1 The LTTng Documentation
2 =======================
3 Philippe Proulx <pproulx@efficios.com>
7 include::../common/copyright.txt[]
10 include::../common/warning-not-maintained.txt[]
13 include::../common/welcome.txt[]
16 include::../common/audience.txt[]
20 === What's in this documentation?
22 The LTTng Documentation is divided into the following sections:
24 * **<<nuts-and-bolts,Nuts and bolts>>** explains the
25 rudiments of software tracing and the rationale behind the
28 You can skip this section if you’re familiar with software tracing and
29 with the LTTng project.
31 * **<<installing-lttng,Installation>>** describes the steps to
32 install the LTTng packages on common Linux distributions and from
35 You can skip this section if you already properly installed LTTng on
38 * **<<getting-started,Quick start>>** is a concise guide to
39 getting started quickly with LTTng kernel and user space tracing.
41 We recommend this section if you're new to LTTng or to software tracing
44 You can skip this section if you're not new to LTTng.
46 * **<<core-concepts,Core concepts>>** explains the concepts at
49 It's a good idea to become familiar with the core concepts
50 before attempting to use the toolkit.
52 * **<<plumbing,Components of LTTng>>** describes the various components
53 of the LTTng machinery, like the daemons, the libraries, and the
54 command-line interface.
55 * **<<instrumenting,Instrumentation>>** shows different ways to
56 instrument user applications and the Linux kernel.
58 Instrumenting source code is essential to provide a meaningful
61 You can skip this section if you do not have a programming background.
63 * **<<controlling-tracing,Tracing control>>** is divided into topics
64 which demonstrate how to use the vast array of features that
65 LTTng{nbsp}{revision} offers.
66 * **<<reference,Reference>>** contains reference tables.
67 * **<<glossary,Glossary>>** is a specialized dictionary of terms related
68 to LTTng or to the field of software tracing.
71 include::../common/convention.txt[]
74 include::../common/acknowledgements.txt[]
78 == What's new in LTTng {revision}?
80 LTTng{nbsp}{revision} bears the name _Isseki Nicho_. The result of a
81 collaboration between http://www.dieuduciel.com/[Dieu du Ciel!] and
82 Nagano-based Shiga Kogen,
83 https://www.beeradvocate.com/beer/profile/1141/53111/[_**Isseki
84 Nicho**_] is a strong Imperial Dark Saison offering a rich roasted malt
85 flavor combined with a complex fruity finish typical of Saison yeasts.
87 New features and changes in LTTng{nbsp}{revision}:
89 * **Tracing control**:
90 ** You can attach <<java-application-context,Java application-specific
91 context fields>> to a <<channel,channel>> with the
92 man:lttng-add-context(1) command:
97 $ lttng add-context --jul --type='$app.retriever:cur_msg_id'
101 Here, `$app` is the prefix of all application-specific context fields,
102 `retriever` names a _context information retriever_ defined at the
103 application level, and `cur_msg_id` names a context field read from this
106 Both the `java.util.logging` and Apache log4j <<domain,tracing domains>>
109 ** You can use Java application-specific <<adding-context,context>>
110 fields in the <<enabling-disabling-events,filter expression>> of an
111 <<event,event rule>>:
116 $ lttng enable-event --log4j my_logger \
117 --filter='$app.retriever:cur_msg_id == 23'
121 ** New `lttng status` command which is the equivalent of +lttng list
122 __CUR__+, where +__CUR__+ is the name of the current
123 <<tracing-session,tracing session>>.
125 See man:lttng-status(1).
127 ** New `lttng metadata regenerate` command to
128 <<metadata-regenerate,regenerate the metadata file of an LTTng
129 trace>> at any moment. This command is meant to be used to resample
130 the wall time following a major
131 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
132 so that a system which boots with an incorrect wall time can be
133 traced before its wall time is NTP-corrected.
135 See man:lttng-metadata(1).
137 ** New command-line interface warnings when <<event,event records>> or
138 whole sub-buffers are
139 <<channel-overwrite-mode-vs-discard-mode,lost>>. The warning messages
140 are printed when a <<tracing-session,tracing session>> is
141 <<basic-tracing-session-control,stopped>> (man:lttng-stop(1)
144 * **User space tracing**:
145 ** Shared object base address dump in order to map <<event,event
146 records>> to original source location (file and line number).
