1 The LTTng Documentation
2 =======================
3 Philippe Proulx <pproulx@efficios.com>
4 v2.12, 25 February 2021
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 Skip this section if you’re familiar with software tracing and with the
28 * **<<installing-lttng,Installation>>** describes the steps to
29 install the LTTng packages on common Linux distributions and from
32 Skip this section if you already properly installed LTTng on your target
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 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 Skip this section if you don't 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{nbsp}{revision}?
77 LTTng{nbsp}{revision} bears the name _Ta Meilleure_, a Northeast IPA
78 beer brewed by https://lagabiere.com/[Lagabière]. Translating to ``Your
79 best one'', this beer gives out strong aromas of passion fruit, lemon,
80 and peaches. Tastewise, expect a lot of fruit, a creamy texture, and a
81 smooth lingering hop bitterness.
83 New features and changes in LTTng{nbsp}{revision}:
87 * Clear the contents of one or more <<tracing-session,tracing sessions>>
88 without having to destroy and reconfigure them
89 with the new man:lttng-clear(1) command.
91 This is especially useful to clear the tracing data of a tracing session
92 between attempts to reproduce a problem.
94 See <<clear,Clear a tracing session>>.
96 * Before LTTng{nbsp}{revision}, the man:lttng-track(1) and
97 man:lttng-untrack(1) commands used to add and remove process IDs
98 (PIDs) to a whitelist so that LTTng would only trace processes with
101 LTTng{nbsp}{revision} adds Unix user IDs (UIDs) and Unix group IDs
102 (GIDs) to the available <<pid-tracking,process attributes to track>>.
103 You can specify numeric user/group IDs and user/group names to track,
108 $ lttng track --userspace --vuid=http,999 --vgid=mysql,9
111 While you can also track UIDs and GIDs with the
112 opt:lttng-enable-event(1):--filter option of the `enable-event` command,
113 this dedicated process attribute tracking approach reduces tracing
114 overhead and prevents the creation of <<def-sub-buffer,sub-buffers>> for
115 the users and groups which LTTng doesn't track.
117 In the command manual pages, the term ``whitelist'' is renamed to
118 ``inclusion set'' to clarify the concept.
120 * The <<lttng-relayd,relay daemon>> can now maintain many files
121 virtually opened without using as many file descriptors (FD). It does
122 so by closing and reopening FDs as needed.
124 This feature is meant as a workaround for users who can't bump the
125 system limit because of permission restrictions.
127 The new opt:lttng-relayd(8):--fd-pool-size relay daemon option
128 sets the maximum number of simultaneously opened file descriptors
129 (using the soft `RLIMIT_NOFILE` resource limit of the process by
130 default; see man:getrlimit(2)).
132 * By default, the relay daemon writes its traces under a predefined
134 +$LTTNG_HOME/lttng-traces/__host__/__session__/__domain__+, with:
141 <<tracing-session,Tracing session>> name.
144 <<domain,Tracing domain>> name (`ust` or `kernel`).
147 Change this hierarchy to group traces by tracing session name rather
149 (+$LTTNG_HOME/lttng-traces/__session__/__host__/__domain__+) with the
150 new opt:lttng-relayd(8):--group-output-by-session option of the
153 This feature is especially useful if you're tracing two or more hosts,
154 having different hostnames, which share the same tracing session name as
155 part of their configuration. In this scenario, you can use a descriptive
156 tracing session name (for example, `connection-hang`) across a fleet of
157 machines streaming to a single relay daemon.
159 * The relay daemon has a new opt:lttng-relayd(8):--working-directory
160 option to override its working directory.
162 Linux kernel tracing::
164 * New instrumentation hooks to trace the entry and exit tracepoints of
165 the network reception code paths of the Linux kernel.
167 Use the resulting event records to identify the bounds of a network
168 reception and link the events that occur in the interim (for example,
169 wake-ups) to a specific network reception instance. You can also
170 analyze the latency of the network stack thanks to those event records.
172 * The `thread` field of the `irqaction` structure, which specifies the
173 process to wake up when a threaded interrupt request (IRQ) occurs, is
174 now part of the `lttng_statedump_interrupt` event record.
176 Use this information to discover which processes handle the various
177 IRQs. You can also associate the events occurring in the context of
178 those processes with their respective IRQ.
180 * New `lttng_statedump_cpu_topology` tracepoint to record the active
183 Use this information to discover which CPUs are SMT siblings or part of
184 the same socket. As of LTTng{nbsp}{revision}, only the x86 architecture
185 is supported since all architectures describe their topologies
188 The `architecture` field of the tracepoint is statically defined and
189 exists for all architecture implementations. Analysis tools can
190 therefore anticipate the layout of the event record.
192 Event record example:
196 lttng_statedump_cpu_topology:
202 model_name: Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz
208 * New product UUID metadata environment field, `product_uuid`,
209 which LTTng copies from the
210 https://en.wikipedia.org/wiki/Desktop_Management_Interface[Desktop
211 Management Interface] (DMI).
213 Use this environment field to uniquely identify a machine (virtual or
214 physical) in order to correlate traces from multiple virtual machines.
220 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
221 generation_ is a modern toolkit for tracing Linux systems and
222 applications. So your first question might be:
229 As the history of software engineering progressed and led to what
230 we now take for granted--complex, numerous and
231 interdependent software applications running in parallel on
232 sophisticated operating systems like Linux--the authors of such
233 components, software developers, began feeling a natural
234 urge to have tools that would ensure the robustness and good performance
235 of their masterpieces.
237 One major achievement in this field is, inarguably, the
238 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
239 an essential tool for developers to find and fix bugs. But even the best
240 debugger won't help make your software run faster, and nowadays, faster
241 software means either more work done by the same hardware, or cheaper
242 hardware for the same work.
244 A _profiler_ is often the tool of choice to identify performance
245 bottlenecks. Profiling is suitable to identify _where_ performance is
246 lost in a given software. The profiler outputs a profile, a statistical
247 summary of observed events, which you may use to discover which
248 functions took the most time to execute. However, a profiler won't
249 report _why_ some identified functions are the bottleneck. Bottlenecks
250 might only occur when specific conditions are met, conditions that are
251 sometimes impossible to capture by a statistical profiler, or impossible
252 to reproduce with an application altered by the overhead of an
253 event-based profiler. For a thorough investigation of software
254 performance issues, a history of execution is essential, with the
255 recorded values of variables and context fields you choose, and
256 with as little influence as possible on the instrumented software. This
257 is where tracing comes in handy.
259 _Tracing_ is a technique used to understand what goes on in a running
260 software system. The software used for tracing is called a _tracer_,
261 which is conceptually similar to a tape recorder. When recording,
262 specific instrumentation points placed in the software source code
263 generate events that are saved on a giant tape: a _trace_ file. You
264 can trace user applications and the operating system at the same time,
265 opening the possibility of resolving a wide range of problems that would
266 otherwise be extremely challenging.
268 Tracing is often compared to _logging_. However, tracers and loggers are
269 two different tools, serving two different purposes. Tracers are
270 designed to record much lower-level events that occur much more
271 frequently than log messages, often in the range of thousands per
272 second, with very little execution overhead. Logging is more appropriate
273 for a very high-level analysis of less frequent events: user accesses,
274 exceptional conditions (errors and warnings, for example), database
275 transactions, instant messaging communications, and such. Simply put,
276 logging is one of the many use cases that can be satisfied with tracing.
278 The list of recorded events inside a trace file can be read manually
279 like a log file for the maximum level of detail, but it is generally
280 much more interesting to perform application-specific analyses to
281 produce reduced statistics and graphs that are useful to resolve a
282 given problem. Trace viewers and analyzers are specialized tools
285 In the end, this is what LTTng is: a powerful, open source set of
286 tools to trace the Linux kernel and user applications at the same time.
287 LTTng is composed of several components actively maintained and
288 developed by its link:/community/#where[community].
291 [[lttng-alternatives]]
292 === Alternatives to noch:{LTTng}
294 Excluding proprietary solutions, a few competing software tracers
297 https://github.com/dtrace4linux/linux[dtrace4linux]::
298 A port of Sun Microsystems' DTrace to Linux.
300 The cmd:dtrace tool interprets user scripts and is responsible for
301 loading code into the Linux kernel for further execution and collecting
304 https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF]::
305 A subsystem in the Linux kernel in which a virtual machine can
306 execute programs passed from the user space to the kernel.
308 You can attach such programs to tracepoints and kprobes thanks to a
309 system call, and they can output data to the user space when executed
310 thanks to different mechanisms (pipe, VM register values, and eBPF maps,
313 https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]::
314 The de facto function tracer of the Linux kernel.
316 Its user interface is a set of special files in sysfs.
318 https://perf.wiki.kernel.org/[perf]::
319 A performance analysis tool for Linux which supports hardware
320 performance counters, tracepoints, as well as other counters and
323 The controlling utility of perf is the cmd:perf command line/text UI
326 http://linux.die.net/man/1/strace[strace]::
327 A command-line utility which records system calls made by a
328 user process, as well as signal deliveries and changes of process
331 strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace] to
332 fulfill its function.
334 http://www.sysdig.org/[sysdig]::
335 Like SystemTap, uses scripts to analyze Linux kernel events.
337 You write scripts, or _chisels_ in the jargon of sysdig, in Lua and
338 sysdig executes them while it traces the system or afterwards. The
339 interface of sysdig is the cmd:sysdig command-line tool as well as the
340 text UI-based cmd:csysdig tool.
342 https://sourceware.org/systemtap/[SystemTap]::
343 A Linux kernel and user space tracer which uses custom user scripts
344 to produce plain text traces.
346 SystemTap converts the scripts to the C language, and then compiles them
347 as Linux kernel modules which are loaded to produce trace data. The
348 primary user interface of SystemTap is the cmd:stap command-line tool.
350 The main distinctive features of LTTng is that it produces correlated
351 kernel and user space traces, as well as doing so with the lowest
352 overhead amongst other solutions. It produces trace files in the
353 http://diamon.org/ctf[CTF] format, a file format optimized
354 for the production and analyses of multi-gigabyte data.
356 LTTng is the result of more than 10{nbsp}years of active open source
357 development by a community of passionate developers. LTTng is currently
358 available on major desktop and server Linux distributions.
360 The main interface for tracing control is a single command-line tool
361 named cmd:lttng. The latter can create several tracing sessions, enable
362 and disable events on the fly, filter events efficiently with custom
363 user expressions, start and stop tracing, and much more. LTTng can
364 record the traces on the file system or send them over the network, and
365 keep them totally or partially. You can view the traces once tracing
366 becomes inactive or in real-time.
368 <<installing-lttng,Install LTTng now>> and
369 <<getting-started,start tracing>>!
375 **LTTng** is a set of software <<plumbing,components>> which interact to
376 <<instrumenting,instrument>> the Linux kernel and user applications, and
377 to <<controlling-tracing,control tracing>> (start and stop
378 tracing, enable and disable event rules, and the rest). Those
379 components are bundled into the following packages:
382 Libraries and command-line interface to control tracing.
385 Linux kernel modules to instrument and trace the kernel.
388 Libraries and Java/Python packages to instrument and trace user
391 Most distributions mark the LTTng-modules and LTTng-UST packages as
392 optional when installing LTTng-tools (which is always required). In the
393 following sections, we always provide the steps to install all three,
396 * You only need to install LTTng-modules if you intend to trace the
398 * You only need to install LTTng-UST if you intend to trace user
402 .Availability of LTTng{nbsp}{revision} for major Linux distributions as of 25{nbsp}February{nbsp}2021.
404 |Distribution |Available in releases
406 |https://www.ubuntu.com/[Ubuntu]
407 |<<ubuntu,Ubuntu 20.10 _Groovy Gorilla_>>.
409 Ubuntu{nbsp}16.04 _Xenial Xerus_, Ubuntu{nbsp}18.04 _Bionic Beaver_,
410 and Ubuntu{nbsp}20.04 _Focal Fossa_:
411 <<ubuntu-ppa,use the LTTng Stable{nbsp}{revision} PPA>>.
413 |https://www.debian.org/[Debian]
414 |<<debian,Debian "bullseye" (testing)>>.
416 |https://getfedora.org/[Fedora]
417 |xref:fedora[Fedora{nbsp}33, Fedora{nbsp}34, and Fedora{nbsp}35].
419 |https://www.archlinux.org/[Arch Linux]
420 |<<arch-linux,_Community_ repository and AUR>>.
422 |https://www.redhat.com/[RHEL] and https://www.suse.com/[SLES]
423 |See http://packages.efficios.com/[EfficiOS Enterprise Packages].
425 |https://alpinelinux.org/[Alpine Linux]
426 |<<alpine-linux,Alpine Linux{nbsp}3.12 and Alpine Linux{nbsp}3.13>>.
428 |https://buildroot.org/[Buildroot]
429 |xref:buildroot[Buildroot{nbsp}2020.08 and Buildroot{nbsp}2020.11].
431 |https://www.openembedded.org/wiki/Main_Page[OpenEmbedded] and
432 https://www.yoctoproject.org/[Yocto]
433 |<<oe-yocto,Yocto Project{nbsp}3.2 _Gatesgarth_>>
434 (`openembedded-core` layer).
439 === [[ubuntu-official-repositories]]Ubuntu
441 LTTng{nbsp}{revision} is available on Ubuntu{nbsp}20.10 _Groovy
442 Gorilla_. For previous supported releases of Ubuntu, <<ubuntu-ppa,use
443 the LTTng Stable{nbsp}{revision} PPA>>.
445 To install LTTng{nbsp}{revision} on Ubuntu{nbsp}20.10 _Groovy Gorilla_:
447 . Install the main LTTng{nbsp}{revision} packages:
452 # apt-get install lttng-tools
453 # apt-get install lttng-modules-dkms
454 # apt-get install liblttng-ust-dev
458 . **If you need to instrument and trace
459 <<java-application,Java applications>>**, install the LTTng-UST
465 # apt-get install liblttng-ust-agent-java
469 . **If you need to instrument and trace
470 <<python-application,Python{nbsp}3 applications>>**, install the
471 LTTng-UST Python agent:
476 # apt-get install python3-lttngust
482 === Ubuntu: noch:{LTTng} Stable {revision} PPA
484 The https://launchpad.net/~lttng/+archive/ubuntu/stable-{revision}[LTTng
485 Stable{nbsp}{revision} PPA] offers the latest stable
486 LTTng{nbsp}{revision} packages for Ubuntu{nbsp}16.04 _Xenial Xerus_,
487 Ubuntu{nbsp}18.04 _Bionic Beaver_, and Ubuntu{nbsp}20.04 _Focal Fossa_.
489 To install LTTng{nbsp}{revision} from the LTTng Stable{nbsp}{revision}
492 . Add the LTTng Stable{nbsp}{revision} PPA repository and update the
498 # apt-add-repository ppa:lttng/stable-2.12
503 . Install the main LTTng{nbsp}{revision} packages:
508 # apt-get install lttng-tools
509 # apt-get install lttng-modules-dkms
510 # apt-get install liblttng-ust-dev
514 . **If you need to instrument and trace
515 <<java-application,Java applications>>**, install the LTTng-UST
521 # apt-get install liblttng-ust-agent-java
525 . **If you need to instrument and trace
526 <<python-application,Python{nbsp}3 applications>>**, install the
527 LTTng-UST Python agent:
532 # apt-get install python3-lttngust
540 To install LTTng{nbsp}{revision} on Debian "bullseye" (testing):
542 . Install the main LTTng{nbsp}{revision} packages:
547 # apt-get install lttng-modules-dkms
548 # apt-get install liblttng-ust-dev
549 # apt-get install lttng-tools
553 . **If you need to instrument and trace <<java-application,Java
554 applications>>**, install the LTTng-UST Java agent:
559 # apt-get install liblttng-ust-agent-java
563 . **If you need to instrument and trace <<python-application,Python
564 applications>>**, install the LTTng-UST Python agent:
569 # apt-get install python3-lttngust
577 To install LTTng{nbsp}{revision} on Fedora{nbsp}33, Fedora{nbsp}34, or
580 . Install the LTTng-tools{nbsp}{revision} and LTTng-UST{nbsp}{revision}
586 # yum install lttng-tools
587 # yum install lttng-ust
591 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
597 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.12.tar.bz2 &&
598 tar -xf lttng-modules-latest-2.12.tar.bz2 &&
599 cd lttng-modules-2.12.* &&
601 sudo make modules_install &&
607 .Java and Python application instrumentation and tracing
609 If you need to instrument and trace <<java-application,Java
610 applications>> on Fedora, you need to build and install
611 LTTng-UST{nbsp}{revision} <<building-from-source,from source>> and pass
612 the `--enable-java-agent-jul`, `--enable-java-agent-log4j`, or
613 `--enable-java-agent-all` options to the `configure` script, depending
614 on which Java logging framework you use.
616 If you need to instrument and trace <<python-application,Python
617 applications>> on Fedora, you need to build and install
618 LTTng-UST{nbsp}{revision} from source and pass the
619 `--enable-python-agent` option to the `configure` script.
626 LTTng-UST{nbsp}{revision} is available in the _community_
627 repository of Arch Linux, while LTTng-tools{nbsp}{revision} and
628 LTTng-modules{nbsp}{revision} are available in the
629 https://aur.archlinux.org/[AUR].
631 To install LTTng{nbsp}{revision} on Arch Linux, using
632 https://github.com/Jguer/yay[yay] for the AUR packages:
634 . Install the main LTTng{nbsp}{revision} packages:
639 # pacman -Sy lttng-ust
640 $ yay -Sy lttng-tools
641 $ yay -Sy lttng-modules
645 . **If you need to instrument and trace <<python-application,Python
646 applications>>**, install the LTTng-UST Python agent:
651 # pacman -Sy python-lttngust
652 # pacman -Sy python2-lttngust
660 To install LTTng-tools{nbsp}{revision} and LTTng-UST{nbsp}{revision} on
661 Alpine Linux{nbsp}3.12 or Alpine Linux{nbsp}3.13:
663 . Add the LTTng packages:
668 # apk add lttng-tools
669 # apk add lttng-ust-dev
673 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
679 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.12.tar.bz2 &&
680 tar -xf lttng-modules-latest-2.12.tar.bz2 &&
681 cd lttng-modules-2.12.* &&
683 sudo make modules_install &&
692 To install LTTng{nbsp}{revision} on Buildroot{nbsp}2020.08 or
693 Buildroot{nbsp}2020.11:
695 . Launch the Buildroot configuration tool:
704 . In **Kernel**, check **Linux kernel**.
705 . In **Toolchain**, check **Enable WCHAR support**.
706 . In **Target packages**{nbsp}→ **Debugging, profiling and benchmark**,
707 check **lttng-modules** and **lttng-tools**.
708 . In **Target packages**{nbsp}→ **Libraries**{nbsp}→
709 **Other**, check **lttng-libust**.
713 === OpenEmbedded and Yocto
715 LTTng{nbsp}{revision} recipes are available in the
716 https://layers.openembedded.org/layerindex/branch/master/layer/openembedded-core/[`openembedded-core`]
717 layer for Yocto Project{nbsp}3.2 _Gatesgarth_ under the following names:
723 With BitBake, the simplest way to include LTTng recipes in your target
724 image is to add them to `IMAGE_INSTALL_append` in path:{conf/local.conf}:
727 IMAGE_INSTALL_append = " lttng-tools lttng-modules lttng-ust"
732 . Select a machine and an image recipe.
733 . Click **Edit image recipe**.
734 . Under the **All recipes** tab, search for **lttng**.
735 . Check the desired LTTng recipes.
738 [[building-from-source]]
739 === Build from source
741 To build and install LTTng{nbsp}{revision} from source:
743 . Using the package manager of your distribution, or from source,
744 install the following dependencies of LTTng-tools and LTTng-UST:
747 * https://sourceforge.net/projects/libuuid/[libuuid]
748 * http://directory.fsf.org/wiki/Popt[popt]
749 * http://liburcu.org/[Userspace RCU]
750 * http://www.xmlsoft.org/[libxml2]
751 * **Optional**: https://github.com/numactl/numactl[numactl]
754 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
760 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.12.tar.bz2 &&
761 tar -xf lttng-modules-latest-2.12.tar.bz2 &&
762 cd lttng-modules-2.12.* &&
764 sudo make modules_install &&
769 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
775 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.12.tar.bz2 &&
776 tar -xf lttng-ust-latest-2.12.tar.bz2 &&
777 cd lttng-ust-2.12.* &&
785 Add `--disable-numa` to `./configure` if you don't have
786 https://github.com/numactl/numactl[numactl].
790 .Java and Python application tracing
792 If you need to instrument and trace <<java-application,Java
793 applications>>, pass the `--enable-java-agent-jul`,
794 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
795 `configure` script, depending on which Java logging framework you use.
797 If you need to instrument and trace <<python-application,Python
798 applications>>, pass the `--enable-python-agent` option to the
799 `configure` script. You can set the `PYTHON` environment variable to the
800 path to the Python interpreter for which to install the LTTng-UST Python
808 By default, LTTng-UST libraries are installed to
809 dir:{/usr/local/lib}, which is the de facto directory in which to
810 keep self-compiled and third-party libraries.
812 When <<building-tracepoint-providers-and-user-application,linking an
813 instrumented user application with `liblttng-ust`>>:
815 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
817 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
818 man:gcc(1), man:g++(1), or man:clang(1).
822 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
828 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.12.tar.bz2 &&
829 tar -xf lttng-tools-latest-2.12.tar.bz2 &&
830 cd lttng-tools-2.12.* &&
838 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
839 previous steps automatically for a given version of LTTng and confine
840 the installed files in a specific directory. This can be useful to test
841 LTTng without installing it on your system.
847 This is a short guide to get started quickly with LTTng kernel and user
850 Before you follow this guide, make sure to <<installing-lttng,install>>
853 This tutorial walks you through the steps to:
855 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
856 . <<tracing-your-own-user-application,Trace a user application>> written
858 . <<viewing-and-analyzing-your-traces,View and analyze the
862 [[tracing-the-linux-kernel]]
863 === Trace the Linux kernel
865 The following command lines start with the `#` prompt because you need
866 root privileges to trace the Linux kernel. You can also trace the kernel
867 as a regular user if your Unix user is a member of the
868 <<tracing-group,tracing group>>.
870 . Create a <<tracing-session,tracing session>> which writes its traces
871 to dir:{/tmp/my-kernel-trace}:
876 # lttng create my-kernel-session --output=/tmp/my-kernel-trace
880 . List the available kernel tracepoints and system calls:
885 # lttng list --kernel
886 # lttng list --kernel --syscall
890 . Create <<event,event rules>> which match the desired instrumentation
891 point names, for example the `sched_switch` and `sched_process_fork`
892 tracepoints, and the man:open(2) and man:close(2) system calls:
897 # lttng enable-event --kernel sched_switch,sched_process_fork
898 # lttng enable-event --kernel --syscall open,close
902 Create an event rule which matches _all_ the Linux kernel
903 tracepoints with the opt:lttng-enable-event(1):--all option
904 (this will generate a lot of data when tracing):
909 # lttng enable-event --kernel --all
913 . <<basic-tracing-session-control,Start tracing>>:
922 . Do some operation on your system for a few seconds. For example,
923 load a website, or list the files of a directory.
924 . <<creating-destroying-tracing-sessions,Destroy>> the current
934 The man:lttng-destroy(1) command doesn't destroy the trace data; it
935 only destroys the state of the tracing session.
937 The man:lttng-destroy(1) command also runs the man:lttng-stop(1) command
938 implicitly (see <<basic-tracing-session-control,Start and stop a tracing
939 session>>). You need to stop tracing to make LTTng flush the remaining
940 trace data and make the trace readable.
942 . For the sake of this example, make the recorded trace accessible to
948 # chown -R $(whoami) /tmp/my-kernel-trace
952 See <<viewing-and-analyzing-your-traces,View and analyze the
953 recorded events>> to view the recorded events.
956 [[tracing-your-own-user-application]]
957 === Trace a user application
959 This section steps you through a simple example to trace a
960 _Hello world_ program written in C.
