1 The LTTng Documentation
2 =======================
3 Philippe Proulx <pproulx@efficios.com>
7 include::../common/copyright.txt[]
10 include::../common/welcome.txt[]
13 include::../common/audience.txt[]
17 === What's in this documentation?
19 The LTTng Documentation is divided into the following sections:
21 * **<<nuts-and-bolts,Nuts and bolts>>** explains the
22 rudiments of software tracing and the rationale behind the
25 You can skip this section if you’re familiar with software tracing and
26 with the LTTng project.
28 * **<<installing-lttng,Installation>>** describes the steps to
29 install the LTTng packages on common Linux distributions and from
32 You can skip this section if you already properly installed LTTng on
35 * **<<getting-started,Quick start>>** is a concise guide to
36 getting started quickly with LTTng kernel and user space tracing.
38 We recommend this section if you're new to LTTng or to software tracing
41 You can skip this section if you're not new to LTTng.
43 * **<<core-concepts,Core concepts>>** explains the concepts at
46 It's a good idea to become familiar with the core concepts
47 before attempting to use the toolkit.
49 * **<<plumbing,Components of LTTng>>** describes the various components
50 of the LTTng machinery, like the daemons, the libraries, and the
51 command-line interface.
52 * **<<instrumenting,Instrumentation>>** shows different ways to
53 instrument user applications and the Linux kernel.
55 Instrumenting source code is essential to provide a meaningful
58 You can skip this section if you do not have a programming background.
60 * **<<controlling-tracing,Tracing control>>** is divided into topics
61 which demonstrate how to use the vast array of features that
62 LTTng{nbsp}{revision} offers.
63 * **<<reference,Reference>>** contains reference tables.
64 * **<<glossary,Glossary>>** is a specialized dictionary of terms related
65 to LTTng or to the field of software tracing.
68 include::../common/convention.txt[]
71 include::../common/acknowledgements.txt[]
75 == What's new in LTTng {revision}?
77 LTTng{nbsp}{revision} bears the name _Joannès_. A Berliner Weisse style
78 beer from the http://letreflenoir.com/[Trèfle Noir] microbrewery in
79 https://en.wikipedia.org/wiki/Rouyn-Noranda[Rouyn-Noranda], the
80 https://www.beeradvocate.com/beer/profile/20537/238967/[_**Joannès**_]
81 is a tangy beer with a distinct pink dress and intense fruit flavor,
82 thanks to the presence of fresh blackcurrant grown in Témiscamingue.
84 New features and changes in LTTng{nbsp}{revision}:
86 * **Tracing control**:
87 ** You can override the name or the URL of a tracing session
88 configuration when you use man:lttng-load(1) thanks to the new
89 opt:lttng-load(1):--override-name and
90 opt:lttng-load(1):--override-url options.
91 ** The new `lttng regenerate` command replaces the now deprecated
92 `lttng metadata` command of LTTng 2.8. man:lttng-regenerate(1) can
93 also <<regenerate-statedump,generate the state dump event records>>
94 of a given tracing session on demand, a handy feature when
95 <<taking-a-snapshot,taking a snapshot>>.
96 ** You can add PMU counters by raw ID with man:lttng-add-context(1):
101 $ lttng add-context --kernel --type=perf:cpu:raw:r0013c:x86unhalted
105 The format of the raw ID is the same as used with man:perf-record(1).
106 See <<adding-context,Add context fields to a channel>> for more
109 ** The LTTng <<lttng-relayd,relay daemon>> is now supported on
110 OS{nbsp}X and macOS for a smoother integration within a trace
111 analysis workflow, regardless of the platform used.
113 * **User space tracing**:
114 ** Improved performance (tested on x86-64 and ARMv7-A
115 (https://en.wikipedia.org/wiki/Cubieboard[Cubieboard])
117 ** New helper library (`liblttng-ust-fd`) to help with
118 <<liblttng-ust-fd,applications which close file descriptors that
119 don't belong to them>>, for example, in a loop which closes file
120 descriptors after man:fork(2), or BSD's `closeall()`.
121 ** More accurate <<liblttng-ust-dl,dynamic linker instrumentation>> and
122 state dump event records, especially when a dynamically loaded
123 library manually loads its own dependencies.
124 ** New `ctf_*()` field definition macros (see man:lttng-ust(3)):
125 *** `ctf_array_hex()`
126 *** `ctf_array_network()`
127 *** `ctf_array_network_hex()`
128 *** `ctf_sequence_hex()`
129 *** `ctf_sequence_network()`
130 *** `ctf_sequence_network_hex()`
131 ** New `lttng_ust_loaded` weak symbol defined by `liblttng-ust` for
132 an application to know if the LTTng-UST shared library is loaded
140 int lttng_ust_loaded __attribute__((weak));
144 if (lttng_ust_loaded) {
145 puts("LTTng-UST is loaded!");
147 puts("LTTng-UST is not loaded!");
155 ** LTTng-UST thread names have the `-ust` suffix.
157 * **Linux kernel tracing**:
158 ** Improved performance (tested on x86-64 and ARMv7-A
159 (https://en.wikipedia.org/wiki/Cubieboard[Cubieboard])
161 ** New enumeration <<lttng-modules-tp-fields,field definition macros>>:
162 `ctf_enum()` and `ctf_user_enum()`.
163 ** IPv4, IPv6, and TCP header data is recorded in the event records
164 produced by tracepoints starting with `net_`.
165 ** Detailed system call event records: `select`, `pselect6`, `poll`,
166 `ppoll`, `epoll_wait`, `epoll_pwait`, and `epoll_ctl` on all
167 architectures supported by LTTng-modules, and `accept4` on x86-64.
168 ** New I²C instrumentation: the `extract_sensitive_payload` parameter
169 of the new `lttng-probe-i2c` LTTng module controls whether or not
170 the payloads of I²C messages are recorded in I²C event records, since
171 they may contain sensitive data (for example, keystrokes).
172 ** When the LTTng kernel modules are built into the Linux kernel image,
173 the `CONFIG_TRACEPOINTS` configuration option is automatically
180 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
181 generation_ is a modern toolkit for tracing Linux systems and
182 applications. So your first question might be:
189 As the history of software engineering progressed and led to what
190 we now take for granted--complex, numerous and
191 interdependent software applications running in parallel on
192 sophisticated operating systems like Linux--the authors of such
193 components, software developers, began feeling a natural
194 urge to have tools that would ensure the robustness and good performance
195 of their masterpieces.
197 One major achievement in this field is, inarguably, the
198 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
199 an essential tool for developers to find and fix bugs. But even the best
200 debugger won't help make your software run faster, and nowadays, faster
201 software means either more work done by the same hardware, or cheaper
202 hardware for the same work.
204 A _profiler_ is often the tool of choice to identify performance
205 bottlenecks. Profiling is suitable to identify _where_ performance is
206 lost in a given software. The profiler outputs a profile, a statistical
207 summary of observed events, which you may use to discover which
208 functions took the most time to execute. However, a profiler won't
209 report _why_ some identified functions are the bottleneck. Bottlenecks
210 might only occur when specific conditions are met, conditions that are
211 sometimes impossible to capture by a statistical profiler, or impossible
212 to reproduce with an application altered by the overhead of an
213 event-based profiler. For a thorough investigation of software
214 performance issues, a history of execution is essential, with the
215 recorded values of variables and context fields you choose, and
216 with as little influence as possible on the instrumented software. This
217 is where tracing comes in handy.
219 _Tracing_ is a technique used to understand what goes on in a running
220 software system. The software used for tracing is called a _tracer_,
221 which is conceptually similar to a tape recorder. When recording,
222 specific instrumentation points placed in the software source code
223 generate events that are saved on a giant tape: a _trace_ file. You
224 can trace user applications and the operating system at the same time,
225 opening the possibility of resolving a wide range of problems that would
226 otherwise be extremely challenging.
228 Tracing is often compared to _logging_. However, tracers and loggers are
229 two different tools, serving two different purposes. Tracers are
230 designed to record much lower-level events that occur much more
231 frequently than log messages, often in the range of thousands per
232 second, with very little execution overhead. Logging is more appropriate
233 for a very high-level analysis of less frequent events: user accesses,
234 exceptional conditions (errors and warnings, for example), database
235 transactions, instant messaging communications, and such. Simply put,
236 logging is one of the many use cases that can be satisfied with tracing.
238 The list of recorded events inside a trace file can be read manually
239 like a log file for the maximum level of detail, but it is generally
240 much more interesting to perform application-specific analyses to
241 produce reduced statistics and graphs that are useful to resolve a
242 given problem. Trace viewers and analyzers are specialized tools
245 In the end, this is what LTTng is: a powerful, open source set of
246 tools to trace the Linux kernel and user applications at the same time.
247 LTTng is composed of several components actively maintained and
248 developed by its link:/community/#where[community].
251 [[lttng-alternatives]]
252 === Alternatives to noch:{LTTng}
254 Excluding proprietary solutions, a few competing software tracers
257 * https://github.com/dtrace4linux/linux[dtrace4linux] is a port of
258 Sun Microsystems's DTrace to Linux. The cmd:dtrace tool interprets
259 user scripts and is responsible for loading code into the
260 Linux kernel for further execution and collecting the outputted data.
261 * https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF] is a
262 subsystem in the Linux kernel in which a virtual machine can execute
263 programs passed from the user space to the kernel. You can attach
264 such programs to tracepoints and KProbes thanks to a system call, and
265 they can output data to the user space when executed thanks to
266 different mechanisms (pipe, VM register values, and eBPF maps, to name
268 * https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]
269 is the de facto function tracer of the Linux kernel. Its user
270 interface is a set of special files in sysfs.
271 * https://perf.wiki.kernel.org/[perf] is
272 a performance analyzing tool for Linux which supports hardware
273 performance counters, tracepoints, as well as other counters and
274 types of probes. perf's controlling utility is the cmd:perf command
276 * http://linux.die.net/man/1/strace[strace]
277 is a command-line utility which records system calls made by a
278 user process, as well as signal deliveries and changes of process
279 state. strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace]
280 to fulfill its function.
281 * http://www.sysdig.org/[sysdig], like SystemTap, uses scripts to
282 analyze Linux kernel events. You write scripts, or _chisels_ in
283 sysdig's jargon, in Lua and sysdig executes them while the system is
284 being traced or afterwards. sysdig's interface is the cmd:sysdig
285 command-line tool as well as the curses-based cmd:csysdig tool.
286 * https://sourceware.org/systemtap/[SystemTap] is a Linux kernel and
287 user space tracer which uses custom user scripts to produce plain text
288 traces. SystemTap converts the scripts to the C language, and then
289 compiles them as Linux kernel modules which are loaded to produce
290 trace data. SystemTap's primary user interface is the cmd:stap
293 The main distinctive features of LTTng is that it produces correlated
294 kernel and user space traces, as well as doing so with the lowest
295 overhead amongst other solutions. It produces trace files in the
296 http://diamon.org/ctf[CTF] format, a file format optimized
297 for the production and analyses of multi-gigabyte data.
299 LTTng is the result of more than 10 years of active open source
300 development by a community of passionate developers.
301 LTTng{nbsp}{revision} is currently available on major desktop and server
304 The main interface for tracing control is a single command-line tool
305 named cmd:lttng. The latter can create several tracing sessions, enable
306 and disable events on the fly, filter events efficiently with custom
307 user expressions, start and stop tracing, and much more. LTTng can
308 record the traces on the file system or send them over the network, and
309 keep them totally or partially. You can view the traces once tracing
310 becomes inactive or in real-time.
312 <<installing-lttng,Install LTTng now>> and
313 <<getting-started,start tracing>>!
319 **LTTng** is a set of software <<plumbing,components>> which interact to
320 <<instrumenting,instrument>> the Linux kernel and user applications, and
321 to <<controlling-tracing,control tracing>> (start and stop
322 tracing, enable and disable event rules, and the rest). Those
323 components are bundled into the following packages:
325 * **LTTng-tools**: Libraries and command-line interface to
327 * **LTTng-modules**: Linux kernel modules to instrument and
329 * **LTTng-UST**: Libraries and Java/Python packages to instrument and
330 trace user applications.
332 Most distributions mark the LTTng-modules and LTTng-UST packages as
333 optional when installing LTTng-tools (which is always required). In the
334 following sections, we always provide the steps to install all three,
337 * You only need to install LTTng-modules if you intend to trace the
339 * You only need to install LTTng-UST if you intend to trace user
343 .Availability of LTTng{nbsp}{revision} for major Linux distributions as of 14 March 2017.
345 |Distribution |Available in releases |Alternatives
347 |https://www.ubuntu.com/[Ubuntu]
348 |Ubuntu{nbsp}14.04 _Trusty Tahr_ and Ubuntu{nbsp}16.04 _Xenial Xerus_:
349 <<ubuntu-ppa,use the LTTng Stable{nbsp}{revision} PPA>>.
350 |link:/docs/v2.8#doc-ubuntu[LTTng{nbsp}2.8 for Ubuntu{nbsp}16.10 _Yakkety Yak_].
352 LTTng{nbsp}{revision} for Ubuntu{nbsp}17.04 _Zesty Zapus_ (not released
355 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
356 other Ubuntu releases.
358 |https://getfedora.org/[Fedora]
360 |link:/docs/v2.8#doc-fedora[LTTng{nbsp}2.8 for Fedora 25].
362 <<building-from-source,Build LTTng-modules{nbsp}{revision} from
365 |https://www.debian.org/[Debian]
366 |<<debian,Debian "stretch" (testing) and Debian "sid" (unstable)>>.
367 |<<building-from-source,Build LTTng{nbsp}{revision} from source>> for
368 other Debian releases.
370 |https://www.opensuse.org/[openSUSE]
372 |<<building-from-source,Build LTTng-modules{nbsp}{revision} from
375 |https://www.archlinux.org/[Arch Linux]
376 |<<arch-linux,Latest AUR packages>>.
379 |https://alpinelinux.org/[Alpine Linux]
380 |<<alpine-linux,Alpine Linux "edge">>.
381 |<<building-from-source,Build LTTng{nbsp}{revision} from source>> for
382 other Alpine Linux releases.
384 |https://www.redhat.com/[RHEL] and https://www.suse.com/[SLES]
385 |See http://packages.efficios.com/[EfficiOS Enterprise Packages].
388 |https://buildroot.org/[Buildroot]
389 |<<buildroot,Buildroot 2017.02>>.
390 |link:/docs/v2.8#doc-buildroot[LTTng{nbsp}2.8 for Buildroot 2016.11].
392 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
393 other Buildroot releases.
395 |http://www.openembedded.org/wiki/Main_Page[OpenEmbedded] and
396 https://www.yoctoproject.org/[Yocto]
398 |link:/docs/v2.8#doc-oe-yocto[LTTng{nbsp}2.8 for Yocto Project{nbsp}2.2 _Morty_]
399 (`openembedded-core` layer).
401 LTTng{nbsp}{revision} for Yocto Project{nbsp}2.3 _Pyro_
402 (`openembedded-core` layer) (not released yet).
404 <<building-from-source,Build LTTng{nbsp}{revision} from source>> for
405 other OpenEmbedded releases.
410 === [[ubuntu-official-repositories]]Ubuntu
413 ==== noch:{LTTng} Stable {revision} PPA
415 The https://launchpad.net/~lttng/+archive/ubuntu/stable-{revision}[LTTng
416 Stable{nbsp}{revision} PPA] offers the latest stable
417 LTTng{nbsp}{revision} packages for:
419 * Ubuntu{nbsp}14.04 _Trusty Tahr_
420 * Ubuntu{nbsp}16.04 _Xenial Xerus_
422 To install LTTng{nbsp}{revision} from the LTTng Stable{nbsp}{revision} PPA:
424 . Add the LTTng Stable{nbsp}{revision} PPA repository and update the
430 # apt-add-repository ppa:lttng/stable-2.9
435 . Install the main LTTng{nbsp}{revision} packages:
440 # apt-get install lttng-tools
441 # apt-get install lttng-modules-dkms
442 # apt-get install liblttng-ust-dev
446 . **If you need to instrument and trace
447 <<java-application,Java applications>>**, install the LTTng-UST
453 # apt-get install liblttng-ust-agent-java
457 . **If you need to instrument and trace
458 <<python-application,Python{nbsp}3 applications>>**, install the
459 LTTng-UST Python agent:
464 # apt-get install python3-lttngust
472 To install LTTng{nbsp}{revision} on Debian "stretch" (testing) or
473 Debian "sid" (unstable):
475 . Install the main LTTng{nbsp}{revision} packages:
480 # apt-get install lttng-modules-dkms
481 # apt-get install liblttng-ust-dev
482 # apt-get install lttng-tools
486 . **If you need to instrument and trace <<java-application,Java
487 applications>>**, install the LTTng-UST Java agent:
492 # apt-get install liblttng-ust-agent-java
496 . **If you need to instrument and trace <<python-application,Python
497 applications>>**, install the LTTng-UST Python agent:
502 # apt-get install python3-lttngust
510 To install LTTng{nbsp}{revision} on Arch Linux using
511 https://archlinux.fr/yaourt-en[Yaourt]:
513 . Install the main LTTng{nbsp}{revision} packages:
518 $ yaourt -S lttng-tools
519 $ yaourt -S lttng-ust
520 $ yaourt -S lttng-modules
524 . **If you need to instrument and trace <<python-application,Python
525 applications>>**, install the LTTng-UST Python agent:
530 $ yaourt -S python-lttngust
531 $ yaourt -S python2-lttngust
539 To install LTTng-tools{nbsp}{revision} and LTTng-UST{nbsp}{revision} on
542 . Make sure your system is
543 https://wiki.alpinelinux.org/wiki/Edge[configured for "edge"].
544 . Enable the _testing_ repository by uncommenting the corresponding
545 line in path:{/etc/apk/repositories}.
546 . Add the LTTng packages:
551 # apk add lttng-tools
552 # apk add lttng-ust-dev
556 To install LTTng-modules{nbsp}{revision} (Linux kernel tracing support)
557 on Alpine Linux "edge":
559 . Add the vanilla Linux kernel:
564 # apk add linux-vanilla linux-vanilla-dev
568 . Reboot with the vanilla Linux kernel.
569 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
575 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
576 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
577 cd lttng-modules-2.8.* &&
579 sudo make modules_install &&
585 [[enterprise-distributions]]
586 === RHEL, SUSE, and other enterprise distributions
588 To install LTTng on enterprise Linux distributions, such as Red Hat
589 Enterprise Linux (RHEL) and SUSE Linux Enterprise Server (SUSE), please
590 see http://packages.efficios.com/[EfficiOS Enterprise Packages].
596 To install LTTng{nbsp}{revision} on Buildroot{nbsp}2017.02:
598 . Launch the Buildroot configuration tool:
607 . In **Kernel**, check **Linux kernel**.
608 . In **Toolchain**, check **Enable WCHAR support**.
609 . In **Target packages**{nbsp}→ **Debugging, profiling and benchmark**,
610 check **lttng-modules** and **lttng-tools**.
611 . In **Target packages**{nbsp}→ **Libraries**{nbsp}→
612 **Other**, check **lttng-libust**.
615 [[building-from-source]]
616 === Build from source
618 To build and install LTTng{nbsp}{revision} from source:
620 . Using your distribution's package manager, or from source, install
621 the following dependencies of LTTng-tools and LTTng-UST:
624 * https://sourceforge.net/projects/libuuid/[libuuid]
625 * http://directory.fsf.org/wiki/Popt[popt]
626 * http://liburcu.org/[Userspace RCU]
627 * http://www.xmlsoft.org/[libxml2]
630 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
636 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.9.tar.bz2 &&
637 tar -xf lttng-modules-latest-2.9.tar.bz2 &&
638 cd lttng-modules-2.9.* &&
640 sudo make modules_install &&
645 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
651 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.9.tar.bz2 &&
652 tar -xf lttng-ust-latest-2.9.tar.bz2 &&
653 cd lttng-ust-2.9.* &&
663 .Java and Python application tracing
665 If you need to instrument and trace <<java-application,Java
666 applications>>, pass the `--enable-java-agent-jul`,
667 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
668 `configure` script, depending on which Java logging framework you use.
670 If you need to instrument and trace <<python-application,Python
671 applications>>, pass the `--enable-python-agent` option to the
672 `configure` script. You can set the `PYTHON` environment variable to the
673 path to the Python interpreter for which to install the LTTng-UST Python
681 By default, LTTng-UST libraries are installed to
682 dir:{/usr/local/lib}, which is the de facto directory in which to
683 keep self-compiled and third-party libraries.
685 When <<building-tracepoint-providers-and-user-application,linking an
686 instrumented user application with `liblttng-ust`>>:
688 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
690 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
691 man:gcc(1), man:g++(1), or man:clang(1).