148 If you attach the `ip` and `vpid` <<adding-context,context fields>> to a
149 user space <<channel,channel>> and if you use the
150 <<liblttng-ust-dl,path:{liblttng-ust-dl.so} helper>>, you can retrieve
151 the source location where a given event record was generated.
153 The http://diamon.org/babeltrace/[Babeltrace] trace viewer supports this
154 state dump and those context fields since version 1.4 to print the
155 source location of a given event record. http://tracecompass.org/[Trace
156 Compass] also supports this since version 2.0.
158 ** A <<java-application,Java application>> which uses
159 `java.util.logging` now adds an LTTng-UST log handler to the desired
162 The previous workflow was to initialize the LTTng-UST Java agent
163 by calling `LTTngAgent.getLTTngAgent()`. This had the effect of adding
164 an LTTng-UST log handler to the root loggers.
166 ** A <<java-application,Java application>> which uses Apache log4j now
167 adds an LTTng-UST log appender to the desired log4j loggers.
169 The previous workflow was to initialize the LTTng-UST Java agent
170 by calling `LTTngAgent.getLTTngAgent()`. This had the effect of adding
171 an LTTng-UST appender to the root loggers.
173 ** Any <<java-application,Java application>> can provide
174 <<java-application-context,dynamic context fields>> while running
175 thanks to a new API provided by the <<lttng-ust-agents,LTTng-UST Java
176 agent>>. You can require LTTng to record specific context fields in
177 event records, and you can use them in the filter expression of
178 <<event,event rules>>.
180 * **Linux kernel tracing**:
181 ** The LTTng kernel modules can now be built into a Linux kernel image,
182 that is, not as loadable modules.
185 https://github.com/lttng/lttng-modules/blob/stable-{revision}/README.md#kernel-built-in-support[`README.md`]
188 ** New instrumentation:
189 *** ARM64 architecture support.
191 *** x86 `irq_vectors`.
192 ** New <<adding-context,context fields>>:
195 *** `need_reschedule`
196 *** `migratable` (specific to RT-Preempt)
197 ** Clock source plugin support for advanced cases where a custom source
198 of time is needed to timestamp LTTng event records.
200 See https://github.com/lttng/lttng-modules/blob/stable-{revision}/lttng-clock.h[`lttng-clock.h`]
201 for an overview of the small API.
204 ** The link:/man[man pages] of the man:lttng(1) command-line tool are
205 split into one man page per command (à la Git), for example:
210 $ man lttng-enable-event
214 You can also use the `--help` option of any man:lttng(1) command to
217 The content and formatting of all the LTTng man pages has improved
224 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
225 generation_ is a modern toolkit for tracing Linux systems and
226 applications. So your first question might be:
233 As the history of software engineering progressed and led to what
234 we now take for granted--complex, numerous and
235 interdependent software applications running in parallel on
236 sophisticated operating systems like Linux--the authors of such
237 components, software developers, began feeling a natural
238 urge to have tools that would ensure the robustness and good performance
239 of their masterpieces.
241 One major achievement in this field is, inarguably, the
242 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
243 an essential tool for developers to find and fix bugs. But even the best
244 debugger won't help make your software run faster, and nowadays, faster
245 software means either more work done by the same hardware, or cheaper
246 hardware for the same work.
248 A _profiler_ is often the tool of choice to identify performance
249 bottlenecks. Profiling is suitable to identify _where_ performance is
250 lost in a given software. The profiler outputs a profile, a statistical
251 summary of observed events, which you may use to discover which
252 functions took the most time to execute. However, a profiler won't
253 report _why_ some identified functions are the bottleneck. Bottlenecks
254 might only occur when specific conditions are met, conditions that are
255 sometimes impossible to capture by a statistical profiler, or impossible
256 to reproduce with an application altered by the overhead of an
257 event-based profiler. For a thorough investigation of software
258 performance issues, a history of execution is essential, with the
259 recorded values of variables and context fields you choose, and
260 with as little influence as possible on the instrumented software. This
261 is where tracing comes in handy.
