . Configure the tracing session as usual with the man:lttng(1)
6465 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6467 You can list the available live tracing sessions with Babeltrace:
6471 $ babeltrace --input-format=lttng-live net://localhost
6474 You can start the relay daemon on another system. In this case, you need
6475 to specify the relay daemon's URL when you create the tracing session
6476 with the opt:lttng-create(1):--set-url option. You also need to replace
6477 `localhost` in the procedure above with the host name of the system on
6478 which the relay daemon is running.
6480 See man:lttng-create(1) and man:lttng-relayd(8) for the complete list of
6481 command-line options.
6485 [[taking-a-snapshot]]
6486 === Take a snapshot of the current sub-buffers of a tracing session
6488 The normal behavior of LTTng is to append full sub-buffers to growing
6489 trace data files. This is ideal to keep a full history of the events
6490 that occurred on the target system, but it can
6491 represent too much data in some situations. For example, you may wish
6492 to trace your application continuously until some critical situation
6493 happens, in which case you only need the latest few recorded
6494 events to perform the desired analysis, not multi-gigabyte trace files.
6496 With the man:lttng-snapshot(1) command, you can take a snapshot of the
6497 current sub-buffers of a given <<tracing-session,tracing session>>.
6498 LTTng can write the snapshot to the local file system or send it over
6503 . Create a tracing session in _snapshot mode_:
6508 $ lttng create my-session --snapshot
6512 The <<channel-overwrite-mode-vs-discard-mode,event loss mode>> of
6513 <<channel,channels>> created in this mode is automatically set to
6514 _overwrite_ (flight recorder mode).
6516 . Configure the tracing session as usual with the man:lttng(1)
6517 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6519 . **Optional**: When you need to take a snapshot,
6520 <<basic-tracing-session-control,stop tracing>>.
6522 You can take a snapshot when the tracers are active, but if you stop
6523 them first, you are sure that the data in the sub-buffers does not
6524 change before you actually take the snapshot.
6531 $ lttng snapshot record --name=my-first-snapshot
6535 LTTng writes the current sub-buffers of all the current tracing
6536 session's channels to trace files on the local file system. Those trace
6537 files have `my-first-snapshot` in their name.
6539 There is no difference between the format of a normal trace file and the
6540 format of a snapshot: viewers of LTTng traces also support LTTng
6543 By default, LTTng writes snapshot files to the path shown by
6544 `lttng snapshot list-output`. You can change this path or decide to send
6545 snapshots over the network using either:
6547 . An output path or URL that you specify when you create the
6549 . An snapshot output path or URL that you add using
6550 `lttng snapshot add-output`
6551 . An output path or URL that you provide directly to the
6552 `lttng snapshot record` command.
6554 Method 3 overrides method 2, which overrides method 1. When you
6555 specify a URL, a relay daemon must listen on a remote system (see
6556 <<sending-trace-data-over-the-network,Send trace data over the network>>).
6561 === Use the machine interface
6563 With any command of the man:lttng(1) command-line tool, you can set the
6564 opt:lttng(1):--mi option to `xml` (before the command name) to get an
6565 XML machine interface output, for example:
6569 $ lttng --mi=xml enable-event --kernel --syscall open
6572 A schema definition (XSD) is
6573 https://github.com/lttng/lttng-tools/blob/stable-2.10/src/common/mi-lttng-3.0.xsd[available]
6574 to ease the integration with external tools as much as possible.
6578 [[metadata-regenerate]]
6579 === Regenerate the metadata of an LTTng trace
6581 An LTTng trace, which is a http://diamon.org/ctf[CTF] trace, has both
6582 data stream files and a metadata file. This metadata file contains,
6583 amongst other things, information about the offset of the clock sources
6584 used to timestamp <<event,event records>> when tracing.
6586 If, once a <<tracing-session,tracing session>> is
6587 <<basic-tracing-session-control,started>>, a major
6588 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
6589 happens, the trace's clock offset also needs to be updated. You
6590 can use the `metadata` item of the man:lttng-regenerate(1) command
6593 The main use case of this command is to allow a system to boot with
6594 an incorrect wall time and trace it with LTTng before its wall time
6595 is corrected. Once the system is known to be in a state where its
6596 wall time is correct, it can run `lttng regenerate metadata`.