962 To create the traceable user application:
964 . Create the tracepoint provider header file, which defines the
965 tracepoints and the events they can generate:
971 #undef TRACEPOINT_PROVIDER
972 #define TRACEPOINT_PROVIDER hello_world
974 #undef TRACEPOINT_INCLUDE
975 #define TRACEPOINT_INCLUDE "./hello-tp.h"
977 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
980 #include <lttng/tracepoint.h>
990 ctf_string(my_string_field, my_string_arg)
991 ctf_integer(int, my_integer_field, my_integer_arg)
995 #endif /* _HELLO_TP_H */
997 #include <lttng/tracepoint-event.h>
1001 . Create the tracepoint provider package source file:
1007 #define TRACEPOINT_CREATE_PROBES
1008 #define TRACEPOINT_DEFINE
1010 #include "hello-tp.h"
1014 . Build the tracepoint provider package:
1019 $ gcc -c -I. hello-tp.c
1023 . Create the _Hello World_ application source file:
1030 #include "hello-tp.h"
1032 int main(int argc, char *argv[])
1036 puts("Hello, World!\nPress Enter to continue...");
1039 * The following getchar() call is only placed here for the purpose
1040 * of this demonstration, to pause the application in order for
1041 * you to have time to list its tracepoints. It's not needed
1047 * A tracepoint() call.
1049 * Arguments, as defined in hello-tp.h:
1051 * 1. Tracepoint provider name (required)
1052 * 2. Tracepoint name (required)
1053 * 3. my_integer_arg (first user-defined argument)
1054 * 4. my_string_arg (second user-defined argument)
1056 * Notice the tracepoint provider and tracepoint names are
1057 * NOT strings: they are in fact parts of variables that the
1058 * macros in hello-tp.h create.
1060 tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
1062 for (x = 0; x < argc; ++x) {
1063 tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
1066 puts("Quitting now!");
1067 tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
1074 . Build the application:
1083 . Link the application with the tracepoint provider package,
1084 `liblttng-ust`, and `libdl`:
1089 $ gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
1093 Here's the whole build process:
1096 .Build steps of the user space tracing tutorial.
1097 image::ust-flow.png[]
1099 To trace the user application:
1101 . Run the application with a few arguments:
1106 $ ./hello world and beyond
1115 Press Enter to continue...
1119 . Start an LTTng <<lttng-sessiond,session daemon>>:
1124 $ lttng-sessiond --daemonize
1128 Note that a session daemon might already be running, for example as a
1129 service that the service manager of the distribution started.
1131 . List the available user space tracepoints:
1136 $ lttng list --userspace
1140 You see the `hello_world:my_first_tracepoint` tracepoint listed
1141 under the `./hello` process.
1143 . Create a <<tracing-session,tracing session>>:
1148 $ lttng create my-user-space-session
1152 . Create an <<event,event rule>> which matches the
1153 `hello_world:my_first_tracepoint` event name:
1158 $ lttng enable-event --userspace hello_world:my_first_tracepoint
1162 . <<basic-tracing-session-control,Start tracing>>:
1171 . Go back to the running `hello` application and press Enter. The
1172 program executes all `tracepoint()` instrumentation points and exits.
1173 . <<creating-destroying-tracing-sessions,Destroy>> the current
1183 The man:lttng-destroy(1) command doesn't destroy the trace data; it
1184 only destroys the state of the tracing session.
1186 The man:lttng-destroy(1) command also runs the man:lttng-stop(1) command
1187 implicitly (see <<basic-tracing-session-control,Start and stop a tracing
1188 session>>). You need to stop tracing to make LTTng flush the remaining
1189 trace data and make the trace readable.
1191 By default, LTTng saves the traces in
1192 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
1193 where +__name__+ is the tracing session name. The
1194 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
1196 See <<viewing-and-analyzing-your-traces,View and analyze the
1197 recorded events>> to view the recorded events.
1200 [[viewing-and-analyzing-your-traces]]
1201 === View and analyze the recorded events
1203 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
1204 kernel>> and <<tracing-your-own-user-application,Trace a user
1205 application>> tutorials, you can inspect the recorded events.
1207 There are many tools you can use to read LTTng traces:
1209 https://babeltrace.org/[Babeltrace{nbsp}2]::
1210 A rich, flexible trace manipulation toolkit which includes
1211 a versatile command-line interface
1212 (https://babeltrace.org/docs/v2.0/man1/babeltrace2.1/[cmd:babeltrace2]),
1213 a https://babeltrace.org/docs/v2.0/libbabeltrace2/[C library],
1214 and https://babeltrace.org/docs/v2.0/python/bt2/[Python{nbsp}3 bindings]
1215 so that you can easily process or convert an LTTng trace with
1218 The Babeltrace{nbsp}2 project ships with a
1219 https://babeltrace.org/docs/v2.0/man7/babeltrace2-plugin-ctf.7/[plugin]
1220 which supports the format of the traces which LTTng produces,
1221 https://diamon.org/ctf/[CTF].
1223 http://tracecompass.org/[Trace Compass]::
1224 A graphical user interface for viewing and analyzing any type of
1225 logs or traces, including those of LTTng.
1227 https://github.com/lttng/lttng-analyses[**LTTng analyses**]::
1228 An experimental project which includes many high-level analyses of
1229 LTTng kernel traces, like scheduling statistics, interrupt
1230 frequency distribution, top CPU usage, and more.
1232 NOTE: This section assumes that LTTng saved the traces it recorded
1233 during the previous tutorials to their default location, in the
1234 dir:{$LTTNG_HOME/lttng-traces} directory. The env:LTTNG_HOME
1235 environment variable defaults to `$HOME` if not set.
1238 [[viewing-and-analyzing-your-traces-bt]]
1239 ==== Use the cmd:babeltrace2 command-line tool
1241 The simplest way to list all the recorded events of an LTTng trace is to
1243 https://babeltrace.org/docs/v2.0/man1/babeltrace2.1/[cmd:babeltrace2]
1248 $ babeltrace2 ~/lttng-traces/my-user-space-session*
1251 cmd:babeltrace2 finds all traces recursively within the given path and
1252 prints all their events, sorting them chronologically.
1254 Pipe the output of cmd:babeltrace2 into a tool like man:grep(1) for
1259 $ babeltrace2 /tmp/my-kernel-trace | grep _switch
1262 Pipe the output of cmd:babeltrace2 into a tool like man:wc(1) to count
1263 the recorded events:
1267 $ babeltrace2 /tmp/my-kernel-trace | grep _open | wc --lines
1271 [[viewing-and-analyzing-your-traces-bt-python]]
1272 ==== Use the Babeltrace{nbsp}2 Python bindings
1274 The <<viewing-and-analyzing-your-traces-bt,text output of
1275 cmd:babeltrace2>> is useful to isolate events by simple matching using
1276 man:grep(1) and similar utilities. However, more elaborate filters, such
1277 as keeping only event records with a field value falling within a
1278 specific range, are not trivial to write using a shell. Moreover,
1279 reductions and even the most basic computations involving multiple event
1280 records are virtually impossible to implement.
1282 Fortunately, Babeltrace{nbsp}2 ships with
1283 https://babeltrace.org/docs/v2.0/python/bt2/[Python{nbsp}3 bindings]
1284 which make it easy to read the event records of an LTTng trace
1285 sequentially and compute the desired information.
1287 The following script accepts an LTTng Linux kernel trace path as its
1288 first argument and prints the short names of the top five running
1289 processes on CPU{nbsp}0 during the whole trace:
1299 # Get the trace path from the first command-line argument.
1300 it = bt2.TraceCollectionMessageIterator(sys.argv[1])
1302 # This counter dictionary will hold execution times:
1304 # Task command name -> Total execution time (ns)
1305 exec_times = collections.Counter()
1307 # This holds the last `sched_switch` timestamp.
1311 # We only care about event messages.
1312 if type(msg) is not bt2._EventMessageConst:
1315 # Event of the event message.
1318 # Keep only `sched_switch` events.
1319 if event.cls.name != 'sched_switch':
1322 # Keep only events which occurred on CPU 0.
1323 if event.packet.context_field['cpu_id'] != 0:
1326 # Event timestamp (ns).
1327 cur_ts = msg.default_clock_snapshot.ns_from_origin
1333 # (Short) name of the previous task command.
1334 prev_comm = str(event.payload_field['prev_comm'])
1336 # Initialize an entry in our dictionary if not yet done.
1337 if prev_comm not in exec_times:
1338 exec_times[prev_comm] = 0
1340 # Compute previous command execution time.
1341 diff = cur_ts - last_ts
1343 # Update execution time of this command.
1344 exec_times[prev_comm] += diff
1346 # Update last timestamp.
1350 for name, ns in exec_times.most_common(5):
1351 print('{:20}{} s'.format(name, ns / 1e9))
1354 if __name__ == '__main__':
1362 $ python3 top5proc.py /tmp/my-kernel-trace/kernel
1368 swapper/0 48.607245889 s
1369 chromium 7.192738188 s
1370 pavucontrol 0.709894415 s
1371 Compositor 0.660867933 s
1372 Xorg.bin 0.616753786 s
1375 Note that `swapper/0` is the ``idle'' process of CPU{nbsp}0 on Linux;
1376 since we weren't using the CPU that much when tracing, its first
1377 position in the list makes sense.
1381 == [[understanding-lttng]]Core concepts
1383 From a user's perspective, the LTTng system is built on a few concepts,
1384 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
1385 operates by sending commands to the <<lttng-sessiond,session daemon>>.
1386 Understanding how those objects relate to eachother is key in mastering
1389 The core concepts are:
1391 * <<tracing-session,Tracing session>>
1392 * <<domain,Tracing domain>>
1393 * <<channel,Channel and ring buffer>>
1394 * <<"event","Instrumentation point, event rule, event, and event record">>
1400 A _tracing session_ is a stateful dialogue between you and
1401 a <<lttng-sessiond,session daemon>>. You can
1402 <<creating-destroying-tracing-sessions,create a new tracing
1403 session>> with the `lttng create` command.
1405 Most of what you do when you control LTTng tracers happens within a
1406 tracing session. In particular, a tracing session:
1409 * Has its own set of trace files.
1410 * Has its own state of activity (started or stopped).
1411 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
1413 * Has its own <<channel,channels>> to which are associated their own
1414 <<event,event rules>>.
1415 * Has its own <<pid-tracking,process attribute tracking>> inclusion
1419 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
1420 image::concepts.png[]
1422 Those attributes and objects are completely isolated between different
1425 A tracing session is analogous to a cash machine session:
1426 the operations you do on the banking system through the cash machine do
1427 not alter the data of other users of the same system. In the case of
1428 the cash machine, a session lasts as long as your bank card is inside.
1429 In the case of LTTng, a tracing session lasts from the `lttng create`
1430 command to the `lttng destroy` command.
1433 .Each Unix user has its own set of tracing sessions.
1434 image::many-sessions.png[]
1437 [[tracing-session-mode]]
1438 ==== Tracing session mode
1440 LTTng can send the generated trace data to different locations. The
1441 _tracing session mode_ dictates where to send it. The following modes
1442 are available in LTTng{nbsp}{revision}:
1444 [[local-mode]]Local mode::
1445 LTTng writes the traces to the file system of the machine it traces
1448 [[net-streaming-mode]]Network streaming mode::
1449 LTTng sends the traces over the network to a
1450 <<lttng-relayd,relay daemon>> running on a remote system.
1453 LTTng doesn't write the traces by default.
1455 Instead, you can request LTTng to <<taking-a-snapshot,take a snapshot>>,
1456 that is, a copy of the current sub-buffers of the tracing session, and
1457 to write it to the file system of the target or to send it over the
1458 network to a <<lttng-relayd,relay daemon>> running on a remote system.
1460 [[live-mode]]Live mode::
1461 This mode is similar to the network streaming mode, but a live
1462 trace viewer can connect to the distant relay daemon to
1463 <<lttng-live,view event records as LTTng generates them>>.
1469 A _tracing domain_ is a namespace for event sources. A tracing domain
1470 has its own properties and features.
1472 There are currently five available tracing domains:
1476 * `java.util.logging` (JUL)
1480 You must specify a tracing domain when using some commands to avoid
1481 ambiguity. For example, since all the domains support named tracepoints
1482 as event sources (instrumentation points that you manually insert in the
1483 source code), you need to specify a tracing domain when
1484 <<enabling-disabling-events,creating an event rule>> because all the
1485 tracing domains could have tracepoints with the same names.
1487 You can create <<channel,channels>> in the Linux kernel and user space
1488 tracing domains. The other tracing domains have a single default
1493 === Channel and ring buffer
1495 A _channel_ is an object which is responsible for a set of ring buffers.
1496 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1497 tracer emits an event, it can record it to one or more
1498 sub-buffers. The attributes of a channel determine what to do when
1499 there's no space left for a new event record because all sub-buffers
1500 are full, where to send a full sub-buffer, and other behaviours.
1502 A channel is always associated to a <<domain,tracing domain>>. The
1503 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1504 a default channel which you can't configure.
1506 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1507 an event, it records it to the sub-buffers of all
1508 the enabled channels with a satisfied event rule, as long as those
1509 channels are part of active <<tracing-session,tracing sessions>>.
1512 [[channel-buffering-schemes]]
1513 ==== Per-user vs. per-process buffering schemes
1515 A channel has at least one ring buffer _per CPU_. LTTng always
1516 records an event to the ring buffer associated to the CPU on which it
1519 Two _buffering schemes_ are available when you
1520 <<enabling-disabling-channels,create a channel>> in the
1521 user space <<domain,tracing domain>>:
1523 Per-user buffering::
1524 Allocate one set of ring buffers--one per CPU--shared by all the
1525 instrumented processes of each Unix user.
1529 .Per-user buffering scheme.
1530 image::per-user-buffering.png[]
1533 Per-process buffering::
1534 Allocate one set of ring buffers--one per CPU--for each
1535 instrumented process.
1539 .Per-process buffering scheme.
1540 image::per-process-buffering.png[]
1543 The per-process buffering scheme tends to consume more memory than the
1544 per-user option because systems generally have more instrumented
1545 processes than Unix users running instrumented processes. However, the
1546 per-process buffering scheme ensures that one process having a high
1547 event throughput won't fill all the shared sub-buffers of the same
1550 The Linux kernel tracing domain has only one available buffering scheme
1551 which is to allocate a single set of ring buffers for the whole system.
1552 This scheme is similar to the per-user option, but with a single, global
1553 user ``running'' the kernel.
1556 [[channel-overwrite-mode-vs-discard-mode]]
1557 ==== Overwrite vs. discard event record loss modes
1559 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1560 arc in the following animations) of the ring buffer of a specific
1561 channel. When there's no space left in a sub-buffer, the tracer marks it
1562 as consumable (red) and another, empty sub-buffer starts receiving the
1563 following event records. A <<lttng-consumerd,consumer daemon>>
1564 eventually consumes the marked sub-buffer (returns to white).
1567 [role="docsvg-channel-subbuf-anim"]
1572 In an ideal world, sub-buffers are consumed faster than they are filled,
1573 as it is the case in the previous animation. In the real world,
1574 however, all sub-buffers can be full at some point, leaving no space to
1575 record the following events.
1577 By default, LTTng-modules and LTTng-UST are _non-blocking_ tracers: when
1578 no empty sub-buffer is available, it is acceptable to lose event records
1579 when the alternative would be to cause substantial delays in the
1580 execution of the instrumented application. LTTng privileges performance
1581 over integrity; it aims at perturbing the target system as little as
1582 possible in order to make tracing of subtle race conditions and rare
1583 interrupt cascades possible.
1585 Since LTTng{nbsp}2.10, the LTTng user space tracer, LTTng-UST, supports
1586 a _blocking mode_. See the <<blocking-timeout-example,blocking timeout
1587 example>> to learn how to use the blocking mode.
1589 When it comes to losing event records because no empty sub-buffer is
1590 available, or because the <<opt-blocking-timeout,blocking timeout>> is
1591 reached, the _event record loss mode_ of the channel determines what to
1592 do. The available event record loss modes are:
1595 Drop the newest event records until the tracer releases a sub-buffer.
1597 This is the only available mode when you specify a
1598 <<opt-blocking-timeout,blocking timeout>>.
1601 Clear the sub-buffer containing the oldest event records and start
1602 writing the newest event records there.
1604 This mode is sometimes called _flight recorder mode_ because it's
1606 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1607 always keep a fixed amount of the latest data.
1609 Which mechanism you should choose depends on your context: prioritize
1610 the newest or the oldest event records in the ring buffer?
1612 Beware that, in overwrite mode, the tracer abandons a _whole sub-buffer_
1613 as soon as a there's no space left for a new event record, whereas in
1614 discard mode, the tracer only discards the event record that doesn't
1617 In discard mode, LTTng increments a count of lost event records when an
1618 event record is lost and saves this count to the trace. Since
1619 LTTng{nbsp}2.8, in overwrite mode, LTTng writes to a given sub-buffer
1620 its sequence number within its data stream. With a <<local-mode,local>>,
1621 <<net-streaming-mode,network streaming>>, or <<live-mode,live>>
1622 <<tracing-session,tracing session>>, a trace reader can use such
1623 sequence numbers to report lost packets. In overwrite mode, LTTng
1624 doesn't write to the trace the exact number of lost event records in
1625 those lost sub-buffers.
1627 Trace analyses can use saved discarded event record and sub-buffer
1628 (packet) counts of the trace to decide whether or not to perform the
1629 analyses even if trace data is known to be missing.
1631 There are a few ways to decrease your probability of losing event
1633 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1634 how to fine-tune the sub-buffer count and size of a channel to virtually
1635 stop losing event records, though at the cost of greater memory usage.
1638 [[channel-subbuf-size-vs-subbuf-count]]
1639 ==== Sub-buffer count and size
1641 When you <<enabling-disabling-channels,create a channel>>, you can
1642 set its number of sub-buffers and their size.
1644 Note that there is noticeable CPU overhead introduced when
1645 switching sub-buffers (marking a full one as consumable and switching
1646 to an empty one for the following events to be recorded). Knowing this,
1647 the following list presents a few practical situations along with how
1648 to configure the sub-buffer count and size for them:
1650 * **High event throughput**: In general, prefer bigger sub-buffers to
1651 lower the risk of losing event records.
1653 Having bigger sub-buffers also ensures a lower
1654 <<channel-switch-timer,sub-buffer switching frequency>>.
1656 The number of sub-buffers is only meaningful if you create the channel
1657 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1658 other sub-buffers are left unaltered.
1660 * **Low event throughput**: In general, prefer smaller sub-buffers
1661 since the risk of losing event records is low.
1663 Because events occur less frequently, the sub-buffer switching frequency
1664 should remain low and thus the overhead of the tracer shouldn't be a
1667 * **Low memory system**: If your target system has a low memory
1668 limit, prefer fewer first, then smaller sub-buffers.
1670 Even if the system is limited in memory, you want to keep the
1671 sub-buffers as big as possible to avoid a high sub-buffer switching
1674 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1675 which means event data is very compact. For example, the average
1676 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1677 sub-buffer size of 1{nbsp}MiB is considered big.
1679 The previous situations highlight the major trade-off between a few big
1680 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1681 frequency vs. how much data is lost in overwrite mode. Assuming a
1682 constant event throughput and using the overwrite mode, the two
1683 following configurations have the same ring buffer total size:
1686 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1691 * **Two sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1692 switching frequency, but if a sub-buffer overwrite happens, half of
1693 the event records so far (4{nbsp}MiB) are definitely lost.
1694 * **Eight sub-buffers of 1{nbsp}MiB each**: Expect four times the
1695 overhead of the tracer as the previous configuration, but if a
1696 sub-buffer overwrite happens, only the eighth of event records so far
1697 are definitely lost.
1699 In discard mode, the sub-buffers count parameter is pointless: use two
1700 sub-buffers and set their size according to the requirements of your
1704 [[channel-switch-timer]]
1705 ==== Switch timer period
1707 The _switch timer period_ is an important configurable attribute of
1708 a channel to ensure periodic sub-buffer flushing.
1710 When the _switch timer_ expires, a sub-buffer switch happens. Set
1711 the switch timer period attribute when you
1712 <<enabling-disabling-channels,create a channel>> to ensure that LTTng
1713 consumes and commits trace data to trace files or to a distant relay
1714 daemon periodically in case of a low event throughput.
1717 [role="docsvg-channel-switch-timer"]
1722 This attribute is also convenient when you use big sub-buffers to cope
1723 with a sporadic high event throughput, even if the throughput is
1727 [[channel-read-timer]]
1728 ==== Read timer period
1730 By default, the LTTng tracers use a notification mechanism to signal a
1731 full sub-buffer so that a consumer daemon can consume it. When such
1732 notifications must be avoided, for example in real-time applications,
1733 use the _read timer_ of the channel instead. When the read timer fires,
1734 the <<lttng-consumerd,consumer daemon>> checks for full, consumable
1738 [[tracefile-rotation]]
1739 ==== Trace file count and size
1741 By default, trace files can grow as large as needed. Set the maximum
1742 size of each trace file that a channel writes when you
1743 <<enabling-disabling-channels,create a channel>>. When the size of a
1744 trace file reaches the fixed maximum size of the channel, LTTng creates
1745 another file to contain the next event records. LTTng appends a file
1746 count to each trace file name in this case.
1748 If you set the trace file size attribute when you create a channel, the
1749 maximum number of trace files that LTTng creates is _unlimited_ by
1750 default. To limit them, set a maximum number of trace files. When the
1751 number of trace files reaches the fixed maximum count of the channel,
1752 the oldest trace file is overwritten. This mechanism is called _trace
1757 Even if you don't limit the trace file count, you can't assume that
1758 LTTng doesn't manage any trace file.
1760 In other words, there is no safe way to know if LTTng still holds a
1761 given trace file open with the trace file rotation feature.
1763 The only way to obtain an unmanaged, self-contained LTTng trace before
1764 you <<creating-destroying-tracing-sessions,destroy>> the tracing session
1765 is with the <<session-rotation,tracing session rotation>> feature
1766 (available since LTTng{nbsp}2.11).
1771 === Instrumentation point, event rule, event, and event record
1773 An _event rule_ is a set of conditions which must be **all** satisfied
1774 for LTTng to record an occuring event.
1776 You set the conditions when you <<enabling-disabling-events,create
1779 You always attach an event rule to a <<channel,channel>> when you create
1782 When an event passes the conditions of an event rule, LTTng records it
1783 in one of the sub-buffers of the attached channel.
1785 The available conditions, as of LTTng{nbsp}{revision}, are:
1787 * The event rule _is enabled_.
1788 * The type of the instrumentation point _is{nbsp}T_.
1789 * The name of the instrumentation point (sometimes called _event name_)
1790 _matches{nbsp}N_, but _isn't{nbsp}E_.
1791 * The log level of the instrumentation point _is as severe as{nbsp}L_, or
1792 _is exactly{nbsp}L_.
1793 * The fields of the payload of the event _satisfy_ a filter
1794 expression{nbsp}__F__.
1796 As you can see, all the conditions but the dynamic filter are related to
1797 the status of the event rule or to the instrumentation point, not to the
1798 occurring events. This is why, without a filter, checking if an event
1799 passes an event rule isn't a dynamic task: when you create or modify an
1800 event rule, all the tracers of its tracing domain enable or disable the
1801 instrumentation points themselves once. This is possible because the
1802 attributes of an instrumentation point (type, name, and log level) are
1803 defined statically. In other words, without a dynamic filter, the tracer
1804 _doesn't evaluate_ the arguments of an instrumentation point unless it
1805 matches an enabled event rule.
1807 Note that, for LTTng to record an event, the <<channel,channel>> to
1808 which a matching event rule is attached must also be enabled, and the
1809 <<tracing-session,tracing session>> owning this channel must be active
1813 .Logical path from an instrumentation point to an event record.
1814 image::event-rule.png[]
1816 .Event, event record, or event rule?
1818 With so many similar terms, it's easy to get confused.
1820 An **event** is the consequence of the execution of an _instrumentation
1821 point_, like a tracepoint that you manually place in some source code,
1822 or a Linux kernel kprobe. An event is said to _occur_ at a specific
1823 time. Different actions can be taken upon the occurrence of an event,
1824 like record the payload of the event to a buffer.
1826 An **event record** is the representation of an event in a sub-buffer. A
1827 tracer is responsible for capturing the payload of an event, current
1828 context variables, the ID of the event, and its timestamp. LTTng
1829 can append this sub-buffer to a trace file.