695 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
701 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.9.tar.bz2 &&
702 tar -xf lttng-tools-latest-2.9.tar.bz2 &&
703 cd lttng-tools-2.9.* &&
711 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
712 previous steps automatically for a given version of LTTng and confine
713 the installed files in a specific directory. This can be useful to test
714 LTTng without installing it on your system.
720 This is a short guide to get started quickly with LTTng kernel and user
723 Before you follow this guide, make sure to <<installing-lttng,install>>
726 This tutorial walks you through the steps to:
728 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
729 . <<tracing-your-own-user-application,Trace a user application>> written
731 . <<viewing-and-analyzing-your-traces,View and analyze the
735 [[tracing-the-linux-kernel]]
736 === Trace the Linux kernel
738 The following command lines start with the `#` prompt because you need
739 root privileges to trace the Linux kernel. You can also trace the kernel
740 as a regular user if your Unix user is a member of the
741 <<tracing-group,tracing group>>.
743 . Create a <<tracing-session,tracing session>> which writes its traces
744 to dir:{/tmp/my-kernel-trace}:
749 # lttng create my-kernel-session --output=/tmp/my-kernel-trace
753 . List the available kernel tracepoints and system calls:
758 # lttng list --kernel
759 # lttng list --kernel --syscall
763 . Create <<event,event rules>> which match the desired instrumentation
764 point names, for example the `sched_switch` and `sched_process_fork`
765 tracepoints, and the man:open(2) and man:close(2) system calls:
770 # lttng enable-event --kernel sched_switch,sched_process_fork
771 # lttng enable-event --kernel --syscall open,close
775 You can also create an event rule which matches _all_ the Linux kernel
776 tracepoints (this will generate a lot of data when tracing):
781 # lttng enable-event --kernel --all
785 . <<basic-tracing-session-control,Start tracing>>:
794 . Do some operation on your system for a few seconds. For example,
795 load a website, or list the files of a directory.
796 . <<basic-tracing-session-control,Stop tracing>> and destroy the
807 The man:lttng-destroy(1) command does not destroy the trace data; it
808 only destroys the state of the tracing session.
810 . For the sake of this example, make the recorded trace accessible to
816 # chown -R $(whoami) /tmp/my-kernel-trace
820 See <<viewing-and-analyzing-your-traces,View and analyze the
821 recorded events>> to view the recorded events.
824 [[tracing-your-own-user-application]]
825 === Trace a user application
827 This section steps you through a simple example to trace a
828 _Hello world_ program written in C.
830 To create the traceable user application:
832 . Create the tracepoint provider header file, which defines the
833 tracepoints and the events they can generate:
839 #undef TRACEPOINT_PROVIDER
840 #define TRACEPOINT_PROVIDER hello_world
842 #undef TRACEPOINT_INCLUDE
843 #define TRACEPOINT_INCLUDE "./hello-tp.h"
845 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
848 #include <lttng/tracepoint.h>
858 ctf_string(my_string_field, my_string_arg)
859 ctf_integer(int, my_integer_field, my_integer_arg)
863 #endif /* _HELLO_TP_H */
865 #include <lttng/tracepoint-event.h>
869 . Create the tracepoint provider package source file:
875 #define TRACEPOINT_CREATE_PROBES
876 #define TRACEPOINT_DEFINE
878 #include "hello-tp.h"
882 . Build the tracepoint provider package:
887 $ gcc -c -I. hello-tp.c
891 . Create the _Hello World_ application source file:
898 #include "hello-tp.h"
900 int main(int argc, char *argv[])
904 puts("Hello, World!\nPress Enter to continue...");
907 * The following getchar() call is only placed here for the purpose
908 * of this demonstration, to pause the application in order for
909 * you to have time to list its tracepoints. It is not
915 * A tracepoint() call.
917 * Arguments, as defined in hello-tp.h:
919 * 1. Tracepoint provider name (required)
920 * 2. Tracepoint name (required)
921 * 3. my_integer_arg (first user-defined argument)
922 * 4. my_string_arg (second user-defined argument)
924 * Notice the tracepoint provider and tracepoint names are
925 * NOT strings: they are in fact parts of variables that the
926 * macros in hello-tp.h create.
928 tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
930 for (x = 0; x < argc; ++x) {
931 tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
934 puts("Quitting now!");
935 tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
942 . Build the application:
951 . Link the application with the tracepoint provider package,
952 `liblttng-ust`, and `libdl`:
957 $ gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
961 Here's the whole build process:
964 .User space tracing tutorial's build steps.
965 image::ust-flow.png[]
967 To trace the user application:
969 . Run the application with a few arguments:
974 $ ./hello world and beyond
983 Press Enter to continue...
987 . Start an LTTng <<lttng-sessiond,session daemon>>:
992 $ lttng-sessiond --daemonize
996 Note that a session daemon might already be running, for example as
997 a service that the distribution's service manager started.
999 . List the available user space tracepoints:
1004 $ lttng list --userspace
1008 You see the `hello_world:my_first_tracepoint` tracepoint listed
1009 under the `./hello` process.
1011 . Create a <<tracing-session,tracing session>>:
1016 $ lttng create my-user-space-session
1020 . Create an <<event,event rule>> which matches the
1021 `hello_world:my_first_tracepoint` event name:
1026 $ lttng enable-event --userspace hello_world:my_first_tracepoint
1030 . <<basic-tracing-session-control,Start tracing>>:
1039 . Go back to the running `hello` application and press Enter. The
1040 program executes all `tracepoint()` instrumentation points and exits.
1041 . <<basic-tracing-session-control,Stop tracing>> and destroy the
1052 The man:lttng-destroy(1) command does not destroy the trace data; it
1053 only destroys the state of the tracing session.
1055 By default, LTTng saves the traces in
1056 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
1057 where +__name__+ is the tracing session name. The
1058 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
1060 See <<viewing-and-analyzing-your-traces,View and analyze the
1061 recorded events>> to view the recorded events.
1064 [[viewing-and-analyzing-your-traces]]
1065 === View and analyze the recorded events
1067 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
1068 kernel>> and <<tracing-your-own-user-application,Trace a user
1069 application>> tutorials, you can inspect the recorded events.
1071 Many tools are available to read LTTng traces:
1073 * **cmd:babeltrace** is a command-line utility which converts trace
1074 formats; it supports the format that LTTng produces, CTF, as well as a
1075 basic text output which can be ++grep++ed. The cmd:babeltrace command
1076 is part of the http://diamon.org/babeltrace[Babeltrace] project.
1077 * Babeltrace also includes
1078 **https://www.python.org/[Python] bindings** so
1079 that you can easily open and read an LTTng trace with your own script,
1080 benefiting from the power of Python.
1081 * http://tracecompass.org/[**Trace Compass**]
1082 is a graphical user interface for viewing and analyzing any type of
1083 logs or traces, including LTTng's.
1084 * https://github.com/lttng/lttng-analyses[**LTTng analyses**] is a
1085 project which includes many high-level analyses of LTTng kernel
1086 traces, like scheduling statistics, interrupt frequency distribution,
1087 top CPU usage, and more.
1089 NOTE: This section assumes that the traces recorded during the previous
1090 tutorials were saved to their default location, in the
1091 dir:{$LTTNG_HOME/lttng-traces} directory. The env:LTTNG_HOME
1092 environment variable defaults to `$HOME` if not set.
1095 [[viewing-and-analyzing-your-traces-bt]]
1096 ==== Use the cmd:babeltrace command-line tool
1098 The simplest way to list all the recorded events of a trace is to pass
1099 its path to cmd:babeltrace with no options:
1103 $ babeltrace ~/lttng-traces/my-user-space-session*
1106 cmd:babeltrace finds all traces recursively within the given path and
1107 prints all their events, merging them in chronological order.
1109 You can pipe the output of cmd:babeltrace into a tool like man:grep(1) for
1114 $ babeltrace /tmp/my-kernel-trace | grep _switch
1117 You can pipe the output of cmd:babeltrace into a tool like man:wc(1) to
1118 count the recorded events:
1122 $ babeltrace /tmp/my-kernel-trace | grep _open | wc --lines
1126 [[viewing-and-analyzing-your-traces-bt-python]]
1127 ==== Use the Babeltrace Python bindings
1129 The <<viewing-and-analyzing-your-traces-bt,text output of cmd:babeltrace>>
1130 is useful to isolate events by simple matching using man:grep(1) and
1131 similar utilities. However, more elaborate filters, such as keeping only
1132 event records with a field value falling within a specific range, are
1133 not trivial to write using a shell. Moreover, reductions and even the
1134 most basic computations involving multiple event records are virtually
1135 impossible to implement.
1137 Fortunately, Babeltrace ships with Python 3 bindings which makes it easy
1138 to read the event records of an LTTng trace sequentially and compute the
1139 desired information.
1141 The following script accepts an LTTng Linux kernel trace path as its
1142 first argument and prints the short names of the top 5 running processes
1143 on CPU 0 during the whole trace:
1148 from collections import Counter
1154 if len(sys.argv) != 2:
1155 msg = 'Usage: python3 {} TRACEPATH'.format(sys.argv[0])
1156 print(msg, file=sys.stderr)
1159 # A trace collection contains one or more traces
1160 col = babeltrace.TraceCollection()
1162 # Add the trace provided by the user (LTTng traces always have
1164 if col.add_trace(sys.argv[1], 'ctf') is None:
1165 raise RuntimeError('Cannot add trace')
1167 # This counter dict contains execution times:
1169 # task command name -> total execution time (ns)
1170 exec_times = Counter()
1172 # This contains the last `sched_switch` timestamp
1176 for event in col.events:
1177 # Keep only `sched_switch` events
1178 if event.name != 'sched_switch':
1181 # Keep only events which happened on CPU 0
1182 if event['cpu_id'] != 0:
1186 cur_ts = event.timestamp
1192 # Previous task command (short) name
1193 prev_comm = event['prev_comm']
1195 # Initialize entry in our dict if not yet done
1196 if prev_comm not in exec_times:
1197 exec_times[prev_comm] = 0
1199 # Compute previous command execution time
1200 diff = cur_ts - last_ts
1202 # Update execution time of this command
1203 exec_times[prev_comm] += diff
1205 # Update last timestamp
1209 for name, ns in exec_times.most_common(5):
1211 print('{:20}{} s'.format(name, s))
1216 if __name__ == '__main__':
1217 sys.exit(0 if top5proc() else 1)
1224 $ python3 top5proc.py /tmp/my-kernel-trace/kernel
1230 swapper/0 48.607245889 s
1231 chromium 7.192738188 s
1232 pavucontrol 0.709894415 s
1233 Compositor 0.660867933 s
1234 Xorg.bin 0.616753786 s
1237 Note that `swapper/0` is the "idle" process of CPU 0 on Linux; since we
1238 weren't using the CPU that much when tracing, its first position in the
1243 == [[understanding-lttng]]Core concepts
1245 From a user's perspective, the LTTng system is built on a few concepts,
1246 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
1247 operates by sending commands to the <<lttng-sessiond,session daemon>>.
1248 Understanding how those objects relate to eachother is key in mastering
1251 The core concepts are:
1253 * <<tracing-session,Tracing session>>
1254 * <<domain,Tracing domain>>
1255 * <<channel,Channel and ring buffer>>
1256 * <<"event","Instrumentation point, event rule, event, and event record">>
1262 A _tracing session_ is a stateful dialogue between you and
1263 a <<lttng-sessiond,session daemon>>. You can
1264 <<creating-destroying-tracing-sessions,create a new tracing
1265 session>> with the `lttng create` command.
1267 Anything that you do when you control LTTng tracers happens within a
1268 tracing session. In particular, a tracing session:
1271 * Has its own set of trace files.
1272 * Has its own state of activity (started or stopped).
1273 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
1275 * Has its own <<channel,channels>> which have their own
1276 <<event,event rules>>.
1279 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
1280 image::concepts.png[]
1282 Those attributes and objects are completely isolated between different
1285 A tracing session is analogous to a cash machine session:
1286 the operations you do on the banking system through the cash machine do
1287 not alter the data of other users of the same system. In the case of
1288 the cash machine, a session lasts as long as your bank card is inside.
1289 In the case of LTTng, a tracing session lasts from the `lttng create`
1290 command to the `lttng destroy` command.
1293 .Each Unix user has its own set of tracing sessions.
1294 image::many-sessions.png[]
1297 [[tracing-session-mode]]
1298 ==== Tracing session mode
1300 LTTng can send the generated trace data to different locations. The
1301 _tracing session mode_ dictates where to send it. The following modes
1302 are available in LTTng{nbsp}{revision}:
1305 LTTng writes the traces to the file system of the machine being traced
1308 Network streaming mode::
1309 LTTng sends the traces over the network to a
1310 <<lttng-relayd,relay daemon>> running on a remote system.
1313 LTTng does not write the traces by default. Instead, you can request
1314 LTTng to <<taking-a-snapshot,take a snapshot>>, that is, a copy of the
1315 current tracing buffers, and to write it to the target's file system
1316 or to send it over the network to a <<lttng-relayd,relay daemon>>
1317 running on a remote system.
1320 This mode is similar to the network streaming mode, but a live
1321 trace viewer can connect to the distant relay daemon to
1322 <<lttng-live,view event records as LTTng generates them>> by
1329 A _tracing domain_ is a namespace for event sources. A tracing domain
1330 has its own properties and features.
1332 There are currently five available tracing domains:
1336 * `java.util.logging` (JUL)
1340 You must specify a tracing domain when using some commands to avoid
1341 ambiguity. For example, since all the domains support named tracepoints
1342 as event sources (instrumentation points that you manually insert in the
1343 source code), you need to specify a tracing domain when
1344 <<enabling-disabling-events,creating an event rule>> because all the
1345 tracing domains could have tracepoints with the same names.
1347 Some features are reserved to specific tracing domains. Dynamic function
1348 entry and return instrumentation points, for example, are currently only
1349 supported in the Linux kernel tracing domain, but support for other
1350 tracing domains could be added in the future.
1352 You can create <<channel,channels>> in the Linux kernel and user space
1353 tracing domains. The other tracing domains have a single default
1358 === Channel and ring buffer
1360 A _channel_ is an object which is responsible for a set of ring buffers.
1361 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1362 tracer emits an event, it can record it to one or more
1363 sub-buffers. The attributes of a channel determine what to do when
1364 there's no space left for a new event record because all sub-buffers
1365 are full, where to send a full sub-buffer, and other behaviours.
1367 A channel is always associated to a <<domain,tracing domain>>. The
1368 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1369 a default channel which you cannot configure.
1371 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1372 an event, it records it to the sub-buffers of all
1373 the enabled channels with a satisfied event rule, as long as those
1374 channels are part of active <<tracing-session,tracing sessions>>.
1377 [[channel-buffering-schemes]]
1378 ==== Per-user vs. per-process buffering schemes
1380 A channel has at least one ring buffer _per CPU_. LTTng always
1381 records an event to the ring buffer associated to the CPU on which it
1384 Two _buffering schemes_ are available when you
1385 <<enabling-disabling-channels,create a channel>> in the
1386 user space <<domain,tracing domain>>:
1388 Per-user buffering::
1389 Allocate one set of ring buffers--one per CPU--shared by all the
1390 instrumented processes of each Unix user.
1394 .Per-user buffering scheme.
1395 image::per-user-buffering.png[]
1398 Per-process buffering::
1399 Allocate one set of ring buffers--one per CPU--for each
1400 instrumented process.
1404 .Per-process buffering scheme.
1405 image::per-process-buffering.png[]
1408 The per-process buffering scheme tends to consume more memory than the
1409 per-user option because systems generally have more instrumented
1410 processes than Unix users running instrumented processes. However, the
1411 per-process buffering scheme ensures that one process having a high
1412 event throughput won't fill all the shared sub-buffers of the same
1415 The Linux kernel tracing domain has only one available buffering scheme
1416 which is to allocate a single set of ring buffers for the whole system.
1417 This scheme is similar to the per-user option, but with a single, global
1418 user "running" the kernel.
1421 [[channel-overwrite-mode-vs-discard-mode]]
1422 ==== Overwrite vs. discard event loss modes
1424 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1425 arc in the following animation) of a specific channel's ring buffer.
1426 When there's no space left in a sub-buffer, the tracer marks it as
1427 consumable (red) and another, empty sub-buffer starts receiving the
1428 following event records. A <<lttng-consumerd,consumer daemon>>
1429 eventually consumes the marked sub-buffer (returns to white).
1432 [role="docsvg-channel-subbuf-anim"]
1437 In an ideal world, sub-buffers are consumed faster than they are filled,
1438 as is the case in the previous animation. In the real world,
1439 however, all sub-buffers can be full at some point, leaving no space to
1440 record the following events.
1442 By design, LTTng is a _non-blocking_ tracer: when no empty sub-buffer is
1443 available, it is acceptable to lose event records when the alternative
1444 would be to cause substantial delays in the instrumented application's
1445 execution. LTTng privileges performance over integrity; it aims at
1446 perturbing the traced system as little as possible in order to make
1447 tracing of subtle race conditions and rare interrupt cascades possible.
1449 When it comes to losing event records because no empty sub-buffer is
1450 available, the channel's _event loss mode_ determines what to do. The
1451 available event loss modes are:
1454 Drop the newest event records until a the tracer
1455 releases a sub-buffer.
1458 Clear the sub-buffer containing the oldest event records and start
1459 writing the newest event records there.
1461 This mode is sometimes called _flight recorder mode_ because it's
1463 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1464 always keep a fixed amount of the latest data.
1466 Which mechanism you should choose depends on your context: prioritize
1467 the newest or the oldest event records in the ring buffer?
1469 Beware that, in overwrite mode, the tracer abandons a whole sub-buffer
1470 as soon as a there's no space left for a new event record, whereas in
1471 discard mode, the tracer only discards the event record that doesn't
1474 In discard mode, LTTng increments a count of lost event records when
1475 an event record is lost and saves this count to the trace. In
1476 overwrite mode, LTTng keeps no information when it overwrites a
1477 sub-buffer before consuming it.
1479 There are a few ways to decrease your probability of losing event
1481 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1482 how you can fine-une the sub-buffer count and size of a channel to
1483 virtually stop losing event records, though at the cost of greater
1487 [[channel-subbuf-size-vs-subbuf-count]]
1488 ==== Sub-buffer count and size
1490 When you <<enabling-disabling-channels,create a channel>>, you can
1491 set its number of sub-buffers and their size.
1493 Note that there is noticeable CPU overhead introduced when
1494 switching sub-buffers (marking a full one as consumable and switching
1495 to an empty one for the following events to be recorded). Knowing this,
1496 the following list presents a few practical situations along with how
1497 to configure the sub-buffer count and size for them:
1499 * **High event throughput**: In general, prefer bigger sub-buffers to
1500 lower the risk of losing event records.
1502 Having bigger sub-buffers also ensures a lower sub-buffer switching
1505 The number of sub-buffers is only meaningful if you create the channel
1506 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1507 other sub-buffers are left unaltered.
1509 * **Low event throughput**: In general, prefer smaller sub-buffers
1510 since the risk of losing event records is low.
1512 Because events occur less frequently, the sub-buffer switching frequency
1513 should remain low and thus the tracer's overhead should not be a
1516 * **Low memory system**: If your target system has a low memory
1517 limit, prefer fewer first, then smaller sub-buffers.
1519 Even if the system is limited in memory, you want to keep the
1520 sub-buffers as big as possible to avoid a high sub-buffer switching
1523 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1524 which means event data is very compact. For example, the average
1525 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1526 sub-buffer size of 1{nbsp}MiB is considered big.
1528 The previous situations highlight the major trade-off between a few big
1529 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1530 frequency vs. how much data is lost in overwrite mode. Assuming a
1531 constant event throughput and using the overwrite mode, the two
1532 following configurations have the same ring buffer total size:
1535 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1540 * **2 sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1541 switching frequency, but if a sub-buffer overwrite happens, half of
1542 the event records so far (4{nbsp}MiB) are definitely lost.
1543 * **8 sub-buffers of 1{nbsp}MiB each**: Expect 4{nbsp}times the tracer's
1544 overhead as the previous configuration, but if a sub-buffer
1545 overwrite happens, only the eighth of event records so far are
1548 In discard mode, the sub-buffers count parameter is pointless: use two
1549 sub-buffers and set their size according to the requirements of your
1553 [[channel-switch-timer]]
1554 ==== Switch timer period
1556 The _switch timer period_ is an important configurable attribute of
1557 a channel to ensure periodic sub-buffer flushing.
1559 When the _switch timer_ expires, a sub-buffer switch happens. You can
1560 set the switch timer period attribute when you
1561 <<enabling-disabling-channels,create a channel>> to ensure that event
1562 data is consumed and committed to trace files or to a distant relay
1563 daemon periodically in case of a low event throughput.
1566 [role="docsvg-channel-switch-timer"]
1571 This attribute is also convenient when you use big sub-buffers to cope
1572 with a sporadic high event throughput, even if the throughput is
1576 [[channel-read-timer]]
1577 ==== Read timer period
1579 By default, the LTTng tracers use a notification mechanism to signal a
1580 full sub-buffer so that a consumer daemon can consume it. When such
1581 notifications must be avoided, for example in real-time applications,
1582 you can use the channel's _read timer_ instead. When the read timer
1583 fires, the <<lttng-consumerd,consumer daemon>> checks for full,
1584 consumable sub-buffers.