6598 To regenerate the metadata of an LTTng trace:
6600 * Use the `metadata` item of the man:lttng-regenerate(1) command:
6605 $ lttng regenerate metadata
6611 `lttng regenerate metadata` has the following limitations:
6613 * Tracing session <<creating-destroying-tracing-sessions,created>>
6615 * User space <<channel,channels>>, if any, are using
6616 <<channel-buffering-schemes,per-user buffering>>.
6621 [[regenerate-statedump]]
6622 === Regenerate the state dump of a tracing session
6624 The LTTng kernel and user space tracers generate state dump
6625 <<event,event records>> when the application starts or when you
6626 <<basic-tracing-session-control,start a tracing session>>. An analysis
6627 can use the state dump event records to set an initial state before it
6628 builds the rest of the state from the following event records.
6629 http://tracecompass.org/[Trace Compass] is a notable example of an
6630 application which uses the state dump of an LTTng trace.
6632 When you <<taking-a-snapshot,take a snapshot>>, it's possible that the
6633 state dump event records are not included in the snapshot because they
6634 were recorded to a sub-buffer that has been consumed or overwritten
6637 You can use the `lttng regenerate statedump` command to emit the state
6638 dump event records again.
6640 To regenerate the state dump of the current tracing session, provided
6641 create it in snapshot mode, before you take a snapshot:
6643 . Use the `statedump` item of the man:lttng-regenerate(1) command:
6648 $ lttng regenerate statedump
6652 . <<basic-tracing-session-control,Stop the tracing session>>:
6661 . <<taking-a-snapshot,Take a snapshot>>:
6666 $ lttng snapshot record --name=my-snapshot
6670 Depending on the event throughput, you should run steps 1 and 2
6671 as closely as possible.
6673 NOTE: To record the state dump events, you need to
6674 <<enabling-disabling-events,create event rules>> which enable them.
6675 LTTng-UST state dump tracepoints start with `lttng_ust_statedump:`.
6676 LTTng-modules state dump tracepoints start with `lttng_statedump_`.
6680 [[persistent-memory-file-systems]]
6681 === Record trace data on persistent memory file systems
6683 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
6684 (NVRAM) is random-access memory that retains its information when power
6685 is turned off (non-volatile). Systems with such memory can store data
6686 structures in RAM and retrieve them after a reboot, without flushing
6687 to typical _storage_.
6689 Linux supports NVRAM file systems thanks to either
6690 http://pramfs.sourceforge.net/[PRAMFS] or
6691 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
6692 (requires Linux 4.1+).
6694 This section does not describe how to operate such file systems;
6695 we assume that you have a working persistent memory file system.
6697 When you create a <<tracing-session,tracing session>>, you can specify
6698 the path of the shared memory holding the sub-buffers. If you specify a
6699 location on an NVRAM file system, then you can retrieve the latest
6700 recorded trace data when the system reboots after a crash.
6702 To record trace data on a persistent memory file system and retrieve the
6703 trace data after a system crash:
6705 . Create a tracing session with a sub-buffer shared memory path located
6706 on an NVRAM file system:
6711 $ lttng create my-session --shm-path=/path/to/shm
6715 . Configure the tracing session as usual with the man:lttng(1)
6716 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6718 . After a system crash, use the man:lttng-crash(1) command-line tool to
6719 view the trace data recorded on the NVRAM file system:
6724 $ lttng-crash /path/to/shm
6728 The binary layout of the ring buffer files is not exactly the same as
6729 the trace files layout. This is why you need to use man:lttng-crash(1)
6730 instead of your preferred trace viewer directly.
6732 To convert the ring buffer files to LTTng trace files:
6734 * Use the opt:lttng-crash(1):--extract option of man:lttng-crash(1):
6739 $ lttng-crash --extract=/path/to/trace /path/to/shm
6745 [[notif-trigger-api]]
6746 === Get notified when a channel's buffer usage is too high or too low
6748 With LTTng's $$C/C++$$ notification and trigger API, your user
6749 application can get notified when the buffer usage of one or more
6750 <<channel,channels>> becomes too low or too high. You can use this API
6751 and enable or disable <<event,event rules>> during tracing to avoid
6752 <<channel-overwrite-mode-vs-discard-mode,discarded event records>>.