1831 An **event rule** is a set of conditions which must _all_ be satisfied
1832 for LTTng to record an occuring event. Events still occur without
1833 satisfying event rules, but LTTng doesn't record them.
1838 == Components of noch:{LTTng}
1840 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1841 to call LTTng a simple _tool_ since it is composed of multiple
1842 interacting components. This section describes those components,
1843 explains their respective roles, and shows how they connect together to
1844 form the LTTng ecosystem.
1846 The following diagram shows how the most important components of LTTng
1847 interact with user applications, the Linux kernel, and you:
1850 .Control and trace data paths between LTTng components.
1851 image::plumbing.png[]
1853 The LTTng project incorporates:
1855 * **LTTng-tools**: Libraries and command-line interface to
1856 control tracing sessions.
1857 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1858 ** <<lttng-consumerd,Consumer daemon>> (cmd:lttng-consumerd).
1859 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1860 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1861 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1862 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1864 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1865 headers to instrument and trace any native user application.
1866 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1867 *** `liblttng-ust-libc-wrapper`
1868 *** `liblttng-ust-pthread-wrapper`
1869 *** `liblttng-ust-cyg-profile`
1870 *** `liblttng-ust-cyg-profile-fast`
1871 *** `liblttng-ust-dl`
1872 ** User space tracepoint provider source files generator command-line
1873 tool (man:lttng-gen-tp(1)).
1874 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1875 Java applications using `java.util.logging` or
1876 Apache log4j{nbsp}1.2 logging.
1877 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1878 Python applications using the standard `logging` package.
1879 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1881 ** LTTng kernel tracer module.
1882 ** Tracing ring buffer kernel modules.
1883 ** Probe kernel modules.
1884 ** LTTng logger kernel module.
1888 === Tracing control command-line interface
1891 .The tracing control command-line interface.
1892 image::plumbing-lttng-cli.png[]
1894 The _man:lttng(1) command-line tool_ is the standard user interface to
1895 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1896 is part of LTTng-tools.
1898 The cmd:lttng tool is linked with
1899 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1900 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1902 The cmd:lttng tool has a Git-like interface:
1906 $ lttng <GENERAL OPTIONS> <COMMAND> <COMMAND OPTIONS>
1909 The <<controlling-tracing,Tracing control>> section explores the
1910 available features of LTTng using the cmd:lttng tool.
1913 [[liblttng-ctl-lttng]]
1914 === Tracing control library
1917 .The tracing control library.
1918 image::plumbing-liblttng-ctl.png[]
1920 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1921 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1922 underlying details of the protocol. `liblttng-ctl` is part of LTTng-tools.
1924 The <<lttng-cli,cmd:lttng command-line tool>>
1925 is linked with `liblttng-ctl`.
1927 Use `liblttng-ctl` in C or $$C++$$ source code by including its
1932 #include <lttng/lttng.h>
1935 Some objects are referenced by name (C string), such as tracing
1936 sessions, but most of them require to create a handle first using
1937 `lttng_create_handle()`.
1939 As of LTTng{nbsp}{revision}, the best available developer documentation for
1940 `liblttng-ctl` is its installed header files. Every function and structure is
1941 thoroughly documented.
1945 === User space tracing library
1948 .The user space tracing library.
1949 image::plumbing-liblttng-ust.png[]
1951 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1952 is the LTTng user space tracer. It receives commands from a
1953 <<lttng-sessiond,session daemon>>, for example to
1954 enable and disable specific instrumentation points, and writes event
1955 records to ring buffers shared with a
1956 <<lttng-consumerd,consumer daemon>>.
1957 `liblttng-ust` is part of LTTng-UST.
1959 Public C header files are installed beside `liblttng-ust` to
1960 instrument any <<c-application,C or $$C++$$ application>>.
1962 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1963 packages, use their own library providing tracepoints which is
1964 linked with `liblttng-ust`.
1966 An application or library doesn't have to initialize `liblttng-ust`
1967 manually: its constructor does the necessary tasks to properly register
1968 to a session daemon. The initialization phase also enables the
1969 instrumentation points matching the <<event,event rules>> that you
1973 [[lttng-ust-agents]]
1974 === User space tracing agents
1977 .The user space tracing agents.
1978 image::plumbing-lttng-ust-agents.png[]
1980 The _LTTng-UST Java and Python agents_ are regular Java and Python
1981 packages which add LTTng tracing capabilities to the
1982 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1984 In the case of Java, the
1985 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1986 core logging facilities] and
1987 https://logging.apache.org/log4j/1.2/[Apache log4j{nbsp}1.2] are supported.
1988 Note that Apache Log4{nbsp}2 isn't supported.
1990 In the case of Python, the standard
1991 https://docs.python.org/3/library/logging.html[`logging`] package
1992 is supported. Both Python{nbsp}2 and Python{nbsp}3 modules can import the
1993 LTTng-UST Python agent package.
1995 The applications using the LTTng-UST agents are in the
1996 `java.util.logging` (JUL),
1997 log4j, and Python <<domain,tracing domains>>.
1999 Both agents use the same mechanism to trace the log statements. When an
2000 agent initializes, it creates a log handler that attaches to the root
2001 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
2002 When the application executes a log statement, the root logger passes it
2003 to the log handler of the agent. The log handler of the agent calls a
2004 native function in a tracepoint provider package shared library linked
2005 with <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
2006 other fields, like its logger name and its log level. This native
2007 function contains a user space instrumentation point, hence tracing the
2010 The log level condition of an
2011 <<event,event rule>> is considered when tracing
2012 a Java or a Python application, and it's compatible with the standard
2013 JUL, log4j, and Python log levels.
2017 === LTTng kernel modules
2020 .The LTTng kernel modules.
2021 image::plumbing-lttng-modules.png[]
2023 The _LTTng kernel modules_ are a set of Linux kernel modules
2024 which implement the kernel tracer of the LTTng project. The LTTng
2025 kernel modules are part of LTTng-modules.
2027 The LTTng kernel modules include:
2029 * A set of _probe_ modules.
2031 Each module attaches to a specific subsystem
2032 of the Linux kernel using its tracepoint instrument points. There are
2033 also modules to attach to the entry and return points of the Linux
2034 system call functions.
2036 * _Ring buffer_ modules.
2038 A ring buffer implementation is provided as kernel modules. The LTTng
2039 kernel tracer writes to the ring buffer; a
2040 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
2042 * The _LTTng kernel tracer_ module.
2043 * The _LTTng logger_ module.
2045 The LTTng logger module implements the special path:{/proc/lttng-logger}
2046 (and path:{/dev/lttng-logger} since LTTng{nbsp}2.11) files so that any
2047 executable can generate LTTng events by opening and writing to those
2050 See <<proc-lttng-logger-abi,LTTng logger>>.
2052 Generally, you don't have to load the LTTng kernel modules manually
2053 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
2054 daemon>> loads the necessary modules when starting. If you have extra
2055 probe modules, you can specify to load them to the session daemon on
2058 The LTTng kernel modules are installed in
2059 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
2060 the kernel release (see `uname --kernel-release`).
2067 .The session daemon.
2068 image::plumbing-sessiond.png[]
2070 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
2071 managing tracing sessions and for controlling the various components of
2072 LTTng. The session daemon is part of LTTng-tools.
2074 The session daemon sends control requests to and receives control
2077 * The <<lttng-ust,user space tracing library>>.
2079 Any instance of the user space tracing library first registers to
2080 a session daemon. Then, the session daemon can send requests to
2081 this instance, such as:
2084 ** Get the list of tracepoints.
2085 ** Share an <<event,event rule>> so that the user space tracing library
2086 can enable or disable tracepoints. Amongst the possible conditions
2087 of an event rule is a filter expression which `liblttng-ust` evalutes
2088 when an event occurs.
2089 ** Share <<channel,channel>> attributes and ring buffer locations.
2092 The session daemon and the user space tracing library use a Unix
2093 domain socket for their communication.
2095 * The <<lttng-ust-agents,user space tracing agents>>.
2097 Any instance of a user space tracing agent first registers to
2098 a session daemon. Then, the session daemon can send requests to
2099 this instance, such as:
2102 ** Get the list of loggers.
2103 ** Enable or disable a specific logger.
2106 The session daemon and the user space tracing agent use a TCP connection
2107 for their communication.
2109 * The <<lttng-modules,LTTng kernel tracer>>.
2110 * The <<lttng-consumerd,consumer daemon>>.
2112 The session daemon sends requests to the consumer daemon to instruct
2113 it where to send the trace data streams, amongst other information.
2115 * The <<lttng-relayd,relay daemon>>.
2117 The session daemon receives commands from the
2118 <<liblttng-ctl-lttng,tracing control library>>.
2120 The root session daemon loads the appropriate
2121 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
2122 a <<lttng-consumerd,consumer daemon>> as soon as you create
2123 an <<event,event rule>>.
2125 The session daemon doesn't send and receive trace data: this is the
2126 role of the <<lttng-consumerd,consumer daemon>> and
2127 <<lttng-relayd,relay daemon>>. It does, however, generate the
2128 http://diamon.org/ctf/[CTF] metadata stream.
2130 Each Unix user can have its own session daemon instance. The
2131 tracing sessions which different session daemons manage are completely
2134 The root user's session daemon is the only one which is
2135 allowed to control the LTTng kernel tracer, and its spawned consumer
2136 daemon is the only one which is allowed to consume trace data from the
2137 LTTng kernel tracer. Note, however, that any Unix user which is a member
2138 of the <<tracing-group,tracing group>> is allowed
2139 to create <<channel,channels>> in the
2140 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
2143 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
2144 session daemon when using its `create` command if none is currently
2145 running. You can also start the session daemon manually.
2152 .The consumer daemon.
2153 image::plumbing-consumerd.png[]
2155 The _consumer daemon_, cmd:lttng-consumerd, is a daemon which shares
2156 ring buffers with user applications or with the LTTng kernel modules to
2157 collect trace data and send it to some location (on disk or to a
2158 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
2159 is part of LTTng-tools.
2161 You don't start a consumer daemon manually: a consumer daemon is always
2162 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
2163 <<event,event rule>>, that is, before you start tracing. When you kill
2164 its owner session daemon, the consumer daemon also exits because it is
2165 the child process of the session daemon. Command-line options of
2166 man:lttng-sessiond(8) target the consumer daemon process.
2168 There are up to two running consumer daemons per Unix user, whereas only
2169 one session daemon can run per user. This is because each process can be
2170 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
2171 and 64-bit processes, it is more efficient to have separate
2172 corresponding 32-bit and 64-bit consumer daemons. The root user is an
2173 exception: it can have up to _three_ running consumer daemons: 32-bit
2174 and 64-bit instances for its user applications, and one more
2175 reserved for collecting kernel trace data.
2183 image::plumbing-relayd.png[]
2185 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
2186 between remote session and consumer daemons, local trace files, and a
2187 remote live trace viewer. The relay daemon is part of LTTng-tools.
2189 The main purpose of the relay daemon is to implement a receiver of
2190 <<sending-trace-data-over-the-network,trace data over the network>>.
2191 This is useful when the target system doesn't have much file system
2192 space to record trace files locally.
2194 The relay daemon is also a server to which a
2195 <<lttng-live,live trace viewer>> can
2196 connect. The live trace viewer sends requests to the relay daemon to
2197 receive trace data as the target system emits events. The
2198 communication protocol is named _LTTng live_; it is used over TCP
2201 Note that you can start the relay daemon on the target system directly.
2202 This is the setup of choice when the use case is to view events as
2203 the target system emits them without the need of a remote system.
2207 == [[using-lttng]]Instrumentation
2209 There are many examples of tracing and monitoring in our everyday life:
2211 * You have access to real-time and historical weather reports and
2212 forecasts thanks to weather stations installed around the country.
2213 * You know your heart is safe thanks to an electrocardiogram.
2214 * You make sure not to drive your car too fast and to have enough fuel
2215 to reach your destination thanks to gauges visible on your dashboard.
2217 All the previous examples have something in common: they rely on
2218 **instruments**. Without the electrodes attached to the surface of your
2219 body skin, cardiac monitoring is futile.
2221 LTTng, as a tracer, is no different from those real life examples. If
2222 you're about to trace a software system or, in other words, record its
2223 history of execution, you better have **instrumentation points** in the
2224 subject you're tracing, that is, the actual software.
2226 Various ways were developed to instrument a piece of software for LTTng
2227 tracing. The most straightforward one is to manually place
2228 instrumentation points, called _tracepoints_, in the source code of the
2229 software. It is also possible to add instrumentation points dynamically
2230 in the Linux kernel <<domain,tracing domain>>.
2232 If you're only interested in tracing the Linux kernel, your
2233 instrumentation needs are probably already covered by the built-in
2234 <<lttng-modules,Linux kernel tracepoints>> of LTTng. You may also wish
2235 to trace a user application which is already instrumented for LTTng
2236 tracing. In such cases, skip this whole section and read the topics of
2237 the <<controlling-tracing,Tracing control>> section.
2239 Many methods are available to instrument a piece of software for LTTng
2242 * <<c-application,User space instrumentation for C and $$C++$$
2244 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
2245 * <<java-application,User space Java agent>>.
2246 * <<python-application,User space Python agent>>.
2247 * <<proc-lttng-logger-abi,LTTng logger>>.
2248 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
2252 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
2254 The procedure to instrument a C or $$C++$$ user application with
2255 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
2257 . <<tracepoint-provider,Create the source files of a tracepoint provider
2259 . <<probing-the-application-source-code,Add tracepoints to
2260 the source code of the application>>.
2261 . <<building-tracepoint-providers-and-user-application,Build and link
2262 a tracepoint provider package and the user application>>.
2264 If you need quick, man:printf(3)-like instrumentation, skip
2265 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
2268 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
2269 instrument a user application with `liblttng-ust`.
2272 [[tracepoint-provider]]
2273 ==== Create the source files of a tracepoint provider package
2275 A _tracepoint provider_ is a set of compiled functions which provide
2276 **tracepoints** to an application, the type of instrumentation point
2277 supported by LTTng-UST. Those functions can emit events with
2278 user-defined fields and serialize those events as event records to one
2279 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
2280 macro, which you <<probing-the-application-source-code,insert in the
2281 source code of a user application>>, calls those functions.
2283 A _tracepoint provider package_ is an object file (`.o`) or a shared
2284 library (`.so`) which contains one or more tracepoint providers.
2285 Its source files are:
2287 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
2288 * A <<tpp-source,tracepoint provider package source>> (`.c`).
2290 A tracepoint provider package is dynamically linked with `liblttng-ust`,
2291 the LTTng user space tracer, at run time.
2294 .User application linked with `liblttng-ust` and containing a tracepoint provider.
2295 image::ust-app.png[]
2297 NOTE: If you need quick, man:printf(3)-like instrumentation,
2298 skip creating and using a tracepoint provider and use
2299 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
2303 ===== Create a tracepoint provider header file template
2305 A _tracepoint provider header file_ contains the tracepoint
2306 definitions of a tracepoint provider.
2308 To create a tracepoint provider header file:
2310 . Start from this template:
2314 .Tracepoint provider header file template (`.h` file extension).
2316 #undef TRACEPOINT_PROVIDER
2317 #define TRACEPOINT_PROVIDER provider_name
2319 #undef TRACEPOINT_INCLUDE
2320 #define TRACEPOINT_INCLUDE "./tp.h"
2322 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
2325 #include <lttng/tracepoint.h>
2328 * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
2329 * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
2334 #include <lttng/tracepoint-event.h>
2340 * `provider_name` with the name of your tracepoint provider.
2341 * `"tp.h"` with the name of your tracepoint provider header file.
2343 . Below the `#include <lttng/tracepoint.h>` line, put your
2344 <<defining-tracepoints,tracepoint definitions>>.
2346 Your tracepoint provider name must be unique amongst all the possible
2347 tracepoint provider names used on the same target system. We
2348 suggest to include the name of your project or company in the name,
2349 for example, `org_lttng_my_project_tpp`.
2351 TIP: [[lttng-gen-tp]]Use the man:lttng-gen-tp(1) tool to create
2352 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
2353 write are the <<defining-tracepoints,tracepoint definitions>>.
2356 [[defining-tracepoints]]
2357 ===== Create a tracepoint definition
2359 A _tracepoint definition_ defines, for a given tracepoint:
2361 * Its **input arguments**. They are the macro parameters that the
2362 `tracepoint()` macro accepts for this particular tracepoint
2363 in the source code of the user application.
2364 * Its **output event fields**. They are the sources of event fields
2365 that form the payload of any event that the execution of the
2366 `tracepoint()` macro emits for this particular tracepoint.
2368 Create a tracepoint definition by using the
2369 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
2371 <<tpp-header,tracepoint provider header file template>>.
2373 The syntax of the `TRACEPOINT_EVENT()` macro is:
2376 .`TRACEPOINT_EVENT()` macro syntax.
2379 /* Tracepoint provider name */
2382 /* Tracepoint name */
2385 /* Input arguments */
2390 /* Output event fields */
2399 * `provider_name` with your tracepoint provider name.
2400 * `tracepoint_name` with your tracepoint name.
2401 * `arguments` with the <<tpp-def-input-args,input arguments>>.
2402 * `fields` with the <<tpp-def-output-fields,output event field>>
2405 This tracepoint emits events named `provider_name:tracepoint_name`.
2408 .Event name length limitation
2410 The concatenation of the tracepoint provider name and the
2411 tracepoint name must not exceed **254{nbsp}characters**. If it does, the
2412 instrumented application compiles and runs, but LTTng throws multiple
2413 warnings and you could experience serious issues.
2416 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
2419 .`TP_ARGS()` macro syntax.
2428 * `type` with the C type of the argument.
2429 * `arg_name` with the argument name.
2431 You can repeat `type` and `arg_name` up to 10{nbsp}times to have more
2434 .`TP_ARGS()` usage with three arguments.
2446 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
2447 tracepoint definition with no input arguments.
2449 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
2450 `ctf_*()` macros. Each `ctf_*()` macro defines one event field. See
2451 man:lttng-ust(3) for a complete description of the available `ctf_*()`
2452 macros. A `ctf_*()` macro specifies the type, size, and byte order of
2455 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
2456 C expression that the tracer evalutes at the `tracepoint()` macro site
2457 in the source code of the application. This expression provides the
2458 source of data of a field. The argument expression can include input
2459 argument names listed in the `TP_ARGS()` macro.
2461 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
2462 must be unique within a given tracepoint definition.
2464 Here's a complete tracepoint definition example:
2466 .Tracepoint definition.
2468 The following tracepoint definition defines a tracepoint which takes
2469 three input arguments and has four output event fields.
2473 #include "my-custom-structure.h"
2479 const struct my_custom_structure*, my_custom_structure,
2484 ctf_string(query_field, query)
2485 ctf_float(double, ratio_field, ratio)
2486 ctf_integer(int, recv_size, my_custom_structure->recv_size)
2487 ctf_integer(int, send_size, my_custom_structure->send_size)
2492 Refer to this tracepoint definition with the `tracepoint()` macro in
2493 the source code of your application like this:
2497 tracepoint(my_provider, my_tracepoint,
2498 my_structure, some_ratio, the_query);
2502 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2503 if they satisfy an enabled <<event,event rule>>.
2506 [[using-tracepoint-classes]]
2507 ===== Use a tracepoint class
2509 A _tracepoint class_ is a class of tracepoints which share the same
2510 output event field definitions. A _tracepoint instance_ is one
2511 instance of such a defined tracepoint class, with its own tracepoint
2514 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2515 shorthand which defines both a tracepoint class and a tracepoint
2516 instance at the same time.
2518 When you build a tracepoint provider package, the C or $$C++$$ compiler
2519 creates one serialization function for each **tracepoint class**. A
2520 serialization function is responsible for serializing the event fields
2521 of a tracepoint to a sub-buffer when tracing.
2523 For various performance reasons, when your situation requires multiple
2524 tracepoint definitions with different names, but with the same event
2525 fields, we recommend that you manually create a tracepoint class
2526 and instantiate as many tracepoint instances as needed. One positive
2527 effect of such a design, amongst other advantages, is that all
2528 tracepoint instances of the same tracepoint class reuse the same
2529 serialization function, thus reducing
2530 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2532 .Use a tracepoint class and tracepoint instances.
2534 Consider the following three tracepoint definitions:
2546 ctf_integer(int, userid, userid)
2547 ctf_integer(size_t, len, len)
2559 ctf_integer(int, userid, userid)
2560 ctf_integer(size_t, len, len)
2572 ctf_integer(int, userid, userid)
2573 ctf_integer(size_t, len, len)
2578 In this case, we create three tracepoint classes, with one implicit
2579 tracepoint instance for each of them: `get_account`, `get_settings`, and
2580 `get_transaction`. However, they all share the same event field names
2581 and types. Hence three identical, yet independent serialization
2582 functions are created when you build the tracepoint provider package.
2584 A better design choice is to define a single tracepoint class and three
2585 tracepoint instances:
2589 /* The tracepoint class */
2590 TRACEPOINT_EVENT_CLASS(
2591 /* Tracepoint provider name */
2594 /* Tracepoint class name */
2597 /* Input arguments */
2603 /* Output event fields */
2605 ctf_integer(int, userid, userid)
2606 ctf_integer(size_t, len, len)
2610 /* The tracepoint instances */
2611 TRACEPOINT_EVENT_INSTANCE(
2612 /* Tracepoint provider name */
2615 /* Tracepoint class name */
2618 /* Tracepoint name */
2621 /* Input arguments */
2627 TRACEPOINT_EVENT_INSTANCE(
2636 TRACEPOINT_EVENT_INSTANCE(
2649 [[assigning-log-levels]]
2650 ===== Assign a log level to a tracepoint definition
2652 Assign a _log level_ to a <<defining-tracepoints,tracepoint definition>>
2653 with the `TRACEPOINT_LOGLEVEL()` macro.
2655 Assigning different levels of severity to tracepoint definitions can
2656 be useful: when you <<enabling-disabling-events,create an event rule>>,
2657 you can target tracepoints having a log level as severe as a specific
2660 The concept of LTTng-UST log levels is similar to the levels found
2661 in typical logging frameworks:
2663 * In a logging framework, the log level is given by the function
2664 or method name you use at the log statement site: `debug()`,
2665 `info()`, `warn()`, `error()`, and so on.
2666 * In LTTng-UST, you statically assign the log level to a tracepoint
2667 definition; any `tracepoint()` macro invocation which refers to
2668 this definition has this log level.
2670 You must use `TRACEPOINT_LOGLEVEL()` _after_ the
2671 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2672 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2675 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2678 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2680 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2685 * `provider_name` with the tracepoint provider name.
2686 * `tracepoint_name` with the tracepoint name.
2687 * `log_level` with the log level to assign to the tracepoint
2688 definition named `tracepoint_name` in the `provider_name`
2689 tracepoint provider.
2691 See man:lttng-ust(3) for a list of available log level names.
2693 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2697 /* Tracepoint definition */
2706 ctf_integer(int, userid, userid)
2707 ctf_integer(size_t, len, len)
2711 /* Log level assignment */
2712 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2718 ===== Create a tracepoint provider package source file
2720 A _tracepoint provider package source file_ is a C source file which
2721 includes a <<tpp-header,tracepoint provider header file>> to expand its
2722 macros into event serialization and other functions.
2724 Use the following tracepoint provider package source file template:
2727 .Tracepoint provider package source file template.
2729 #define TRACEPOINT_CREATE_PROBES
2734 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2735 header file>> name. You may also include more than one tracepoint
2736 provider header file here to create a tracepoint provider package
2737 holding more than one tracepoint providers.
2740 [[probing-the-application-source-code]]
2741 ==== Add tracepoints to the source code of an application
2743 Once you <<tpp-header,create a tracepoint provider header file>>, use
2744 the `tracepoint()` macro in the source code of your application to
2745 insert the tracepoints that this header
2746 <<defining-tracepoints,defines>>.
2748 The `tracepoint()` macro takes at least two parameters: the tracepoint
2749 provider name and the tracepoint name. The corresponding tracepoint
2750 definition defines the other parameters.
2752 .`tracepoint()` usage.
2754 The following <<defining-tracepoints,tracepoint definition>> defines a
2755 tracepoint which takes two input arguments and has two output event
2759 .Tracepoint provider header file.