1587 [[tracefile-rotation]]
1588 ==== Trace file count and size
1590 By default, trace files can grow as large as needed. You can set the
1591 maximum size of each trace file that a channel writes when you
1592 <<enabling-disabling-channels,create a channel>>. When the size of
1593 a trace file reaches the channel's fixed maximum size, LTTng creates
1594 another file to contain the next event records. LTTng appends a file
1595 count to each trace file name in this case.
1597 If you set the trace file size attribute when you create a channel, the
1598 maximum number of trace files that LTTng creates is _unlimited_ by
1599 default. To limit them, you can also set a maximum number of trace
1600 files. When the number of trace files reaches the channel's fixed
1601 maximum count, the oldest trace file is overwritten. This mechanism is
1602 called _trace file rotation_.
1606 === Instrumentation point, event rule, event, and event record
1608 An _event rule_ is a set of conditions which must be **all** satisfied
1609 for LTTng to record an occuring event.
1611 You set the conditions when you <<enabling-disabling-events,create
1614 You always attach an event rule to <<channel,channel>> when you create
1617 When an event passes the conditions of an event rule, LTTng records it
1618 in one of the attached channel's sub-buffers.
1620 The available conditions, as of LTTng{nbsp}{revision}, are:
1622 * The event rule _is enabled_.
1623 * The instrumentation point's type _is{nbsp}T_.
1624 * The instrumentation point's name (sometimes called _event name_)
1625 _matches{nbsp}N_, but _is not{nbsp}E_.
1626 * The instrumentation point's log level _is as severe as{nbsp}L_, or
1627 _is exactly{nbsp}L_.
1628 * The fields of the event's payload _satisfy_ a filter
1629 expression{nbsp}__F__.
1631 As you can see, all the conditions but the dynamic filter are related to
1632 the event rule's status or to the instrumentation point, not to the
1633 occurring events. This is why, without a filter, checking if an event
1634 passes an event rule is not a dynamic task: when you create or modify an
1635 event rule, all the tracers of its tracing domain enable or disable the
1636 instrumentation points themselves once. This is possible because the
1637 attributes of an instrumentation point (type, name, and log level) are
1638 defined statically. In other words, without a dynamic filter, the tracer
1639 _does not evaluate_ the arguments of an instrumentation point unless it
1640 matches an enabled event rule.
1642 Note that, for LTTng to record an event, the <<channel,channel>> to
1643 which a matching event rule is attached must also be enabled, and the
1644 tracing session owning this channel must be active.
1647 .Logical path from an instrumentation point to an event record.
1648 image::event-rule.png[]
1650 .Event, event record, or event rule?
1652 With so many similar terms, it's easy to get confused.
1654 An **event** is the consequence of the execution of an _instrumentation
1655 point_, like a tracepoint that you manually place in some source code,
1656 or a Linux kernel KProbe. An event is said to _occur_ at a specific
1657 time. Different actions can be taken upon the occurance of an event,
1658 like record the event's payload to a buffer.
1660 An **event record** is the representation of an event in a sub-buffer. A
1661 tracer is responsible for capturing the payload of an event, current
1662 context variables, the event's ID, and the event's timestamp. LTTng
1663 can append this sub-buffer to a trace file.
1665 An **event rule** is a set of conditions which must all be satisfied for
1666 LTTng to record an occuring event. Events still occur without
1667 satisfying event rules, but LTTng does not record them.
1672 == Components of noch:{LTTng}
1674 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1675 to call LTTng a simple _tool_ since it is composed of multiple
1676 interacting components. This section describes those components,
1677 explains their respective roles, and shows how they connect together to
1678 form the LTTng ecosystem.
1680 The following diagram shows how the most important components of LTTng
1681 interact with user applications, the Linux kernel, and you:
1684 .Control and trace data paths between LTTng components.
1685 image::plumbing.png[]
1687 The LTTng project incorporates:
1689 * **LTTng-tools**: Libraries and command-line interface to
1690 control tracing sessions.
1691 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1692 ** <<lttng-consumerd,Consumer daemon>> (man:lttng-consumerd(8)).
1693 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1694 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1695 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1696 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1698 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1699 headers to instrument and trace any native user application.
1700 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1701 *** `liblttng-ust-libc-wrapper`
1702 *** `liblttng-ust-pthread-wrapper`
1703 *** `liblttng-ust-cyg-profile`
1704 *** `liblttng-ust-cyg-profile-fast`
1705 *** `liblttng-ust-dl`
1706 ** User space tracepoint provider source files generator command-line
1707 tool (man:lttng-gen-tp(1)).
1708 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1709 Java applications using `java.util.logging` or
1710 Apache log4j 1.2 logging.
1711 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1712 Python applications using the standard `logging` package.
1713 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1715 ** LTTng kernel tracer module.
1716 ** Tracing ring buffer kernel modules.
1717 ** Probe kernel modules.
1718 ** LTTng logger kernel module.
1722 === Tracing control command-line interface
1725 .The tracing control command-line interface.
1726 image::plumbing-lttng-cli.png[]
1728 The _man:lttng(1) command-line tool_ is the standard user interface to
1729 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1730 is part of LTTng-tools.
1732 The cmd:lttng tool is linked with
1733 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1734 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1736 The cmd:lttng tool has a Git-like interface:
1740 $ lttng <GENERAL OPTIONS> <COMMAND> <COMMAND OPTIONS>
1743 The <<controlling-tracing,Tracing control>> section explores the
1744 available features of LTTng using the cmd:lttng tool.
1747 [[liblttng-ctl-lttng]]
1748 === Tracing control library
1751 .The tracing control library.
1752 image::plumbing-liblttng-ctl.png[]
1754 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1755 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1756 underlying protocol's details. `liblttng-ctl` is part of LTTng-tools.
1758 The <<lttng-cli,cmd:lttng command-line tool>>
1759 is linked with `liblttng-ctl`.
1761 You can use `liblttng-ctl` in C or $$C++$$ source code by including its
1766 #include <lttng/lttng.h>
1769 Some objects are referenced by name (C string), such as tracing
1770 sessions, but most of them require to create a handle first using
1771 `lttng_create_handle()`.
1773 The best available developer documentation for `liblttng-ctl` is, as of
1774 LTTng{nbsp}{revision}, its installed header files. Every function and
1775 structure is thoroughly documented.
1779 === User space tracing library
1782 .The user space tracing library.
1783 image::plumbing-liblttng-ust.png[]
1785 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1786 is the LTTng user space tracer. It receives commands from a
1787 <<lttng-sessiond,session daemon>>, for example to
1788 enable and disable specific instrumentation points, and writes event
1789 records to ring buffers shared with a
1790 <<lttng-consumerd,consumer daemon>>.
1791 `liblttng-ust` is part of LTTng-UST.
1793 Public C header files are installed beside `liblttng-ust` to
1794 instrument any <<c-application,C or $$C++$$ application>>.
1796 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1797 packages, use their own library providing tracepoints which is
1798 linked with `liblttng-ust`.
1800 An application or library does not have to initialize `liblttng-ust`
1801 manually: its constructor does the necessary tasks to properly register
1802 to a session daemon. The initialization phase also enables the
1803 instrumentation points matching the <<event,event rules>> that you
1807 [[lttng-ust-agents]]
1808 === User space tracing agents
1811 .The user space tracing agents.
1812 image::plumbing-lttng-ust-agents.png[]
1814 The _LTTng-UST Java and Python agents_ are regular Java and Python
1815 packages which add LTTng tracing capabilities to the
1816 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1818 In the case of Java, the
1819 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1820 core logging facilities] and
1821 https://logging.apache.org/log4j/1.2/[Apache log4j 1.2] are supported.
1822 Note that Apache Log4{nbsp}2 is not supported.
1824 In the case of Python, the standard
1825 https://docs.python.org/3/library/logging.html[`logging`] package
1826 is supported. Both Python 2 and Python 3 modules can import the
1827 LTTng-UST Python agent package.
1829 The applications using the LTTng-UST agents are in the
1830 `java.util.logging` (JUL),
1831 log4j, and Python <<domain,tracing domains>>.
1833 Both agents use the same mechanism to trace the log statements. When an
1834 agent is initialized, it creates a log handler that attaches to the root
1835 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
1836 When the application executes a log statement, it is passed to the
1837 agent's log handler by the root logger. The agent's log handler calls a
1838 native function in a tracepoint provider package shared library linked
1839 with <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
1840 other fields, like its logger name and its log level. This native
1841 function contains a user space instrumentation point, hence tracing the
1844 The log level condition of an
1845 <<event,event rule>> is considered when tracing
1846 a Java or a Python application, and it's compatible with the standard
1847 JUL, log4j, and Python log levels.
1851 === LTTng kernel modules
1854 .The LTTng kernel modules.
1855 image::plumbing-lttng-modules.png[]
1857 The _LTTng kernel modules_ are a set of Linux kernel modules
1858 which implement the kernel tracer of the LTTng project. The LTTng
1859 kernel modules are part of LTTng-modules.
1861 The LTTng kernel modules include:
1863 * A set of _probe_ modules.
1865 Each module attaches to a specific subsystem
1866 of the Linux kernel using its tracepoint instrument points. There are
1867 also modules to attach to the entry and return points of the Linux
1868 system call functions.
1870 * _Ring buffer_ modules.
1872 A ring buffer implementation is provided as kernel modules. The LTTng
1873 kernel tracer writes to the ring buffer; a
1874 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
1876 * The _LTTng kernel tracer_ module.
1877 * The _LTTng logger_ module.
1879 The LTTng logger module implements the special path:{/proc/lttng-logger}
1880 file so that any executable can generate LTTng events by opening and
1881 writing to this file.
1883 See <<proc-lttng-logger-abi,LTTng logger>>.
1885 Generally, you do not have to load the LTTng kernel modules manually
1886 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
1887 daemon>> loads the necessary modules when starting. If you have extra
1888 probe modules, you can specify to load them to the session daemon on
1891 The LTTng kernel modules are installed in
1892 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
1893 the kernel release (see `uname --kernel-release`).
1900 .The session daemon.
1901 image::plumbing-sessiond.png[]
1903 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
1904 managing tracing sessions and for controlling the various components of
1905 LTTng. The session daemon is part of LTTng-tools.
1907 The session daemon sends control requests to and receives control
1910 * The <<lttng-ust,user space tracing library>>.
1912 Any instance of the user space tracing library first registers to
1913 a session daemon. Then, the session daemon can send requests to
1914 this instance, such as:
1917 ** Get the list of tracepoints.
1918 ** Share an <<event,event rule>> so that the user space tracing library
1919 can enable or disable tracepoints. Amongst the possible conditions
1920 of an event rule is a filter expression which `liblttng-ust` evalutes
1921 when an event occurs.
1922 ** Share <<channel,channel>> attributes and ring buffer locations.
1925 The session daemon and the user space tracing library use a Unix
1926 domain socket for their communication.
1928 * The <<lttng-ust-agents,user space tracing agents>>.
1930 Any instance of a user space tracing agent first registers to
1931 a session daemon. Then, the session daemon can send requests to
1932 this instance, such as:
1935 ** Get the list of loggers.
1936 ** Enable or disable a specific logger.
1939 The session daemon and the user space tracing agent use a TCP connection
1940 for their communication.
1942 * The <<lttng-modules,LTTng kernel tracer>>.
1943 * The <<lttng-consumerd,consumer daemon>>.
1945 The session daemon sends requests to the consumer daemon to instruct
1946 it where to send the trace data streams, amongst other information.
1948 * The <<lttng-relayd,relay daemon>>.
1950 The session daemon receives commands from the
1951 <<liblttng-ctl-lttng,tracing control library>>.
1953 The root session daemon loads the appropriate
1954 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
1955 a <<lttng-consumerd,consumer daemon>> as soon as you create
1956 an <<event,event rule>>.
1958 The session daemon does not send and receive trace data: this is the
1959 role of the <<lttng-consumerd,consumer daemon>> and
1960 <<lttng-relayd,relay daemon>>. It does, however, generate the
1961 http://diamon.org/ctf/[CTF] metadata stream.
1963 Each Unix user can have its own session daemon instance. The
1964 tracing sessions managed by different session daemons are completely
1967 The root user's session daemon is the only one which is
1968 allowed to control the LTTng kernel tracer, and its spawned consumer
1969 daemon is the only one which is allowed to consume trace data from the
1970 LTTng kernel tracer. Note, however, that any Unix user which is a member
1971 of the <<tracing-group,tracing group>> is allowed
1972 to create <<channel,channels>> in the
1973 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
1976 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
1977 session daemon when using its `create` command if none is currently
1978 running. You can also start the session daemon manually.
1985 .The consumer daemon.
1986 image::plumbing-consumerd.png[]
1988 The _consumer daemon_, man:lttng-consumerd(8), is a daemon which shares
1989 ring buffers with user applications or with the LTTng kernel modules to
1990 collect trace data and send it to some location (on disk or to a
1991 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
1992 is part of LTTng-tools.
1994 You do not start a consumer daemon manually: a consumer daemon is always
1995 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
1996 <<event,event rule>>, that is, before you start tracing. When you kill
1997 its owner session daemon, the consumer daemon also exits because it is
1998 the session daemon's child process. Command-line options of
1999 man:lttng-sessiond(8) target the consumer daemon process.
2001 There are up to two running consumer daemons per Unix user, whereas only
2002 one session daemon can run per user. This is because each process can be
2003 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
2004 and 64-bit processes, it is more efficient to have separate
2005 corresponding 32-bit and 64-bit consumer daemons. The root user is an
2006 exception: it can have up to _three_ running consumer daemons: 32-bit
2007 and 64-bit instances for its user applications, and one more
2008 reserved for collecting kernel trace data.
2016 image::plumbing-relayd.png[]
2018 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
2019 between remote session and consumer daemons, local trace files, and a
2020 remote live trace viewer. The relay daemon is part of LTTng-tools.
2022 The main purpose of the relay daemon is to implement a receiver of
2023 <<sending-trace-data-over-the-network,trace data over the network>>.
2024 This is useful when the target system does not have much file system
2025 space to record trace files locally.
2027 The relay daemon is also a server to which a
2028 <<lttng-live,live trace viewer>> can
2029 connect. The live trace viewer sends requests to the relay daemon to
2030 receive trace data as the target system emits events. The
2031 communication protocol is named _LTTng live_; it is used over TCP
2034 Note that you can start the relay daemon on the target system directly.
2035 This is the setup of choice when the use case is to view events as
2036 the target system emits them without the need of a remote system.
2040 == [[using-lttng]]Instrumentation
2042 There are many examples of tracing and monitoring in our everyday life:
2044 * You have access to real-time and historical weather reports and
2045 forecasts thanks to weather stations installed around the country.
2046 * You know your heart is safe thanks to an electrocardiogram.
2047 * You make sure not to drive your car too fast and to have enough fuel
2048 to reach your destination thanks to gauges visible on your dashboard.
2050 All the previous examples have something in common: they rely on
2051 **instruments**. Without the electrodes attached to the surface of your
2052 body's skin, cardiac monitoring is futile.
2054 LTTng, as a tracer, is no different from those real life examples. If
2055 you're about to trace a software system or, in other words, record its
2056 history of execution, you better have **instrumentation points** in the
2057 subject you're tracing, that is, the actual software.
2059 Various ways were developed to instrument a piece of software for LTTng
2060 tracing. The most straightforward one is to manually place
2061 instrumentation points, called _tracepoints_, in the software's source
2062 code. It is also possible to add instrumentation points dynamically in
2063 the Linux kernel <<domain,tracing domain>>.
2065 If you're only interested in tracing the Linux kernel, your
2066 instrumentation needs are probably already covered by LTTng's built-in
2067 <<lttng-modules,Linux kernel tracepoints>>. You may also wish to trace a
2068 user application which is already instrumented for LTTng tracing.
2069 In such cases, you can skip this whole section and read the topics of
2070 the <<controlling-tracing,Tracing control>> section.
2072 Many methods are available to instrument a piece of software for LTTng
2075 * <<c-application,User space instrumentation for C and $$C++$$
2077 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
2078 * <<java-application,User space Java agent>>.
2079 * <<python-application,User space Python agent>>.
2080 * <<proc-lttng-logger-abi,LTTng logger>>.
2081 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
2085 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
2087 The procedure to instrument a C or $$C++$$ user application with
2088 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
2090 . <<tracepoint-provider,Create the source files of a tracepoint provider
2092 . <<probing-the-application-source-code,Add tracepoints to
2093 the application's source code>>.
2094 . <<building-tracepoint-providers-and-user-application,Build and link
2095 a tracepoint provider package and the user application>>.
2097 If you need quick, man:printf(3)-like instrumentation, you can skip
2098 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
2101 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
2102 instrument a user application with `liblttng-ust`.
2105 [[tracepoint-provider]]
2106 ==== Create the source files of a tracepoint provider package
2108 A _tracepoint provider_ is a set of compiled functions which provide
2109 **tracepoints** to an application, the type of instrumentation point
2110 supported by LTTng-UST. Those functions can emit events with
2111 user-defined fields and serialize those events as event records to one
2112 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
2113 macro, which you <<probing-the-application-source-code,insert in a user
2114 application's source code>>, calls those functions.
2116 A _tracepoint provider package_ is an object file (`.o`) or a shared
2117 library (`.so`) which contains one or more tracepoint providers.
2118 Its source files are:
2120 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
2121 * A <<tpp-source,tracepoint provider package source>> (`.c`).
2123 A tracepoint provider package is dynamically linked with `liblttng-ust`,
2124 the LTTng user space tracer, at run time.
2127 .User application linked with `liblttng-ust` and containing a tracepoint provider.
2128 image::ust-app.png[]
2130 NOTE: If you need quick, man:printf(3)-like instrumentation, you can
2131 skip creating and using a tracepoint provider and use
2132 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
2136 ===== Create a tracepoint provider header file template
2138 A _tracepoint provider header file_ contains the tracepoint
2139 definitions of a tracepoint provider.
2141 To create a tracepoint provider header file:
2143 . Start from this template:
2147 .Tracepoint provider header file template (`.h` file extension).
2149 #undef TRACEPOINT_PROVIDER
2150 #define TRACEPOINT_PROVIDER provider_name
2152 #undef TRACEPOINT_INCLUDE
2153 #define TRACEPOINT_INCLUDE "./tp.h"
2155 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
2158 #include <lttng/tracepoint.h>
2161 * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
2162 * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
2167 #include <lttng/tracepoint-event.h>
2173 * `provider_name` with the name of your tracepoint provider.
2174 * `"tp.h"` with the name of your tracepoint provider header file.
2176 . Below the `#include <lttng/tracepoint.h>` line, put your
2177 <<defining-tracepoints,tracepoint definitions>>.
2179 Your tracepoint provider name must be unique amongst all the possible
2180 tracepoint provider names used on the same target system. We
2181 suggest to include the name of your project or company in the name,
2182 for example, `org_lttng_my_project_tpp`.
2184 TIP: [[lttng-gen-tp]]You can use the man:lttng-gen-tp(1) tool to create
2185 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
2186 write are the <<defining-tracepoints,tracepoint definitions>>.
2189 [[defining-tracepoints]]
2190 ===== Create a tracepoint definition
2192 A _tracepoint definition_ defines, for a given tracepoint:
2194 * Its **input arguments**. They are the macro parameters that the
2195 `tracepoint()` macro accepts for this particular tracepoint
2196 in the user application's source code.
2197 * Its **output event fields**. They are the sources of event fields
2198 that form the payload of any event that the execution of the
2199 `tracepoint()` macro emits for this particular tracepoint.
2201 You can create a tracepoint definition by using the
2202 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
2204 <<tpp-header,tracepoint provider header file template>>.
2206 The syntax of the `TRACEPOINT_EVENT()` macro is:
2209 .`TRACEPOINT_EVENT()` macro syntax.
2212 /* Tracepoint provider name */
2215 /* Tracepoint name */
2218 /* Input arguments */
2223 /* Output event fields */
2232 * `provider_name` with your tracepoint provider name.
2233 * `tracepoint_name` with your tracepoint name.
2234 * `arguments` with the <<tpp-def-input-args,input arguments>>.
2235 * `fields` with the <<tpp-def-output-fields,output event field>>
2238 This tracepoint emits events named `provider_name:tracepoint_name`.
2241 .Event name's length limitation
2243 The concatenation of the tracepoint provider name and the
2244 tracepoint name must not exceed **254 characters**. If it does, the
2245 instrumented application compiles and runs, but LTTng throws multiple
2246 warnings and you could experience serious issues.
2249 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
2252 .`TP_ARGS()` macro syntax.
2261 * `type` with the C type of the argument.
2262 * `arg_name` with the argument name.
2264 You can repeat `type` and `arg_name` up to 10 times to have
2265 more than one argument.
2267 .`TP_ARGS()` usage with three arguments.
2279 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
2280 tracepoint definition with no input arguments.