263 _Tracing_ is a technique used to understand what goes on in a running
264 software system. The software used for tracing is called a _tracer_,
265 which is conceptually similar to a tape recorder. When recording,
266 specific instrumentation points placed in the software source code
267 generate events that are saved on a giant tape: a _trace_ file. You
268 can trace user applications and the operating system at the same time,
269 opening the possibility of resolving a wide range of problems that would
270 otherwise be extremely challenging.
272 Tracing is often compared to _logging_. However, tracers and loggers are
273 two different tools, serving two different purposes. Tracers are
274 designed to record much lower-level events that occur much more
275 frequently than log messages, often in the range of thousands per
276 second, with very little execution overhead. Logging is more appropriate
277 for a very high-level analysis of less frequent events: user accesses,
278 exceptional conditions (errors and warnings, for example), database
279 transactions, instant messaging communications, and such. Simply put,
280 logging is one of the many use cases that can be satisfied with tracing.
282 The list of recorded events inside a trace file can be read manually
283 like a log file for the maximum level of detail, but it is generally
284 much more interesting to perform application-specific analyses to
285 produce reduced statistics and graphs that are useful to resolve a
286 given problem. Trace viewers and analyzers are specialized tools
289 In the end, this is what LTTng is: a powerful, open source set of
290 tools to trace the Linux kernel and user applications at the same time.
291 LTTng is composed of several components actively maintained and
292 developed by its link:/community/#where[community].
295 [[lttng-alternatives]]
296 === Alternatives to noch:{LTTng}
298 Excluding proprietary solutions, a few competing software tracers
301 * https://github.com/dtrace4linux/linux[dtrace4linux] is a port of
302 Sun Microsystems's DTrace to Linux. The cmd:dtrace tool interprets
303 user scripts and is responsible for loading code into the
304 Linux kernel for further execution and collecting the outputted data.
305 * https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF] is a
306 subsystem in the Linux kernel in which a virtual machine can execute
307 programs passed from the user space to the kernel. You can attach
308 such programs to tracepoints and KProbes thanks to a system call, and
309 they can output data to the user space when executed thanks to
310 different mechanisms (pipe, VM register values, and eBPF maps, to name
312 * https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]
313 is the de facto function tracer of the Linux kernel. Its user
314 interface is a set of special files in sysfs.
315 * https://perf.wiki.kernel.org/[perf] is
316 a performance analyzing tool for Linux which supports hardware
317 performance counters, tracepoints, as well as other counters and
318 types of probes. perf's controlling utility is the cmd:perf command
320 * http://linux.die.net/man/1/strace[strace]
321 is a command-line utility which records system calls made by a
322 user process, as well as signal deliveries and changes of process
323 state. strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace]
324 to fulfill its function.
325 * http://www.sysdig.org/[sysdig], like SystemTap, uses scripts to
326 analyze Linux kernel events. You write scripts, or _chisels_ in
327 sysdig's jargon, in Lua and sysdig executes them while the system is
328 being traced or afterwards. sysdig's interface is the cmd:sysdig
329 command-line tool as well as the curses-based cmd:csysdig tool.
330 * https://sourceware.org/systemtap/[SystemTap] is a Linux kernel and
331 user space tracer which uses custom user scripts to produce plain text
332 traces. SystemTap converts the scripts to the C language, and then
333 compiles them as Linux kernel modules which are loaded to produce
334 trace data. SystemTap's primary user interface is the cmd:stap
337 The main distinctive features of LTTng is that it produces correlated
338 kernel and user space traces, as well as doing so with the lowest
339 overhead amongst other solutions. It produces trace files in the
340 http://diamon.org/ctf[CTF] format, a file format optimized
341 for the production and analyses of multi-gigabyte data.
343 LTTng is the result of more than 10 years of active open source
344 development by a community of passionate developers.
345 LTTng{nbsp}{revision} is currently available on major desktop and server
348 The main interface for tracing control is a single command-line tool
349 named cmd:lttng. The latter can create several tracing sessions, enable
350 and disable events on the fly, filter events efficiently with custom
351 user expressions, start and stop tracing, and much more. LTTng can
352 record the traces on the file system or send them over the network, and
353 keep them totally or partially. You can view the traces once tracing
354 becomes inactive or in real-time.