2761 #include "my-custom-structure.h"
2768 const char*, cmd_name
2771 ctf_string(cmd_name, cmd_name)
2772 ctf_integer(int, number_of_args, argc)
2777 Refer to this tracepoint definition with the `tracepoint()` macro in
2778 the source code of your application like this:
2781 .Application source file.
2785 int main(int argc, char* argv[])
2787 tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2793 Note how the source code of the application includes
2794 the tracepoint provider header file containing the tracepoint
2795 definitions to use, path:{tp.h}.
2798 .`tracepoint()` usage with a complex tracepoint definition.
2800 Consider this complex tracepoint definition, where multiple event
2801 fields refer to the same input arguments in their argument expression
2805 .Tracepoint provider header file.
2807 /* For `struct stat` */
2808 #include <sys/types.h>
2809 #include <sys/stat.h>
2821 ctf_integer(int, my_constant_field, 23 + 17)
2822 ctf_integer(int, my_int_arg_field, my_int_arg)
2823 ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2824 ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2825 my_str_arg[2] + my_str_arg[3])
2826 ctf_string(my_str_arg_field, my_str_arg)
2827 ctf_integer_hex(off_t, size_field, st->st_size)
2828 ctf_float(double, size_dbl_field, (double) st->st_size)
2829 ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2830 size_t, strlen(my_str_arg) / 2)
2835 Refer to this tracepoint definition with the `tracepoint()` macro in
2836 the source code of your application like this:
2839 .Application source file.
2841 #define TRACEPOINT_DEFINE
2848 stat("/etc/fstab", &s);
2849 tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2855 If you look at the event record that LTTng writes when tracing this
2856 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2857 it should look like this:
2859 .Event record fields
2861 |Field name |Field value
2862 |`my_constant_field` |40
2863 |`my_int_arg_field` |23
2864 |`my_int_arg_field2` |529
2866 |`my_str_arg_field` |`Hello, World!`
2867 |`size_field` |0x12d
2868 |`size_dbl_field` |301.0
2869 |`half_my_str_arg_field` |`Hello,`
2873 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2874 compute--they use the call stack, for example. To avoid this computation
2875 when the tracepoint is disabled, use the `tracepoint_enabled()` and
2876 `do_tracepoint()` macros.
2878 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2882 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2884 tracepoint_enabled(provider_name, tracepoint_name)
2885 do_tracepoint(provider_name, tracepoint_name, ...)
2890 * `provider_name` with the tracepoint provider name.
2891 * `tracepoint_name` with the tracepoint name.
2893 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2894 `tracepoint_name` from the provider named `provider_name` is enabled
2897 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2898 if the tracepoint is enabled. Using `tracepoint()` with
2899 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2900 the `tracepoint_enabled()` check, thus a race condition is
2901 possible in this situation:
2904 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2906 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2907 stuff = prepare_stuff();
2910 tracepoint(my_provider, my_tracepoint, stuff);
2913 If the tracepoint is enabled after the condition, then `stuff` isn't
2914 prepared: the emitted event will either contain wrong data, or the whole
2915 application could crash (segmentation fault, for example).
2917 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2918 `STAP_PROBEV()` call. If you need it, you must emit
2922 [[building-tracepoint-providers-and-user-application]]
2923 ==== Build and link a tracepoint provider package and an application
2925 Once you have one or more <<tpp-header,tracepoint provider header
2926 files>> and a <<tpp-source,tracepoint provider package source file>>,
2927 create the tracepoint provider package by compiling its source
2928 file. From here, multiple build and run scenarios are possible. The
2929 following table shows common application and library configurations
2930 along with the required command lines to achieve them.
2932 In the following diagrams, we use the following file names:
2935 Executable application.
2938 Application object file.
2941 Tracepoint provider package object file.
2944 Tracepoint provider package archive file.
2947 Tracepoint provider package shared object file.
2950 User library object file.
2953 User library shared object file.
2955 We use the following symbols in the diagrams of table below:
2958 .Symbols used in the build scenario diagrams.
2959 image::ust-sit-symbols.png[]
2961 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2962 variable in the following instructions.
2964 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2965 .Common tracepoint provider package scenarios.
2967 |Scenario |Instructions
2970 The instrumented application is statically linked with
2971 the tracepoint provider package object.
2973 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2976 include::../common/ust-sit-step-tp-o.txt[]
2978 To build the instrumented application:
2980 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2985 #define TRACEPOINT_DEFINE
2989 . Compile the application source file:
2998 . Build the application:
3003 $ gcc -o app app.o tpp.o -llttng-ust -ldl
3007 To run the instrumented application:
3009 * Start the application:
3019 The instrumented application is statically linked with the
3020 tracepoint provider package archive file.
3022 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
3025 To create the tracepoint provider package archive file:
3027 . Compile the <<tpp-source,tracepoint provider package source file>>:
3036 . Create the tracepoint provider package archive file:
3041 $ ar rcs tpp.a tpp.o
3045 To build the instrumented application:
3047 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3052 #define TRACEPOINT_DEFINE
3056 . Compile the application source file:
3065 . Build the application:
3070 $ gcc -o app app.o tpp.a -llttng-ust -ldl
3074 To run the instrumented application:
3076 * Start the application:
3086 The instrumented application is linked with the tracepoint provider
3087 package shared object.
3089 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
3092 include::../common/ust-sit-step-tp-so.txt[]
3094 To build the instrumented application:
3096 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3101 #define TRACEPOINT_DEFINE
3105 . Compile the application source file:
3114 . Build the application:
3119 $ gcc -o app app.o -ldl -L. -ltpp
3123 To run the instrumented application:
3125 * Start the application:
3135 The tracepoint provider package shared object is preloaded before the
3136 instrumented application starts.
3138 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
3141 include::../common/ust-sit-step-tp-so.txt[]
3143 To build the instrumented application:
3145 . In path:{app.c}, before including path:{tpp.h}, add the
3151 #define TRACEPOINT_DEFINE
3152 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3156 . Compile the application source file:
3165 . Build the application:
3170 $ gcc -o app app.o -ldl
3174 To run the instrumented application with tracing support:
3176 * Preload the tracepoint provider package shared object and
3177 start the application:
3182 $ LD_PRELOAD=./libtpp.so ./app
3186 To run the instrumented application without tracing support:
3188 * Start the application:
3198 The instrumented application dynamically loads the tracepoint provider
3199 package shared object.
3201 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
3204 include::../common/ust-sit-step-tp-so.txt[]
3206 To build the instrumented application:
3208 . In path:{app.c}, before including path:{tpp.h}, add the
3214 #define TRACEPOINT_DEFINE
3215 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3219 . Compile the application source file:
3228 . Build the application:
3233 $ gcc -o app app.o -ldl
3237 To run the instrumented application:
3239 * Start the application:
3249 The application is linked with the instrumented user library.
3251 The instrumented user library is statically linked with the tracepoint
3252 provider package object file.
3254 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
3257 include::../common/ust-sit-step-tp-o-fpic.txt[]
3259 To build the instrumented user library:
3261 . In path:{emon.c}, before including path:{tpp.h}, add the
3267 #define TRACEPOINT_DEFINE
3271 . Compile the user library source file:
3276 $ gcc -I. -fpic -c emon.c
3280 . Build the user library shared object:
3285 $ gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
3289 To build the application:
3291 . Compile the application source file:
3300 . Build the application:
3305 $ gcc -o app app.o -L. -lemon
3309 To run the application:
3311 * Start the application:
3321 The application is linked with the instrumented user library.
3323 The instrumented user library is linked with the tracepoint provider
3324 package shared object.
3326 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3329 include::../common/ust-sit-step-tp-so.txt[]
3331 To build the instrumented user library:
3333 . In path:{emon.c}, before including path:{tpp.h}, add the
3339 #define TRACEPOINT_DEFINE
3343 . Compile the user library source file:
3348 $ gcc -I. -fpic -c emon.c
3352 . Build the user library shared object:
3357 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3361 To build the application:
3363 . Compile the application source file:
3372 . Build the application:
3377 $ gcc -o app app.o -L. -lemon
3381 To run the application:
3383 * Start the application:
3393 The tracepoint provider package shared object is preloaded before the
3396 The application is linked with the instrumented user library.
3398 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
3401 include::../common/ust-sit-step-tp-so.txt[]
3403 To build the instrumented user library:
3405 . In path:{emon.c}, before including path:{tpp.h}, add the
3411 #define TRACEPOINT_DEFINE
3412 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3416 . Compile the user library source file:
3421 $ gcc -I. -fpic -c emon.c
3425 . Build the user library shared object:
3430 $ gcc -shared -o libemon.so emon.o -ldl
3434 To build the application:
3436 . Compile the application source file:
3445 . Build the application:
3450 $ gcc -o app app.o -L. -lemon
3454 To run the application with tracing support:
3456 * Preload the tracepoint provider package shared object and
3457 start the application:
3462 $ LD_PRELOAD=./libtpp.so ./app
3466 To run the application without tracing support:
3468 * Start the application:
3478 The application is linked with the instrumented user library.
3480 The instrumented user library dynamically loads the tracepoint provider
3481 package shared object.
3483 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3486 include::../common/ust-sit-step-tp-so.txt[]
3488 To build the instrumented user library:
3490 . In path:{emon.c}, before including path:{tpp.h}, add the
3496 #define TRACEPOINT_DEFINE
3497 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3501 . Compile the user library source file:
3506 $ gcc -I. -fpic -c emon.c
3510 . Build the user library shared object:
3515 $ gcc -shared -o libemon.so emon.o -ldl
3519 To build the application:
3521 . Compile the application source file:
3530 . Build the application:
3535 $ gcc -o app app.o -L. -lemon
3539 To run the application:
3541 * Start the application:
3551 The application dynamically loads the instrumented user library.
3553 The instrumented user library is linked with the tracepoint provider
3554 package shared object.
3556 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3559 include::../common/ust-sit-step-tp-so.txt[]
3561 To build the instrumented user library:
3563 . In path:{emon.c}, before including path:{tpp.h}, add the
3569 #define TRACEPOINT_DEFINE
3573 . Compile the user library source file:
3578 $ gcc -I. -fpic -c emon.c
3582 . Build the user library shared object:
3587 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3591 To build the application:
3593 . Compile the application source file:
3602 . Build the application:
3607 $ gcc -o app app.o -ldl -L. -lemon
3611 To run the application:
3613 * Start the application:
3623 The application dynamically loads the instrumented user library.
3625 The instrumented user library dynamically loads the tracepoint provider
3626 package shared object.
3628 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3631 include::../common/ust-sit-step-tp-so.txt[]
3633 To build the instrumented user library:
3635 . In path:{emon.c}, before including path:{tpp.h}, add the
3641 #define TRACEPOINT_DEFINE
3642 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3646 . Compile the user library source file:
3651 $ gcc -I. -fpic -c emon.c
3655 . Build the user library shared object:
3660 $ gcc -shared -o libemon.so emon.o -ldl
3664 To build the application:
3666 . Compile the application source file:
3675 . Build the application:
3680 $ gcc -o app app.o -ldl -L. -lemon
3684 To run the application:
3686 * Start the application:
3696 The tracepoint provider package shared object is preloaded before the
3699 The application dynamically loads the instrumented user library.
3701 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3704 include::../common/ust-sit-step-tp-so.txt[]
3706 To build the instrumented user library:
3708 . In path:{emon.c}, before including path:{tpp.h}, add the
3714 #define TRACEPOINT_DEFINE
3715 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3719 . Compile the user library source file:
3724 $ gcc -I. -fpic -c emon.c
3728 . Build the user library shared object:
3733 $ gcc -shared -o libemon.so emon.o -ldl
3737 To build the application:
3739 . Compile the application source file:
3748 . Build the application:
3753 $ gcc -o app app.o -L. -lemon
3757 To run the application with tracing support:
3759 * Preload the tracepoint provider package shared object and
3760 start the application:
3765 $ LD_PRELOAD=./libtpp.so ./app
3769 To run the application without tracing support:
3771 * Start the application:
3781 The application is statically linked with the tracepoint provider
3782 package object file.
3784 The application is linked with the instrumented user library.
3786 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3789 include::../common/ust-sit-step-tp-o.txt[]
3791 To build the instrumented user library:
3793 . In path:{emon.c}, before including path:{tpp.h}, add the
3799 #define TRACEPOINT_DEFINE
3803 . Compile the user library source file:
3808 $ gcc -I. -fpic -c emon.c
3812 . Build the user library shared object:
3817 $ gcc -shared -o libemon.so emon.o
3821 To build the application:
3823 . Compile the application source file:
3832 . Build the application:
3837 $ gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3841 To run the instrumented application:
3843 * Start the application:
3853 The application is statically linked with the tracepoint provider
3854 package object file.
3856 The application dynamically loads the instrumented user library.
3858 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3861 include::../common/ust-sit-step-tp-o.txt[]
3863 To build the application:
3865 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3870 #define TRACEPOINT_DEFINE
3874 . Compile the application source file:
3883 . Build the application:
3888 $ gcc -Wl,--export-dynamic -o app app.o tpp.o \
3893 The `--export-dynamic` option passed to the linker is necessary for the
3894 dynamically loaded library to ``see'' the tracepoint symbols defined in
3897 To build the instrumented user library:
3899 . Compile the user library source file:
3904 $ gcc -I. -fpic -c emon.c
3908 . Build the user library shared object:
3913 $ gcc -shared -o libemon.so emon.o
3917 To run the application:
3919 * Start the application:
3930 [[using-lttng-ust-with-daemons]]
3931 ===== Use noch:{LTTng-UST} with daemons
3933 If your instrumented application calls man:fork(2), man:clone(2),
3934 or BSD's man:rfork(2), without a following man:exec(3)-family
3935 system call, you must preload the path:{liblttng-ust-fork.so} shared
3936 object when you start the application.
3940 $ LD_PRELOAD=liblttng-ust-fork.so ./my-app
3943 If your tracepoint provider package is
3944 a shared library which you also preload, you must put both
3945 shared objects in env:LD_PRELOAD:
3949 $ LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3955 ===== Use noch:{LTTng-UST} with applications which close file descriptors that don't belong to them
3957 If your instrumented application closes one or more file descriptors
3958 which it did not open itself, you must preload the
3959 path:{liblttng-ust-fd.so} shared object when you start the application:
3963 $ LD_PRELOAD=liblttng-ust-fd.so ./my-app
3966 Typical use cases include closing all the file descriptors after
3967 man:fork(2) or man:rfork(2) and buggy applications doing
3971 [[lttng-ust-pkg-config]]
3972 ===== Use noch:{pkg-config}
3974 On some distributions, LTTng-UST ships with a
3975 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3976 metadata file. If this is your case, then use cmd:pkg-config to
3977 build an application on the command line:
3981 $ gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3985 [[instrumenting-32-bit-app-on-64-bit-system]]
3986 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3988 In order to trace a 32-bit application running on a 64-bit system,
3989 LTTng must use a dedicated 32-bit
3990 <<lttng-consumerd,consumer daemon>>.
3992 The following steps show how to build and install a 32-bit consumer
3993 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3994 build and install the 32-bit LTTng-UST libraries, and how to build and
3995 link an instrumented 32-bit application in that context.
3997 To build a 32-bit instrumented application for a 64-bit target system,
3998 assuming you have a fresh target system with no installed Userspace RCU
4001 . Download, build, and install a 32-bit version of Userspace RCU:
4006 $ cd $(mktemp -d) &&
4007 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
4008 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
4009 cd userspace-rcu-0.9.* &&
4010 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
4012 sudo make install &&
4017 . Using the package manager of your distribution, or from source,
4018 install the following 32-bit versions of the following dependencies of
4019 LTTng-tools and LTTng-UST:
4022 * https://sourceforge.net/projects/libuuid/[libuuid]
4023 * http://directory.fsf.org/wiki/Popt[popt]
4024 * http://www.xmlsoft.org/[libxml2]
4027 . Download, build, and install a 32-bit version of the latest
4028 LTTng-UST{nbsp}{revision}:
4033 $ cd $(mktemp -d) &&
4034 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.12.tar.bz2 &&
4035 tar -xf lttng-ust-latest-2.12.tar.bz2 &&
4036 cd lttng-ust-2.12.* &&
4037 ./configure --libdir=/usr/local/lib32 \
4038 CFLAGS=-m32 CXXFLAGS=-m32 \
4039 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
4041 sudo make install &&
4048 Depending on your distribution,
4049 32-bit libraries could be installed at a different location than
4050 `/usr/lib32`. For example, Debian is known to install
4051 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
4053 In this case, make sure to set `LDFLAGS` to all the
4054 relevant 32-bit library paths, for example:
4058 $ LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
4062 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
4063 the 32-bit consumer daemon:
4068 $ cd $(mktemp -d) &&
4069 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.12.tar.bz2 &&
4070 tar -xf lttng-tools-latest-2.12.tar.bz2 &&
4071 cd lttng-tools-2.12.* &&
4072 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
4073 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' \
4074 --disable-bin-lttng --disable-bin-lttng-crash \
4075 --disable-bin-lttng-relayd --disable-bin-lttng-sessiond &&
4077 cd src/bin/lttng-consumerd &&
4078 sudo make install &&
4083 . From your distribution or from source,
4084 <<installing-lttng,install>> the 64-bit versions of
4085 LTTng-UST and Userspace RCU.
4086 . Download, build, and install the 64-bit version of the
4087 latest LTTng-tools{nbsp}{revision}:
4092 $ cd $(mktemp -d) &&
4093 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.12.tar.bz2 &&
4094 tar -xf lttng-tools-latest-2.12.tar.bz2 &&
4095 cd lttng-tools-2.12.* &&
4096 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
4097 --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
4099 sudo make install &&
4104 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
4105 when linking your 32-bit application:
4108 -m32 -L/usr/lib32 -L/usr/local/lib32 \
4109 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
4112 For example, let's rebuild the quick start example in
4113 <<tracing-your-own-user-application,Trace a user application>> as an
4114 instrumented 32-bit application:
4119 $ gcc -m32 -c -I. hello-tp.c
4120 $ gcc -m32 -c hello.c
4121 $ gcc -m32 -o hello hello.o hello-tp.o \
4122 -L/usr/lib32 -L/usr/local/lib32 \
4123 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
4128 No special action is required to execute the 32-bit application and
4129 to trace it: use the command-line man:lttng(1) tool as usual.
4136 man:tracef(3) is a small LTTng-UST API designed for quick,
4137 man:printf(3)-like instrumentation without the burden of
4138 <<tracepoint-provider,creating>> and
4139 <<building-tracepoint-providers-and-user-application,building>>
4140 a tracepoint provider package.
4142 To use `tracef()` in your application:
4144 . In the C or C++ source files where you need to use `tracef()`,
4145 include `<lttng/tracef.h>`:
4150 #include <lttng/tracef.h>
4154 . In the source code of the application, use `tracef()` like you would
4162 tracef("my message: %d (%s)", my_integer, my_string);
4168 . Link your application with `liblttng-ust`:
4173 $ gcc -o app app.c -llttng-ust
4177 To trace the events that `tracef()` calls emit:
4179 * <<enabling-disabling-events,Create an event rule>> which matches the
4180 `lttng_ust_tracef:*` event name:
4185 $ lttng enable-event --userspace 'lttng_ust_tracef:*'
4190 .Limitations of `tracef()`
4192 The `tracef()` utility function was developed to make user space tracing
4193 super simple, albeit with notable disadvantages compared to
4194 <<defining-tracepoints,user-defined tracepoints>>:
4196 * All the emitted events have the same tracepoint provider and
4197 tracepoint names, respectively `lttng_ust_tracef` and `event`.
4198 * There is no static type checking.
4199 * The only event record field you actually get, named `msg`, is a string
4200 potentially containing the values you passed to `tracef()`
4201 using your own format string. This also means that you can't filter
4202 events with a custom expression at run time because there are no
4204 * Since `tracef()` uses the man:vasprintf(3) function of the
4205 C{nbsp}standard library behind the scenes to format the strings at run
4206 time, its expected performance is lower than with user-defined
4207 tracepoints, which don't require a conversion to a string.
4209 Taking this into consideration, `tracef()` is useful for some quick
4210 prototyping and debugging, but you shouldn't consider it for any
4211 permanent and serious applicative instrumentation.
4217 ==== Use `tracelog()`
4219 The man:tracelog(3) API is very similar to <<tracef,`tracef()`>>, with
4220 the difference that it accepts an additional log level parameter.
4222 The goal of `tracelog()` is to ease the migration from logging to
4225 To use `tracelog()` in your application:
4227 . In the C or C++ source files where you need to use `tracelog()`,
4228 include `<lttng/tracelog.h>`:
4233 #include <lttng/tracelog.h>
4237 . In the source code of the application, use `tracelog()` like you would
4238 use man:printf(3), except for the first parameter which is the log
4246 tracelog(TRACE_WARNING, "my message: %d (%s)",
4247 my_integer, my_string);
4253 See man:lttng-ust(3) for a list of available log level names.
4255 . Link your application with `liblttng-ust`:
4260 $ gcc -o app app.c -llttng-ust
4264 To trace the events that `tracelog()` calls emit with a log level
4265 _as severe as_ a specific log level:
4267 * <<enabling-disabling-events,Create an event rule>> which matches the
4268 `lttng_ust_tracelog:*` event name and a minimum level
4274 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4275 --loglevel=TRACE_WARNING
4279 To trace the events that `tracelog()` calls emit with a
4280 _specific log level_:
4282 * Create an event rule which matches the `lttng_ust_tracelog:*`
4283 event name and a specific log level:
4288 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4289 --loglevel-only=TRACE_INFO
4294 [[prebuilt-ust-helpers]]
4295 === Prebuilt user space tracing helpers
4297 The LTTng-UST package provides a few helpers in the form of preloadable
4298 shared objects which automatically instrument system functions and
4301 The helper shared objects are normally found in dir:{/usr/lib}. If you
4302 built LTTng-UST <<building-from-source,from source>>, they are probably
4303 located in dir:{/usr/local/lib}.
4305 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
4308 path:{liblttng-ust-libc-wrapper.so}::
4309 path:{liblttng-ust-pthread-wrapper.so}::
4310 <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
4311 memory and POSIX threads function tracing>>.
4313 path:{liblttng-ust-cyg-profile.so}::
4314 path:{liblttng-ust-cyg-profile-fast.so}::
4315 <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
4317 path:{liblttng-ust-dl.so}::
4318 <<liblttng-ust-dl,Dynamic linker tracing>>.
4320 To use a user space tracing helper with any user application:
4322 * Preload the helper shared object when you start the application:
4327 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
4331 You can preload more than one helper:
4336 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
4342 [[liblttng-ust-libc-pthread-wrapper]]
4343 ==== Instrument C standard library memory and POSIX threads functions
4345 The path:{liblttng-ust-libc-wrapper.so} and
4346 path:{liblttng-ust-pthread-wrapper.so} helpers
4347 add instrumentation to some C standard library and POSIX
4351 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
4353 |TP provider name |TP name |Instrumented function
4355 .6+|`lttng_ust_libc` |`malloc` |man:malloc(3)
4356 |`calloc` |man:calloc(3)
4357 |`realloc` |man:realloc(3)
4358 |`free` |man:free(3)
4359 |`memalign` |man:memalign(3)
4360 |`posix_memalign` |man:posix_memalign(3)
4364 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
4366 |TP provider name |TP name |Instrumented function
4368 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
4369 |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
4370 |`pthread_mutex_trylock` |man:pthread_mutex_trylock(3p)
4371 |`pthread_mutex_unlock` |man:pthread_mutex_unlock(3p)
4374 When you preload the shared object, it replaces the functions listed
4375 in the previous tables by wrappers which contain tracepoints and call
4376 the replaced functions.
4379 [[liblttng-ust-cyg-profile]]
4380 ==== Instrument function entry and exit
4382 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
4383 to the entry and exit points of functions.
4385 man:gcc(1) and man:clang(1) have an option named
4386 https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html[`-finstrument-functions`]
4387 which generates instrumentation calls for entry and exit to functions.
4388 The LTTng-UST function tracing helpers,
4389 path:{liblttng-ust-cyg-profile.so} and
4390 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
4391 to add tracepoints to the two generated functions (which contain
4392 `cyg_profile` in their names, hence the name of the helper).