2282 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
2283 `ctf_*()` macros. Each `ctf_*()` macro defines one event field. See
2284 man:lttng-ust(3) for a complete description of the available `ctf_*()`
2285 macros. A `ctf_*()` macro specifies the type, size, and byte order of
2288 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
2289 C expression that the tracer evalutes at the `tracepoint()` macro site
2290 in the application's source code. This expression provides a field's
2291 source of data. The argument expression can include input argument names
2292 listed in the `TP_ARGS()` macro.
2294 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
2295 must be unique within a given tracepoint definition.
2297 Here's a complete tracepoint definition example:
2299 .Tracepoint definition.
2301 The following tracepoint definition defines a tracepoint which takes
2302 three input arguments and has four output event fields.
2306 #include "my-custom-structure.h"
2312 const struct my_custom_structure*, my_custom_structure,
2317 ctf_string(query_field, query)
2318 ctf_float(double, ratio_field, ratio)
2319 ctf_integer(int, recv_size, my_custom_structure->recv_size)
2320 ctf_integer(int, send_size, my_custom_structure->send_size)
2325 You can refer to this tracepoint definition with the `tracepoint()`
2326 macro in your application's source code like this:
2330 tracepoint(my_provider, my_tracepoint,
2331 my_structure, some_ratio, the_query);
2335 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2336 if they satisfy an enabled <<event,event rule>>.
2339 [[using-tracepoint-classes]]
2340 ===== Use a tracepoint class
2342 A _tracepoint class_ is a class of tracepoints which share the same
2343 output event field definitions. A _tracepoint instance_ is one
2344 instance of such a defined tracepoint class, with its own tracepoint
2347 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2348 shorthand which defines both a tracepoint class and a tracepoint
2349 instance at the same time.
2351 When you build a tracepoint provider package, the C or $$C++$$ compiler
2352 creates one serialization function for each **tracepoint class**. A
2353 serialization function is responsible for serializing the event fields
2354 of a tracepoint to a sub-buffer when tracing.
2356 For various performance reasons, when your situation requires multiple
2357 tracepoint definitions with different names, but with the same event
2358 fields, we recommend that you manually create a tracepoint class
2359 and instantiate as many tracepoint instances as needed. One positive
2360 effect of such a design, amongst other advantages, is that all
2361 tracepoint instances of the same tracepoint class reuse the same
2362 serialization function, thus reducing
2363 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2365 .Use a tracepoint class and tracepoint instances.
2367 Consider the following three tracepoint definitions:
2379 ctf_integer(int, userid, userid)
2380 ctf_integer(size_t, len, len)
2392 ctf_integer(int, userid, userid)
2393 ctf_integer(size_t, len, len)
2405 ctf_integer(int, userid, userid)
2406 ctf_integer(size_t, len, len)
2411 In this case, we create three tracepoint classes, with one implicit
2412 tracepoint instance for each of them: `get_account`, `get_settings`, and
2413 `get_transaction`. However, they all share the same event field names
2414 and types. Hence three identical, yet independent serialization
2415 functions are created when you build the tracepoint provider package.
2417 A better design choice is to define a single tracepoint class and three
2418 tracepoint instances:
2422 /* The tracepoint class */
2423 TRACEPOINT_EVENT_CLASS(
2424 /* Tracepoint provider name */
2427 /* Tracepoint class name */
2430 /* Input arguments */
2436 /* Output event fields */
2438 ctf_integer(int, userid, userid)
2439 ctf_integer(size_t, len, len)
2443 /* The tracepoint instances */
2444 TRACEPOINT_EVENT_INSTANCE(
2445 /* Tracepoint provider name */
2448 /* Tracepoint class name */
2451 /* Tracepoint name */
2454 /* Input arguments */
2460 TRACEPOINT_EVENT_INSTANCE(
2469 TRACEPOINT_EVENT_INSTANCE(
2482 [[assigning-log-levels]]
2483 ===== Assign a log level to a tracepoint definition
2485 You can assign an optional _log level_ to a
2486 <<defining-tracepoints,tracepoint definition>>.
2488 Assigning different levels of severity to tracepoint definitions can
2489 be useful: when you <<enabling-disabling-events,create an event rule>>,
2490 you can target tracepoints having a log level as severe as a specific
2493 The concept of LTTng-UST log levels is similar to the levels found
2494 in typical logging frameworks:
2496 * In a logging framework, the log level is given by the function
2497 or method name you use at the log statement site: `debug()`,
2498 `info()`, `warn()`, `error()`, and so on.
2499 * In LTTng-UST, you statically assign the log level to a tracepoint
2500 definition; any `tracepoint()` macro invocation which refers to
2501 this definition has this log level.
2503 You can assign a log level to a tracepoint definition with the
2504 `TRACEPOINT_LOGLEVEL()` macro. You must use this macro _after_ the
2505 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2506 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2509 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2512 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2514 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2519 * `provider_name` with the tracepoint provider name.
2520 * `tracepoint_name` with the tracepoint name.
2521 * `log_level` with the log level to assign to the tracepoint
2522 definition named `tracepoint_name` in the `provider_name`
2523 tracepoint provider.
2525 See man:lttng-ust(3) for a list of available log level names.
2527 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2531 /* Tracepoint definition */
2540 ctf_integer(int, userid, userid)
2541 ctf_integer(size_t, len, len)
2545 /* Log level assignment */
2546 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2552 ===== Create a tracepoint provider package source file
2554 A _tracepoint provider package source file_ is a C source file which
2555 includes a <<tpp-header,tracepoint provider header file>> to expand its
2556 macros into event serialization and other functions.
2558 You can always use the following tracepoint provider package source
2562 .Tracepoint provider package source file template.
2564 #define TRACEPOINT_CREATE_PROBES
2569 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2570 header file>> name. You may also include more than one tracepoint
2571 provider header file here to create a tracepoint provider package
2572 holding more than one tracepoint providers.
2575 [[probing-the-application-source-code]]
2576 ==== Add tracepoints to an application's source code
2578 Once you <<tpp-header,create a tracepoint provider header file>>, you
2579 can use the `tracepoint()` macro in your application's
2580 source code to insert the tracepoints that this header
2581 <<defining-tracepoints,defines>>.
2583 The `tracepoint()` macro takes at least two parameters: the tracepoint
2584 provider name and the tracepoint name. The corresponding tracepoint
2585 definition defines the other parameters.
2587 .`tracepoint()` usage.
2589 The following <<defining-tracepoints,tracepoint definition>> defines a
2590 tracepoint which takes two input arguments and has two output event
2594 .Tracepoint provider header file.
2596 #include "my-custom-structure.h"
2603 const char*, cmd_name
2606 ctf_string(cmd_name, cmd_name)
2607 ctf_integer(int, number_of_args, argc)
2612 You can refer to this tracepoint definition with the `tracepoint()`
2613 macro in your application's source code like this:
2616 .Application's source file.
2620 int main(int argc, char* argv[])
2622 tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2628 Note how the application's source code includes
2629 the tracepoint provider header file containing the tracepoint
2630 definitions to use, path:{tp.h}.
2633 .`tracepoint()` usage with a complex tracepoint definition.
2635 Consider this complex tracepoint definition, where multiple event
2636 fields refer to the same input arguments in their argument expression
2640 .Tracepoint provider header file.
2642 /* For `struct stat` */
2643 #include <sys/types.h>
2644 #include <sys/stat.h>
2656 ctf_integer(int, my_constant_field, 23 + 17)
2657 ctf_integer(int, my_int_arg_field, my_int_arg)
2658 ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2659 ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2660 my_str_arg[2] + my_str_arg[3])
2661 ctf_string(my_str_arg_field, my_str_arg)
2662 ctf_integer_hex(off_t, size_field, st->st_size)
2663 ctf_float(double, size_dbl_field, (double) st->st_size)
2664 ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2665 size_t, strlen(my_str_arg) / 2)
2670 You can refer to this tracepoint definition with the `tracepoint()`
2671 macro in your application's source code like this:
2674 .Application's source file.
2676 #define TRACEPOINT_DEFINE
2683 stat("/etc/fstab", &s);
2684 tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2690 If you look at the event record that LTTng writes when tracing this
2691 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2692 it should look like this:
2694 .Event record fields
2696 |Field's name |Field's value
2697 |`my_constant_field` |40
2698 |`my_int_arg_field` |23
2699 |`my_int_arg_field2` |529
2701 |`my_str_arg_field` |`Hello, World!`
2702 |`size_field` |0x12d
2703 |`size_dbl_field` |301.0
2704 |`half_my_str_arg_field` |`Hello,`
2708 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2709 compute--they use the call stack, for example. To avoid this
2710 computation when the tracepoint is disabled, you can use the
2711 `tracepoint_enabled()` and `do_tracepoint()` macros.
2713 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2717 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2719 tracepoint_enabled(provider_name, tracepoint_name)
2720 do_tracepoint(provider_name, tracepoint_name, ...)
2725 * `provider_name` with the tracepoint provider name.
2726 * `tracepoint_name` with the tracepoint name.
2728 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2729 `tracepoint_name` from the provider named `provider_name` is enabled
2732 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2733 if the tracepoint is enabled. Using `tracepoint()` with
2734 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2735 the `tracepoint_enabled()` check, thus a race condition is
2736 possible in this situation:
2739 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2741 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2742 stuff = prepare_stuff();
2745 tracepoint(my_provider, my_tracepoint, stuff);
2748 If the tracepoint is enabled after the condition, then `stuff` is not
2749 prepared: the emitted event will either contain wrong data, or the whole
2750 application could crash (segmentation fault, for example).
2752 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2753 `STAP_PROBEV()` call. If you need it, you must emit
2757 [[building-tracepoint-providers-and-user-application]]
2758 ==== Build and link a tracepoint provider package and an application
2760 Once you have one or more <<tpp-header,tracepoint provider header
2761 files>> and a <<tpp-source,tracepoint provider package source file>>,
2762 you can create the tracepoint provider package by compiling its source
2763 file. From here, multiple build and run scenarios are possible. The
2764 following table shows common application and library configurations
2765 along with the required command lines to achieve them.
2767 In the following diagrams, we use the following file names:
2770 Executable application.
2773 Application's object file.
2776 Tracepoint provider package object file.
2779 Tracepoint provider package archive file.
2782 Tracepoint provider package shared object file.
2785 User library object file.
2788 User library shared object file.
2790 We use the following symbols in the diagrams of table below:
2793 .Symbols used in the build scenario diagrams.
2794 image::ust-sit-symbols.png[]
2796 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2797 variable in the following instructions.
2799 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2800 .Common tracepoint provider package scenarios.
2802 |Scenario |Instructions
2805 The instrumented application is statically linked with
2806 the tracepoint provider package object.
2808 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2811 include::../common/ust-sit-step-tp-o.txt[]
2813 To build the instrumented application:
2815 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2820 #define TRACEPOINT_DEFINE
2824 . Compile the application source file:
2833 . Build the application:
2838 $ gcc -o app app.o tpp.o -llttng-ust -ldl
2842 To run the instrumented application:
2844 * Start the application:
2854 The instrumented application is statically linked with the
2855 tracepoint provider package archive file.
2857 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
2860 To create the tracepoint provider package archive file:
2862 . Compile the <<tpp-source,tracepoint provider package source file>>:
2871 . Create the tracepoint provider package archive file:
2876 $ ar rcs tpp.a tpp.o
2880 To build the instrumented application:
2882 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2887 #define TRACEPOINT_DEFINE
2891 . Compile the application source file:
2900 . Build the application:
2905 $ gcc -o app app.o tpp.a -llttng-ust -ldl
2909 To run the instrumented application:
2911 * Start the application:
2921 The instrumented application is linked with the tracepoint provider
2922 package shared object.
2924 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
2927 include::../common/ust-sit-step-tp-so.txt[]
2929 To build the instrumented application:
2931 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2936 #define TRACEPOINT_DEFINE
2940 . Compile the application source file:
2949 . Build the application:
2954 $ gcc -o app app.o -ldl -L. -ltpp
2958 To run the instrumented application:
2960 * Start the application:
2970 The tracepoint provider package shared object is preloaded before the
2971 instrumented application starts.
2973 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
2976 include::../common/ust-sit-step-tp-so.txt[]
2978 To build the instrumented application:
2980 . In path:{app.c}, before including path:{tpp.h}, add the
2986 #define TRACEPOINT_DEFINE
2987 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2991 . Compile the application source file:
3000 . Build the application:
3005 $ gcc -o app app.o -ldl
3009 To run the instrumented application with tracing support:
3011 * Preload the tracepoint provider package shared object and
3012 start the application:
3017 $ LD_PRELOAD=./libtpp.so ./app
3021 To run the instrumented application without tracing support:
3023 * Start the application:
3033 The instrumented application dynamically loads the tracepoint provider
3034 package shared object.
3036 See the <<dlclose-warning,warning about `dlclose()`>>.
3038 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
3041 include::../common/ust-sit-step-tp-so.txt[]
3043 To build the instrumented application:
3045 . In path:{app.c}, before including path:{tpp.h}, add the
3051 #define TRACEPOINT_DEFINE
3052 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3056 . Compile the application source file:
3065 . Build the application:
3070 $ gcc -o app app.o -ldl
3074 To run the instrumented application:
3076 * Start the application:
3086 The application is linked with the instrumented user library.
3088 The instrumented user library is statically linked with the tracepoint
3089 provider package object file.
3091 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
3094 include::../common/ust-sit-step-tp-o-fpic.txt[]
3096 To build the instrumented user library:
3098 . In path:{emon.c}, before including path:{tpp.h}, add the
3104 #define TRACEPOINT_DEFINE
3108 . Compile the user library source file:
3113 $ gcc -I. -fpic -c emon.c
3117 . Build the user library shared object:
3122 $ gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
3126 To build the application:
3128 . Compile the application source file:
3137 . Build the application:
3142 $ gcc -o app app.o -L. -lemon
3146 To run the application:
3148 * Start the application:
3158 The application is linked with the instrumented user library.
3160 The instrumented user library is linked with the tracepoint provider
3161 package shared object.
3163 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3166 include::../common/ust-sit-step-tp-so.txt[]
3168 To build the instrumented user library:
3170 . In path:{emon.c}, before including path:{tpp.h}, add the
3176 #define TRACEPOINT_DEFINE
3180 . Compile the user library source file:
3185 $ gcc -I. -fpic -c emon.c
3189 . Build the user library shared object:
3194 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3198 To build the application:
3200 . Compile the application source file:
3209 . Build the application:
3214 $ gcc -o app app.o -L. -lemon
3218 To run the application:
3220 * Start the application:
3230 The tracepoint provider package shared object is preloaded before the
3233 The application is linked with the instrumented user library.
3235 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
3238 include::../common/ust-sit-step-tp-so.txt[]
3240 To build the instrumented user library:
3242 . In path:{emon.c}, before including path:{tpp.h}, add the
3248 #define TRACEPOINT_DEFINE
3249 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3253 . Compile the user library source file:
3258 $ gcc -I. -fpic -c emon.c
3262 . Build the user library shared object:
3267 $ gcc -shared -o libemon.so emon.o -ldl
3271 To build the application:
3273 . Compile the application source file:
3282 . Build the application:
3287 $ gcc -o app app.o -L. -lemon
3291 To run the application with tracing support:
3293 * Preload the tracepoint provider package shared object and
3294 start the application:
3299 $ LD_PRELOAD=./libtpp.so ./app
3303 To run the application without tracing support:
3305 * Start the application:
3315 The application is linked with the instrumented user library.
3317 The instrumented user library dynamically loads the tracepoint provider
3318 package shared object.
3320 See the <<dlclose-warning,warning about `dlclose()`>>.
3322 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3325 include::../common/ust-sit-step-tp-so.txt[]
3327 To build the instrumented user library:
3329 . In path:{emon.c}, before including path:{tpp.h}, add the
3335 #define TRACEPOINT_DEFINE
3336 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3340 . Compile the user library source file:
3345 $ gcc -I. -fpic -c emon.c
3349 . Build the user library shared object:
3354 $ gcc -shared -o libemon.so emon.o -ldl
3358 To build the application:
3360 . Compile the application source file:
3369 . Build the application:
3374 $ gcc -o app app.o -L. -lemon
3378 To run the application:
3380 * Start the application:
3390 The application dynamically loads the instrumented user library.
3392 The instrumented user library is linked with the tracepoint provider
3393 package shared object.
3395 See the <<dlclose-warning,warning about `dlclose()`>>.
3397 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3400 include::../common/ust-sit-step-tp-so.txt[]
3402 To build the instrumented user library:
3404 . In path:{emon.c}, before including path:{tpp.h}, add the
3410 #define TRACEPOINT_DEFINE
3414 . Compile the user library source file:
3419 $ gcc -I. -fpic -c emon.c
3423 . Build the user library shared object:
3428 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3432 To build the application:
3434 . Compile the application source file:
3443 . Build the application:
3448 $ gcc -o app app.o -ldl -L. -lemon
3452 To run the application:
3454 * Start the application:
3464 The application dynamically loads the instrumented user library.
3466 The instrumented user library dynamically loads the tracepoint provider
3467 package shared object.
3469 See the <<dlclose-warning,warning about `dlclose()`>>.
3471 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3474 include::../common/ust-sit-step-tp-so.txt[]
3476 To build the instrumented user library:
3478 . In path:{emon.c}, before including path:{tpp.h}, add the
3484 #define TRACEPOINT_DEFINE
3485 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3489 . Compile the user library source file:
3494 $ gcc -I. -fpic -c emon.c
3498 . Build the user library shared object:
3503 $ gcc -shared -o libemon.so emon.o -ldl
3507 To build the application:
3509 . Compile the application source file:
3518 . Build the application:
3523 $ gcc -o app app.o -ldl -L. -lemon
3527 To run the application:
3529 * Start the application:
3539 The tracepoint provider package shared object is preloaded before the
3542 The application dynamically loads the instrumented user library.
3544 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3547 include::../common/ust-sit-step-tp-so.txt[]
3549 To build the instrumented user library:
3551 . In path:{emon.c}, before including path:{tpp.h}, add the
3557 #define TRACEPOINT_DEFINE
3558 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3562 . Compile the user library source file:
3567 $ gcc -I. -fpic -c emon.c
3571 . Build the user library shared object:
3576 $ gcc -shared -o libemon.so emon.o -ldl
3580 To build the application:
3582 . Compile the application source file:
3591 . Build the application:
3596 $ gcc -o app app.o -L. -lemon
3600 To run the application with tracing support:
3602 * Preload the tracepoint provider package shared object and
3603 start the application:
3608 $ LD_PRELOAD=./libtpp.so ./app
3612 To run the application without tracing support:
3614 * Start the application:
3624 The application is statically linked with the tracepoint provider
3625 package object file.
3627 The application is linked with the instrumented user library.
3629 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3632 include::../common/ust-sit-step-tp-o.txt[]
3634 To build the instrumented user library:
3636 . In path:{emon.c}, before including path:{tpp.h}, add the
3642 #define TRACEPOINT_DEFINE
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
3664 To build the application:
3666 . Compile the application source file:
3675 . Build the application:
3680 $ gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3684 To run the instrumented application:
3686 * Start the application:
3696 The application is statically linked with the tracepoint provider
3697 package object file.
3699 The application dynamically loads the instrumented user library.
3701 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3704 include::../common/ust-sit-step-tp-o.txt[]
3706 To build the application:
3708 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3713 #define TRACEPOINT_DEFINE
3717 . Compile the application source file:
3726 . Build the application:
3731 $ gcc -Wl,--export-dynamic -o app app.o tpp.o \
3736 The `--export-dynamic` option passed to the linker is necessary for the
3737 dynamically loaded library to ``see'' the tracepoint symbols defined in
3740 To build the instrumented user library:
3742 . Compile the user library source file:
3747 $ gcc -I. -fpic -c emon.c
3751 . Build the user library shared object:
3756 $ gcc -shared -o libemon.so emon.o
3760 To run the application:
3762 * Start the application:
3774 .Do not use man:dlclose(3) on a tracepoint provider package
3776 Never use man:dlclose(3) on any shared object which:
3778 * Is linked with, statically or dynamically, a tracepoint provider
3780 * Calls man:dlopen(3) itself to dynamically open a tracepoint provider
3781 package shared object.
3783 This is currently considered **unsafe** due to a lack of reference
3784 counting from LTTng-UST to the shared object.
3786 A known workaround (available since glibc 2.2) is to use the
3787 `RTLD_NODELETE` flag when calling man:dlopen(3) initially. This has the
3788 effect of not unloading the loaded shared object, even if man:dlclose(3)
3791 You can also preload the tracepoint provider package shared object with
3792 the env:LD_PRELOAD environment variable to overcome this limitation.
3796 [[using-lttng-ust-with-daemons]]
3797 ===== Use noch:{LTTng-UST} with daemons
3799 If your instrumented application calls man:fork(2), man:clone(2),
3800 or BSD's man:rfork(2), without a following man:exec(3)-family
3801 system call, you must preload the path:{liblttng-ust-fork.so} shared
3802 object when you start the application.