6754 .Have a user application get notified when an LTTng channel's buffer usage is too high.
6756 In this example, we create and build an application which gets notified
6757 when the buffer usage of a specific LTTng channel is higher than
6758 75{nbsp}%. We only print that it is the case in the example, but we
6759 could as well use the API of <<liblttng-ctl-lttng,`liblttng-ctl`>> to
6760 disable event rules when this happens.
6762 . Create the application's C source file:
6770 #include <lttng/domain.h>
6771 #include <lttng/action/action.h>
6772 #include <lttng/action/notify.h>
6773 #include <lttng/condition/condition.h>
6774 #include <lttng/condition/buffer-usage.h>
6775 #include <lttng/condition/evaluation.h>
6776 #include <lttng/notification/channel.h>
6777 #include <lttng/notification/notification.h>
6778 #include <lttng/trigger/trigger.h>
6779 #include <lttng/endpoint.h>
6781 int main(int argc, char *argv[])
6783 int exit_status = 0;
6784 struct lttng_notification_channel *notification_channel;
6785 struct lttng_condition *condition;
6786 struct lttng_action *action;
6787 struct lttng_trigger *trigger;
6788 const char *tracing_session_name;
6789 const char *channel_name;
6792 tracing_session_name = argv[1];
6793 channel_name = argv[2];
6796 * Create a notification channel. A notification channel
6797 * connects the user application to the LTTng session daemon.
6798 * This notification channel can be used to listen to various
6799 * types of notifications.
6801 notification_channel = lttng_notification_channel_create(
6802 lttng_session_daemon_notification_endpoint);
6805 * Create a "high buffer usage" condition. In this case, the
6806 * condition is reached when the buffer usage is greater than or
6807 * equal to 75 %. We create the condition for a specific tracing
6808 * session name, channel name, and for the user space tracing
6811 * The "low buffer usage" condition type also exists.
6813 condition = lttng_condition_buffer_usage_high_create();
6814 lttng_condition_buffer_usage_set_threshold_ratio(condition, .75);
6815 lttng_condition_buffer_usage_set_session_name(
6816 condition, tracing_session_name);
6817 lttng_condition_buffer_usage_set_channel_name(condition,
6819 lttng_condition_buffer_usage_set_domain_type(condition,
6823 * Create an action (get a notification) to take when the
6824 * condition created above is reached.
6826 action = lttng_action_notify_create();
6829 * Create a trigger. A trigger associates a condition to an
6830 * action: the action is executed when the condition is reached.
6832 trigger = lttng_trigger_create(condition, action);
6834 /* Register the trigger to LTTng. */
6835 lttng_register_trigger(trigger);
6838 * Now that we have registered a trigger, a notification will be
6839 * emitted everytime its condition is met. To receive this
6840 * notification, we must subscribe to notifications that match
6841 * the same condition.
6843 lttng_notification_channel_subscribe(notification_channel,
6847 * Notification loop. You can put this in a dedicated thread to
6848 * avoid blocking the main thread.
6851 struct lttng_notification *notification;
6852 enum lttng_notification_channel_status status;
6853 const struct lttng_evaluation *notification_evaluation;
6854 const struct lttng_condition *notification_condition;
6855 double buffer_usage;
6857 /* Receive the next notification. */
6858 status = lttng_notification_channel_get_next_notification(
6859 notification_channel, ¬ification);
6862 case LTTNG_NOTIFICATION_CHANNEL_STATUS_OK:
6864 case LTTNG_NOTIFICATION_CHANNEL_STATUS_NOTIFICATIONS_DROPPED:
6866 * The session daemon can drop notifications if
6867 * a monitoring application is not consuming the
6868 * notifications fast enough.
6871 case LTTNG_NOTIFICATION_CHANNEL_STATUS_CLOSED:
6873 * The notification channel has been closed by the
6874 * session daemon. This is typically caused by a session
6875 * daemon shutting down.