4394 To use the LTTng-UST function tracing helper, the source files to
4395 instrument must be built using the `-finstrument-functions` compiler
4398 There are two versions of the LTTng-UST function tracing helper:
4400 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
4401 that you should only use when it can be _guaranteed_ that the
4402 complete event stream is recorded without any lost event record.
4403 Any kind of duplicate information is left out.
4405 Assuming no event record is lost, having only the function addresses on
4406 entry is enough to create a call graph, since an event record always
4407 contains the ID of the CPU that generated it.
4409 Use a tool like man:addr2line(1) to convert function addresses back to
4410 source file names and line numbers.
4412 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
4413 which also works in use cases where event records might get discarded or
4414 not recorded from application startup.
4415 In these cases, the trace analyzer needs more information to be
4416 able to reconstruct the program flow.
4418 See man:lttng-ust-cyg-profile(3) to learn more about the instrumentation
4419 points of this helper.
4421 All the tracepoints that this helper provides have the
4422 log level `TRACE_DEBUG_FUNCTION` (see man:lttng-ust(3)).
4424 TIP: It's sometimes a good idea to limit the number of source files that
4425 you compile with the `-finstrument-functions` option to prevent LTTng
4426 from writing an excessive amount of trace data at run time. When using
4428 `-finstrument-functions-exclude-function-list` option to avoid
4429 instrument entries and exits of specific function names.
4434 ==== Instrument the dynamic linker
4436 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4437 man:dlopen(3) and man:dlclose(3) function calls.
4439 See man:lttng-ust-dl(3) to learn more about the instrumentation points
4444 [[java-application]]
4445 === User space Java agent
4447 You can instrument any Java application which uses one of the following
4450 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4451 (JUL) core logging facilities.
4452 * http://logging.apache.org/log4j/1.2/[**Apache log4j{nbsp}1.2**], since
4453 LTTng{nbsp}2.6. Note that Apache Log4j{nbsp}2 isn't supported.
4456 .LTTng-UST Java agent imported by a Java application.
4457 image::java-app.png[]
4459 Note that the methods described below are new in LTTng{nbsp}2.8.
4460 Previous LTTng versions use another technique.
4462 NOTE: We use http://openjdk.java.net/[OpenJDK]{nbsp}8 for development
4463 and https://ci.lttng.org/[continuous integration], thus this version is
4464 directly supported. However, the LTTng-UST Java agent is also tested
4465 with OpenJDK{nbsp}7.
4470 ==== Use the LTTng-UST Java agent for `java.util.logging`
4472 To use the LTTng-UST Java agent in a Java application which uses
4473 `java.util.logging` (JUL):
4475 . In the source code of the Java application, import the LTTng-UST log
4476 handler package for `java.util.logging`:
4481 import org.lttng.ust.agent.jul.LttngLogHandler;
4485 . Create an LTTng-UST JUL log handler:
4490 Handler lttngUstLogHandler = new LttngLogHandler();
4494 . Add this handler to the JUL loggers which should emit LTTng events:
4499 Logger myLogger = Logger.getLogger("some-logger");
4501 myLogger.addHandler(lttngUstLogHandler);
4505 . Use `java.util.logging` log statements and configuration as usual.
4506 The loggers with an attached LTTng-UST log handler can emit
4509 . Before exiting the application, remove the LTTng-UST log handler from
4510 the loggers attached to it and call its `close()` method:
4515 myLogger.removeHandler(lttngUstLogHandler);
4516 lttngUstLogHandler.close();
4520 This isn't strictly necessary, but it is recommended for a clean
4521 disposal of the resources of the handler.
4523 . Include the common and JUL-specific JAR files of the LTTng-UST Java agent,
4524 path:{lttng-ust-agent-common.jar} and path:{lttng-ust-agent-jul.jar},
4526 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4527 path] when you build the Java application.
4529 The JAR files are typically located in dir:{/usr/share/java}.
4531 IMPORTANT: The LTTng-UST Java agent must be
4532 <<installing-lttng,installed>> for the logging framework your
4535 .Use the LTTng-UST Java agent for `java.util.logging`.
4540 import java.io.IOException;
4541 import java.util.logging.Handler;
4542 import java.util.logging.Logger;
4543 import org.lttng.ust.agent.jul.LttngLogHandler;
4547 private static final int answer = 42;
4549 public static void main(String[] argv) throws Exception
4552 Logger logger = Logger.getLogger("jello");
4554 // Create an LTTng-UST log handler
4555 Handler lttngUstLogHandler = new LttngLogHandler();
4557 // Add the LTTng-UST log handler to our logger
4558 logger.addHandler(lttngUstLogHandler);
4561 logger.info("some info");
4562 logger.warning("some warning");
4564 logger.finer("finer information; the answer is " + answer);
4566 logger.severe("error!");
4568 // Not mandatory, but cleaner
4569 logger.removeHandler(lttngUstLogHandler);
4570 lttngUstLogHandler.close();
4579 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4582 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4583 <<enabling-disabling-events,create an event rule>> matching the
4584 `jello` JUL logger, and <<basic-tracing-session-control,start tracing>>:
4589 $ lttng enable-event --jul jello
4593 Run the compiled class:
4597 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4600 <<basic-tracing-session-control,Stop tracing>> and inspect the
4610 In the resulting trace, an <<event,event record>> generated by a Java
4611 application using `java.util.logging` is named `lttng_jul:event` and
4612 has the following fields:
4621 Name of the class in which the log statement was executed.
4624 Name of the method in which the log statement was executed.
4627 Logging time (timestamp in milliseconds).
4630 Log level integer value.
4633 ID of the thread in which the log statement was executed.
4635 Use the opt:lttng-enable-event(1):--loglevel or
4636 opt:lttng-enable-event(1):--loglevel-only option of the
4637 man:lttng-enable-event(1) command to target a range of JUL log levels
4638 or a specific JUL log level.
4643 ==== Use the LTTng-UST Java agent for Apache log4j
4645 To use the LTTng-UST Java agent in a Java application which uses
4646 Apache log4j{nbsp}1.2:
4648 . In the source code of the Java application, import the LTTng-UST log
4649 appender package for Apache log4j:
4654 import org.lttng.ust.agent.log4j.LttngLogAppender;
4658 . Create an LTTng-UST log4j log appender:
4663 Appender lttngUstLogAppender = new LttngLogAppender();
4667 . Add this appender to the log4j loggers which should emit LTTng events:
4672 Logger myLogger = Logger.getLogger("some-logger");
4674 myLogger.addAppender(lttngUstLogAppender);
4678 . Use Apache log4j log statements and configuration as usual. The
4679 loggers with an attached LTTng-UST log appender can emit LTTng events.
4681 . Before exiting the application, remove the LTTng-UST log appender from
4682 the loggers attached to it and call its `close()` method:
4687 myLogger.removeAppender(lttngUstLogAppender);
4688 lttngUstLogAppender.close();
4692 This isn't strictly necessary, but it is recommended for a clean
4693 disposal of the resources of the appender.
4695 . Include the common and log4j-specific JAR
4696 files of the LTTng-UST Java agent, path:{lttng-ust-agent-common.jar} and
4697 path:{lttng-ust-agent-log4j.jar}, in the
4698 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4699 path] when you build the Java application.
4701 The JAR files are typically located in dir:{/usr/share/java}.
4703 IMPORTANT: The LTTng-UST Java agent must be
4704 <<installing-lttng,installed>> for the logging framework your
4707 .Use the LTTng-UST Java agent for Apache log4j.
4712 import org.apache.log4j.Appender;
4713 import org.apache.log4j.Logger;
4714 import org.lttng.ust.agent.log4j.LttngLogAppender;
4718 private static final int answer = 42;
4720 public static void main(String[] argv) throws Exception
4723 Logger logger = Logger.getLogger("jello");
4725 // Create an LTTng-UST log appender
4726 Appender lttngUstLogAppender = new LttngLogAppender();
4728 // Add the LTTng-UST log appender to our logger
4729 logger.addAppender(lttngUstLogAppender);
4732 logger.info("some info");
4733 logger.warn("some warning");
4735 logger.debug("debug information; the answer is " + answer);
4737 logger.fatal("error!");
4739 // Not mandatory, but cleaner
4740 logger.removeAppender(lttngUstLogAppender);
4741 lttngUstLogAppender.close();
4747 Build this example (`$LOG4JPATH` is the path to the Apache log4j JAR
4752 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH Test.java
4755 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4756 <<enabling-disabling-events,create an event rule>> matching the
4757 `jello` log4j logger, and <<basic-tracing-session-control,start tracing>>:
4762 $ lttng enable-event --log4j jello
4766 Run the compiled class:
4770 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH:. Test
4773 <<basic-tracing-session-control,Stop tracing>> and inspect the
4783 In the resulting trace, an <<event,event record>> generated by a Java
4784 application using log4j is named `lttng_log4j:event` and
4785 has the following fields:
4794 Name of the class in which the log statement was executed.
4797 Name of the method in which the log statement was executed.
4800 Name of the file in which the executed log statement is located.
4803 Line number at which the log statement was executed.
4809 Log level integer value.
4812 Name of the Java thread in which the log statement was executed.
4814 Use the opt:lttng-enable-event(1):--loglevel or
4815 opt:lttng-enable-event(1):--loglevel-only option of the
4816 man:lttng-enable-event(1) command to target a range of Apache log4j
4817 log levels or a specific log4j log level.
4821 [[java-application-context]]
4822 ==== Provide application-specific context fields in a Java application
4824 A Java application-specific context field is a piece of state provided
4825 by the application which <<adding-context,you can add>>, using the
4826 man:lttng-add-context(1) command, to each <<event,event record>>
4827 produced by the log statements of this application.
4829 For example, a given object might have a current request ID variable.
4830 You can create a context information retriever for this object and
4831 assign a name to this current request ID. You can then, using the
4832 man:lttng-add-context(1) command, add this context field by name to
4833 the JUL or log4j <<channel,channel>>.
4835 To provide application-specific context fields in a Java application:
4837 . In the source code of the Java application, import the LTTng-UST
4838 Java agent context classes and interfaces:
4843 import org.lttng.ust.agent.context.ContextInfoManager;
4844 import org.lttng.ust.agent.context.IContextInfoRetriever;
4848 . Create a context information retriever class, that is, a class which
4849 implements the `IContextInfoRetriever` interface:
4854 class MyContextInfoRetriever implements IContextInfoRetriever
4857 public Object retrieveContextInfo(String key)
4859 if (key.equals("intCtx")) {
4861 } else if (key.equals("strContext")) {
4862 return "context value!";
4871 This `retrieveContextInfo()` method is the only member of the
4872 `IContextInfoRetriever` interface. Its role is to return the current
4873 value of a state by name to create a context field. The names of the
4874 context fields and which state variables they return depends on your
4877 All primitive types and objects are supported as context fields.
4878 When `retrieveContextInfo()` returns an object, the context field
4879 serializer calls its `toString()` method to add a string field to
4880 event records. The method can also return `null`, which means that
4881 no context field is available for the required name.
4883 . Register an instance of your context information retriever class to
4884 the context information manager singleton:
4889 IContextInfoRetriever cir = new MyContextInfoRetriever();
4890 ContextInfoManager cim = ContextInfoManager.getInstance();
4891 cim.registerContextInfoRetriever("retrieverName", cir);
4895 . Before exiting the application, remove your context information
4896 retriever from the context information manager singleton:
4901 ContextInfoManager cim = ContextInfoManager.getInstance();
4902 cim.unregisterContextInfoRetriever("retrieverName");
4906 This isn't strictly necessary, but it is recommended for a clean
4907 disposal of some resources of the manager.
4909 . Build your Java application with LTTng-UST Java agent support as
4910 usual, following the procedure for either the <<jul,JUL>> or
4911 <<log4j,Apache log4j>> framework.
4914 .Provide application-specific context fields in a Java application.
4919 import java.util.logging.Handler;
4920 import java.util.logging.Logger;
4921 import org.lttng.ust.agent.jul.LttngLogHandler;
4922 import org.lttng.ust.agent.context.ContextInfoManager;
4923 import org.lttng.ust.agent.context.IContextInfoRetriever;
4927 // Our context information retriever class
4928 private static class MyContextInfoRetriever
4929 implements IContextInfoRetriever
4932 public Object retrieveContextInfo(String key) {
4933 if (key.equals("intCtx")) {
4935 } else if (key.equals("strContext")) {
4936 return "context value!";
4943 private static final int answer = 42;
4945 public static void main(String args[]) throws Exception
4947 // Get the context information manager instance
4948 ContextInfoManager cim = ContextInfoManager.getInstance();
4950 // Create and register our context information retriever
4951 IContextInfoRetriever cir = new MyContextInfoRetriever();
4952 cim.registerContextInfoRetriever("myRetriever", cir);
4955 Logger logger = Logger.getLogger("jello");
4957 // Create an LTTng-UST log handler
4958 Handler lttngUstLogHandler = new LttngLogHandler();
4960 // Add the LTTng-UST log handler to our logger
4961 logger.addHandler(lttngUstLogHandler);
4964 logger.info("some info");
4965 logger.warning("some warning");
4967 logger.finer("finer information; the answer is " + answer);
4969 logger.severe("error!");
4971 // Not mandatory, but cleaner
4972 logger.removeHandler(lttngUstLogHandler);
4973 lttngUstLogHandler.close();
4974 cim.unregisterContextInfoRetriever("myRetriever");
4983 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4986 <<creating-destroying-tracing-sessions,Create a tracing session>>
4987 and <<enabling-disabling-events,create an event rule>> matching the
4993 $ lttng enable-event --jul jello
4996 <<adding-context,Add the application-specific context fields>> to the
5001 $ lttng add-context --jul --type='$app.myRetriever:intCtx'
5002 $ lttng add-context --jul --type='$app.myRetriever:strContext'
5005 <<basic-tracing-session-control,Start tracing>>:
5012 Run the compiled class:
5016 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
5019 <<basic-tracing-session-control,Stop tracing>> and inspect the
5031 [[python-application]]
5032 === User space Python agent
5034 You can instrument a Python{nbsp}2 or Python{nbsp}3 application which
5036 https://docs.python.org/3/library/logging.html[`logging`] package.
5038 Each log statement emits an LTTng event once the
5039 application module imports the
5040 <<lttng-ust-agents,LTTng-UST Python agent>> package.
5043 .A Python application importing the LTTng-UST Python agent.
5044 image::python-app.png[]
5046 To use the LTTng-UST Python agent:
5048 . In the source code of the Python application, import the LTTng-UST
5058 The LTTng-UST Python agent automatically adds its logging handler to the
5059 root logger at import time.
5061 Any log statement that the application executes before this import does
5062 not emit an LTTng event.
5064 IMPORTANT: The LTTng-UST Python agent must be
5065 <<installing-lttng,installed>>.
5067 . Use log statements and logging configuration as usual.
5068 Since the LTTng-UST Python agent adds a handler to the _root_
5069 logger, you can trace any log statement from any logger.
5071 .Use the LTTng-UST Python agent.
5082 logging.basicConfig()
5083 logger = logging.getLogger('my-logger')
5086 logger.debug('debug message')
5087 logger.info('info message')
5088 logger.warn('warn message')
5089 logger.error('error message')
5090 logger.critical('critical message')
5094 if __name__ == '__main__':
5098 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
5099 logging handler which prints to the standard error stream, isn't
5100 strictly required for LTTng-UST tracing to work, but in versions of
5101 Python preceding{nbsp}3.2, you could see a warning message which indicates
5102 that no handler exists for the logger `my-logger`.
5104 <<creating-destroying-tracing-sessions,Create a tracing session>>,
5105 <<enabling-disabling-events,create an event rule>> matching the
5106 `my-logger` Python logger, and <<basic-tracing-session-control,start
5112 $ lttng enable-event --python my-logger
5116 Run the Python script:
5123 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
5133 In the resulting trace, an <<event,event record>> generated by a Python
5134 application is named `lttng_python:event` and has the following fields:
5137 Logging time (string).
5146 Name of the function in which the log statement was executed.
5149 Line number at which the log statement was executed.
5152 Log level integer value.
5155 ID of the Python thread in which the log statement was executed.
5158 Name of the Python thread in which the log statement was executed.
5160 Use the opt:lttng-enable-event(1):--loglevel or
5161 opt:lttng-enable-event(1):--loglevel-only option of the
5162 man:lttng-enable-event(1) command to target a range of Python log levels
5163 or a specific Python log level.
5165 When an application imports the LTTng-UST Python agent, the agent tries
5166 to register to a <<lttng-sessiond,session daemon>>. Note that you must
5167 <<start-sessiond,start the session daemon>> _before_ you run the Python
5168 application. If a session daemon is found, the agent tries to register
5169 to it during five seconds, after which the application continues
5170 without LTTng tracing support. Override this timeout value with
5171 the env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
5174 If the session daemon stops while a Python application with an imported
5175 LTTng-UST Python agent runs, the agent retries to connect and to
5176 register to a session daemon every three seconds. Override this
5177 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
5182 [[proc-lttng-logger-abi]]
5185 The `lttng-tracer` Linux kernel module, part of
5186 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger files
5187 path:{/proc/lttng-logger} and path:{/dev/lttng-logger} (since
5188 LTTng{nbsp}2.11) when it's loaded. Any application can write text data
5189 to any of those files to emit an LTTng event.
5192 .An application writes to the LTTng logger file to emit an LTTng event.
5193 image::lttng-logger.png[]
5195 The LTTng logger is the quickest method--not the most efficient,
5196 however--to add instrumentation to an application. It is designed
5197 mostly to instrument shell scripts:
5201 $ echo "Some message, some $variable" > /dev/lttng-logger
5204 Any event that the LTTng logger emits is named `lttng_logger` and
5205 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
5206 other instrumentation points in the kernel tracing domain, **any Unix
5207 user** can <<enabling-disabling-events,create an event rule>> which
5208 matches its event name, not only the root user or users in the
5209 <<tracing-group,tracing group>>.
5211 To use the LTTng logger:
5213 * From any application, write text data to the path:{/dev/lttng-logger}
5216 The `msg` field of `lttng_logger` event records contains the
5219 NOTE: The maximum message length of an LTTng logger event is
5220 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
5221 than one event to contain the remaining data.
5223 You shouldn't use the LTTng logger to trace a user application which
5224 can be instrumented in a more efficient way, namely:
5226 * <<c-application,C and $$C++$$ applications>>.
5227 * <<java-application,Java applications>>.
5228 * <<python-application,Python applications>>.
5230 .Use the LTTng logger.
5235 echo 'Hello, World!' > /dev/lttng-logger
5237 df --human-readable --print-type / > /dev/lttng-logger
5240 <<creating-destroying-tracing-sessions,Create a tracing session>>,
5241 <<enabling-disabling-events,create an event rule>> matching the
5242 `lttng_logger` Linux kernel tracepoint, and
5243 <<basic-tracing-session-control,start tracing>>:
5248 $ lttng enable-event --kernel lttng_logger
5252 Run the Bash script:
5259 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
5270 [[instrumenting-linux-kernel]]
5271 === LTTng kernel tracepoints
5273 NOTE: This section shows how to _add_ instrumentation points to the
5274 Linux kernel. The subsystems of the kernel are already thoroughly
5275 instrumented at strategic places for LTTng when you
5276 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
5280 There are two methods to instrument the Linux kernel:
5282 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
5283 tracepoint which uses the `TRACE_EVENT()` API.
5285 Choose this if you want to instrumentation a Linux kernel tree with an
5286 instrumentation point compatible with ftrace, perf, and SystemTap.
5288 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
5289 instrument an out-of-tree kernel module.
5291 Choose this if you don't need ftrace, perf, or SystemTap support.
5295 [[linux-add-lttng-layer]]
5296 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
5298 This section shows how to add an LTTng layer to existing ftrace
5299 instrumentation using the `TRACE_EVENT()` API.
5301 This section doesn't document the `TRACE_EVENT()` macro. Read the
5302 following articles to learn more about this API:
5304 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part{nbsp}1)]
5305 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part{nbsp}2)]
5306 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part{nbsp}3)]
5308 The following procedure assumes that your ftrace tracepoints are
5309 correctly defined in their own header and that they are created in
5310 one source file using the `CREATE_TRACE_POINTS` definition.
5312 To add an LTTng layer over an existing ftrace tracepoint:
5314 . Make sure the following kernel configuration options are
5320 * `CONFIG_HIGH_RES_TIMERS`
5321 * `CONFIG_TRACEPOINTS`
5324 . Build the Linux source tree with your custom ftrace tracepoints.
5325 . Boot the resulting Linux image on your target system.
5327 Confirm that the tracepoints exist by looking for their names in the
5328 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
5329 is your subsystem name.
5331 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
5336 $ cd $(mktemp -d) &&
5337 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.12.tar.bz2 &&
5338 tar -xf lttng-modules-latest-2.12.tar.bz2 &&
5339 cd lttng-modules-2.12.*
5343 . In dir:{instrumentation/events/lttng-module}, relative to the root
5344 of the LTTng-modules source tree, create a header file named
5345 +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
5346 LTTng-modules tracepoint definitions using the LTTng-modules
5349 Start with this template:
5353 .path:{instrumentation/events/lttng-module/my_subsys.h}
5356 #define TRACE_SYSTEM my_subsys
5358 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
5359 #define _LTTNG_MY_SUBSYS_H
5361 #include "../../../probes/lttng-tracepoint-event.h"
5362 #include <linux/tracepoint.h>
5364 LTTNG_TRACEPOINT_EVENT(
5366 * Format is identical to the TRACE_EVENT() version for the three
5367 * following macro parameters:
5370 TP_PROTO(int my_int, const char *my_string),
5371 TP_ARGS(my_int, my_string),
5373 /* LTTng-modules specific macros */
5375 ctf_integer(int, my_int_field, my_int)
5376 ctf_string(my_bar_field, my_bar)
5380 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
5382 #include "../../../probes/define_trace.h"
5386 The entries in the `TP_FIELDS()` section are the list of fields for the
5387 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
5388 the `TRACE_EVENT()` ftrace macro.
5390 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
5391 complete description of the available `ctf_*()` macros.
5393 . Create the kernel module C{nbsp}source file of the LTTng-modules
5394 probe, +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
5399 .path:{probes/lttng-probe-my-subsys.c}
5401 #include <linux/module.h>
5402 #include "../lttng-tracer.h"
5405 * Build-time verification of mismatch between mainline
5406 * TRACE_EVENT() arguments and the LTTng-modules adaptation
5407 * layer LTTNG_TRACEPOINT_EVENT() arguments.
5409 #include <trace/events/my_subsys.h>
5411 /* Create LTTng tracepoint probes */
5412 #define LTTNG_PACKAGE_BUILD
5413 #define CREATE_TRACE_POINTS
5414 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
5416 #include "../instrumentation/events/lttng-module/my_subsys.h"
5418 MODULE_LICENSE("GPL and additional rights");
5419 MODULE_AUTHOR("Your name <your-email>");
5420 MODULE_DESCRIPTION("LTTng my_subsys probes");
5421 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
5422 __stringify(LTTNG_MODULES_MINOR_VERSION) "."
5423 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
5424 LTTNG_MODULES_EXTRAVERSION);
5428 . Edit path:{probes/KBuild} and add your new kernel module object
5429 next to the existing ones:
5433 .path:{probes/KBuild}
5437 obj-m += lttng-probe-module.o
5438 obj-m += lttng-probe-power.o
5440 obj-m += lttng-probe-my-subsys.o
5446 . Build and install the LTTng kernel modules:
5451 $ make KERNELDIR=/path/to/linux
5452 # make modules_install && depmod -a
5456 Replace `/path/to/linux` with the path to the Linux source tree where
5457 you defined and used tracepoints with the `TRACE_EVENT()` ftrace macro.
5459 Note that you can also use the
5460 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
5461 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
5462 C code that need to be executed before the event fields are recorded.
5464 The best way to learn how to use the previous LTTng-modules macros is to
5465 inspect the existing LTTng-modules tracepoint definitions in the
5466 dir:{instrumentation/events/lttng-module} header files. Compare them
5467 with the Linux kernel mainline versions in the
5468 dir:{include/trace/events} directory of the Linux source tree.