3806 $ LD_PRELOAD=liblttng-ust-fork.so ./my-app
3809 If your tracepoint provider package is
3810 a shared library which you also preload, you must put both
3811 shared objects in env:LD_PRELOAD:
3815 $ LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3821 ===== Use noch:{LTTng-UST} with applications which close file descriptors that don't belong to them
3823 If your instrumented application closes one or more file descriptors
3824 which it did not open itself, you must preload the
3825 path:{liblttng-ust-fd.so} shared object when you start the application:
3829 $ LD_PRELOAD=liblttng-ust-fd.so ./my-app
3832 Typical use cases include closing all the file descriptors after
3833 man:fork(2) or man:rfork(2) and buggy applications doing
3837 [[lttng-ust-pkg-config]]
3838 ===== Use noch:{pkg-config}
3840 On some distributions, LTTng-UST ships with a
3841 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3842 metadata file. If this is your case, then you can use cmd:pkg-config to
3843 build an application on the command line:
3847 $ gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3851 [[instrumenting-32-bit-app-on-64-bit-system]]
3852 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3854 In order to trace a 32-bit application running on a 64-bit system,
3855 LTTng must use a dedicated 32-bit
3856 <<lttng-consumerd,consumer daemon>>.
3858 The following steps show how to build and install a 32-bit consumer
3859 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3860 build and install the 32-bit LTTng-UST libraries, and how to build and
3861 link an instrumented 32-bit application in that context.
3863 To build a 32-bit instrumented application for a 64-bit target system,
3864 assuming you have a fresh target system with no installed Userspace RCU
3867 . Download, build, and install a 32-bit version of Userspace RCU:
3872 $ cd $(mktemp -d) &&
3873 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
3874 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
3875 cd userspace-rcu-0.9.* &&
3876 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
3878 sudo make install &&
3883 . Using your distribution's package manager, or from source, install
3884 the following 32-bit versions of the following dependencies of
3885 LTTng-tools and LTTng-UST:
3888 * https://sourceforge.net/projects/libuuid/[libuuid]
3889 * http://directory.fsf.org/wiki/Popt[popt]
3890 * http://www.xmlsoft.org/[libxml2]
3893 . Download, build, and install a 32-bit version of the latest
3894 LTTng-UST{nbsp}{revision}:
3899 $ cd $(mktemp -d) &&
3900 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.9.tar.bz2 &&
3901 tar -xf lttng-ust-latest-2.9.tar.bz2 &&
3902 cd lttng-ust-2.9.* &&
3903 ./configure --libdir=/usr/local/lib32 \
3904 CFLAGS=-m32 CXXFLAGS=-m32 \
3905 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3907 sudo make install &&
3914 Depending on your distribution,
3915 32-bit libraries could be installed at a different location than
3916 `/usr/lib32`. For example, Debian is known to install
3917 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
3919 In this case, make sure to set `LDFLAGS` to all the
3920 relevant 32-bit library paths, for example:
3924 $ LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
3928 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
3929 the 32-bit consumer daemon:
3934 $ cd $(mktemp -d) &&
3935 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.9.tar.bz2 &&
3936 tar -xf lttng-tools-latest-2.9.tar.bz2 &&
3937 cd lttng-tools-2.9.* &&
3938 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
3939 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' \
3940 --disable-bin-lttng --disable-bin-lttng-crash \
3941 --disable-bin-lttng-relayd --disable-bin-lttng-sessiond &&
3943 cd src/bin/lttng-consumerd &&
3944 sudo make install &&
3949 . From your distribution or from source,
3950 <<installing-lttng,install>> the 64-bit versions of
3951 LTTng-UST and Userspace RCU.
3952 . Download, build, and install the 64-bit version of the
3953 latest LTTng-tools{nbsp}{revision}:
3958 $ cd $(mktemp -d) &&
3959 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.9.tar.bz2 &&
3960 tar -xf lttng-tools-latest-2.9.tar.bz2 &&
3961 cd lttng-tools-2.9.* &&
3962 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
3963 --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
3965 sudo make install &&
3970 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
3971 when linking your 32-bit application:
3974 -m32 -L/usr/lib32 -L/usr/local/lib32 \
3975 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
3978 For example, let's rebuild the quick start example in
3979 <<tracing-your-own-user-application,Trace a user application>> as an
3980 instrumented 32-bit application:
3985 $ gcc -m32 -c -I. hello-tp.c
3986 $ gcc -m32 -c hello.c
3987 $ gcc -m32 -o hello hello.o hello-tp.o \
3988 -L/usr/lib32 -L/usr/local/lib32 \
3989 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
3994 No special action is required to execute the 32-bit application and
3995 to trace it: use the command-line man:lttng(1) tool as usual.
4002 man:tracef(3) is a small LTTng-UST API designed for quick,
4003 man:printf(3)-like instrumentation without the burden of
4004 <<tracepoint-provider,creating>> and
4005 <<building-tracepoint-providers-and-user-application,building>>
4006 a tracepoint provider package.
4008 To use `tracef()` in your application:
4010 . In the C or C++ source files where you need to use `tracef()`,
4011 include `<lttng/tracef.h>`:
4016 #include <lttng/tracef.h>
4020 . In the application's source code, use `tracef()` like you would use
4028 tracef("my message: %d (%s)", my_integer, my_string);
4034 . Link your application with `liblttng-ust`:
4039 $ gcc -o app app.c -llttng-ust
4043 To trace the events that `tracef()` calls emit:
4045 * <<enabling-disabling-events,Create an event rule>> which matches the
4046 `lttng_ust_tracef:*` event name:
4051 $ lttng enable-event --userspace 'lttng_ust_tracef:*'
4056 .Limitations of `tracef()`
4058 The `tracef()` utility function was developed to make user space tracing
4059 super simple, albeit with notable disadvantages compared to
4060 <<defining-tracepoints,user-defined tracepoints>>:
4062 * All the emitted events have the same tracepoint provider and
4063 tracepoint names, respectively `lttng_ust_tracef` and `event`.
4064 * There is no static type checking.
4065 * The only event record field you actually get, named `msg`, is a string
4066 potentially containing the values you passed to `tracef()`
4067 using your own format string. This also means that you cannot filter
4068 events with a custom expression at run time because there are no
4070 * Since `tracef()` uses the C standard library's man:vasprintf(3)
4071 function behind the scenes to format the strings at run time, its
4072 expected performance is lower than with user-defined tracepoints,
4073 which do not require a conversion to a string.
4075 Taking this into consideration, `tracef()` is useful for some quick
4076 prototyping and debugging, but you should not consider it for any
4077 permanent and serious applicative instrumentation.
4083 ==== Use `tracelog()`
4085 The man:tracelog(3) API is very similar to <<tracef,`tracef()`>>, with
4086 the difference that it accepts an additional log level parameter.
4088 The goal of `tracelog()` is to ease the migration from logging to
4091 To use `tracelog()` in your application:
4093 . In the C or C++ source files where you need to use `tracelog()`,
4094 include `<lttng/tracelog.h>`:
4099 #include <lttng/tracelog.h>
4103 . In the application's source code, use `tracelog()` like you would use
4104 man:printf(3), except for the first parameter which is the log
4112 tracelog(TRACE_WARNING, "my message: %d (%s)",
4113 my_integer, my_string);
4119 See man:lttng-ust(3) for a list of available log level names.
4121 . Link your application with `liblttng-ust`:
4126 $ gcc -o app app.c -llttng-ust
4130 To trace the events that `tracelog()` calls emit with a log level
4131 _as severe as_ a specific log level:
4133 * <<enabling-disabling-events,Create an event rule>> which matches the
4134 `lttng_ust_tracelog:*` event name and a minimum level
4140 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4141 --loglevel=TRACE_WARNING
4145 To trace the events that `tracelog()` calls emit with a
4146 _specific log level_:
4148 * Create an event rule which matches the `lttng_ust_tracelog:*`
4149 event name and a specific log level:
4154 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4155 --loglevel-only=TRACE_INFO
4160 [[prebuilt-ust-helpers]]
4161 === Prebuilt user space tracing helpers
4163 The LTTng-UST package provides a few helpers in the form or preloadable
4164 shared objects which automatically instrument system functions and
4167 The helper shared objects are normally found in dir:{/usr/lib}. If you
4168 built LTTng-UST <<building-from-source,from source>>, they are probably
4169 located in dir:{/usr/local/lib}.
4171 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
4174 path:{liblttng-ust-libc-wrapper.so}::
4175 path:{liblttng-ust-pthread-wrapper.so}::
4176 <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
4177 memory and POSIX threads function tracing>>.
4179 path:{liblttng-ust-cyg-profile.so}::
4180 path:{liblttng-ust-cyg-profile-fast.so}::
4181 <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
4183 path:{liblttng-ust-dl.so}::
4184 <<liblttng-ust-dl,Dynamic linker tracing>>.
4186 To use a user space tracing helper with any user application:
4188 * Preload the helper shared object when you start the application:
4193 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
4197 You can preload more than one helper:
4202 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
4208 [[liblttng-ust-libc-pthread-wrapper]]
4209 ==== Instrument C standard library memory and POSIX threads functions
4211 The path:{liblttng-ust-libc-wrapper.so} and
4212 path:{liblttng-ust-pthread-wrapper.so} helpers
4213 add instrumentation to some C standard library and POSIX
4217 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
4219 |TP provider name |TP name |Instrumented function
4221 .6+|`lttng_ust_libc` |`malloc` |man:malloc(3)
4222 |`calloc` |man:calloc(3)
4223 |`realloc` |man:realloc(3)
4224 |`free` |man:free(3)
4225 |`memalign` |man:memalign(3)
4226 |`posix_memalign` |man:posix_memalign(3)
4230 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
4232 |TP provider name |TP name |Instrumented function
4234 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
4235 |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
4236 |`pthread_mutex_trylock` |man:pthread_mutex_trylock(3p)
4237 |`pthread_mutex_unlock` |man:pthread_mutex_unlock(3p)
4240 When you preload the shared object, it replaces the functions listed
4241 in the previous tables by wrappers which contain tracepoints and call
4242 the replaced functions.
4245 [[liblttng-ust-cyg-profile]]
4246 ==== Instrument function entry and exit
4248 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
4249 to the entry and exit points of functions.
4251 man:gcc(1) and man:clang(1) have an option named
4252 https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html[`-finstrument-functions`]
4253 which generates instrumentation calls for entry and exit to functions.
4254 The LTTng-UST function tracing helpers,
4255 path:{liblttng-ust-cyg-profile.so} and
4256 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
4257 to add tracepoints to the two generated functions (which contain
4258 `cyg_profile` in their names, hence the helper's name).
4260 To use the LTTng-UST function tracing helper, the source files to
4261 instrument must be built using the `-finstrument-functions` compiler
4264 There are two versions of the LTTng-UST function tracing helper:
4266 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
4267 that you should only use when it can be _guaranteed_ that the
4268 complete event stream is recorded without any lost event record.
4269 Any kind of duplicate information is left out.
4271 Assuming no event record is lost, having only the function addresses on
4272 entry is enough to create a call graph, since an event record always
4273 contains the ID of the CPU that generated it.
4275 You can use a tool like man:addr2line(1) to convert function addresses
4276 back to source file names and line numbers.
4278 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
4279 which also works in use cases where event records might get discarded or
4280 not recorded from application startup.
4281 In these cases, the trace analyzer needs more information to be
4282 able to reconstruct the program flow.
4284 See man:lttng-ust-cyg-profile(3) to learn more about the instrumentation
4285 points of this helper.
4287 All the tracepoints that this helper provides have the
4288 log level `TRACE_DEBUG_FUNCTION` (see man:lttng-ust(3)).
4290 TIP: It's sometimes a good idea to limit the number of source files that
4291 you compile with the `-finstrument-functions` option to prevent LTTng
4292 from writing an excessive amount of trace data at run time. When using
4293 man:gcc(1), you can use the
4294 `-finstrument-functions-exclude-function-list` option to avoid
4295 instrument entries and exits of specific function names.
4300 ==== Instrument the dynamic linker
4302 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4303 man:dlopen(3) and man:dlclose(3) function calls.
4305 See man:lttng-ust-dl(3) to learn more about the instrumentation points
4310 [[java-application]]
4311 === User space Java agent
4313 You can instrument any Java application which uses one of the following
4316 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4317 (JUL) core logging facilities.
4318 * http://logging.apache.org/log4j/1.2/[**Apache log4j 1.2**], since
4319 LTTng 2.6. Note that Apache Log4j{nbsp}2 is not supported.
4322 .LTTng-UST Java agent imported by a Java application.
4323 image::java-app.png[]
4325 Note that the methods described below are new in LTTng{nbsp}{revision}.
4326 Previous LTTng versions use another technique.
4328 NOTE: We use http://openjdk.java.net/[OpenJDK]{nbsp}8 for development
4329 and https://ci.lttng.org/[continuous integration], thus this version is
4330 directly supported. However, the LTTng-UST Java agent is also tested
4331 with OpenJDK{nbsp}7.
4336 ==== Use the LTTng-UST Java agent for `java.util.logging`
4338 To use the LTTng-UST Java agent in a Java application which uses
4339 `java.util.logging` (JUL):
4341 . In the Java application's source code, import the LTTng-UST
4342 log handler package for `java.util.logging`:
4347 import org.lttng.ust.agent.jul.LttngLogHandler;
4351 . Create an LTTng-UST JUL log handler:
4356 Handler lttngUstLogHandler = new LttngLogHandler();
4360 . Add this handler to the JUL loggers which should emit LTTng events:
4365 Logger myLogger = Logger.getLogger("some-logger");
4367 myLogger.addHandler(lttngUstLogHandler);
4371 . Use `java.util.logging` log statements and configuration as usual.
4372 The loggers with an attached LTTng-UST log handler can emit
4375 . Before exiting the application, remove the LTTng-UST log handler from
4376 the loggers attached to it and call its `close()` method:
4381 myLogger.removeHandler(lttngUstLogHandler);
4382 lttngUstLogHandler.close();
4386 This is not strictly necessary, but it is recommended for a clean
4387 disposal of the handler's resources.
4389 . Include the LTTng-UST Java agent's common and JUL-specific JAR files,
4390 path:{lttng-ust-agent-common.jar} and path:{lttng-ust-agent-jul.jar},
4392 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4393 path] when you build the Java application.
4395 The JAR files are typically located in dir:{/usr/share/java}.
4397 IMPORTANT: The LTTng-UST Java agent must be
4398 <<installing-lttng,installed>> for the logging framework your
4401 .Use the LTTng-UST Java agent for `java.util.logging`.
4406 import java.io.IOException;
4407 import java.util.logging.Handler;
4408 import java.util.logging.Logger;
4409 import org.lttng.ust.agent.jul.LttngLogHandler;
4413 private static final int answer = 42;
4415 public static void main(String[] argv) throws Exception
4418 Logger logger = Logger.getLogger("jello");
4420 // Create an LTTng-UST log handler
4421 Handler lttngUstLogHandler = new LttngLogHandler();
4423 // Add the LTTng-UST log handler to our logger
4424 logger.addHandler(lttngUstLogHandler);
4427 logger.info("some info");
4428 logger.warning("some warning");
4430 logger.finer("finer information; the answer is " + answer);
4432 logger.severe("error!");
4434 // Not mandatory, but cleaner
4435 logger.removeHandler(lttngUstLogHandler);
4436 lttngUstLogHandler.close();
4445 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4448 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4449 <<enabling-disabling-events,create an event rule>> matching the
4450 `jello` JUL logger, and <<basic-tracing-session-control,start tracing>>:
4455 $ lttng enable-event --jul jello
4459 Run the compiled class:
4463 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4466 <<basic-tracing-session-control,Stop tracing>> and inspect the
4476 In the resulting trace, an <<event,event record>> generated by a Java
4477 application using `java.util.logging` is named `lttng_jul:event` and
4478 has the following fields:
4481 Log record's message.
4487 Name of the class in which the log statement was executed.
4490 Name of the method in which the log statement was executed.
4493 Logging time (timestamp in milliseconds).
4496 Log level integer value.
4499 ID of the thread in which the log statement was executed.
4501 You can use the opt:lttng-enable-event(1):--loglevel or
4502 opt:lttng-enable-event(1):--loglevel-only option of the
4503 man:lttng-enable-event(1) command to target a range of JUL log levels
4504 or a specific JUL log level.
4509 ==== Use the LTTng-UST Java agent for Apache log4j
4511 To use the LTTng-UST Java agent in a Java application which uses
4514 . In the Java application's source code, import the LTTng-UST
4515 log appender package for Apache log4j:
4520 import org.lttng.ust.agent.log4j.LttngLogAppender;
4524 . Create an LTTng-UST log4j log appender:
4529 Appender lttngUstLogAppender = new LttngLogAppender();
4533 . Add this appender to the log4j loggers which should emit LTTng events:
4538 Logger myLogger = Logger.getLogger("some-logger");
4540 myLogger.addAppender(lttngUstLogAppender);
4544 . Use Apache log4j log statements and configuration as usual. The
4545 loggers with an attached LTTng-UST log appender can emit LTTng events.
4547 . Before exiting the application, remove the LTTng-UST log appender from
4548 the loggers attached to it and call its `close()` method:
4553 myLogger.removeAppender(lttngUstLogAppender);
4554 lttngUstLogAppender.close();
4558 This is not strictly necessary, but it is recommended for a clean
4559 disposal of the appender's resources.
4561 . Include the LTTng-UST Java agent's common and log4j-specific JAR
4562 files, path:{lttng-ust-agent-common.jar} and
4563 path:{lttng-ust-agent-log4j.jar}, in the
4564 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4565 path] when you build the Java application.
4567 The JAR files are typically located in dir:{/usr/share/java}.
4569 IMPORTANT: The LTTng-UST Java agent must be
4570 <<installing-lttng,installed>> for the logging framework your
4573 .Use the LTTng-UST Java agent for Apache log4j.
4578 import org.apache.log4j.Appender;
4579 import org.apache.log4j.Logger;
4580 import org.lttng.ust.agent.log4j.LttngLogAppender;
4584 private static final int answer = 42;
4586 public static void main(String[] argv) throws Exception
4589 Logger logger = Logger.getLogger("jello");
4591 // Create an LTTng-UST log appender
4592 Appender lttngUstLogAppender = new LttngLogAppender();
4594 // Add the LTTng-UST log appender to our logger
4595 logger.addAppender(lttngUstLogAppender);
4598 logger.info("some info");
4599 logger.warn("some warning");
4601 logger.debug("debug information; the answer is " + answer);
4603 logger.fatal("error!");
4605 // Not mandatory, but cleaner
4606 logger.removeAppender(lttngUstLogAppender);
4607 lttngUstLogAppender.close();
4613 Build this example (`$LOG4JPATH` is the path to the Apache log4j JAR
4618 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH Test.java
4621 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4622 <<enabling-disabling-events,create an event rule>> matching the
4623 `jello` log4j logger, and <<basic-tracing-session-control,start tracing>>:
4628 $ lttng enable-event --log4j jello
4632 Run the compiled class:
4636 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH:. Test
4639 <<basic-tracing-session-control,Stop tracing>> and inspect the
4649 In the resulting trace, an <<event,event record>> generated by a Java
4650 application using log4j is named `lttng_log4j:event` and
4651 has the following fields:
4654 Log record's message.
4660 Name of the class in which the log statement was executed.
4663 Name of the method in which the log statement was executed.
4666 Name of the file in which the executed log statement is located.
4669 Line number at which the log statement was executed.
4675 Log level integer value.
4678 Name of the Java thread in which the log statement was executed.
4680 You can use the opt:lttng-enable-event(1):--loglevel or
4681 opt:lttng-enable-event(1):--loglevel-only option of the
4682 man:lttng-enable-event(1) command to target a range of Apache log4j log levels
4683 or a specific log4j log level.
4687 [[java-application-context]]
4688 ==== Provide application-specific context fields in a Java application
4690 A Java application-specific context field is a piece of state provided
4691 by the application which <<adding-context,you can add>>, using the
4692 man:lttng-add-context(1) command, to each <<event,event record>>
4693 produced by the log statements of this application.
4695 For example, a given object might have a current request ID variable.
4696 You can create a context information retriever for this object and
4697 assign a name to this current request ID. You can then, using the
4698 man:lttng-add-context(1) command, add this context field by name to
4699 the JUL or log4j <<channel,channel>>.
4701 To provide application-specific context fields in a Java application:
4703 . In the Java application's source code, import the LTTng-UST
4704 Java agent context classes and interfaces:
4709 import org.lttng.ust.agent.context.ContextInfoManager;
4710 import org.lttng.ust.agent.context.IContextInfoRetriever;
4714 . Create a context information retriever class, that is, a class which
4715 implements the `IContextInfoRetriever` interface:
4720 class MyContextInfoRetriever implements IContextInfoRetriever
4723 public Object retrieveContextInfo(String key)
4725 if (key.equals("intCtx")) {
4727 } else if (key.equals("strContext")) {
4728 return "context value!";
4737 This `retrieveContextInfo()` method is the only member of the
4738 `IContextInfoRetriever` interface. Its role is to return the current
4739 value of a state by name to create a context field. The names of the
4740 context fields and which state variables they return depends on your
4743 All primitive types and objects are supported as context fields.
4744 When `retrieveContextInfo()` returns an object, the context field
4745 serializer calls its `toString()` method to add a string field to
4746 event records. The method can also return `null`, which means that
4747 no context field is available for the required name.