6879 /* Unhandled conditions or errors. */
6885 * A notification provides, amongst other things:
6887 * * The condition that caused this notification to be
6889 * * The condition evaluation, which provides more
6890 * specific information on the evaluation of the
6893 * The condition evaluation provides the buffer usage
6894 * value at the moment the condition was reached.
6896 notification_condition = lttng_notification_get_condition(
6898 notification_evaluation = lttng_notification_get_evaluation(
6901 /* We're subscribed to only one condition. */
6902 assert(lttng_condition_get_type(notification_condition) ==
6903 LTTNG_CONDITION_TYPE_BUFFER_USAGE_HIGH);
6906 * Get the exact sampled buffer usage from the
6907 * condition evaluation.
6909 lttng_evaluation_buffer_usage_get_usage_ratio(
6910 notification_evaluation, &buffer_usage);
6913 * At this point, instead of printing a message, we
6914 * could do something to reduce the channel's buffer
6915 * usage, like disable specific events.
6917 printf("Buffer usage is %f %% in tracing session \"%s\", "
6918 "user space channel \"%s\".\n", buffer_usage * 100,
6919 tracing_session_name, channel_name);
6920 lttng_notification_destroy(notification);
6924 lttng_action_destroy(action);
6925 lttng_condition_destroy(condition);
6926 lttng_trigger_destroy(trigger);
6927 lttng_notification_channel_destroy(notification_channel);
6933 . Build the `notif-app` application, linking it to `liblttng-ctl`:
6938 $ gcc -o notif-app notif-app.c -llttng-ctl
6942 . <<creating-destroying-tracing-sessions,Create a tracing session>>,
6943 <<enabling-disabling-events,create an event rule>> matching all the
6944 user space tracepoints, and
6945 <<basic-tracing-session-control,start tracing>>:
6950 $ lttng create my-session
6951 $ lttng enable-event --userspace --all
6956 If you create the channel manually with the man:lttng-enable-channel(1)
6957 command, you can control how frequently are the current values of the
6958 channel's properties sampled to evaluate user conditions with the
6959 opt:lttng-enable-channel(1):--monitor-timer option.
6961 . Run the `notif-app` application. This program accepts the
6962 <<tracing-session,tracing session>> name and the user space channel
6963 name as its two first arguments. The channel which LTTng automatically
6964 creates with the man:lttng-enable-event(1) command above is named
6970 $ ./notif-app my-session channel0
6974 . In another terminal, run an application with a very high event
6975 throughput so that the 75{nbsp}% buffer usage condition is reached.
6977 In the first terminal, the application should print lines like this:
6980 Buffer usage is 81.45197 % in tracing session "my-session", user space
6984 If you don't see anything, try modifying the condition in
6985 path:{notif-app.c} to a lower value (0.1, for example), rebuilding it
6986 (step 2) and running it again (step 4).
6993 [[lttng-modules-ref]]
6994 === noch:{LTTng-modules}
6998 [[lttng-tracepoint-enum]]
6999 ==== `LTTNG_TRACEPOINT_ENUM()` usage
7001 Use the `LTTNG_TRACEPOINT_ENUM()` macro to define an enumeration:
7005 LTTNG_TRACEPOINT_ENUM(name, TP_ENUM_VALUES(entries))
7010 * `name` with the name of the enumeration (C identifier, unique
7011 amongst all the defined enumerations).
7012 * `entries` with a list of enumeration entries.
7014 The available enumeration entry macros are:
7016 +ctf_enum_value(__name__, __value__)+::
7017 Entry named +__name__+ mapped to the integral value +__value__+.
7019 +ctf_enum_range(__name__, __begin__, __end__)+::
7020 Entry named +__name__+ mapped to the range of integral values between
7021 +__begin__+ (included) and +__end__+ (included).
7023 +ctf_enum_auto(__name__)+::
7024 Entry named +__name__+ mapped to the integral value following the
7025 last mapping's value.