5472 [[lttng-tracepoint-event-code]]
5473 ===== Use custom C code to access the data for tracepoint fields
5475 Although we recommended to always use the
5476 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
5477 the arguments and fields of an LTTng-modules tracepoint when possible,
5478 sometimes you need a more complex process to access the data that the
5479 tracer records as event record fields. In other words, you need local
5480 variables and multiple C{nbsp}statements instead of simple
5481 argument-based expressions that you pass to the
5482 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
5484 Use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
5485 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
5486 a block of C{nbsp}code to be executed before LTTng records the fields.
5487 The structure of this macro is:
5490 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
5492 LTTNG_TRACEPOINT_EVENT_CODE(
5494 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5495 * version for the following three macro parameters:
5498 TP_PROTO(int my_int, const char *my_string),
5499 TP_ARGS(my_int, my_string),
5501 /* Declarations of custom local variables */
5504 unsigned long b = 0;
5505 const char *name = "(undefined)";
5506 struct my_struct *my_struct;
5510 * Custom code which uses both tracepoint arguments
5511 * (in TP_ARGS()) and local variables (in TP_locvar()).
5513 * Local variables are actually members of a structure pointed
5514 * to by the special variable tp_locvar.
5518 tp_locvar->a = my_int + 17;
5519 tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
5520 tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
5521 tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
5522 put_my_struct(tp_locvar->my_struct);
5531 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5532 * version for this, except that tp_locvar members can be
5533 * used in the argument expression parameters of
5534 * the ctf_*() macros.
5537 ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
5538 ctf_integer(int, my_struct_a, tp_locvar->a)
5539 ctf_string(my_string_field, my_string)
5540 ctf_string(my_struct_name, tp_locvar->name)
5545 IMPORTANT: The C code defined in `TP_code()` must not have any side
5546 effects when executed. In particular, the code must not allocate
5547 memory or get resources without deallocating this memory or putting
5548 those resources afterwards.
5551 [[instrumenting-linux-kernel-tracing]]
5552 ==== Load and unload a custom probe kernel module
5554 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
5555 kernel module>> in the kernel before it can emit LTTng events.
5557 To load the default probe kernel modules and a custom probe kernel
5560 * Use the opt:lttng-sessiond(8):--extra-kmod-probes option to give extra
5561 probe modules to load when starting a root <<lttng-sessiond,session
5565 .Load the `my_subsys`, `usb`, and the default probe modules.
5569 # lttng-sessiond --extra-kmod-probes=my_subsys,usb
5574 You only need to pass the subsystem name, not the whole kernel module
5577 To load _only_ a given custom probe kernel module:
5579 * Use the opt:lttng-sessiond(8):--kmod-probes option to give the probe
5580 modules to load when starting a root session daemon:
5583 .Load only the `my_subsys` and `usb` probe modules.
5587 # lttng-sessiond --kmod-probes=my_subsys,usb
5592 To confirm that a probe module is loaded:
5599 $ lsmod | grep lttng_probe_usb
5603 To unload the loaded probe modules:
5605 * Kill the session daemon with `SIGTERM`:
5610 # pkill lttng-sessiond
5614 You can also use the man:modprobe(8) `--remove` option if the session
5615 daemon terminates abnormally.
5618 [[controlling-tracing]]
5621 Once an application or a Linux kernel is
5622 <<instrumenting,instrumented>> for LTTng tracing,
5625 This section is divided in topics on how to use the various
5626 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
5627 command-line tool>>, to _control_ the LTTng daemons and tracers.
5629 NOTE: In the following subsections, we refer to an man:lttng(1) command
5630 using its man page name. For example, instead of _Run the `create`
5631 command to..._, we use _Run the man:lttng-create(1) command to..._.
5635 === Start a session daemon
5637 In some situations, you need to run a <<lttng-sessiond,session daemon>>
5638 (man:lttng-sessiond(8)) _before_ you can use the man:lttng(1)
5641 You will see the following error when you run a command while no session
5645 Error: No session daemon is available
5648 The only command that automatically runs a session daemon is
5649 man:lttng-create(1), which you use to
5650 <<creating-destroying-tracing-sessions,create a tracing session>>. While
5651 this is most of the time the first operation that you do, sometimes it's
5652 not. Some examples are:
5654 * <<list-instrumentation-points,List the available instrumentation points>>.
5655 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
5657 [[tracing-group]] Each Unix user must have its own running session
5658 daemon to trace user applications. The session daemon that the root user
5659 starts is the only one allowed to control the LTTng kernel tracer. Users
5660 that are part of the _tracing group_ can control the root session
5661 daemon. The default tracing group name is `tracing`; set it to something
5662 else with the opt:lttng-sessiond(8):--group option when you start the
5663 root session daemon.
5665 To start a user session daemon:
5667 * Run man:lttng-sessiond(8):
5672 $ lttng-sessiond --daemonize
5676 To start the root session daemon:
5678 * Run man:lttng-sessiond(8) as the root user:
5683 # lttng-sessiond --daemonize
5687 In both cases, remove the opt:lttng-sessiond(8):--daemonize option to
5688 start the session daemon in foreground.
5690 To stop a session daemon, use man:kill(1) on its process ID (standard
5693 Note that some Linux distributions could manage the LTTng session daemon
5694 as a service. In this case, you should use the service manager to
5695 start, restart, and stop session daemons.
5698 [[creating-destroying-tracing-sessions]]
5699 === Create and destroy a tracing session
5701 Almost all the LTTng control operations happen in the scope of
5702 a <<tracing-session,tracing session>>, which is the dialogue between the
5703 <<lttng-sessiond,session daemon>> and you.
5705 To create a tracing session with a generated name:
5707 * Use the man:lttng-create(1) command:
5716 The name of the created tracing session is `auto` followed by the
5719 To create a tracing session with a specific name:
5721 * Use the optional argument of the man:lttng-create(1) command:
5726 $ lttng create my-session
5730 Replace `my-session` with the specific tracing session name.
5732 LTTng appends the creation date to the name of the created tracing
5735 LTTng writes the traces of a tracing session in
5736 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
5737 name of the tracing session. Note that the env:LTTNG_HOME environment
5738 variable defaults to `$HOME` if not set.
5740 To output LTTng traces to a non-default location:
5742 * Use the opt:lttng-create(1):--output option of the man:lttng-create(1) command:
5747 $ lttng create my-session --output=/tmp/some-directory
5751 You may create as many tracing sessions as you wish.
5753 To list all the existing tracing sessions for your Unix user:
5755 * Use the man:lttng-list(1) command:
5764 [[cur-tracing-session]]When you create a tracing session, it is set as
5765 the _current tracing session_. The following man:lttng(1) commands
5766 operate on the current tracing session when you don't specify one:
5768 [role="list-3-cols"]
5769 * man:lttng-add-context(1)
5770 * man:lttng-clear(1)
5771 * man:lttng-destroy(1)
5772 * man:lttng-disable-channel(1)
5773 * man:lttng-disable-event(1)
5774 * man:lttng-disable-rotation(1)
5775 * man:lttng-enable-channel(1)
5776 * man:lttng-enable-event(1)
5777 * man:lttng-enable-rotation(1)
5779 * man:lttng-regenerate(1)
5780 * man:lttng-rotate(1)
5782 * man:lttng-snapshot(1)
5783 * man:lttng-start(1)
5784 * man:lttng-status(1)
5786 * man:lttng-track(1)
5787 * man:lttng-untrack(1)
5790 To change the current tracing session:
5792 * Use the man:lttng-set-session(1) command:
5797 $ lttng set-session new-session
5801 Replace `new-session` by the name of the new current tracing session.
5803 When you're done tracing in a given tracing session, destroy it. This
5804 operation frees the resources taken by the tracing session to destroy;
5805 it doesn't destroy the trace data that LTTng wrote for this tracing
5806 session (see <<clear,Clear a tracing session>> for one way to do this).
5808 To destroy the current tracing session:
5810 * Use the man:lttng-destroy(1) command:
5819 The man:lttng-destroy(1) command also runs the man:lttng-stop(1)
5820 command implicitly (see <<basic-tracing-session-control,Start and stop a
5821 tracing session>>). You need to stop tracing to make LTTng flush the
5822 remaining trace data and make the trace readable.
5825 [[list-instrumentation-points]]
5826 === List the available instrumentation points
5828 The <<lttng-sessiond,session daemon>> can query the running instrumented
5829 user applications and the Linux kernel to get a list of available
5830 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
5831 they are tracepoints and system calls. For the user space tracing
5832 domain, they are tracepoints. For the other tracing domains, they are
5835 To list the available instrumentation points:
5837 * Use the man:lttng-list(1) command with the option of the requested
5838 tracing domain amongst:
5841 opt:lttng-list(1):--kernel::
5842 Linux kernel tracepoints (your Unix user must be a root user, or it
5843 must be a member of the <<tracing-group,tracing group>>).
5845 opt:lttng-list(1):--kernel with opt:lttng-list(1):--syscall::
5846 Linux kernel system calls (your Unix user must be a root user, or it
5847 must be a member of the tracing group).
5849 opt:lttng-list(1):--userspace::
5850 User space tracepoints.
5852 opt:lttng-list(1):--jul::
5853 `java.util.logging` loggers.
5855 opt:lttng-list(1):--log4j::
5856 Apache log4j loggers.
5858 opt:lttng-list(1):--python::
5862 .List the available user space tracepoints.
5866 $ lttng list --userspace
5870 .List the available Linux kernel system call tracepoints.
5874 $ lttng list --kernel --syscall
5879 [[enabling-disabling-events]]
5880 === Create and enable an event rule
5882 Once you <<creating-destroying-tracing-sessions,create a tracing
5883 session>>, you can create <<event,event rules>> with the
5884 man:lttng-enable-event(1) command.
5886 You specify each condition with a command-line option. The available
5887 condition arguments are shown in the following table.
5889 [role="growable",cols="asciidoc,asciidoc,default"]
5890 .Condition command-line arguments for the man:lttng-enable-event(1) command.
5892 |Argument |Description |Applicable tracing domains
5898 . +--probe=__ADDR__+
5899 . +--function=__ADDR__+
5900 . +--userspace-probe=__PATH__:__SYMBOL__+
5901 . +--userspace-probe=sdt:__PATH__:__PROVIDER__:__NAME__+
5904 Instead of using the default _tracepoint_ instrumentation type, use:
5906 . A Linux system call (entry and exit).
5907 . A Linux https://lwn.net/Articles/132196/[kprobe] (symbol or address).
5908 . The entry and return points of a Linux function (symbol or address).
5909 . The entry point of a user application or library function (path to
5910 application/library and symbol).
5911 . A https://www.sourceware.org/systemtap/wiki/AddingUserSpaceProbingToApps[SystemTap
5912 Statically Defined Tracing] (USDT) probe (path to application/library,
5913 provider and probe names).
5917 |First positional argument.
5920 Tracepoint or system call name.
5922 With the opt:lttng-enable-event(1):--probe,
5923 opt:lttng-enable-event(1):--function, and
5924 opt:lttng-enable-event(1):--userspace-probe options, this is a custom
5925 name given to the event rule. With the JUL, log4j, and Python domains,
5926 this is a logger name.
5928 With a tracepoint, logger, or system call name, use the special
5929 `*` globbing character to match anything (for example, `sched_*`,
5937 . +--loglevel=__LEVEL__+
5938 . +--loglevel-only=__LEVEL__+
5941 . Match only tracepoints or log statements with a logging level at
5942 least as severe as +__LEVEL__+.
5943 . Match only tracepoints or log statements with a logging level
5944 equal to +__LEVEL__+.
5946 See man:lttng-enable-event(1) for the list of available logging level
5949 |User space, JUL, log4j, and Python.
5951 |+--exclude=__EXCLUSIONS__+
5954 When you use a `*` character at the end of the tracepoint or logger
5955 name (first positional argument), exclude the specific names in the
5956 comma-delimited list +__EXCLUSIONS__+.
5959 User space, JUL, log4j, and Python.
5961 |+--filter=__EXPR__+
5964 Match only events which satisfy the expression +__EXPR__+.
5966 See man:lttng-enable-event(1) to learn more about the syntax of a
5973 You attach an event rule to a <<channel,channel>> on creation. If you do
5974 not specify the channel with the opt:lttng-enable-event(1):--channel
5975 option, and if the event rule to create is the first in its
5976 <<domain,tracing domain>> for a given tracing session, then LTTng
5977 creates a _default channel_ for you. This default channel is reused in
5978 subsequent invocations of the man:lttng-enable-event(1) command for the
5979 same tracing domain.
5981 An event rule is always enabled at creation time.
5983 The following examples show how to combine the previous
5984 command-line options to create simple to more complex event rules.
5986 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5990 $ lttng enable-event --kernel sched_switch
5994 .Create an event rule matching four Linux kernel system calls (default channel).
5998 $ lttng enable-event --kernel --syscall open,write,read,close
6002 .Create event rules matching tracepoints with filter expressions (default channel).
6006 $ lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
6011 $ lttng enable-event --kernel --all \
6012 --filter='$ctx.tid == 1988 || $ctx.tid == 1534'
6017 $ lttng enable-event --jul my_logger \
6018 --filter='$app.retriever:cur_msg_id > 3'
6021 IMPORTANT: Make sure to always quote the filter string when you
6022 use man:lttng(1) from a shell.
6024 See also <<pid-tracking,Track process attributes>> which offers another,
6025 more efficient filtering mechanism for process ID, user ID, and group
6029 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
6033 $ lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
6036 IMPORTANT: Make sure to always quote the wildcard character when you
6037 use man:lttng(1) from a shell.
6040 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
6044 $ lttng enable-event --python my-app.'*' \
6045 --exclude='my-app.module,my-app.hello'
6049 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
6053 $ lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
6057 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
6061 $ lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
6065 .Create an event rule matching the `malloc` function entry in path:{/usr/lib/libc.so.6}:
6069 $ lttng enable-event --kernel --userspace-probe=/usr/lib/libc.so.6:malloc \
6074 .Create an event rule matching the `server`/`accept_request` https://www.sourceware.org/systemtap/wiki/AddingUserSpaceProbingToApps[USDT probe] in path:{/usr/bin/serv}:
6078 $ lttng enable-event --kernel --userspace-probe=sdt:serv:server:accept_request \
6079 server_accept_request
6083 The event rules of a given channel form a whitelist: as soon as an
6084 emitted event passes one of them, LTTng can record the event. For
6085 example, an event named `my_app:my_tracepoint` emitted from a user space
6086 tracepoint with a `TRACE_ERROR` log level passes both of the following
6091 $ lttng enable-event --userspace my_app:my_tracepoint
6092 $ lttng enable-event --userspace my_app:my_tracepoint \
6093 --loglevel=TRACE_INFO
6096 The second event rule is redundant: the first one includes
6100 [[disable-event-rule]]
6101 === Disable an event rule
6103 To disable an event rule that you <<enabling-disabling-events,created>>
6104 previously, use the man:lttng-disable-event(1) command. This command
6105 disables _all_ the event rules (of a given tracing domain and channel)
6106 which match an instrumentation point. The other conditions aren't
6107 supported as of LTTng{nbsp}{revision}.
6109 The LTTng tracer doesn't record an emitted event which passes
6110 a _disabled_ event rule.
6112 .Disable an event rule matching a Python logger (default channel).
6116 $ lttng disable-event --python my-logger
6120 .Disable an event rule matching all `java.util.logging` loggers (default channel).
6124 $ lttng disable-event --jul '*'
6128 .Disable _all_ the event rules of the default channel.
6130 The opt:lttng-disable-event(1):--all-events option isn't, like the
6131 opt:lttng-enable-event(1):--all option of man:lttng-enable-event(1), the
6132 equivalent of the event name `*` (wildcard): it disables _all_ the event
6133 rules of a given channel.
6137 $ lttng disable-event --jul --all-events
6141 NOTE: You can't delete an event rule once you create it.
6145 === Get the status of a tracing session
6147 To get the status of the <<cur-tracing-session,current tracing
6148 session>>, that is, its parameters, its channels, event rules, and their
6151 * Use the man:lttng-status(1) command:
6160 To get the status of any tracing session:
6162 * Use the man:lttng-list(1) command with the name of the tracing
6168 $ lttng list my-session
6172 Replace `my-session` with the desired tracing session name.
6175 [[basic-tracing-session-control]]
6176 === Start and stop a tracing session
6178 Once you <<creating-destroying-tracing-sessions,create a tracing
6180 <<enabling-disabling-events,create one or more event rules>>,
6181 you can start and stop the tracers for this tracing session.
6183 To start tracing in the <<cur-tracing-session,current tracing session>>:
6185 * Use the man:lttng-start(1) command:
6194 LTTng is very flexible: you can launch user applications before
6195 or after the you start the tracers. The tracers only record the events
6196 if they pass enabled event rules and if they occur while the tracers are
6199 To stop tracing in the current tracing session:
6201 * Use the man:lttng-stop(1) command:
6210 If there were <<channel-overwrite-mode-vs-discard-mode,lost event
6211 records>> or lost sub-buffers since the last time you ran
6212 man:lttng-start(1), warnings are printed when you run the
6213 man:lttng-stop(1) command.
6215 IMPORTANT: You need to stop tracing to make LTTng flush the remaining
6216 trace data and make the trace readable. Note that the
6217 man:lttng-destroy(1) command (see
6218 <<creating-destroying-tracing-sessions,Create and destroy a tracing
6219 session>>) also runs the man:lttng-stop(1) command implicitly.
6223 === Clear a tracing session
6225 You might need to remove all the current tracing data of one or more
6226 <<tracing-session,tracing sessions>> between multiple attempts to
6227 reproduce a problem without interrupting the LTTng tracing activity.
6229 To clear the tracing data of the
6230 <<cur-tracing-session,current tracing session>>:
6232 * Use the man:lttng-clear(1) command:
6241 To clear the tracing data of all the tracing sessions:
6243 * Use the `lttng clear` command with the opt:lttng-clear(1):--all
6254 [[enabling-disabling-channels]]
6255 === Create a channel
6257 Once you create a tracing session, you can create a <<channel,channel>>
6258 with the man:lttng-enable-channel(1) command.
6260 Note that LTTng automatically creates a default channel when, for a
6261 given <<domain,tracing domain>>, no channels exist and you
6262 <<enabling-disabling-events,create>> the first event rule. This default
6263 channel is named `channel0` and its attributes are set to reasonable
6264 values. Therefore, you only need to create a channel when you need
6265 non-default attributes.
6267 You specify each non-default channel attribute with a command-line
6268 option when you use the man:lttng-enable-channel(1) command. The
6269 available command-line options are:
6271 [role="growable",cols="asciidoc,asciidoc"]
6272 .Command-line options for the man:lttng-enable-channel(1) command.
6274 |Option |Description
6280 <<channel-overwrite-mode-vs-discard-mode,event record loss mode>> instead
6281 of the default _discard_ mode.
6283 |`--buffers-pid` (user space tracing domain only)
6286 Use the per-process <<channel-buffering-schemes,buffering scheme>>
6287 instead of the default per-user buffering scheme.
6289 |+--subbuf-size=__SIZE__+
6292 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
6293 either for each Unix user (default), or for each instrumented process.
6295 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
6297 |+--num-subbuf=__COUNT__+
6300 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
6301 for each Unix user (default), or for each instrumented process.
6303 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
6305 |+--tracefile-size=__SIZE__+
6308 Set the maximum size of each trace file that this channel writes within
6309 a stream to +__SIZE__+ bytes instead of no maximum.
6311 See <<tracefile-rotation,Trace file count and size>>.
6313 |+--tracefile-count=__COUNT__+
6316 Limit the number of trace files that this channel creates to
6317 +__COUNT__+ channels instead of no limit.
6319 See <<tracefile-rotation,Trace file count and size>>.
6321 |+--switch-timer=__PERIODUS__+
6324 Set the <<channel-switch-timer,switch timer period>>
6325 to +__PERIODUS__+{nbsp}µs.
6327 |+--read-timer=__PERIODUS__+
6330 Set the <<channel-read-timer,read timer period>>
6331 to +__PERIODUS__+{nbsp}µs.
6333 |[[opt-blocking-timeout]]+--blocking-timeout=__TIMEOUTUS__+
6336 Set the timeout of user space applications which load LTTng-UST
6337 in blocking mode to +__TIMEOUTUS__+:
6340 Never block (non-blocking mode).
6343 Block forever until space is available in a sub-buffer to record
6346 __n__, a positive value::
6347 Wait for at most __n__ µs when trying to write into a sub-buffer.
6349 Note that, for this option to have any effect on an instrumented
6350 user space application, you need to run the application with a set
6351 env:LTTNG_UST_ALLOW_BLOCKING environment variable.
6353 |+--output=__TYPE__+ (Linux kernel tracing domain only)
6356 Set the output type of the channel to +__TYPE__+, either `mmap` or
6361 You can only create a channel in the Linux kernel and user space
6362 <<domain,tracing domains>>: other tracing domains have their own channel
6363 created on the fly when <<enabling-disabling-events,creating event
6368 Because of a current LTTng limitation, you must create all channels
6369 _before_ you <<basic-tracing-session-control,start tracing>> in a given
6370 tracing session, that is, before the first time you run
6373 Since LTTng automatically creates a default channel when you use the
6374 man:lttng-enable-event(1) command with a specific tracing domain, you
6375 can't, for example, create a Linux kernel event rule, start tracing,
6376 and then create a user space event rule, because no user space channel
6377 exists yet and it's too late to create one.
6379 For this reason, make sure to configure your channels properly
6380 before starting the tracers for the first time!
6383 The following examples show how to combine the previous
6384 command-line options to create simple to more complex channels.
6386 .Create a Linux kernel channel with default attributes.
6390 $ lttng enable-channel --kernel my-channel
6394 .Create a user space channel with four sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
6398 $ lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
6399 --buffers-pid my-channel
6403 .[[blocking-timeout-example]]Create a default user space channel with an infinite blocking timeout.
6405 <<creating-destroying-tracing-sessions,Create a tracing-session>>,
6406 create the channel, <<enabling-disabling-events,create an event rule>>,
6407 and <<basic-tracing-session-control,start tracing>>:
6412 $ lttng enable-channel --userspace --blocking-timeout=inf blocking-channel
6413 $ lttng enable-event --userspace --channel=blocking-channel --all
6417 Run an application instrumented with LTTng-UST and allow it to block:
6421 $ LTTNG_UST_ALLOW_BLOCKING=1 my-app
6425 .Create a Linux kernel channel which rotates eight trace files of 4{nbsp}MiB each for each stream
6429 $ lttng enable-channel --kernel --tracefile-count=8 \
6430 --tracefile-size=4194304 my-channel
6434 .Create a user space channel in overwrite (or _flight recorder_) mode.
6438 $ lttng enable-channel --userspace --overwrite my-channel
6442 <<enabling-disabling-events,Create>> the same event rule in
6443 two different channels:
6447 $ lttng enable-event --userspace --channel=my-channel app:tp
6448 $ lttng enable-event --userspace --channel=other-channel app:tp
6451 If both channels are enabled, when a tracepoint named `app:tp` is
6452 reached, LTTng records two events, one for each channel.
6456 === Disable a channel
6458 To disable a specific channel that you <<enabling-disabling-channels,created>>
6459 previously, use the man:lttng-disable-channel(1) command.
6461 .Disable a specific Linux kernel channel.
6465 $ lttng disable-channel --kernel my-channel
6469 The state of a channel precedes the individual states of event rules
6470 attached to it: event rules which belong to a disabled channel, even if
6471 they are enabled, are also considered disabled.
6475 === Add context fields to a channel
6477 Event record fields in trace files provide important information about
6478 events that occured previously, but sometimes some external context may
6479 help you solve a problem faster.
6481 Examples of context fields are:
6483 * The **process ID**, **thread ID**, **process name**, and
6484 **process priority** of the thread in which the event occurs.
6485 * The **hostname** of the system on which the event occurs.
6486 * The Linux kernel and user call stacks (since
6488 * The current values of many possible **performance counters** using
6490 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
6492 ** Branch instructions, misses, and loads.
6494 * Any context defined at the application level (supported for the
6495 JUL and log4j <<domain,tracing domains>>).
6497 To get the full list of available context fields, see
6498 `lttng add-context --list`. Some context fields are reserved for a
6499 specific <<domain,tracing domain>> (Linux kernel or user space).
6501 You add context fields to <<channel,channels>>. All the events
6502 that a channel with added context fields records contain those fields.
6504 To add context fields to one or all the channels of a given tracing
6507 * Use the man:lttng-add-context(1) command.