4749 . Register an instance of your context information retriever class to
4750 the context information manager singleton:
4755 IContextInfoRetriever cir = new MyContextInfoRetriever();
4756 ContextInfoManager cim = ContextInfoManager.getInstance();
4757 cim.registerContextInfoRetriever("retrieverName", cir);
4761 . Before exiting the application, remove your context information
4762 retriever from the context information manager singleton:
4767 ContextInfoManager cim = ContextInfoManager.getInstance();
4768 cim.unregisterContextInfoRetriever("retrieverName");
4772 This is not strictly necessary, but it is recommended for a clean
4773 disposal of some manager's resources.
4775 . Build your Java application with LTTng-UST Java agent support as
4776 usual, following the procedure for either the <<jul,JUL>> or
4777 <<log4j,Apache log4j>> framework.
4780 .Provide application-specific context fields in a Java application.
4785 import java.util.logging.Handler;
4786 import java.util.logging.Logger;
4787 import org.lttng.ust.agent.jul.LttngLogHandler;
4788 import org.lttng.ust.agent.context.ContextInfoManager;
4789 import org.lttng.ust.agent.context.IContextInfoRetriever;
4793 // Our context information retriever class
4794 private static class MyContextInfoRetriever
4795 implements IContextInfoRetriever
4798 public Object retrieveContextInfo(String key) {
4799 if (key.equals("intCtx")) {
4801 } else if (key.equals("strContext")) {
4802 return "context value!";
4809 private static final int answer = 42;
4811 public static void main(String args[]) throws Exception
4813 // Get the context information manager instance
4814 ContextInfoManager cim = ContextInfoManager.getInstance();
4816 // Create and register our context information retriever
4817 IContextInfoRetriever cir = new MyContextInfoRetriever();
4818 cim.registerContextInfoRetriever("myRetriever", cir);
4821 Logger logger = Logger.getLogger("jello");
4823 // Create an LTTng-UST log handler
4824 Handler lttngUstLogHandler = new LttngLogHandler();
4826 // Add the LTTng-UST log handler to our logger
4827 logger.addHandler(lttngUstLogHandler);
4830 logger.info("some info");
4831 logger.warning("some warning");
4833 logger.finer("finer information; the answer is " + answer);
4835 logger.severe("error!");
4837 // Not mandatory, but cleaner
4838 logger.removeHandler(lttngUstLogHandler);
4839 lttngUstLogHandler.close();
4840 cim.unregisterContextInfoRetriever("myRetriever");
4849 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4852 <<creating-destroying-tracing-sessions,Create a tracing session>>
4853 and <<enabling-disabling-events,create an event rule>> matching the
4859 $ lttng enable-event --jul jello
4862 <<adding-context,Add the application-specific context fields>> to the
4867 $ lttng add-context --jul --type='$app.myRetriever:intCtx'
4868 $ lttng add-context --jul --type='$app.myRetriever:strContext'
4871 <<basic-tracing-session-control,Start tracing>>:
4878 Run the compiled class:
4882 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4885 <<basic-tracing-session-control,Stop tracing>> and inspect the
4897 [[python-application]]
4898 === User space Python agent
4900 You can instrument a Python 2 or Python 3 application which uses the
4901 standard https://docs.python.org/3/library/logging.html[`logging`]
4904 Each log statement emits an LTTng event once the
4905 application module imports the
4906 <<lttng-ust-agents,LTTng-UST Python agent>> package.
4909 .A Python application importing the LTTng-UST Python agent.
4910 image::python-app.png[]
4912 To use the LTTng-UST Python agent:
4914 . In the Python application's source code, import the LTTng-UST Python
4924 The LTTng-UST Python agent automatically adds its logging handler to the
4925 root logger at import time.
4927 Any log statement that the application executes before this import does
4928 not emit an LTTng event.
4930 IMPORTANT: The LTTng-UST Python agent must be
4931 <<installing-lttng,installed>>.
4933 . Use log statements and logging configuration as usual.
4934 Since the LTTng-UST Python agent adds a handler to the _root_
4935 logger, you can trace any log statement from any logger.
4937 .Use the LTTng-UST Python agent.
4948 logging.basicConfig()
4949 logger = logging.getLogger('my-logger')
4952 logger.debug('debug message')
4953 logger.info('info message')
4954 logger.warn('warn message')
4955 logger.error('error message')
4956 logger.critical('critical message')
4960 if __name__ == '__main__':
4964 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
4965 logging handler which prints to the standard error stream, is not
4966 strictly required for LTTng-UST tracing to work, but in versions of
4967 Python preceding 3.2, you could see a warning message which indicates
4968 that no handler exists for the logger `my-logger`.
4970 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4971 <<enabling-disabling-events,create an event rule>> matching the
4972 `my-logger` Python logger, and <<basic-tracing-session-control,start
4978 $ lttng enable-event --python my-logger
4982 Run the Python script:
4989 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
4999 In the resulting trace, an <<event,event record>> generated by a Python
5000 application is named `lttng_python:event` and has the following fields:
5003 Logging time (string).
5006 Log record's message.
5012 Name of the function in which the log statement was executed.
5015 Line number at which the log statement was executed.
5018 Log level integer value.
5021 ID of the Python thread in which the log statement was executed.
5024 Name of the Python thread in which the log statement was executed.
5026 You can use the opt:lttng-enable-event(1):--loglevel or
5027 opt:lttng-enable-event(1):--loglevel-only option of the
5028 man:lttng-enable-event(1) command to target a range of Python log levels
5029 or a specific Python log level.
5031 When an application imports the LTTng-UST Python agent, the agent tries
5032 to register to a <<lttng-sessiond,session daemon>>. Note that you must
5033 <<start-sessiond,start the session daemon>> _before_ you run the Python
5034 application. If a session daemon is found, the agent tries to register
5035 to it during 5{nbsp}seconds, after which the application continues
5036 without LTTng tracing support. You can override this timeout value with
5037 the env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
5040 If the session daemon stops while a Python application with an imported
5041 LTTng-UST Python agent runs, the agent retries to connect and to
5042 register to a session daemon every 3{nbsp}seconds. You can override this
5043 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
5048 [[proc-lttng-logger-abi]]
5051 The `lttng-tracer` Linux kernel module, part of
5052 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger file
5053 path:{/proc/lttng-logger} when it's loaded. Any application can write
5054 text data to this file to emit an LTTng event.
5057 .An application writes to the LTTng logger file to emit an LTTng event.
5058 image::lttng-logger.png[]
5060 The LTTng logger is the quickest method--not the most efficient,
5061 however--to add instrumentation to an application. It is designed
5062 mostly to instrument shell scripts:
5066 $ echo "Some message, some $variable" > /proc/lttng-logger
5069 Any event that the LTTng logger emits is named `lttng_logger` and
5070 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
5071 other instrumentation points in the kernel tracing domain, **any Unix
5072 user** can <<enabling-disabling-events,create an event rule>> which
5073 matches its event name, not only the root user or users in the
5074 <<tracing-group,tracing group>>.
5076 To use the LTTng logger:
5078 * From any application, write text data to the path:{/proc/lttng-logger}
5081 The `msg` field of `lttng_logger` event records contains the
5084 NOTE: The maximum message length of an LTTng logger event is
5085 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
5086 than one event to contain the remaining data.
5088 You should not use the LTTng logger to trace a user application which
5089 can be instrumented in a more efficient way, namely:
5091 * <<c-application,C and $$C++$$ applications>>.
5092 * <<java-application,Java applications>>.
5093 * <<python-application,Python applications>>.
5095 .Use the LTTng logger.
5100 echo 'Hello, World!' > /proc/lttng-logger
5102 df --human-readable --print-type / > /proc/lttng-logger
5105 <<creating-destroying-tracing-sessions,Create a tracing session>>,
5106 <<enabling-disabling-events,create an event rule>> matching the
5107 `lttng_logger` Linux kernel tracepoint, and
5108 <<basic-tracing-session-control,start tracing>>:
5113 $ lttng enable-event --kernel lttng_logger
5117 Run the Bash script:
5124 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
5135 [[instrumenting-linux-kernel]]
5136 === LTTng kernel tracepoints
5138 NOTE: This section shows how to _add_ instrumentation points to the
5139 Linux kernel. The kernel's subsystems are already thoroughly
5140 instrumented at strategic places for LTTng when you
5141 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
5145 There are two methods to instrument the Linux kernel:
5147 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
5148 tracepoint which uses the `TRACE_EVENT()` API.
5150 Choose this if you want to instrumentation a Linux kernel tree with an
5151 instrumentation point compatible with ftrace, perf, and SystemTap.
5153 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
5154 instrument an out-of-tree kernel module.
5156 Choose this if you don't need ftrace, perf, or SystemTap support.
5160 [[linux-add-lttng-layer]]
5161 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
5163 This section shows how to add an LTTng layer to existing ftrace
5164 instrumentation using the `TRACE_EVENT()` API.
5166 This section does not document the `TRACE_EVENT()` macro. You can
5167 read the following articles to learn more about this API:
5169 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part 1)]
5170 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part 2)]
5171 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part 3)]
5173 The following procedure assumes that your ftrace tracepoints are
5174 correctly defined in their own header and that they are created in
5175 one source file using the `CREATE_TRACE_POINTS` definition.
5177 To add an LTTng layer over an existing ftrace tracepoint:
5179 . Make sure the following kernel configuration options are
5185 * `CONFIG_HIGH_RES_TIMERS`
5186 * `CONFIG_TRACEPOINTS`
5189 . Build the Linux source tree with your custom ftrace tracepoints.
5190 . Boot the resulting Linux image on your target system.
5192 Confirm that the tracepoints exist by looking for their names in the
5193 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
5194 is your subsystem's name.
5196 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
5201 $ cd $(mktemp -d) &&
5202 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.9.tar.bz2 &&
5203 tar -xf lttng-modules-latest-2.9.tar.bz2 &&
5204 cd lttng-modules-2.9.*
5208 . In dir:{instrumentation/events/lttng-module}, relative to the root
5209 of the LTTng-modules source tree, create a header file named
5210 +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
5211 LTTng-modules tracepoint definitions using the LTTng-modules
5214 Start with this template:
5218 .path:{instrumentation/events/lttng-module/my_subsys.h}
5221 #define TRACE_SYSTEM my_subsys
5223 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
5224 #define _LTTNG_MY_SUBSYS_H
5226 #include "../../../probes/lttng-tracepoint-event.h"
5227 #include <linux/tracepoint.h>
5229 LTTNG_TRACEPOINT_EVENT(
5231 * Format is identical to TRACE_EVENT()'s version for the three
5232 * following macro parameters:
5235 TP_PROTO(int my_int, const char *my_string),
5236 TP_ARGS(my_int, my_string),
5238 /* LTTng-modules specific macros */
5240 ctf_integer(int, my_int_field, my_int)
5241 ctf_string(my_bar_field, my_bar)
5245 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
5247 #include "../../../probes/define_trace.h"
5251 The entries in the `TP_FIELDS()` section are the list of fields for the
5252 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
5253 ftrace's `TRACE_EVENT()` macro.
5255 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
5256 complete description of the available `ctf_*()` macros.
5258 . Create the LTTng-modules probe's kernel module C source file,
5259 +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
5264 .path:{probes/lttng-probe-my-subsys.c}
5266 #include <linux/module.h>
5267 #include "../lttng-tracer.h"
5270 * Build-time verification of mismatch between mainline
5271 * TRACE_EVENT() arguments and the LTTng-modules adaptation
5272 * layer LTTNG_TRACEPOINT_EVENT() arguments.
5274 #include <trace/events/my_subsys.h>
5276 /* Create LTTng tracepoint probes */
5277 #define LTTNG_PACKAGE_BUILD
5278 #define CREATE_TRACE_POINTS
5279 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
5281 #include "../instrumentation/events/lttng-module/my_subsys.h"
5283 MODULE_LICENSE("GPL and additional rights");
5284 MODULE_AUTHOR("Your name <your-email>");
5285 MODULE_DESCRIPTION("LTTng my_subsys probes");
5286 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
5287 __stringify(LTTNG_MODULES_MINOR_VERSION) "."
5288 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
5289 LTTNG_MODULES_EXTRAVERSION);
5293 . Edit path:{probes/Makefile} and add your new kernel module object
5294 next to the existing ones:
5298 .path:{probes/Makefile}
5302 obj-m += lttng-probe-module.o
5303 obj-m += lttng-probe-power.o
5305 obj-m += lttng-probe-my-subsys.o
5311 . Build and install the LTTng kernel modules:
5316 $ make KERNELDIR=/path/to/linux
5317 # make modules_install && depmod -a
5321 Replace `/path/to/linux` with the path to the Linux source tree where
5322 you defined and used tracepoints with ftrace's `TRACE_EVENT()` macro.
5324 Note that you can also use the
5325 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
5326 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
5327 C code that need to be executed before the event fields are recorded.
5329 The best way to learn how to use the previous LTTng-modules macros is to
5330 inspect the existing LTTng-modules tracepoint definitions in the
5331 dir:{instrumentation/events/lttng-module} header files. Compare them
5332 with the Linux kernel mainline versions in the
5333 dir:{include/trace/events} directory of the Linux source tree.
5337 [[lttng-tracepoint-event-code]]
5338 ===== Use custom C code to access the data for tracepoint fields
5340 Although we recommended to always use the
5341 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
5342 the arguments and fields of an LTTng-modules tracepoint when possible,
5343 sometimes you need a more complex process to access the data that the
5344 tracer records as event record fields. In other words, you need local
5345 variables and multiple C{nbsp}statements instead of simple
5346 argument-based expressions that you pass to the
5347 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
5349 You can use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
5350 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
5351 a block of C{nbsp}code to be executed before LTTng records the fields.
5352 The structure of this macro is:
5355 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
5357 LTTNG_TRACEPOINT_EVENT_CODE(
5359 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5360 * version for the following three macro parameters:
5363 TP_PROTO(int my_int, const char *my_string),
5364 TP_ARGS(my_int, my_string),
5366 /* Declarations of custom local variables */
5369 unsigned long b = 0;
5370 const char *name = "(undefined)";
5371 struct my_struct *my_struct;
5375 * Custom code which uses both tracepoint arguments
5376 * (in TP_ARGS()) and local variables (in TP_locvar()).
5378 * Local variables are actually members of a structure pointed
5379 * to by the special variable tp_locvar.
5383 tp_locvar->a = my_int + 17;
5384 tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
5385 tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
5386 tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
5387 put_my_struct(tp_locvar->my_struct);
5396 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5397 * version for this, except that tp_locvar members can be
5398 * used in the argument expression parameters of
5399 * the ctf_*() macros.
5402 ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
5403 ctf_integer(int, my_struct_a, tp_locvar->a)
5404 ctf_string(my_string_field, my_string)
5405 ctf_string(my_struct_name, tp_locvar->name)
5410 IMPORTANT: The C code defined in `TP_code()` must not have any side
5411 effects when executed. In particular, the code must not allocate
5412 memory or get resources without deallocating this memory or putting
5413 those resources afterwards.
5416 [[instrumenting-linux-kernel-tracing]]
5417 ==== Load and unload a custom probe kernel module
5419 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
5420 kernel module>> in the kernel before it can emit LTTng events.
5422 To load the default probe kernel modules and a custom probe kernel
5425 * Use the opt:lttng-sessiond(8):--extra-kmod-probes option to give extra
5426 probe modules to load when starting a root <<lttng-sessiond,session
5430 .Load the `my_subsys`, `usb`, and the default probe modules.
5434 # lttng-sessiond --extra-kmod-probes=my_subsys,usb
5439 You only need to pass the subsystem name, not the whole kernel module
5442 To load _only_ a given custom probe kernel module:
5444 * Use the opt:lttng-sessiond(8):--kmod-probes option to give the probe
5445 modules to load when starting a root session daemon:
5448 .Load only the `my_subsys` and `usb` probe modules.
5452 # lttng-sessiond --kmod-probes=my_subsys,usb
5457 To confirm that a probe module is loaded:
5464 $ lsmod | grep lttng_probe_usb
5468 To unload the loaded probe modules:
5470 * Kill the session daemon with `SIGTERM`:
5475 # pkill lttng-sessiond
5479 You can also use man:modprobe(8)'s `--remove` option if the session
5480 daemon terminates abnormally.
5483 [[controlling-tracing]]
5486 Once an application or a Linux kernel is
5487 <<instrumenting,instrumented>> for LTTng tracing,
5490 This section is divided in topics on how to use the various
5491 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
5492 command-line tool>>, to _control_ the LTTng daemons and tracers.
5494 NOTE: In the following subsections, we refer to an man:lttng(1) command
5495 using its man page name. For example, instead of _Run the `create`
5496 command to..._, we use _Run the man:lttng-create(1) command to..._.
5500 === Start a session daemon
5502 In some situations, you need to run a <<lttng-sessiond,session daemon>>
5503 (man:lttng-sessiond(8)) _before_ you can use the man:lttng(1)
5506 You will see the following error when you run a command while no session
5510 Error: No session daemon is available
5513 The only command that automatically runs a session daemon is
5514 man:lttng-create(1), which you use to
5515 <<creating-destroying-tracing-sessions,create a tracing session>>. While
5516 this is most of the time the first operation that you do, sometimes it's
5517 not. Some examples are:
5519 * <<list-instrumentation-points,List the available instrumentation points>>.
5520 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
5522 [[tracing-group]] Each Unix user must have its own running session
5523 daemon to trace user applications. The session daemon that the root user
5524 starts is the only one allowed to control the LTTng kernel tracer. Users
5525 that are part of the _tracing group_ can control the root session
5526 daemon. The default tracing group name is `tracing`; you can set it to
5527 something else with the opt:lttng-sessiond(8):--group option when you
5528 start the root session daemon.
5530 To start a user session daemon:
5532 * Run man:lttng-sessiond(8):
5537 $ lttng-sessiond --daemonize
5541 To start the root session daemon:
5543 * Run man:lttng-sessiond(8) as the root user:
5548 # lttng-sessiond --daemonize
5552 In both cases, remove the opt:lttng-sessiond(8):--daemonize option to
5553 start the session daemon in foreground.
5555 To stop a session daemon, use man:kill(1) on its process ID (standard
5558 Note that some Linux distributions could manage the LTTng session daemon
5559 as a service. In this case, you should use the service manager to
5560 start, restart, and stop session daemons.
5563 [[creating-destroying-tracing-sessions]]
5564 === Create and destroy a tracing session
5566 Almost all the LTTng control operations happen in the scope of
5567 a <<tracing-session,tracing session>>, which is the dialogue between the
5568 <<lttng-sessiond,session daemon>> and you.
5570 To create a tracing session with a generated name:
5572 * Use the man:lttng-create(1) command:
5581 The created tracing session's name is `auto` followed by the
5584 To create a tracing session with a specific name:
5586 * Use the optional argument of the man:lttng-create(1) command:
5591 $ lttng create my-session
5595 Replace `my-session` with the specific tracing session name.
5597 LTTng appends the creation date to the created tracing session's name.
5599 LTTng writes the traces of a tracing session in
5600 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
5601 name of the tracing session. Note that the env:LTTNG_HOME environment
5602 variable defaults to `$HOME` if not set.
5604 To output LTTng traces to a non-default location:
5606 * Use the opt:lttng-create(1):--output option of the man:lttng-create(1) command:
5611 $ lttng create my-session --output=/tmp/some-directory
5615 You may create as many tracing sessions as you wish.
5617 To list all the existing tracing sessions for your Unix user:
5619 * Use the man:lttng-list(1) command:
5628 When you create a tracing session, it is set as the _current tracing
5629 session_. The following man:lttng(1) commands operate on the current
5630 tracing session when you don't specify one:
5632 [role="list-3-cols"]
5649 To change the current tracing session:
5651 * Use the man:lttng-set-session(1) command:
5656 $ lttng set-session new-session
5660 Replace `new-session` by the name of the new current tracing session.
5662 When you are done tracing in a given tracing session, you can destroy
5663 it. This operation frees the resources taken by the tracing session
5664 to destroy; it does not destroy the trace data that LTTng wrote for
5665 this tracing session.
5667 To destroy the current tracing session:
5669 * Use the man:lttng-destroy(1) command:
5679 [[list-instrumentation-points]]
5680 === List the available instrumentation points
5682 The <<lttng-sessiond,session daemon>> can query the running instrumented
5683 user applications and the Linux kernel to get a list of available
5684 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
5685 they are tracepoints and system calls. For the user space tracing
5686 domain, they are tracepoints. For the other tracing domains, they are
5689 To list the available instrumentation points:
5691 * Use the man:lttng-list(1) command with the requested tracing domain's
5695 * opt:lttng-list(1):--kernel: Linux kernel tracepoints (your Unix user
5696 must be a root user, or it must be a member of the
5697 <<tracing-group,tracing group>>).