7027 The last value of a `ctf_enum_value()` entry is its +__value__+
7030 The last value of a `ctf_enum_range()` entry is its +__end__+ parameter.
7032 If `ctf_enum_auto()` is the first entry in the list, its integral
7035 Use the `ctf_enum()` <<lttng-modules-tp-fields,field definition macro>>
7036 to use a defined enumeration as a tracepoint field.
7038 .Define an enumeration with `LTTNG_TRACEPOINT_ENUM()`.
7042 LTTNG_TRACEPOINT_ENUM(
7045 ctf_enum_auto("AUTO: EXPECT 0")
7046 ctf_enum_value("VALUE: 23", 23)
7047 ctf_enum_value("VALUE: 27", 27)
7048 ctf_enum_auto("AUTO: EXPECT 28")
7049 ctf_enum_range("RANGE: 101 TO 303", 101, 303)
7050 ctf_enum_auto("AUTO: EXPECT 304")
7058 [[lttng-modules-tp-fields]]
7059 ==== Tracepoint fields macros (for `TP_FIELDS()`)
7061 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
7062 tracepoint fields, which must be listed within `TP_FIELDS()` in
7063 `LTTNG_TRACEPOINT_EVENT()`, are:
7065 [role="func-desc growable",cols="asciidoc,asciidoc"]
7066 .Available macros to define LTTng-modules tracepoint fields
7068 |Macro |Description and parameters
7071 +ctf_integer(__t__, __n__, __e__)+
7073 +ctf_integer_nowrite(__t__, __n__, __e__)+
7075 +ctf_user_integer(__t__, __n__, __e__)+
7077 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
7079 Standard integer, displayed in base 10.
7082 Integer C type (`int`, `long`, `size_t`, ...).
7088 Argument expression.
7091 +ctf_integer_hex(__t__, __n__, __e__)+
7093 +ctf_user_integer_hex(__t__, __n__, __e__)+
7095 Standard integer, displayed in base 16.
7104 Argument expression.
7106 |+ctf_integer_oct(__t__, __n__, __e__)+
7108 Standard integer, displayed in base 8.
7117 Argument expression.
7120 +ctf_integer_network(__t__, __n__, __e__)+
7122 +ctf_user_integer_network(__t__, __n__, __e__)+
7124 Integer in network byte order (big-endian), displayed in base 10.
7133 Argument expression.
7136 +ctf_integer_network_hex(__t__, __n__, __e__)+
7138 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
7140 Integer in network byte order, displayed in base 16.
7149 Argument expression.
7152 +ctf_enum(__N__, __t__, __n__, __e__)+
7154 +ctf_enum_nowrite(__N__, __t__, __n__, __e__)+
7156 +ctf_user_enum(__N__, __t__, __n__, __e__)+
7158 +ctf_user_enum_nowrite(__N__, __t__, __n__, __e__)+
7163 Name of a <<lttng-tracepoint-enum,previously defined enumeration>>.
7166 Integer C type (`int`, `long`, `size_t`, ...).
7172 Argument expression.
7175 +ctf_string(__n__, __e__)+
7177 +ctf_string_nowrite(__n__, __e__)+
7179 +ctf_user_string(__n__, __e__)+
7181 +ctf_user_string_nowrite(__n__, __e__)+
7183 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
7189 Argument expression.
7192 +ctf_array(__t__, __n__, __e__, __s__)+
7194 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
7196 +ctf_user_array(__t__, __n__, __e__, __s__)+
7198 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
7200 Statically-sized array of integers.
7203 Array element C type.
7209 Argument expression.
7215 +ctf_array_bitfield(__t__, __n__, __e__, __s__)+
7217 +ctf_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7219 +ctf_user_array_bitfield(__t__, __n__, __e__, __s__)+
7221 +ctf_user_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7223 Statically-sized array of bits.
7225 The type of +__e__+ must be an integer type. +__s__+ is the number
7226 of elements of such type in +__e__+, not the number of bits.
7229 Array element C type.
7235 Argument expression.
7241 +ctf_array_text(__t__, __n__, __e__, __s__)+
7243 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
7245 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
7247 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
7249 Statically-sized array, printed as text.
7251 The string does not need to be null-terminated.
7254 Array element C type (always `char`).
7260 Argument expression.
7266 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
7268 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7270 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
7272 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7274 Dynamically-sized array of integers.