6509 .Add context fields to all the channels of the current tracing session.
6511 The following command line adds the virtual process identifier and
6512 the per-thread CPU cycles count fields to all the user space channels
6514 <<cur-tracing-session,current tracing session>>.
6518 $ lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
6522 .Add performance counter context fields by raw ID
6524 See man:lttng-add-context(1) for the exact format of the context field
6525 type, which is partly compatible with the format used in
6530 $ lttng add-context --userspace --type=perf:thread:raw:r0110:test
6531 $ lttng add-context --kernel --type=perf:cpu:raw:r0013c:x86unhalted
6535 .Add context fields to a specific channel.
6537 The following command line adds the thread identifier and user call
6538 stack context fields to the Linux kernel channel named `my-channel` in
6539 the current tracing session.
6543 $ lttng add-context --kernel --channel=my-channel \
6544 --type=tid --type=callstack-user
6548 .Add an application-specific context field to a specific channel.
6550 The following command line adds the `cur_msg_id` context field of the
6551 `retriever` context retriever for all the instrumented
6552 <<java-application,Java applications>> recording <<event,event records>>
6553 in the channel named `my-channel`:
6557 $ lttng add-context --kernel --channel=my-channel \
6558 --type='$app:retriever:cur_msg_id'
6561 IMPORTANT: Make sure to always quote the `$` character when you
6562 use man:lttng-add-context(1) from a shell.
6565 NOTE: You can't remove context fields from a channel once you add it.
6570 === Track process attributes
6572 It's often useful to only allow processes with specific attributes to
6573 emit events. For example, you may wish to record all the system calls
6574 which a given process makes (à la
6575 http://linux.die.net/man/1/strace[strace]).
6577 The man:lttng-track(1) and man:lttng-untrack(1) commands serve this
6578 purpose. Both commands operate on _inclusion sets_ of process attribute
6579 values. The available process attribute types are:
6581 Linux kernel <<domain,tracing domain>> only::
6585 * Virtual process ID (VPID).
6587 This is the PID as seen by the application.
6589 * Unix user ID (UID) (since LTTng{nbsp}2.12).
6591 * Virtual Unix user ID (VUID) (since LTTng{nbsp}2.12).
6593 This is the UID as seen by the application.
6595 * Unix group ID (GID) (since LTTng{nbsp}2.12).
6597 * Virtual Unix group ID (VGID) (since LTTng{nbsp}2.12).
6599 This is the GID as seen by the application.
6602 User space tracing domain::
6605 * VUID (since LTTng{nbsp}2.12).
6606 * VGID (since LTTng{nbsp}2.12).
6608 Each tracing domain has one inclusion set per process attribute type:
6609 the Linux kernel tracing domain has six while the user space tracing
6612 For a given event which passes an enabled <<event,event rule>> to be
6613 recorded, _all_ the attributes of its executing process must be part of
6614 the inclusion sets of the tracing domain of the event rule.
6616 Add entries to an inclusion set with the man:lttng-track(1) command and
6617 remove entries with the man:lttng-untrack(1) command. A process
6618 attribute is _tracked_ when it's part of an inclusion set and
6619 _untracked_ otherwise.
6623 The process attribute values are _numeric_.
6625 Should a process with a given tracked process ID, for example, exit, and
6626 then a new process be given this ID, then the latter would also be
6627 allowed to emit events.
6629 With the `lttng track` command, you can add Unix user and group _names_
6630 to the user and group inclusion sets: the <<lttng-sessiond,session
6631 daemon>> finds the corresponding UID, VUID, GID, or VGID once on
6632 _addition_ to the inclusion set. This means that if you rename the user
6633 or group after you run `lttng track`, its user/group ID remains tracked.
6636 .Track and untrack virtual process IDs.
6638 For the sake of the following example, assume the target system has
6639 16{nbsp}possible VPIDs.
6642 <<creating-destroying-tracing-sessions,create a tracing session>>,
6643 the user space VPID inclusion set contains _all_ the possible VPIDs:
6646 .All VPIDs are tracked.
6647 image::track-all.png[]
6649 When the inclusion set is full and you use the man:lttng-track(1)
6650 command to specify some VPIDs to track, LTTng first clears the inclusion
6651 set, and then it adds the specific VPIDs to track. After:
6655 $ lttng track --userspace --vpid=3,4,7,10,13
6658 the VPID inclusion set is:
6661 .VPIDs 3, 4, 7, 10, and 13 are tracked.
6662 image::track-3-4-7-10-13.png[]
6664 Add more VPIDs to the inclusion set afterwards:
6668 $ lttng track --userspace --vpid=1,15,16
6674 .VPIDs 1, 15, and 16 are added to the inclusion set.
6675 image::track-1-3-4-7-10-13-15-16.png[]
6677 The man:lttng-untrack(1) command removes entries from process attribute
6678 inclusion sets. Given the previous example, the following command:
6682 $ lttng untrack --userspace --vpid=3,7,10,13
6685 leads to this VPID inclusion set:
6688 .VPIDs 3, 7, 10, and 13 are removed from the inclusion set.
6689 image::track-1-4-15-16.png[]
6691 LTTng can track all the possible VPIDs again using the
6692 opt:lttng-track(1):--all option:
6696 $ lttng track --userspace --vpid --all
6699 The result is, again:
6702 .All VPIDs are tracked.
6703 image::track-all.png[]
6706 .Track only specific process attributes.
6708 A typical use case with process attribute tracking is to start with an
6709 empty inclusion set, then <<basic-tracing-session-control,start the
6710 tracers>>, and then add entries manually while the tracers are active.
6712 Use the opt:lttng-untrack(1):--all option of the
6713 man:lttng-untrack(1) command to clear the inclusion set after you
6714 <<creating-destroying-tracing-sessions,create a tracing session>>, for
6715 example (with UIDs):
6719 $ lttng untrack --kernel --uid --all
6725 .No UIDs are tracked.
6726 image::untrack-all.png[]
6728 If you trace with this inclusion set configuration, the LTTng kernel
6729 tracer records no events within the <<cur-tracing-session,current
6730 tracing session>> because it doesn't track any UID. Use the
6731 man:lttng-track(1) command as usual to track specific UIDs when you need
6736 $ lttng track --kernel --uid=http,11
6742 .UIDs 6 (`http`) and 11 are tracked.
6743 image::track-6-11.png[]
6748 [[saving-loading-tracing-session]]
6749 === Save and load tracing session configurations
6751 Configuring a <<tracing-session,tracing session>> can be long. Some of
6752 the tasks involved are:
6754 * <<enabling-disabling-channels,Create channels>> with
6755 specific attributes.
6756 * <<adding-context,Add context fields>> to specific channels.
6757 * <<enabling-disabling-events,Create event rules>> with specific log
6758 level and filter conditions.
6760 If you use LTTng to solve real world problems, chances are you have to
6761 record events using the same tracing session setup over and over,
6762 modifying a few variables each time in your instrumented program
6763 or environment. To avoid constant tracing session reconfiguration,
6764 the man:lttng(1) command-line tool can save and load tracing session
6765 configurations to/from XML files.
6767 To save a given tracing session configuration:
6769 * Use the man:lttng-save(1) command:
6774 $ lttng save my-session
6778 Replace `my-session` with the name of the tracing session to save.
6780 LTTng saves tracing session configurations to
6781 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
6782 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
6783 the opt:lttng-save(1):--output-path option to change this destination
6786 LTTng saves all configuration parameters, for example:
6788 * The tracing session name.
6789 * The trace data output path.
6790 * The channels with their state and all their attributes.
6791 * The context fields you added to channels.
6792 * The event rules with their state, log level and filter conditions.
6794 To load a tracing session:
6796 * Use the man:lttng-load(1) command:
6801 $ lttng load my-session
6805 Replace `my-session` with the name of the tracing session to load.
6807 When LTTng loads a configuration, it restores your saved tracing session
6808 as if you just configured it manually.
6810 See man:lttng-load(1) for the complete list of command-line options. You
6811 can also save and load many sessions at a time, and decide in which
6812 directory to output the XML files.
6815 [[sending-trace-data-over-the-network]]
6816 === Send trace data over the network
6818 LTTng can send the recorded trace data to a remote system over the
6819 network instead of writing it to the local file system.
6821 To send the trace data over the network:
6823 . On the _remote_ system (which can also be the target system),
6824 start an LTTng <<lttng-relayd,relay daemon>> (man:lttng-relayd(8)):
6833 . On the _target_ system, create a tracing session configured to
6834 send trace data over the network:
6839 $ lttng create my-session --set-url=net://remote-system
6843 Replace `remote-system` by the host name or IP address of the
6844 remote system. See man:lttng-create(1) for the exact URL format.
6846 . On the target system, use the man:lttng(1) command-line tool as usual.
6847 When tracing is active, the consumer daemon of the target sends
6848 sub-buffers to the relay daemon running on the remote system instead
6849 of flushing them to the local file system. The relay daemon writes the
6850 received packets to the local file system.
6852 The relay daemon writes trace files to
6853 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
6854 +__hostname__+ is the host name of the target system and +__session__+
6855 is the tracing session name. Note that the env:LTTNG_HOME environment
6856 variable defaults to `$HOME` if not set. Use the
6857 opt:lttng-relayd(8):--output option of man:lttng-relayd(8) to write
6858 trace files to another base directory.
6863 === View events as LTTng emits them (noch:{LTTng} live)
6865 LTTng live is a network protocol implemented by the <<lttng-relayd,relay
6866 daemon>> (man:lttng-relayd(8)) to allow compatible trace viewers to
6867 display events as LTTng emits them on the target system while tracing is
6870 The relay daemon creates a _tee_: it forwards the trace data to both
6871 the local file system and to connected live viewers:
6874 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
6879 . On the _target system_, create a <<tracing-session,tracing session>>
6885 $ lttng create my-session --live
6889 This spawns a local relay daemon.
6891 . Start the live viewer and configure it to connect to the relay
6892 daemon. For example, with
6893 https://babeltrace.org/docs/v2.0/man1/babeltrace2.1/[cmd:babeltrace2]:
6898 $ babeltrace2 net://localhost/host/hostname/my-session
6905 * `hostname` with the host name of the target system.
6906 * `my-session` with the name of the tracing session to view.
6909 . Configure the tracing session as usual with the man:lttng(1)
6910 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6912 List the available live tracing sessions with Babeltrace{nbsp}2:
6916 $ babeltrace2 net://localhost
6919 You can start the relay daemon on another system. In this case, you need
6920 to specify the URL of the relay daemon when you create the tracing
6921 session with the opt:lttng-create(1):--set-url option. You also need to
6922 replace `localhost` in the procedure above with the host name of the
6923 system on which the relay daemon is running.
6925 See man:lttng-create(1) and man:lttng-relayd(8) for the complete list of
6926 command-line options.
6930 [[taking-a-snapshot]]
6931 === Take a snapshot of the current sub-buffers of a tracing session
6933 The normal behavior of LTTng is to append full sub-buffers to growing
6934 trace data files. This is ideal to keep a full history of the events
6935 that occurred on the target system, but it can
6936 represent too much data in some situations. For example, you may wish
6937 to trace your application continuously until some critical situation
6938 happens, in which case you only need the latest few recorded
6939 events to perform the desired analysis, not multi-gigabyte trace files.
6941 With the man:lttng-snapshot(1) command, you can take a snapshot of the
6942 current sub-buffers of a given <<tracing-session,tracing session>>.
6943 LTTng can write the snapshot to the local file system or send it over
6947 .A snapshot is a copy of the current sub-buffers, which aren't cleared after the operation.
6948 image::snapshot.png[]
6950 If you wish to create unmanaged, self-contained, non-overlapping
6951 trace chunk archives instead of a simple copy of the current
6952 sub-buffers, see the <<session-rotation,tracing session rotation>>
6953 feature (available since LTTng{nbsp}2.11).
6957 . Create a tracing session in _snapshot mode_:
6962 $ lttng create my-session --snapshot
6966 The <<channel-overwrite-mode-vs-discard-mode,event record loss mode>> of
6967 <<channel,channels>> created in this mode is automatically set to
6968 _overwrite_ (flight recorder mode).
6970 . Configure the tracing session as usual with the man:lttng(1)
6971 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6973 . **Optional**: When you need to take a snapshot,
6974 <<basic-tracing-session-control,stop tracing>>.
6976 You can take a snapshot when the tracers are active, but if you stop
6977 them first, you're sure that the data in the sub-buffers doesn't
6978 change before you actually take the snapshot.
6985 $ lttng snapshot record --name=my-first-snapshot
6989 LTTng writes the current sub-buffers of all the channels of the
6990 <<cur-tracing-session,current tracing session>> to
6991 trace files on the local file system. Those trace files have
6992 `my-first-snapshot` in their name.
6994 There is no difference between the format of a normal trace file and the
6995 format of a snapshot: viewers of LTTng traces also support LTTng
6998 By default, LTTng writes snapshot files to the path shown by
6999 `lttng snapshot list-output`. You can change this path or decide to send
7000 snapshots over the network using either:
7002 . An output path or URL that you specify when you
7003 <<creating-destroying-tracing-sessions,create the tracing session>>.
7004 . A snapshot output path or URL that you add using
7005 `lttng snapshot add-output`.
7006 . An output path or URL that you provide directly to the
7007 `lttng snapshot record` command.
7009 Method{nbsp}3 overrides method{nbsp}2, which overrides method 1. When
7010 you specify a URL, a relay daemon must listen on a remote system (see
7011 <<sending-trace-data-over-the-network,Send trace data over the
7016 [[session-rotation]]
7017 === Archive the current trace chunk (rotate a tracing session)
7019 The <<taking-a-snapshot,snapshot user guide>> shows how to dump the
7020 current sub-buffers of a tracing session to the file system or send them
7021 over the network. When you take a snapshot, LTTng doesn't clear the ring
7022 buffers of the tracing session: if you take another snapshot immediately
7023 after, both snapshots could contain overlapping trace data.
7025 Inspired by https://en.wikipedia.org/wiki/Log_rotation[log rotation],
7026 _tracing session rotation_ is a feature which appends the content of the
7027 ring buffers to what's already on the file system or sent over the
7028 network since the creation of the tracing session or since the last
7029 rotation, and then clears those ring buffers to avoid trace data
7032 What LTTng is about to write when performing a tracing session rotation
7033 is called the _current trace chunk_. When this current trace chunk is
7034 written to the file system or sent over the network, it becomes a _trace
7035 chunk archive_. Therefore, a tracing session rotation _archives_ the
7036 current trace chunk.
7039 .A tracing session rotation operation _archives_ the current trace chunk.
7040 image::rotation.png[]
7042 A trace chunk archive is a self-contained LTTng trace which LTTng
7043 doesn't manage anymore: you can read it, modify it, move it, or remove
7046 There are two methods to perform a tracing session rotation: immediately
7047 or with a rotation schedule.
7049 To perform an immediate tracing session rotation:
7051 . <<creating-destroying-tracing-sessions,Create a tracing session>>
7052 in _normal mode_ or _network streaming mode_
7053 (only those two creation modes support tracing session rotation):
7058 $ lttng create my-session
7062 . <<enabling-disabling-events,Create one or more event rules>>
7063 and <<basic-tracing-session-control,start tracing>>:
7068 $ lttng enable-event --kernel sched_'*'
7073 . When needed, immediately rotate the
7074 <<cur-tracing-session,current tracing session>>:
7083 The cmd:lttng-rotate command prints the path to the created trace
7084 chunk archive. See man:lttng-rotate(1) to learn about the format
7085 of trace chunk archive directory names.
7087 Perform other immediate rotations while the tracing session is
7088 active. It is guaranteed that all the trace chunk archives don't
7089 contain overlapping trace data. You can also perform an immediate
7090 rotation once you have <<basic-tracing-session-control,stopped>> the
7093 . When you're done tracing,
7094 <<creating-destroying-tracing-sessions,destroy the current tracing
7104 The tracing session destruction operation creates one last trace
7105 chunk archive from the current trace chunk.
7107 A tracing session rotation schedule is a planned rotation which LTTng
7108 performs automatically based on one of the following conditions:
7110 * A timer with a configured period times out.
7112 * The total size of the flushed part of the current trace chunk
7113 becomes greater than or equal to a configured value.
7115 To schedule a tracing session rotation, set a _rotation schedule_:
7117 . <<creating-destroying-tracing-sessions,Create a tracing session>>
7118 in _normal mode_ or _network streaming mode_
7119 (only those two creation modes support tracing session rotation):
7124 $ lttng create my-session
7128 . <<enabling-disabling-events,Create one or more event rules>>:
7133 $ lttng enable-event --kernel sched_'*'
7137 . Set a tracing session rotation schedule:
7142 $ lttng enable-rotation --timer=10s
7146 In this example, we set a rotation schedule so that LTTng performs a
7147 tracing session rotation every ten seconds.
7149 See man:lttng-enable-rotation(1) to learn more about other ways to set a
7152 . <<basic-tracing-session-control,Start tracing>>:
7161 LTTng performs tracing session rotations automatically while the tracing
7162 session is active thanks to the rotation schedule.
7164 . When you're done tracing,
7165 <<creating-destroying-tracing-sessions,destroy the current tracing
7175 The tracing session destruction operation creates one last trace chunk
7176 archive from the current trace chunk.
7178 Use man:lttng-disable-rotation(1) to unset a tracing session
7181 NOTE: man:lttng-rotate(1) and man:lttng-enable-rotation(1) list
7182 limitations regarding those two commands.
7187 === Use the machine interface
7189 With any command of the man:lttng(1) command-line tool, set the
7190 opt:lttng(1):--mi option to `xml` (before the command name) to get an
7191 XML machine interface output, for example:
7195 $ lttng --mi=xml enable-event --kernel --syscall open
7198 A schema definition (XSD) is
7199 https://github.com/lttng/lttng-tools/blob/stable-{revision}/src/common/src/common/mi-lttng-4.0.xsd[available]
7200 to ease the integration with external tools as much as possible.
7204 [[metadata-regenerate]]
7205 === Regenerate the metadata of an LTTng trace
7207 An LTTng trace, which is a http://diamon.org/ctf[CTF] trace, has both
7208 data stream files and a metadata file. This metadata file contains,
7209 amongst other things, information about the offset of the clock sources
7210 used to timestamp <<event,event records>> when tracing.
7212 If, once a <<tracing-session,tracing session>> is
7213 <<basic-tracing-session-control,started>>, a major
7214 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
7215 happens, the clock offset of the trace also needs to be updated. Use
7216 the `metadata` item of the man:lttng-regenerate(1) command to do so.
7218 The main use case of this command is to allow a system to boot with
7219 an incorrect wall time and trace it with LTTng before its wall time
7220 is corrected. Once the system is known to be in a state where its
7221 wall time is correct, it can run `lttng regenerate metadata`.
7223 To regenerate the metadata of an LTTng trace:
7225 * Use the `metadata` item of the man:lttng-regenerate(1) command:
7230 $ lttng regenerate metadata
7236 `lttng regenerate metadata` has the following limitations:
7238 * Tracing session <<creating-destroying-tracing-sessions,created>>
7240 * User space <<channel,channels>>, if any, are using
7241 <<channel-buffering-schemes,per-user buffering>>.
7246 [[regenerate-statedump]]
7247 === Regenerate the state dump of a tracing session
7249 The LTTng kernel and user space tracers generate state dump
7250 <<event,event records>> when the application starts or when you
7251 <<basic-tracing-session-control,start a tracing session>>. An analysis
7252 can use the state dump event records to set an initial state before it
7253 builds the rest of the state from the following event records.
7254 http://tracecompass.org/[Trace Compass] is a notable example of an
7255 application which uses the state dump of an LTTng trace.
7257 When you <<taking-a-snapshot,take a snapshot>>, it's possible that the
7258 state dump event records aren't included in the snapshot because they
7259 were recorded to a sub-buffer that has been consumed or overwritten
7262 Use the `lttng regenerate statedump` command to emit the state
7263 dump event records again.
7265 To regenerate the state dump of the current tracing session, provided
7266 create it in snapshot mode, before you take a snapshot:
7268 . Use the `statedump` item of the man:lttng-regenerate(1) command:
7273 $ lttng regenerate statedump
7277 . <<basic-tracing-session-control,Stop the tracing session>>:
7286 . <<taking-a-snapshot,Take a snapshot>>:
7291 $ lttng snapshot record --name=my-snapshot
7295 Depending on the event throughput, you should run steps 1 and 2
7296 as closely as possible.
7298 NOTE: To record the state dump events, you need to
7299 <<enabling-disabling-events,create event rules>> which enable them.
7300 LTTng-UST state dump tracepoints start with `lttng_ust_statedump:`.
7301 LTTng-modules state dump tracepoints start with `lttng_statedump_`.
7305 [[persistent-memory-file-systems]]
7306 === Record trace data on persistent memory file systems
7308 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
7309 (NVRAM) is random-access memory that retains its information when power
7310 is turned off (non-volatile). Systems with such memory can store data
7311 structures in RAM and retrieve them after a reboot, without flushing
7312 to typical _storage_.
7314 Linux supports NVRAM file systems thanks to either
7315 http://pramfs.sourceforge.net/[PRAMFS] or
7316 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
7317 (requires Linux{nbsp}4.1+).
7319 This section doesn't describe how to operate such file systems;
7320 we assume that you have a working persistent memory file system.
7322 When you create a <<tracing-session,tracing session>>, you can specify
7323 the path of the shared memory holding the sub-buffers. If you specify a
7324 location on an NVRAM file system, then you can retrieve the latest
7325 recorded trace data when the system reboots after a crash.
7327 To record trace data on a persistent memory file system and retrieve the
7328 trace data after a system crash:
7330 . Create a tracing session with a sub-buffer shared memory path located
7331 on an NVRAM file system:
7336 $ lttng create my-session --shm-path=/path/to/shm
7340 . Configure the tracing session as usual with the man:lttng(1)
7341 command-line tool, and <<basic-tracing-session-control,start tracing>>.
7343 . After a system crash, use the man:lttng-crash(1) command-line tool to
7344 view the trace data recorded on the NVRAM file system:
7349 $ lttng-crash /path/to/shm
7353 The binary layout of the ring buffer files isn't exactly the same as
7354 the trace files layout. This is why you need to use man:lttng-crash(1)
7355 instead of your preferred trace viewer directly.
7357 To convert the ring buffer files to LTTng trace files:
7359 * Use the opt:lttng-crash(1):--extract option of man:lttng-crash(1):
7364 $ lttng-crash --extract=/path/to/trace /path/to/shm
7370 [[notif-trigger-api]]
7371 === Get notified when the buffer usage of a channel is too high or too low
7373 With the $$C/C++$$ notification and trigger API of LTTng, your user
7374 application can get notified when the buffer usage of one or more
7375 <<channel,channels>> becomes too low or too high. Use this API
7376 and enable or disable <<event,event rules>> during tracing to avoid
7377 <<channel-overwrite-mode-vs-discard-mode,discarded event records>>.
7379 .Have a user application get notified when the buffer usage of an LTTng channel is too high.
7381 In this example, we create and build an application which gets notified
7382 when the buffer usage of a specific LTTng channel is higher than
7383 75{nbsp}%. We only print that it is the case in the example, but we
7384 could as well use the API of <<liblttng-ctl-lttng,`liblttng-ctl`>> to
7385 disable event rules when this happens.
7387 . Create the C{nbsp}source file of application:
7395 #include <lttng/domain.h>
7396 #include <lttng/action/action.h>
7397 #include <lttng/action/notify.h>
7398 #include <lttng/condition/condition.h>
7399 #include <lttng/condition/buffer-usage.h>
7400 #include <lttng/condition/evaluation.h>
7401 #include <lttng/notification/channel.h>
7402 #include <lttng/notification/notification.h>
7403 #include <lttng/trigger/trigger.h>
7404 #include <lttng/endpoint.h>
7406 int main(int argc, char *argv[])
7408 int exit_status = 0;
7409 struct lttng_notification_channel *notification_channel;
7410 struct lttng_condition *condition;
7411 struct lttng_action *action;
7412 struct lttng_trigger *trigger;
7413 const char *tracing_session_name;
7414 const char *channel_name;
7417 tracing_session_name = argv[1];
7418 channel_name = argv[2];
7421 * Create a notification channel. A notification channel
7422 * connects the user application to the LTTng session daemon.