5698 * opt:lttng-list(1):--kernel with opt:lttng-list(1):--syscall: Linux
5699 kernel system calls (your Unix user must be a root user, or it must be
5700 a member of the tracing group).
5701 * opt:lttng-list(1):--userspace: user space tracepoints.
5702 * opt:lttng-list(1):--jul: `java.util.logging` loggers.
5703 * opt:lttng-list(1):--log4j: Apache log4j loggers.
5704 * opt:lttng-list(1):--python: Python loggers.
5707 .List the available user space tracepoints.
5711 $ lttng list --userspace
5715 .List the available Linux kernel system call tracepoints.
5719 $ lttng list --kernel --syscall
5724 [[enabling-disabling-events]]
5725 === Create and enable an event rule
5727 Once you <<creating-destroying-tracing-sessions,create a tracing
5728 session>>, you can create <<event,event rules>> with the
5729 man:lttng-enable-event(1) command.
5731 You specify each condition with a command-line option. The available
5732 condition options are shown in the following table.
5734 [role="growable",cols="asciidoc,asciidoc,default"]
5735 .Condition command-line options for the man:lttng-enable-event(1) command.
5737 |Option |Description |Applicable tracing domains
5743 . +--probe=__ADDR__+
5744 . +--function=__ADDR__+
5747 Instead of using the default _tracepoint_ instrumentation type, use:
5749 . A Linux system call.
5750 . A Linux https://lwn.net/Articles/132196/[KProbe] (symbol or address).
5751 . The entry and return points of a Linux function (symbol or address).
5755 |First positional argument.
5758 Tracepoint or system call name. In the case of a Linux KProbe or
5759 function, this is a custom name given to the event rule. With the
5760 JUL, log4j, and Python domains, this is a logger name.
5762 With a tracepoint, logger, or system call name, the last character
5763 can be `*` to match anything that remains.
5770 . +--loglevel=__LEVEL__+
5771 . +--loglevel-only=__LEVEL__+
5774 . Match only tracepoints or log statements with a logging level at
5775 least as severe as +__LEVEL__+.
5776 . Match only tracepoints or log statements with a logging level
5777 equal to +__LEVEL__+.
5779 See man:lttng-enable-event(1) for the list of available logging level
5782 |User space, JUL, log4j, and Python.
5784 |+--exclude=__EXCLUSIONS__+
5787 When you use a `*` character at the end of the tracepoint or logger
5788 name (first positional argument), exclude the specific names in the
5789 comma-delimited list +__EXCLUSIONS__+.
5792 User space, JUL, log4j, and Python.
5794 |+--filter=__EXPR__+
5797 Match only events which satisfy the expression +__EXPR__+.
5799 See man:lttng-enable-event(1) to learn more about the syntax of a
5806 You attach an event rule to a <<channel,channel>> on creation. If you do
5807 not specify the channel with the opt:lttng-enable-event(1):--channel
5808 option, and if the event rule to create is the first in its
5809 <<domain,tracing domain>> for a given tracing session, then LTTng
5810 creates a _default channel_ for you. This default channel is reused in
5811 subsequent invocations of the man:lttng-enable-event(1) command for the
5812 same tracing domain.
5814 An event rule is always enabled at creation time.
5816 The following examples show how you can combine the previous
5817 command-line options to create simple to more complex event rules.
5819 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5823 $ lttng enable-event --kernel sched_switch
5827 .Create an event rule matching four Linux kernel system calls (default channel).
5831 $ lttng enable-event --kernel --syscall open,write,read,close
5835 .Create event rules matching tracepoints with filter expressions (default channel).
5839 $ lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
5844 $ lttng enable-event --kernel --all \
5845 --filter='$ctx.tid == 1988 || $ctx.tid == 1534'
5850 $ lttng enable-event --jul my_logger \
5851 --filter='$app.retriever:cur_msg_id > 3'
5854 IMPORTANT: Make sure to always quote the filter string when you
5855 use man:lttng(1) from a shell.
5858 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
5862 $ lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
5865 IMPORTANT: Make sure to always quote the wildcard character when you
5866 use man:lttng(1) from a shell.
5869 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
5873 $ lttng enable-event --python my-app.'*' \
5874 --exclude='my-app.module,my-app.hello'
5878 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
5882 $ lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
5886 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
5890 $ lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
5894 The event rules of a given channel form a whitelist: as soon as an
5895 emitted event passes one of them, LTTng can record the event. For
5896 example, an event named `my_app:my_tracepoint` emitted from a user space
5897 tracepoint with a `TRACE_ERROR` log level passes both of the following
5902 $ lttng enable-event --userspace my_app:my_tracepoint
5903 $ lttng enable-event --userspace my_app:my_tracepoint \
5904 --loglevel=TRACE_INFO
5907 The second event rule is redundant: the first one includes
5911 [[disable-event-rule]]
5912 === Disable an event rule
5914 To disable an event rule that you <<enabling-disabling-events,created>>
5915 previously, use the man:lttng-disable-event(1) command. This command
5916 disables _all_ the event rules (of a given tracing domain and channel)
5917 which match an instrumentation point. The other conditions are not
5918 supported as of LTTng{nbsp}{revision}.
5920 The LTTng tracer does not record an emitted event which passes
5921 a _disabled_ event rule.
5923 .Disable an event rule matching a Python logger (default channel).
5927 $ lttng disable-event --python my-logger
5931 .Disable an event rule matching all `java.util.logging` loggers (default channel).
5935 $ lttng disable-event --jul '*'
5939 .Disable _all_ the event rules of the default channel.
5941 The opt:lttng-disable-event(1):--all-events option is not, like the
5942 opt:lttng-enable-event(1):--all option of man:lttng-enable-event(1), the
5943 equivalent of the event name `*` (wildcard): it disables _all_ the event
5944 rules of a given channel.
5948 $ lttng disable-event --jul --all-events
5952 NOTE: You cannot delete an event rule once you create it.
5956 === Get the status of a tracing session
5958 To get the status of the current tracing session, that is, its
5959 parameters, its channels, event rules, and their attributes:
5961 * Use the man:lttng-status(1) command:
5971 To get the status of any tracing session:
5973 * Use the man:lttng-list(1) command with the tracing session's name:
5978 $ lttng list my-session
5982 Replace `my-session` with the desired tracing session's name.
5985 [[basic-tracing-session-control]]
5986 === Start and stop a tracing session
5988 Once you <<creating-destroying-tracing-sessions,create a tracing
5990 <<enabling-disabling-events,create one or more event rules>>,
5991 you can start and stop the tracers for this tracing session.
5993 To start tracing in the current tracing session:
5995 * Use the man:lttng-start(1) command:
6004 LTTng is very flexible: you can launch user applications before
6005 or after the you start the tracers. The tracers only record the events
6006 if they pass enabled event rules and if they occur while the tracers are
6009 To stop tracing in the current tracing session:
6011 * Use the man:lttng-stop(1) command:
6020 If there were <<channel-overwrite-mode-vs-discard-mode,lost event
6021 records>> or lost sub-buffers since the last time you ran
6022 man:lttng-start(1), warnings are printed when you run the
6023 man:lttng-stop(1) command.
6026 [[enabling-disabling-channels]]
6027 === Create a channel
6029 Once you create a tracing session, you can create a <<channel,channel>>
6030 with the man:lttng-enable-channel(1) command.
6032 Note that LTTng automatically creates a default channel when, for a
6033 given <<domain,tracing domain>>, no channels exist and you
6034 <<enabling-disabling-events,create>> the first event rule. This default
6035 channel is named `channel0` and its attributes are set to reasonable
6036 values. Therefore, you only need to create a channel when you need
6037 non-default attributes.
6039 You specify each non-default channel attribute with a command-line
6040 option when you use the man:lttng-enable-channel(1) command. The
6041 available command-line options are:
6043 [role="growable",cols="asciidoc,asciidoc"]
6044 .Command-line options for the man:lttng-enable-channel(1) command.
6046 |Option |Description
6052 <<channel-overwrite-mode-vs-discard-mode,event loss mode>> instead of
6053 the default _discard_ mode.
6055 |`--buffers-pid` (user space tracing domain only)
6058 Use the per-process <<channel-buffering-schemes,buffering scheme>>
6059 instead of the default per-user buffering scheme.
6061 |+--subbuf-size=__SIZE__+
6064 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
6065 either for each Unix user (default), or for each instrumented process.
6067 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
6069 |+--num-subbuf=__COUNT__+
6072 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
6073 for each Unix user (default), or for each instrumented process.
6075 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
6077 |+--tracefile-size=__SIZE__+
6080 Set the maximum size of each trace file that this channel writes within
6081 a stream to +__SIZE__+ bytes instead of no maximum.
6083 See <<tracefile-rotation,Trace file count and size>>.
6085 |+--tracefile-count=__COUNT__+
6088 Limit the number of trace files that this channel creates to
6089 +__COUNT__+ channels instead of no limit.
6091 See <<tracefile-rotation,Trace file count and size>>.
6093 |+--switch-timer=__PERIODUS__+
6096 Set the <<channel-switch-timer,switch timer period>>
6097 to +__PERIODUS__+{nbsp}µs.
6099 |+--read-timer=__PERIODUS__+
6102 Set the <<channel-read-timer,read timer period>>
6103 to +__PERIODUS__+{nbsp}µs.
6105 |+--output=__TYPE__+ (Linux kernel tracing domain only)
6108 Set the channel's output type to +__TYPE__+, either `mmap` or `splice`.
6112 You can only create a channel in the Linux kernel and user space
6113 <<domain,tracing domains>>: other tracing domains have their own channel
6114 created on the fly when <<enabling-disabling-events,creating event
6119 Because of a current LTTng limitation, you must create all channels
6120 _before_ you <<basic-tracing-session-control,start tracing>> in a given
6121 tracing session, that is, before the first time you run
6124 Since LTTng automatically creates a default channel when you use the
6125 man:lttng-enable-event(1) command with a specific tracing domain, you
6126 cannot, for example, create a Linux kernel event rule, start tracing,
6127 and then create a user space event rule, because no user space channel
6128 exists yet and it's too late to create one.
6130 For this reason, make sure to configure your channels properly
6131 before starting the tracers for the first time!
6134 The following examples show how you can combine the previous
6135 command-line options to create simple to more complex channels.
6137 .Create a Linux kernel channel with default attributes.
6141 $ lttng enable-channel --kernel my-channel
6145 .Create a user space channel with 4 sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
6149 $ lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
6150 --buffers-pid my-channel
6154 .Create a Linux kernel channel which rotates 8 trace files of 4{nbsp}MiB each for each stream
6158 $ lttng enable-channel --kernel --tracefile-count=8 \
6159 --tracefile-size=4194304 my-channel
6163 .Create a user space channel in overwrite (or _flight recorder_) mode.
6167 $ lttng enable-channel --userspace --overwrite my-channel
6171 You can <<enabling-disabling-events,create>> the same event rule in
6172 two different channels:
6176 $ lttng enable-event --userspace --channel=my-channel app:tp
6177 $ lttng enable-event --userspace --channel=other-channel app:tp
6180 If both channels are enabled, when a tracepoint named `app:tp` is
6181 reached, LTTng records two events, one for each channel.
6185 === Disable a channel
6187 To disable a specific channel that you <<enabling-disabling-channels,created>>
6188 previously, use the man:lttng-disable-channel(1) command.
6190 .Disable a specific Linux kernel channel.
6194 $ lttng disable-channel --kernel my-channel
6198 The state of a channel precedes the individual states of event rules
6199 attached to it: event rules which belong to a disabled channel, even if
6200 they are enabled, are also considered disabled.
6204 === Add context fields to a channel
6206 Event record fields in trace files provide important information about
6207 events that occured previously, but sometimes some external context may
6208 help you solve a problem faster. Examples of context fields are:
6210 * The **process ID**, **thread ID**, **process name**, and
6211 **process priority** of the thread in which the event occurs.
6212 * The **hostname** of the system on which the event occurs.
6213 * The current values of many possible **performance counters** using
6215 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
6217 ** Branch instructions, misses, and loads.
6219 * Any context defined at the application level (supported for the
6220 JUL and log4j <<domain,tracing domains>>).
6222 To get the full list of available context fields, see
6223 `lttng add-context --list`. Some context fields are reserved for a
6224 specific <<domain,tracing domain>> (Linux kernel or user space).
6226 You add context fields to <<channel,channels>>. All the events
6227 that a channel with added context fields records contain those fields.
6229 To add context fields to one or all the channels of a given tracing
6232 * Use the man:lttng-add-context(1) command.
6234 .Add context fields to all the channels of the current tracing session.
6236 The following command line adds the virtual process identifier and
6237 the per-thread CPU cycles count fields to all the user space channels
6238 of the current tracing session.
6242 $ lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
6246 .Add performance counter context fields by raw ID
6248 See man:lttng-add-context(1) for the exact format of the context field
6249 type, which is partly compatible with the format used in
6254 $ lttng add-context --userspace --type=perf:thread:raw:r0110:test
6255 $ lttng add-context --kernel --type=perf:cpu:raw:r0013c:x86unhalted
6259 .Add a context field to a specific channel.
6261 The following command line adds the thread identifier context field
6262 to the Linux kernel channel named `my-channel` in the current
6267 $ lttng add-context --kernel --channel=my-channel --type=tid
6271 .Add an application-specific context field to a specific channel.
6273 The following command line adds the `cur_msg_id` context field of the
6274 `retriever` context retriever for all the instrumented
6275 <<java-application,Java applications>> recording <<event,event records>>
6276 in the channel named `my-channel`:
6280 $ lttng add-context --kernel --channel=my-channel \
6281 --type='$app:retriever:cur_msg_id'
6284 IMPORTANT: Make sure to always quote the `$` character when you
6285 use man:lttng-add-context(1) from a shell.
6288 NOTE: You cannot remove context fields from a channel once you add it.
6293 === Track process IDs
6295 It's often useful to allow only specific process IDs (PIDs) to emit
6296 events. For example, you may wish to record all the system calls made by
6297 a given process (à la http://linux.die.net/man/1/strace[strace]).
6299 The man:lttng-track(1) and man:lttng-untrack(1) commands serve this
6300 purpose. Both commands operate on a whitelist of process IDs. You _add_
6301 entries to this whitelist with the man:lttng-track(1) command and remove
6302 entries with the man:lttng-untrack(1) command. Any process which has one
6303 of the PIDs in the whitelist is allowed to emit LTTng events which pass
6304 an enabled <<event,event rule>>.
6306 NOTE: The PID tracker tracks the _numeric process IDs_. Should a
6307 process with a given tracked ID exit and another process be given this
6308 ID, then the latter would also be allowed to emit events.
6310 .Track and untrack process IDs.
6312 For the sake of the following example, assume the target system has 16
6316 <<creating-destroying-tracing-sessions,create a tracing session>>,
6317 the whitelist contains all the possible PIDs:
6320 .All PIDs are tracked.
6321 image::track-all.png[]
6323 When the whitelist is full and you use the man:lttng-track(1) command to
6324 specify some PIDs to track, LTTng first clears the whitelist, then it
6325 tracks the specific PIDs. After:
6329 $ lttng track --pid=3,4,7,10,13
6335 .PIDs 3, 4, 7, 10, and 13 are tracked.
6336 image::track-3-4-7-10-13.png[]
6338 You can add more PIDs to the whitelist afterwards:
6342 $ lttng track --pid=1,15,16
6348 .PIDs 1, 15, and 16 are added to the whitelist.
6349 image::track-1-3-4-7-10-13-15-16.png[]
6351 The man:lttng-untrack(1) command removes entries from the PID tracker's
6352 whitelist. Given the previous example, the following command:
6356 $ lttng untrack --pid=3,7,10,13
6359 leads to this whitelist:
6362 .PIDs 3, 7, 10, and 13 are removed from the whitelist.
6363 image::track-1-4-15-16.png[]
6365 LTTng can track all possible PIDs again using the opt:track(1):--all
6370 $ lttng track --pid --all
6373 The result is, again:
6376 .All PIDs are tracked.
6377 image::track-all.png[]
6380 .Track only specific PIDs
6382 A very typical use case with PID tracking is to start with an empty
6383 whitelist, then <<basic-tracing-session-control,start the tracers>>, and
6384 then add PIDs manually while tracers are active. You can accomplish this
6385 by using the opt:lttng-untrack(1):--all option of the
6386 man:lttng-untrack(1) command to clear the whitelist after you
6387 <<creating-destroying-tracing-sessions,create a tracing session>>:
6391 $ lttng untrack --pid --all
6397 .No PIDs are tracked.
6398 image::untrack-all.png[]
6400 If you trace with this whitelist configuration, the tracer records no
6401 events for this <<domain,tracing domain>> because no processes are
6402 tracked. You can use the man:lttng-track(1) command as usual to track
6403 specific PIDs, for example:
6407 $ lttng track --pid=6,11
6413 .PIDs 6 and 11 are tracked.
6414 image::track-6-11.png[]
6419 [[saving-loading-tracing-session]]
6420 === Save and load tracing session configurations
6422 Configuring a <<tracing-session,tracing session>> can be long. Some of
6423 the tasks involved are:
6425 * <<enabling-disabling-channels,Create channels>> with
6426 specific attributes.
6427 * <<adding-context,Add context fields>> to specific channels.
6428 * <<enabling-disabling-events,Create event rules>> with specific log
6429 level and filter conditions.
6431 If you use LTTng to solve real world problems, chances are you have to
6432 record events using the same tracing session setup over and over,
6433 modifying a few variables each time in your instrumented program
6434 or environment. To avoid constant tracing session reconfiguration,
6435 the man:lttng(1) command-line tool can save and load tracing session
6436 configurations to/from XML files.
6438 To save a given tracing session configuration:
6440 * Use the man:lttng-save(1) command:
6445 $ lttng save my-session
6449 Replace `my-session` with the name of the tracing session to save.
6451 LTTng saves tracing session configurations to
6452 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
6453 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
6454 the opt:lttng-save(1):--output-path option to change this destination
6457 LTTng saves all configuration parameters, for example:
6459 * The tracing session name.
6460 * The trace data output path.
6461 * The channels with their state and all their attributes.
6462 * The context fields you added to channels.
6463 * The event rules with their state, log level and filter conditions.
6465 To load a tracing session:
6467 * Use the man:lttng-load(1) command:
6472 $ lttng load my-session
6476 Replace `my-session` with the name of the tracing session to load.
6478 When LTTng loads a configuration, it restores your saved tracing session
6479 as if you just configured it manually.
6481 See man:lttng(1) for the complete list of command-line options. You
6482 can also save and load all many sessions at a time, and decide in which
6483 directory to output the XML files.
6486 [[sending-trace-data-over-the-network]]
6487 === Send trace data over the network
6489 LTTng can send the recorded trace data to a remote system over the
6490 network instead of writing it to the local file system.
6492 To send the trace data over the network:
6494 . On the _remote_ system (which can also be the target system),
6495 start an LTTng <<lttng-relayd,relay daemon>> (man:lttng-relayd(8)):
6504 . On the _target_ system, create a tracing session configured to
6505 send trace data over the network:
6510 $ lttng create my-session --set-url=net://remote-system
6514 Replace `remote-system` by the host name or IP address of the
6515 remote system. See man:lttng-create(1) for the exact URL format.
6517 . On the target system, use the man:lttng(1) command-line tool as usual.
6518 When tracing is active, the target's consumer daemon sends sub-buffers
6519 to the relay daemon running on the remote system instead of flushing
6520 them to the local file system. The relay daemon writes the received
6521 packets to the local file system.
6523 The relay daemon writes trace files to
6524 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
6525 +__hostname__+ is the host name of the target system and +__session__+
6526 is the tracing session name. Note that the env:LTTNG_HOME environment
6527 variable defaults to `$HOME` if not set. Use the
6528 opt:lttng-relayd(8):--output option of man:lttng-relayd(8) to write
6529 trace files to another base directory.
6534 === View events as LTTng emits them (noch:{LTTng} live)
6536 LTTng live is a network protocol implemented by the <<lttng-relayd,relay
6537 daemon>> (man:lttng-relayd(8)) to allow compatible trace viewers to
6538 display events as LTTng emits them on the target system while tracing is
6541 The relay daemon creates a _tee_: it forwards the trace data to both
6542 the local file system and to connected live viewers:
6545 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
6550 . On the _target system_, create a <<tracing-session,tracing session>>
6556 $ lttng create my-session --live
6560 This spawns a local relay daemon.
6562 . Start the live viewer and configure it to connect to the relay
6563 daemon. For example, with http://diamon.org/babeltrace[Babeltrace]:
6568 $ babeltrace --input-format=lttng-live \
6569 net://localhost/host/hostname/my-session
6576 * `hostname` with the host name of the target system.