7276 The type of +__E__+ must be unsigned.
7279 Array element C type.
7285 Argument expression.
7288 Length expression C type.
7294 +ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7296 +ctf_user_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7298 Dynamically-sized array of integers, displayed in base 16.
7300 The type of +__E__+ must be unsigned.
7303 Array element C type.
7309 Argument expression.
7312 Length expression C type.
7317 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
7319 Dynamically-sized array of integers in network byte order (big-endian),
7320 displayed in base 10.
7322 The type of +__E__+ must be unsigned.
7325 Array element C type.
7331 Argument expression.
7334 Length expression C type.
7340 +ctf_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7342 +ctf_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7344 +ctf_user_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7346 +ctf_user_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7348 Dynamically-sized array of bits.
7350 The type of +__e__+ must be an integer type. +__s__+ is the number
7351 of elements of such type in +__e__+, not the number of bits.
7353 The type of +__E__+ must be unsigned.
7356 Array element C type.
7362 Argument expression.
7365 Length expression C type.
7371 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7373 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7375 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7377 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7379 Dynamically-sized array, displayed as text.
7381 The string does not need to be null-terminated.
7383 The type of +__E__+ must be unsigned.
7385 The behaviour is undefined if +__e__+ is `NULL`.
7388 Sequence element C type (always `char`).
7394 Argument expression.
7397 Length expression C type.
7403 Use the `_user` versions when the argument expression, `e`, is
7404 a user space address. In the cases of `ctf_user_integer*()` and
7405 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
7408 The `_nowrite` versions omit themselves from the session trace, but are
7409 otherwise identical. This means the `_nowrite` fields won't be written
7410 in the recorded trace. Their primary purpose is to make some
7411 of the event context available to the
7412 <<enabling-disabling-events,event filters>> without having to
7413 commit the data to sub-buffers.
7419 Terms related to LTTng and to tracing in general:
7422 The http://diamon.org/babeltrace[Babeltrace] project, which includes
7423 the cmd:babeltrace command, some libraries, and Python bindings.
7425 <<channel-buffering-schemes,buffering scheme>>::
7426 A layout of sub-buffers applied to a given channel.
7428 <<channel,channel>>::
7429 An entity which is responsible for a set of ring buffers.
7431 <<event,Event rules>> are always attached to a specific channel.
7434 A reference of time for a tracer.
7436 <<lttng-consumerd,consumer daemon>>::
7437 A process which is responsible for consuming the full sub-buffers
7438 and write them to a file system or send them over the network.
7440 <<channel-overwrite-mode-vs-discard-mode,discard mode>>:: The event loss
7441 mode in which the tracer _discards_ new event records when there's no
7442 sub-buffer space left to store them.
7445 The consequence of the execution of an instrumentation
7446 point, like a tracepoint that you manually place in some source code,
7447 or a Linux kernel KProbe.
7449 An event is said to _occur_ at a specific time. Different actions can
7450 be taken upon the occurrence of an event, like record the event's payload
7453 <<channel-overwrite-mode-vs-discard-mode,event loss mode>>::
7454 The mechanism by which event records of a given channel are lost
7455 (not recorded) when there is no sub-buffer space left to store them.
7457 [[def-event-name]]event name::
7458 The name of an event, which is also the name of the event record.
7459 This is also called the _instrumentation point name_.
7462 A record, in a trace, of the payload of an event which occured.
7464 <<event,event rule>>::
7465 Set of conditions which must be satisfied for one or more occuring
7466 events to be recorded.
7468 `java.util.logging`::
7470 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities].
7472 <<instrumenting,instrumentation>>::
7473 The use of LTTng probes to make a piece of software traceable.
7475 instrumentation point::
7476 A point in the execution path of a piece of software that, when
7477 reached by this execution, can emit an event.
7479 instrumentation point name::
7480 See _<<def-event-name,event name>>_.
7483 A http://logging.apache.org/log4j/1.2/[logging library] for Java
7484 developed by the Apache Software Foundation.