7423 * This notification channel can be used to listen to various
7424 * types of notifications.
7426 notification_channel = lttng_notification_channel_create(
7427 lttng_session_daemon_notification_endpoint);
7430 * Create a "high buffer usage" condition. In this case, the
7431 * condition is reached when the buffer usage is greater than or
7432 * equal to 75 %. We create the condition for a specific tracing
7433 * session name, channel name, and for the user space tracing
7436 * The "low buffer usage" condition type also exists.
7438 condition = lttng_condition_buffer_usage_high_create();
7439 lttng_condition_buffer_usage_set_threshold_ratio(condition, .75);
7440 lttng_condition_buffer_usage_set_session_name(
7441 condition, tracing_session_name);
7442 lttng_condition_buffer_usage_set_channel_name(condition,
7444 lttng_condition_buffer_usage_set_domain_type(condition,
7448 * Create an action (get a notification) to take when the
7449 * condition created above is reached.
7451 action = lttng_action_notify_create();
7454 * Create a trigger. A trigger associates a condition to an
7455 * action: the action is executed when the condition is reached.
7457 trigger = lttng_trigger_create(condition, action);
7459 /* Register the trigger to LTTng. */
7460 lttng_register_trigger(trigger);
7463 * Now that we have registered a trigger, a notification will be
7464 * emitted everytime its condition is met. To receive this
7465 * notification, we must subscribe to notifications that match
7466 * the same condition.
7468 lttng_notification_channel_subscribe(notification_channel,
7472 * Notification loop. Put this in a dedicated thread to avoid
7473 * blocking the main thread.
7476 struct lttng_notification *notification;
7477 enum lttng_notification_channel_status status;
7478 const struct lttng_evaluation *notification_evaluation;
7479 const struct lttng_condition *notification_condition;
7480 double buffer_usage;
7482 /* Receive the next notification. */
7483 status = lttng_notification_channel_get_next_notification(
7484 notification_channel, ¬ification);
7487 case LTTNG_NOTIFICATION_CHANNEL_STATUS_OK:
7489 case LTTNG_NOTIFICATION_CHANNEL_STATUS_NOTIFICATIONS_DROPPED:
7491 * The session daemon can drop notifications if a monitoring
7492 * application isn't consuming the notifications fast
7496 case LTTNG_NOTIFICATION_CHANNEL_STATUS_CLOSED:
7498 * The notification channel has been closed by the
7499 * session daemon. This is typically caused by a session
7500 * daemon shutting down.
7504 /* Unhandled conditions or errors. */
7510 * A notification provides, amongst other things:
7512 * * The condition that caused this notification to be
7514 * * The condition evaluation, which provides more
7515 * specific information on the evaluation of the
7518 * The condition evaluation provides the buffer usage
7519 * value at the moment the condition was reached.
7521 notification_condition = lttng_notification_get_condition(
7523 notification_evaluation = lttng_notification_get_evaluation(
7526 /* We're subscribed to only one condition. */
7527 assert(lttng_condition_get_type(notification_condition) ==
7528 LTTNG_CONDITION_TYPE_BUFFER_USAGE_HIGH);
7531 * Get the exact sampled buffer usage from the
7532 * condition evaluation.
7534 lttng_evaluation_buffer_usage_get_usage_ratio(
7535 notification_evaluation, &buffer_usage);
7538 * At this point, instead of printing a message, we
7539 * could do something to reduce the buffer usage of the channel,
7540 * like disable specific events.
7542 printf("Buffer usage is %f %% in tracing session \"%s\", "
7543 "user space channel \"%s\".\n", buffer_usage * 100,
7544 tracing_session_name, channel_name);
7545 lttng_notification_destroy(notification);
7549 lttng_action_destroy(action);
7550 lttng_condition_destroy(condition);
7551 lttng_trigger_destroy(trigger);
7552 lttng_notification_channel_destroy(notification_channel);
7558 . Build the `notif-app` application, linking it to `liblttng-ctl`:
7563 $ gcc -o notif-app notif-app.c -llttng-ctl
7567 . <<creating-destroying-tracing-sessions,Create a tracing session>>,
7568 <<enabling-disabling-events,create an event rule>> matching all the
7569 user space tracepoints, and
7570 <<basic-tracing-session-control,start tracing>>:
7575 $ lttng create my-session
7576 $ lttng enable-event --userspace --all
7581 If you create the channel manually with the man:lttng-enable-channel(1)
7582 command, control how frequently LTTng samples the current values of the
7583 channel properties to evaluate user conditions with the
7584 opt:lttng-enable-channel(1):--monitor-timer option.
7586 . Run the `notif-app` application. This program accepts the
7587 <<tracing-session,tracing session>> name and the user space channel
7588 name as its two first arguments. The channel which LTTng automatically
7589 creates with the man:lttng-enable-event(1) command above is named
7595 $ ./notif-app my-session channel0
7599 . In another terminal, run an application with a very high event
7600 throughput so that the 75{nbsp}% buffer usage condition is reached.
7602 In the first terminal, the application should print lines like this:
7605 Buffer usage is 81.45197 % in tracing session "my-session", user space
7609 If you don't see anything, try modifying the condition in
7610 path:{notif-app.c} to a lower value (0.1, for example), rebuilding it
7611 (step{nbsp}2) and running it again (step{nbsp}4).
7618 [[lttng-modules-ref]]
7619 === noch:{LTTng-modules}
7623 [[lttng-tracepoint-enum]]
7624 ==== `LTTNG_TRACEPOINT_ENUM()` usage
7626 Use the `LTTNG_TRACEPOINT_ENUM()` macro to define an enumeration:
7630 LTTNG_TRACEPOINT_ENUM(name, TP_ENUM_VALUES(entries))
7635 * `name` with the name of the enumeration (C identifier, unique
7636 amongst all the defined enumerations).
7637 * `entries` with a list of enumeration entries.
7639 The available enumeration entry macros are:
7641 +ctf_enum_value(__name__, __value__)+::
7642 Entry named +__name__+ mapped to the integral value +__value__+.
7644 +ctf_enum_range(__name__, __begin__, __end__)+::
7645 Entry named +__name__+ mapped to the range of integral values between
7646 +__begin__+ (included) and +__end__+ (included).
7648 +ctf_enum_auto(__name__)+::
7649 Entry named +__name__+ mapped to the integral value following the
7652 The last value of a `ctf_enum_value()` entry is its +__value__+
7655 The last value of a `ctf_enum_range()` entry is its +__end__+ parameter.
7657 If `ctf_enum_auto()` is the first entry in the list, its integral
7660 Use the `ctf_enum()` <<lttng-modules-tp-fields,field definition macro>>
7661 to use a defined enumeration as a tracepoint field.
7663 .Define an enumeration with `LTTNG_TRACEPOINT_ENUM()`.
7667 LTTNG_TRACEPOINT_ENUM(
7670 ctf_enum_auto("AUTO: EXPECT 0")
7671 ctf_enum_value("VALUE: 23", 23)
7672 ctf_enum_value("VALUE: 27", 27)
7673 ctf_enum_auto("AUTO: EXPECT 28")
7674 ctf_enum_range("RANGE: 101 TO 303", 101, 303)
7675 ctf_enum_auto("AUTO: EXPECT 304")
7683 [[lttng-modules-tp-fields]]
7684 ==== Tracepoint fields macros (for `TP_FIELDS()`)
7686 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
7687 tracepoint fields, which must be listed within `TP_FIELDS()` in
7688 `LTTNG_TRACEPOINT_EVENT()`, are:
7690 [role="func-desc growable",cols="asciidoc,asciidoc"]
7691 .Available macros to define LTTng-modules tracepoint fields
7693 |Macro |Description and parameters
7696 +ctf_integer(__t__, __n__, __e__)+
7698 +ctf_integer_nowrite(__t__, __n__, __e__)+
7700 +ctf_user_integer(__t__, __n__, __e__)+
7702 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
7704 Standard integer, displayed in base{nbsp}10.
7707 Integer C type (`int`, `long`, `size_t`, ...).
7713 Argument expression.
7716 +ctf_integer_hex(__t__, __n__, __e__)+
7718 +ctf_user_integer_hex(__t__, __n__, __e__)+
7720 Standard integer, displayed in base{nbsp}16.
7729 Argument expression.
7731 |+ctf_integer_oct(__t__, __n__, __e__)+
7733 Standard integer, displayed in base{nbsp}8.
7742 Argument expression.
7745 +ctf_integer_network(__t__, __n__, __e__)+
7747 +ctf_user_integer_network(__t__, __n__, __e__)+
7749 Integer in network byte order (big-endian), displayed in base{nbsp}10.
7758 Argument expression.
7761 +ctf_integer_network_hex(__t__, __n__, __e__)+
7763 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
7765 Integer in network byte order, displayed in base{nbsp}16.
7774 Argument expression.
7777 +ctf_enum(__N__, __t__, __n__, __e__)+
7779 +ctf_enum_nowrite(__N__, __t__, __n__, __e__)+
7781 +ctf_user_enum(__N__, __t__, __n__, __e__)+
7783 +ctf_user_enum_nowrite(__N__, __t__, __n__, __e__)+
7788 Name of a <<lttng-tracepoint-enum,previously defined enumeration>>.
7791 Integer C type (`int`, `long`, `size_t`, ...).
7797 Argument expression.
7800 +ctf_string(__n__, __e__)+
7802 +ctf_string_nowrite(__n__, __e__)+
7804 +ctf_user_string(__n__, __e__)+
7806 +ctf_user_string_nowrite(__n__, __e__)+
7808 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
7814 Argument expression.
7817 +ctf_array(__t__, __n__, __e__, __s__)+
7819 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
7821 +ctf_user_array(__t__, __n__, __e__, __s__)+
7823 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
7825 Statically-sized array of integers.
7828 Array element C type.
7834 Argument expression.
7840 +ctf_array_bitfield(__t__, __n__, __e__, __s__)+
7842 +ctf_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7844 +ctf_user_array_bitfield(__t__, __n__, __e__, __s__)+
7846 +ctf_user_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7848 Statically-sized array of bits.
7850 The type of +__e__+ must be an integer type. +__s__+ is the number
7851 of elements of such type in +__e__+, not the number of bits.
7854 Array element C type.
7860 Argument expression.
7866 +ctf_array_text(__t__, __n__, __e__, __s__)+
7868 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
7870 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
7872 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
7874 Statically-sized array, printed as text.
7876 The string doesn't need to be null-terminated.
7879 Array element C type (always `char`).
7885 Argument expression.
7891 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
7893 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7895 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
7897 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7899 Dynamically-sized array of integers.
7901 The type of +__E__+ must be unsigned.
7904 Array element C type.
7910 Argument expression.
7913 Length expression C type.
7919 +ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7921 +ctf_user_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7923 Dynamically-sized array of integers, displayed in base{nbsp}16.
7925 The type of +__E__+ must be unsigned.
7928 Array element C type.
7934 Argument expression.
7937 Length expression C type.
7942 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
7944 Dynamically-sized array of integers in network byte order (big-endian),
7945 displayed in base{nbsp}10.
7947 The type of +__E__+ must be unsigned.
7950 Array element C type.
7956 Argument expression.
7959 Length expression C type.
7965 +ctf_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7967 +ctf_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7969 +ctf_user_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7971 +ctf_user_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7973 Dynamically-sized array of bits.
7975 The type of +__e__+ must be an integer type. +__s__+ is the number
7976 of elements of such type in +__e__+, not the number of bits.
7978 The type of +__E__+ must be unsigned.
7981 Array element C type.
7987 Argument expression.
7990 Length expression C type.
7996 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7998 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
8000 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
8002 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
8004 Dynamically-sized array, displayed as text.
8006 The string doesn't need to be null-terminated.
8008 The type of +__E__+ must be unsigned.
8010 The behaviour is undefined if +__e__+ is `NULL`.
8013 Sequence element C type (always `char`).
8019 Argument expression.
8022 Length expression C type.
8028 Use the `_user` versions when the argument expression, `e`, is
8029 a user space address. In the cases of `ctf_user_integer*()` and
8030 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
8033 The `_nowrite` versions omit themselves from the session trace, but are
8034 otherwise identical. This means the `_nowrite` fields won't be written
8035 in the recorded trace. Their primary purpose is to make some
8036 of the event context available to the
8037 <<enabling-disabling-events,event filters>> without having to
8038 commit the data to sub-buffers.
8044 Terms related to LTTng and to tracing in general:
8047 The http://diamon.org/babeltrace[Babeltrace] project, which includes:
8050 https://babeltrace.org/docs/v2.0/man1/babeltrace2.1/[cmd:babeltrace2]
8051 command-line interface.
8052 * The libbabeltrace2 library which offers a
8053 https://babeltrace.org/docs/v2.0/libbabeltrace2/[C API].
8054 * https://babeltrace.org/docs/v2.0/python/bt2/[Python{nbsp}3 bindings].
8057 [[def-buffering-scheme]]<<channel-buffering-schemes,buffering scheme>>::
8058 A layout of <<def-sub-buffer,sub-buffers>> applied to a given channel.
8060 [[def-channel]]<<channel,channel>>::
8061 An entity which is responsible for a set of
8062 <<def-ring-buffer,ring buffers>>.
8064 <<def-event-rule,Event rules>> are always attached to a specific
8068 A source of time for a <<def-tracer,tracer>>.
8070 [[def-consumer-daemon]]<<lttng-consumerd,consumer daemon>>::
8071 A process which is responsible for consuming the full
8072 <<def-sub-buffer,sub-buffers>> and write them to a file system or
8073 send them over the network.
8075 [[def-current-trace-chunk]]current trace chunk::
8076 A <<def-trace-chunk,trace chunk>> which includes the current content
8077 of all the <<def-sub-buffer,sub-buffers>> of the
8078 <<def-tracing-session-rotation,tracing session>> and the stream files
8079 produced since the latest event amongst:
8081 * The creation of the <<def-tracing-session,tracing session>>.
8082 * The last tracing session rotation, if any.
8084 <<channel-overwrite-mode-vs-discard-mode,discard mode>>::
8085 The <<def-event-record-loss-mode,event record loss mode>> in which
8086 the <<def-tracer,tracer>> _discards_ new event records when there's no
8087 <<def-sub-buffer,sub-buffer>> space left to store them.
8089 [[def-event]]event::
8090 The consequence of the execution of an
8091 <<def-instrumentation-point,instrumentation point>>, like a
8092 <<def-tracepoint,tracepoint>> that you manually place in some source
8093 code, or a Linux kernel kprobe.
8095 An event is said to _occur_ at a specific time. <<def-lttng,LTTng>> can
8096 take various actions upon the occurrence of an event, like record its
8097 payload to a <<def-sub-buffer,sub-buffer>>.
8099 [[def-event-name]]event name::
8100 The name of an <<def-event,event>>, which is also the name of the
8101 <<def-event-record,event record>>.
8103 This is also called the _instrumentation point name_.
8105 [[def-event-record]]event record::
8106 A record, in a <<def-trace,trace>>, of the payload of an
8107 <<def-event,event>> which occured.
8109 [[def-event-record-loss-mode]]<<channel-overwrite-mode-vs-discard-mode,event record loss mode>>::
8110 The mechanism by which event records of a given
8111 <<def-channel,channel>> are lost (not recorded) when there is no
8112 <<def-sub-buffer,sub-buffer>> space left to store them.
8114 [[def-event-rule]]<<event,event rule>>::
8115 Set of conditions which must be satisfied for one or more occuring
8116 <<def-event,events>> to be recorded.
8118 [[def-incl-set]]inclusion set::
8119 In the <<pid-tracking,process attribute tracking>> context: a
8120 set of <<def-proc-attr,process attributes>> of a given type.
8122 <<instrumenting,instrumentation>>::
8123 The use of <<def-lttng,LTTng>> probes to make a piece of software
8126 [[def-instrumentation-point]]instrumentation point::
8127 A point in the execution path of a piece of software that, when
8128 reached by this execution, can emit an <<def-event,event>>.
8130 instrumentation point name::
8131 See _<<def-event-name,event name>>_.
8133 `java.util.logging`::
8135 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities]
8136 of the Java platform.
8139 A http://logging.apache.org/log4j/1.2/[logging library] for Java
8140 developed by the Apache Software Foundation.
8143 Level of severity of a log statement or user space
8144 <<def-instrumentation-point,instrumentation point>>.
8146 [[def-lttng]]LTTng::
8147 The _Linux Trace Toolkit: next generation_ project.
8149 <<lttng-cli,cmd:lttng>>::
8150 A command-line tool provided by the <<def-lttng-tools,LTTng-tools>>
8151 project which you can use to send and receive control messages to and
8152 from a <<def-session-daemon,session daemon>>.
8155 The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
8156 which is a set of analyzing programs that you can use to obtain a
8157 higher level view of an <<def-lttng,LTTng>> <<def-trace,trace>>.
8159 cmd:lttng-consumerd::
8160 The name of the <<def-consumer-daemon,consumer daemon>> program.
8163 A utility provided by the <<def-lttng-tools,LTTng-tools>> project
8164 which can convert <<def-ring-buffer,ring buffer>> files (usually
8165 <<persistent-memory-file-systems,saved on a persistent memory file
8166 system>>) to <<def-trace,trace>> files.
8168 See man:lttng-crash(1).
8170 LTTng Documentation::
8173 <<lttng-live,LTTng live>>::
8174 A communication protocol between the <<lttng-relayd,relay daemon>> and
8175 live viewers which makes it possible to see <<def-event-record,event
8176 records>> ``live'', as they are received by the
8177 <<def-relay-daemon,relay daemon>>.
8179 <<lttng-modules,LTTng-modules>>::
8180 The https://github.com/lttng/lttng-modules[LTTng-modules] project,
8181 which contains the Linux kernel modules to make the Linux kernel
8182 <<def-instrumentation-point,instrumentation points>> available for
8183 <<def-lttng,LTTng>> tracing.
8186 The name of the <<def-relay-daemon,relay daemon>> program.
8188 cmd:lttng-sessiond::
8189 The name of the <<def-session-daemon,session daemon>> program.
8191 [[def-lttng-tools]]LTTng-tools::
8192 The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
8193 contains the various programs and libraries used to
8194 <<controlling-tracing,control tracing>>.
8196 [[def-lttng-ust]]<<lttng-ust,LTTng-UST>>::
8197 The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
8198 contains libraries to instrument
8199 <<def-user-application,user applications>>.
8201 <<lttng-ust-agents,LTTng-UST Java agent>>::
8202 A Java package provided by the <<def-lttng-ust,LTTng-UST>> project to
8203 allow the LTTng instrumentation of `java.util.logging` and Apache
8204 log4j{nbsp}1.2 logging statements.
8206 <<lttng-ust-agents,LTTng-UST Python agent>>::
8207 A Python package provided by the <<def-lttng-ust,LTTng-UST>> project
8208 to allow the <<def-lttng,LTTng>> instrumentation of Python logging
8211 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
8212 The <<def-event-record-loss-mode,event record loss mode>> in which new
8213 <<def-event-record,event records>> _overwrite_ older event records
8214 when there's no <<def-sub-buffer,sub-buffer>> space left to store
8217 <<channel-buffering-schemes,per-process buffering>>::
8218 A <<def-buffering-scheme,buffering scheme>> in which each instrumented
8219 process has its own <<def-sub-buffer,sub-buffers>> for a given user
8220 space <<def-channel,channel>>.
8222 <<channel-buffering-schemes,per-user buffering>>::
8223 A <<def-buffering-scheme,buffering scheme>> in which all the processes
8224 of a Unix user share the same <<def-sub-buffer,sub-buffers>> for a
8225 given user space <<def-channel,channel>>.
8227 [[def-proc-attr]]process attribute::
8228 In the <<pid-tracking,process attribute tracking>> context:
8231 * A virtual process ID.
8233 * A virtual Unix user ID.
8235 * A virtual Unix group ID.
8237 [[def-relay-daemon]]<<lttng-relayd,relay daemon>>::
8238 A process which is responsible for receiving the <<def-trace,trace>>
8239 data which a distant <<def-consumer-daemon,consumer daemon>> sends.
8241 [[def-ring-buffer]]ring buffer::
8242 A set of <<def-sub-buffer,sub-buffers>>.
8245 See _<<def-tracing-session-rotation,tracing session rotation>>_.
8247 [[def-session-daemon]]<<lttng-sessiond,session daemon>>::
8248 A process which receives control commands from you and orchestrates
8249 the <<def-tracer,tracers>> and various <<def-lttng,LTTng>> daemons.
8251 <<taking-a-snapshot,snapshot>>::
8252 A copy of the current data of all the <<def-sub-buffer,sub-buffers>>
8253 of a given <<def-tracing-session,tracing session>>, saved as
8254 <<def-trace,trace>> files.
8256 [[def-sub-buffer]]sub-buffer::
8257 One part of an <<def-lttng,LTTng>> <<def-ring-buffer,ring buffer>>
8258 which contains <<def-event-record,event records>>.
8261 The time information attached to an <<def-event,event>> when it is
8264 [[def-trace]]trace (_noun_)::
8267 * One http://diamon.org/ctf/[CTF] metadata stream file.
8268 * One or more CTF data stream files which are the concatenations of one
8269 or more flushed <<def-sub-buffer,sub-buffers>>.
8271 [[def-trace-verb]]trace (_verb_)::
8272 The action of recording the <<def-event,events>> emitted by an
8273 application or by a system, or to initiate such recording by
8274 controlling a <<def-tracer,tracer>>.
8276 [[def-trace-chunk]]trace chunk::
8277 A self-contained <<def-trace,trace>> which is part of a
8278 <<def-tracing-session,tracing session>>. Each
8279 <<def-tracing-session-rotation, tracing session rotation>> produces a
8280 <<def-trace-chunk-archive,trace chunk archive>>.
8282 [[def-trace-chunk-archive]]trace chunk archive::
8283 The result of a <<def-tracing-session-rotation, tracing session rotation>>.
8285 <<def-lttng,LTTng>> doesn't manage any trace chunk archive, even if its
8286 containing <<def-tracing-session,tracing session>> is still active: you
8287 are free to read it, modify it, move it, or remove it.
8290 The http://tracecompass.org[Trace Compass] project and application.
8292 [[def-tracepoint]]tracepoint::
8293 An instrumentation point using the tracepoint mechanism of the Linux
8294 kernel or of <<def-lttng-ust,LTTng-UST>>.
8296 tracepoint definition::
8297 The definition of a single <<def-tracepoint,tracepoint>>.
8300 The name of a <<def-tracepoint,tracepoint>>.
8302 [[def-tracepoint-provider]]tracepoint provider::
8303 A set of functions providing <<def-tracepoint,tracepoints>> to an
8304 instrumented <<def-user-application,user application>>.
8306 Not to be confused with a <<def-tracepoint-provider-package,tracepoint
8307 provider package>>: many tracepoint providers can exist within a
8308 tracepoint provider package.
8310 [[def-tracepoint-provider-package]]tracepoint provider package::
8311 One or more <<def-tracepoint-provider,tracepoint providers>> compiled
8312 as an https://en.wikipedia.org/wiki/Object_file[object file] or as a
8313 link:https://en.wikipedia.org/wiki/Library_(computing)#Shared_libraries[shared
8316 [[def-tracer]]tracer::
8317 A software which records emitted <<def-event,events>>.
8319 <<domain,tracing domain>>::
8320 A namespace for <<def-event,event>> sources.
8322 <<tracing-group,tracing group>>::
8323 The Unix group in which a Unix user can be to be allowed to
8324 <<def-trace-verb,trace>> the Linux kernel.
8326 [[def-tracing-session]]<<tracing-session,tracing session>>::
8327 A stateful dialogue between you and a <<lttng-sessiond,session daemon>>.
8329 [[def-tracing-session-rotation]]<<session-rotation,tracing session rotation>>::
8330 The action of archiving the
8331 <<def-current-trace-chunk,current trace chunk>> of a
8332 <<def-tracing-session,tracing session>>.
8334 tracked <<def-proc-attr,process attribute>>::
8335 A process attribute which is part of an <<def-incl-set,inclusion
8338 untracked process attribute::
8339 A process attribute which isn't part of an <<def-incl-set,inclusion
8342 [[def-user-application]]user application::
8343 An application running in user space, as opposed to a Linux kernel
8344 module, for example.