6577 * `my-session` with the name of the tracing session to view.
6580 . Configure the tracing session as usual with the man:lttng(1)
6581 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6583 You can list the available live tracing sessions with Babeltrace:
6587 $ babeltrace --input-format=lttng-live net://localhost
6590 You can start the relay daemon on another system. In this case, you need
6591 to specify the relay daemon's URL when you create the tracing session
6592 with the opt:lttng-create(1):--set-url option. You also need to replace
6593 `localhost` in the procedure above with the host name of the system on
6594 which the relay daemon is running.
6596 See man:lttng-create(1) and man:lttng-relayd(8) for the complete list of
6597 command-line options.
6601 [[taking-a-snapshot]]
6602 === Take a snapshot of the current sub-buffers of a tracing session
6604 The normal behavior of LTTng is to append full sub-buffers to growing
6605 trace data files. This is ideal to keep a full history of the events
6606 that occurred on the target system, but it can
6607 represent too much data in some situations. For example, you may wish
6608 to trace your application continuously until some critical situation
6609 happens, in which case you only need the latest few recorded
6610 events to perform the desired analysis, not multi-gigabyte trace files.
6612 With the man:lttng-snapshot(1) command, you can take a snapshot of the
6613 current sub-buffers of a given <<tracing-session,tracing session>>.
6614 LTTng can write the snapshot to the local file system or send it over
6619 . Create a tracing session in _snapshot mode_:
6624 $ lttng create my-session --snapshot
6628 The <<channel-overwrite-mode-vs-discard-mode,event loss mode>> of
6629 <<channel,channels>> created in this mode is automatically set to
6630 _overwrite_ (flight recorder mode).
6632 . Configure the tracing session as usual with the man:lttng(1)
6633 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6635 . **Optional**: When you need to take a snapshot,
6636 <<basic-tracing-session-control,stop tracing>>.
6638 You can take a snapshot when the tracers are active, but if you stop
6639 them first, you are sure that the data in the sub-buffers does not
6640 change before you actually take the snapshot.
6647 $ lttng snapshot record --name=my-first-snapshot
6651 LTTng writes the current sub-buffers of all the current tracing
6652 session's channels to trace files on the local file system. Those trace
6653 files have `my-first-snapshot` in their name.
6655 There is no difference between the format of a normal trace file and the
6656 format of a snapshot: viewers of LTTng traces also support LTTng
6659 By default, LTTng writes snapshot files to the path shown by
6660 `lttng snapshot list-output`. You can change this path or decide to send
6661 snapshots over the network using either:
6663 . An output path or URL that you specify when you create the
6665 . An snapshot output path or URL that you add using
6666 `lttng snapshot add-output`
6667 . An output path or URL that you provide directly to the
6668 `lttng snapshot record` command.
6670 Method 3 overrides method 2, which overrides method 1. When you
6671 specify a URL, a relay daemon must listen on a remote system (see
6672 <<sending-trace-data-over-the-network,Send trace data over the network>>).
6677 === Use the machine interface
6679 With any command of the man:lttng(1) command-line tool, you can set the
6680 opt:lttng(1):--mi option to `xml` (before the command name) to get an
6681 XML machine interface output, for example:
6685 $ lttng --mi=xml enable-event --kernel --syscall open
6688 A schema definition (XSD) is
6689 https://github.com/lttng/lttng-tools/blob/stable-2.9/src/common/mi-lttng-3.0.xsd[available]
6690 to ease the integration with external tools as much as possible.
6694 [[metadata-regenerate]]
6695 === Regenerate the metadata of an LTTng trace
6697 An LTTng trace, which is a http://diamon.org/ctf[CTF] trace, has both
6698 data stream files and a metadata file. This metadata file contains,
6699 amongst other things, information about the offset of the clock sources
6700 used to timestamp <<event,event records>> when tracing.
6702 If, once a <<tracing-session,tracing session>> is
6703 <<basic-tracing-session-control,started>>, a major
6704 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
6705 happens, the trace's clock offset also needs to be updated. You
6706 can use the `metadata` item of the man:lttng-regenerate(1) command
6709 The main use case of this command is to allow a system to boot with
6710 an incorrect wall time and trace it with LTTng before its wall time
6711 is corrected. Once the system is known to be in a state where its
6712 wall time is correct, it can run `lttng regenerate metadata`.
6714 To regenerate the metadata of an LTTng trace:
6716 * Use the `metadata` item of the man:lttng-regenerate(1) command:
6721 $ lttng regenerate metadata
6727 `lttng regenerate metadata` has the following limitations:
6729 * Tracing session <<creating-destroying-tracing-sessions,created>>
6731 * User space <<channel,channels>>, if any, are using
6732 <<channel-buffering-schemes,per-user buffering>>.
6737 [[regenerate-statedump]]
6738 === Regenerate the state dump of a tracing session
6740 The LTTng kernel and user space tracers generate state dump
6741 <<event,event records>> when the application starts or when you
6742 <<basic-tracing-session-control,start a tracing session>>. An analysis
6743 can use the state dump event records to set an initial state before it
6744 builds the rest of the state from the following event records.
6745 http://tracecompass.org/[Trace Compass] is a notable example of an
6746 application which uses the state dump of an LTTng trace.
6748 When you <<taking-a-snapshot,take a snapshot>>, it's possible that the
6749 state dump event records are not included in the snapshot because they
6750 were recorded to a sub-buffer that has been consumed or overwritten
6753 You can use the `lttng regenerate statedump` command to emit the state
6754 dump event records again.
6756 To regenerate the state dump of the current tracing session, provided
6757 create it in snapshot mode, before you take a snapshot:
6759 . Use the `statedump` item of the man:lttng-regenerate(1) command:
6764 $ lttng regenerate statedump
6768 . <<basic-tracing-session-control,Stop the tracing session>>:
6777 . <<taking-a-snapshot,Take a snapshot>>:
6782 $ lttng snapshot record --name=my-snapshot
6786 Depending on the event throughput, you should run steps 1 and 2
6787 as closely as possible.
6789 NOTE: To record the state dump events, you need to
6790 <<enabling-disabling-events,create event rules>> which enable them.
6791 LTTng-UST state dump tracepoints start with `lttng_ust_statedump:`.
6792 LTTng-modules state dump tracepoints start with `lttng_statedump_`.
6796 [[persistent-memory-file-systems]]
6797 === Record trace data on persistent memory file systems
6799 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
6800 (NVRAM) is random-access memory that retains its information when power
6801 is turned off (non-volatile). Systems with such memory can store data
6802 structures in RAM and retrieve them after a reboot, without flushing
6803 to typical _storage_.
6805 Linux supports NVRAM file systems thanks to either
6806 http://pramfs.sourceforge.net/[PRAMFS] or
6807 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
6808 (requires Linux 4.1+).
6810 This section does not describe how to operate such file systems;
6811 we assume that you have a working persistent memory file system.
6813 When you create a <<tracing-session,tracing session>>, you can specify
6814 the path of the shared memory holding the sub-buffers. If you specify a
6815 location on an NVRAM file system, then you can retrieve the latest
6816 recorded trace data when the system reboots after a crash.
6818 To record trace data on a persistent memory file system and retrieve the
6819 trace data after a system crash:
6821 . Create a tracing session with a sub-buffer shared memory path located
6822 on an NVRAM file system:
6827 $ lttng create my-session --shm-path=/path/to/shm
6831 . Configure the tracing session as usual with the man:lttng(1)
6832 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6834 . After a system crash, use the man:lttng-crash(1) command-line tool to
6835 view the trace data recorded on the NVRAM file system:
6840 $ lttng-crash /path/to/shm
6844 The binary layout of the ring buffer files is not exactly the same as
6845 the trace files layout. This is why you need to use man:lttng-crash(1)
6846 instead of your preferred trace viewer directly.
6848 To convert the ring buffer files to LTTng trace files:
6850 * Use the opt:lttng-crash(1):--extract option of man:lttng-crash(1):
6855 $ lttng-crash --extract=/path/to/trace /path/to/shm
6863 [[lttng-modules-ref]]
6864 === noch:{LTTng-modules}
6868 [[lttng-tracepoint-enum]]
6869 ==== `LTTNG_TRACEPOINT_ENUM()` usage
6871 Use the `LTTNG_TRACEPOINT_ENUM()` macro to define an enumeration:
6875 LTTNG_TRACEPOINT_ENUM(name, TP_ENUM_VALUES(entries))
6880 * `name` with the name of the enumeration (C identifier, unique
6881 amongst all the defined enumerations).
6882 * `entries` with a list of enumeration entries.
6884 The available enumeration entry macros are:
6886 +ctf_enum_value(__name__, __value__)+::
6887 Entry named +__name__+ mapped to the integral value +__value__+.
6889 +ctf_enum_range(__name__, __begin__, __end__)+::
6890 Entry named +__name__+ mapped to the range of integral values between
6891 +__begin__+ (included) and +__end__+ (included).
6893 +ctf_enum_auto(__name__)+::
6894 Entry named +__name__+ mapped to the integral value following the
6895 last mapping's value.
6897 The last value of a `ctf_enum_value()` entry is its +__value__+
6900 The last value of a `ctf_enum_range()` entry is its +__end__+ parameter.
6902 If `ctf_enum_auto()` is the first entry in the list, its integral
6905 Use the `ctf_enum()` <<lttng-modules-tp-fields,field definition macro>>
6906 to use a defined enumeration as a tracepoint field.
6908 .Define an enumeration with `LTTNG_TRACEPOINT_ENUM()`.
6912 LTTNG_TRACEPOINT_ENUM(
6915 ctf_enum_auto("AUTO: EXPECT 0")
6916 ctf_enum_value("VALUE: 23", 23)
6917 ctf_enum_value("VALUE: 27", 27)
6918 ctf_enum_auto("AUTO: EXPECT 28")
6919 ctf_enum_range("RANGE: 101 TO 303", 101, 303)
6920 ctf_enum_auto("AUTO: EXPECT 304")
6928 [[lttng-modules-tp-fields]]
6929 ==== Tracepoint fields macros (for `TP_FIELDS()`)
6931 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
6932 tracepoint fields, which must be listed within `TP_FIELDS()` in
6933 `LTTNG_TRACEPOINT_EVENT()`, are:
6935 [role="func-desc growable",cols="asciidoc,asciidoc"]
6936 .Available macros to define LTTng-modules tracepoint fields
6938 |Macro |Description and parameters
6941 +ctf_integer(__t__, __n__, __e__)+
6943 +ctf_integer_nowrite(__t__, __n__, __e__)+
6945 +ctf_user_integer(__t__, __n__, __e__)+
6947 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
6949 Standard integer, displayed in base 10.
6952 Integer C type (`int`, `long`, `size_t`, ...).
6958 Argument expression.
6961 +ctf_integer_hex(__t__, __n__, __e__)+
6963 +ctf_user_integer_hex(__t__, __n__, __e__)+
6965 Standard integer, displayed in base 16.
6974 Argument expression.
6976 |+ctf_integer_oct(__t__, __n__, __e__)+
6978 Standard integer, displayed in base 8.
6987 Argument expression.
6990 +ctf_integer_network(__t__, __n__, __e__)+
6992 +ctf_user_integer_network(__t__, __n__, __e__)+
6994 Integer in network byte order (big-endian), displayed in base 10.
7003 Argument expression.
7006 +ctf_integer_network_hex(__t__, __n__, __e__)+
7008 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
7010 Integer in network byte order, displayed in base 16.
7019 Argument expression.
7022 +ctf_enum(__N__, __t__, __n__, __e__)+
7024 +ctf_enum_nowrite(__N__, __t__, __n__, __e__)+
7026 +ctf_user_enum(__N__, __t__, __n__, __e__)+
7028 +ctf_user_enum_nowrite(__N__, __t__, __n__, __e__)+
7033 Name of a <<lttng-tracepoint-enum,previously defined enumeration>>.
7036 Integer C type (`int`, `long`, `size_t`, ...).
7042 Argument expression.
7045 +ctf_string(__n__, __e__)+
7047 +ctf_string_nowrite(__n__, __e__)+
7049 +ctf_user_string(__n__, __e__)+
7051 +ctf_user_string_nowrite(__n__, __e__)+
7053 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
7059 Argument expression.
7062 +ctf_array(__t__, __n__, __e__, __s__)+
7064 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
7066 +ctf_user_array(__t__, __n__, __e__, __s__)+
7068 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
7070 Statically-sized array of integers.
7073 Array element C type.
7079 Argument expression.
7085 +ctf_array_bitfield(__t__, __n__, __e__, __s__)+
7087 +ctf_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7089 +ctf_user_array_bitfield(__t__, __n__, __e__, __s__)+
7091 +ctf_user_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7093 Statically-sized array of bits.
7095 The type of +__e__+ must be an integer type. +__s__+ is the number
7096 of elements of such type in +__e__+, not the number of bits.
7099 Array element C type.
7105 Argument expression.
7111 +ctf_array_text(__t__, __n__, __e__, __s__)+
7113 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
7115 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
7117 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
7119 Statically-sized array, printed as text.
7121 The string does not need to be null-terminated.
7124 Array element C type (always `char`).
7130 Argument expression.
7136 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
7138 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7140 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
7142 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7144 Dynamically-sized array of integers.
7146 The type of +__E__+ must be unsigned.
7149 Array element C type.
7155 Argument expression.
7158 Length expression C type.
7164 +ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7166 +ctf_user_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7168 Dynamically-sized array of integers, displayed in base 16.
7170 The type of +__E__+ must be unsigned.
7173 Array element C type.
7179 Argument expression.
7182 Length expression C type.
7187 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
7189 Dynamically-sized array of integers in network byte order (big-endian),
7190 displayed in base 10.
7192 The type of +__E__+ must be unsigned.
7195 Array element C type.
7201 Argument expression.
7204 Length expression C type.
7210 +ctf_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7212 +ctf_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7214 +ctf_user_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7216 +ctf_user_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7218 Dynamically-sized array of bits.
7220 The type of +__e__+ must be an integer type. +__s__+ is the number
7221 of elements of such type in +__e__+, not the number of bits.
7223 The type of +__E__+ must be unsigned.
7226 Array element C type.
7232 Argument expression.
7235 Length expression C type.
7241 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7243 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7245 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7247 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7249 Dynamically-sized array, displayed as text.
7251 The string does not need to be null-terminated.
7253 The type of +__E__+ must be unsigned.
7255 The behaviour is undefined if +__e__+ is `NULL`.
7258 Sequence element C type (always `char`).
7264 Argument expression.
7267 Length expression C type.
7273 Use the `_user` versions when the argument expression, `e`, is
7274 a user space address. In the cases of `ctf_user_integer*()` and
7275 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
7278 The `_nowrite` versions omit themselves from the session trace, but are
7279 otherwise identical. This means the `_nowrite` fields won't be written
7280 in the recorded trace. Their primary purpose is to make some
7281 of the event context available to the
7282 <<enabling-disabling-events,event filters>> without having to
7283 commit the data to sub-buffers.
7289 Terms related to LTTng and to tracing in general:
7292 The http://diamon.org/babeltrace[Babeltrace] project, which includes
7293 the cmd:babeltrace command, some libraries, and Python bindings.
7295 <<channel-buffering-schemes,buffering scheme>>::
7296 A layout of sub-buffers applied to a given channel.
7298 <<channel,channel>>::
7299 An entity which is responsible for a set of ring buffers.
7301 <<event,Event rules>> are always attached to a specific channel.
7304 A reference of time for a tracer.
7306 <<lttng-consumerd,consumer daemon>>::
7307 A process which is responsible for consuming the full sub-buffers
7308 and write them to a file system or send them over the network.
7310 <<channel-overwrite-mode-vs-discard-mode,discard mode>>:: The event loss
7311 mode in which the tracer _discards_ new event records when there's no
7312 sub-buffer space left to store them.
7315 The consequence of the execution of an instrumentation
7316 point, like a tracepoint that you manually place in some source code,
7317 or a Linux kernel KProbe.
7319 An event is said to _occur_ at a specific time. Different actions can
7320 be taken upon the occurance of an event, like record the event's payload
7323 <<channel-overwrite-mode-vs-discard-mode,event loss mode>>::
7324 The mechanism by which event records of a given channel are lost
7325 (not recorded) when there is no sub-buffer space left to store them.
7327 [[def-event-name]]event name::
7328 The name of an event, which is also the name of the event record.
7329 This is also called the _instrumentation point name_.
7332 A record, in a trace, of the payload of an event which occured.
7334 <<event,event rule>>::
7335 Set of conditions which must be satisfied for one or more occuring
7336 events to be recorded.
7338 `java.util.logging`::
7340 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities].
7342 <<instrumenting,instrumentation>>::
7343 The use of LTTng probes to make a piece of software traceable.
7345 instrumentation point::
7346 A point in the execution path of a piece of software that, when
7347 reached by this execution, can emit an event.
7349 instrumentation point name::
7350 See _<<def-event-name,event name>>_.
7353 A http://logging.apache.org/log4j/1.2/[logging library] for Java
7354 developed by the Apache Software Foundation.
7357 Level of severity of a log statement or user space
7358 instrumentation point.
7361 The _Linux Trace Toolkit: next generation_ project.
7363 <<lttng-cli,cmd:lttng>>::
7364 A command-line tool provided by the LTTng-tools project which you
7365 can use to send and receive control messages to and from a
7369 The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
7370 which is a set of analyzing programs that are used to obtain a
7371 higher level view of an LTTng trace.
7373 cmd:lttng-consumerd::
7374 The name of the consumer daemon program.
7377 A utility provided by the LTTng-tools project which can convert
7378 ring buffer files (usually
7379 <<persistent-memory-file-systems,saved on a persistent memory file system>>)
7382 LTTng Documentation::
7385 <<lttng-live,LTTng live>>::
7386 A communication protocol between the relay daemon and live viewers
7387 which makes it possible to see events "live", as they are received by
7390 <<lttng-modules,LTTng-modules>>::
7391 The https://github.com/lttng/lttng-modules[LTTng-modules] project,
7392 which contains the Linux kernel modules to make the Linux kernel
7393 instrumentation points available for LTTng tracing.
7396 The name of the relay daemon program.
7398 cmd:lttng-sessiond::
7399 The name of the session daemon program.
7402 The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
7403 contains the various programs and libraries used to
7404 <<controlling-tracing,control tracing>>.
7406 <<lttng-ust,LTTng-UST>>::
7407 The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
7408 contains libraries to instrument user applications.
7410 <<lttng-ust-agents,LTTng-UST Java agent>>::
7411 A Java package provided by the LTTng-UST project to allow the
7412 LTTng instrumentation of `java.util.logging` and Apache log4j 1.2
7415 <<lttng-ust-agents,LTTng-UST Python agent>>::
7416 A Python package provided by the LTTng-UST project to allow the
7417 LTTng instrumentation of Python logging statements.
7419 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
7420 The event loss mode in which new event records overwrite older
7421 event records when there's no sub-buffer space left to store them.
7423 <<channel-buffering-schemes,per-process buffering>>::
7424 A buffering scheme in which each instrumented process has its own
7425 sub-buffers for a given user space channel.
7427 <<channel-buffering-schemes,per-user buffering>>::
7428 A buffering scheme in which all the processes of a Unix user share the
7429 same sub-buffer for a given user space channel.
7431 <<lttng-relayd,relay daemon>>::
7432 A process which is responsible for receiving the trace data sent by
7433 a distant consumer daemon.
7436 A set of sub-buffers.
7438 <<lttng-sessiond,session daemon>>::
7439 A process which receives control commands from you and orchestrates
7440 the tracers and various LTTng daemons.
7442 <<taking-a-snapshot,snapshot>>::
7443 A copy of the current data of all the sub-buffers of a given tracing
7444 session, saved as trace files.
7447 One part of an LTTng ring buffer which contains event records.
7450 The time information attached to an event when it is emitted.
7453 A set of files which are the concatenations of one or more
7454 flushed sub-buffers.
7457 The action of recording the events emitted by an application
7458 or by a system, or to initiate such recording by controlling
7462 The http://tracecompass.org[Trace Compass] project and application.
7465 An instrumentation point using the tracepoint mechanism of the Linux
7466 kernel or of LTTng-UST.
7468 tracepoint definition::
7469 The definition of a single tracepoint.
7472 The name of a tracepoint.
7474 tracepoint provider::
7475 A set of functions providing tracepoints to an instrumented user
7478 Not to be confused with a _tracepoint provider package_: many tracepoint
7479 providers can exist within a tracepoint provider package.
7481 tracepoint provider package::
7482 One or more tracepoint providers compiled as an object file or as
7486 A software which records emitted events.
7488 <<domain,tracing domain>>::
7489 A namespace for event sources.
7491 <<tracing-group,tracing group>>::
7492 The Unix group in which a Unix user can be to be allowed to trace the
7495 <<tracing-session,tracing session>>::
7496 A stateful dialogue between you and a <<lttng-sessiond,session
7500 An application running in user space, as opposed to a Linux kernel
7501 module, for example.