7487 Level of severity of a log statement or user space
7488 instrumentation point.
7491 The _Linux Trace Toolkit: next generation_ project.
7493 <<lttng-cli,cmd:lttng>>::
7494 A command-line tool provided by the LTTng-tools project which you
7495 can use to send and receive control messages to and from a
7499 The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
7500 which is a set of analyzing programs that are used to obtain a
7501 higher level view of an LTTng trace.
7503 cmd:lttng-consumerd::
7504 The name of the consumer daemon program.
7507 A utility provided by the LTTng-tools project which can convert
7508 ring buffer files (usually
7509 <<persistent-memory-file-systems,saved on a persistent memory file system>>)
7512 LTTng Documentation::
7515 <<lttng-live,LTTng live>>::
7516 A communication protocol between the relay daemon and live viewers
7517 which makes it possible to see events "live", as they are received by
7520 <<lttng-modules,LTTng-modules>>::
7521 The https://github.com/lttng/lttng-modules[LTTng-modules] project,
7522 which contains the Linux kernel modules to make the Linux kernel
7523 instrumentation points available for LTTng tracing.
7526 The name of the relay daemon program.
7528 cmd:lttng-sessiond::
7529 The name of the session daemon program.
7532 The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
7533 contains the various programs and libraries used to
7534 <<controlling-tracing,control tracing>>.
7536 <<lttng-ust,LTTng-UST>>::
7537 The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
7538 contains libraries to instrument user applications.
7540 <<lttng-ust-agents,LTTng-UST Java agent>>::
7541 A Java package provided by the LTTng-UST project to allow the
7542 LTTng instrumentation of `java.util.logging` and Apache log4j 1.2
7545 <<lttng-ust-agents,LTTng-UST Python agent>>::
7546 A Python package provided by the LTTng-UST project to allow the
7547 LTTng instrumentation of Python logging statements.
7549 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
7550 The event loss mode in which new event records overwrite older
7551 event records when there's no sub-buffer space left to store them.
7553 <<channel-buffering-schemes,per-process buffering>>::
7554 A buffering scheme in which each instrumented process has its own
7555 sub-buffers for a given user space channel.
7557 <<channel-buffering-schemes,per-user buffering>>::
7558 A buffering scheme in which all the processes of a Unix user share the
7559 same sub-buffer for a given user space channel.
7561 <<lttng-relayd,relay daemon>>::
7562 A process which is responsible for receiving the trace data sent by
7563 a distant consumer daemon.
7566 A set of sub-buffers.
7568 <<lttng-sessiond,session daemon>>::
7569 A process which receives control commands from you and orchestrates
7570 the tracers and various LTTng daemons.
7572 <<taking-a-snapshot,snapshot>>::
7573 A copy of the current data of all the sub-buffers of a given tracing
7574 session, saved as trace files.
7577 One part of an LTTng ring buffer which contains event records.
7580 The time information attached to an event when it is emitted.
7583 A set of files which are the concatenations of one or more
7584 flushed sub-buffers.
7587 The action of recording the events emitted by an application
7588 or by a system, or to initiate such recording by controlling
7592 The http://tracecompass.org[Trace Compass] project and application.
7595 An instrumentation point using the tracepoint mechanism of the Linux
7596 kernel or of LTTng-UST.
7598 tracepoint definition::
7599 The definition of a single tracepoint.
7602 The name of a tracepoint.
7604 tracepoint provider::
7605 A set of functions providing tracepoints to an instrumented user
7608 Not to be confused with a _tracepoint provider package_: many tracepoint
7609 providers can exist within a tracepoint provider package.
7611 tracepoint provider package::
7612 One or more tracepoint providers compiled as an object file or as
7616 A software which records emitted events.
7618 <<domain,tracing domain>>::
7619 A namespace for event sources.
7621 <<tracing-group,tracing group>>::
7622 The Unix group in which a Unix user can be to be allowed to trace the
7625 <<tracing-session,tracing session>>::
7626 A stateful dialogue between you and a <<lttng-sessiond,session
7630 An application running in user space, as opposed to a Linux kernel
7631 module, for example.