356 <<installing-lttng,Install LTTng now>> and
357 <<getting-started,start tracing>>!
363 include::../common/warning-no-installation.txt[]
365 **LTTng** is a set of software <<plumbing,components>> which interact to
366 <<instrumenting,instrument>> the Linux kernel and user applications, and
367 to <<controlling-tracing,control tracing>> (start and stop
368 tracing, enable and disable event rules, and the rest). Those
369 components are bundled into the following packages:
371 * **LTTng-tools**: Libraries and command-line interface to
373 * **LTTng-modules**: Linux kernel modules to instrument and
375 * **LTTng-UST**: Libraries and Java/Python packages to instrument and
376 trace user applications.
378 Most distributions mark the LTTng-modules and LTTng-UST packages as
379 optional when installing LTTng-tools (which is always required). Note
382 * You only need to install LTTng-modules if you intend to trace the
384 * You only need to install LTTng-UST if you intend to trace user
388 [[building-from-source]]
389 === Build from source
391 To build and install LTTng{nbsp}{revision} from source:
393 . Using your distribution's package manager, or from source, install
394 the following dependencies of LTTng-tools and LTTng-UST:
397 * https://sourceforge.net/projects/libuuid/[libuuid]
398 * http://directory.fsf.org/wiki/Popt[popt]
399 * http://liburcu.org/[Userspace RCU]
400 * http://www.xmlsoft.org/[libxml2]
403 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
409 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
410 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
411 cd lttng-modules-2.8.* &&
413 sudo make modules_install &&
418 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
424 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.8.tar.bz2 &&
425 tar -xf lttng-ust-latest-2.8.tar.bz2 &&
426 cd lttng-ust-2.8.* &&
436 .Java and Python application tracing
438 If you need to instrument and trace <<java-application,Java
439 applications>>, pass the `--enable-java-agent-jul`,
440 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
441 `configure` script, depending on which Java logging framework you use.
443 If you need to instrument and trace <<python-application,Python
444 applications>>, pass the `--enable-python-agent` option to the
445 `configure` script. You can set the `PYTHON` environment variable to the
446 path to the Python interpreter for which to install the LTTng-UST Python
454 By default, LTTng-UST libraries are installed to
455 dir:{/usr/local/lib}, which is the de facto directory in which to
456 keep self-compiled and third-party libraries.
458 When <<building-tracepoint-providers-and-user-application,linking an
459 instrumented user application with `liblttng-ust`>>:
461 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
463 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
464 man:gcc(1), man:g++(1), or man:clang(1).
468 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
474 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.8.tar.bz2 &&
475 tar -xf lttng-tools-latest-2.8.tar.bz2 &&
476 cd lttng-tools-2.8.* &&
484 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
485 previous steps automatically for a given version of LTTng and confine
486 the installed files in a specific directory. This can be useful to test
487 LTTng without installing it on your system.
493 This is a short guide to get started quickly with LTTng kernel and user
496 Before you follow this guide, make sure to <<installing-lttng,install>>
499 This tutorial walks you through the steps to:
501 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
502 . <<tracing-your-own-user-application,Trace a user application>> written
504 . <<viewing-and-analyzing-your-traces,View and analyze the
508 [[tracing-the-linux-kernel]]
509 === Trace the Linux kernel
511 The following command lines start with the `#` prompt because you need
512 root privileges to trace the Linux kernel. You can also trace the kernel
513 as a regular user if your Unix user is a member of the
514 <<tracing-group,tracing group>>.
516 . Create a <<tracing-session,tracing session>> which writes its traces
517 to dir:{/tmp/my-kernel-trace}:
522 # lttng create my-kernel-session --output=/tmp/my-kernel-trace
526 . List the available kernel tracepoints and system calls:
531 # lttng list --kernel
532 # lttng list --kernel --syscall
536 . Create <<event,event rules>> which match the desired instrumentation
537 point names, for example the `sched_switch` and `sched_process_fork`
538 tracepoints, and the man:open(2) and man:close(2) system calls:
543 # lttng enable-event --kernel sched_switch,sched_process_fork
544 # lttng enable-event --kernel --syscall open,close
548 You can also create an event rule which matches _all_ the Linux kernel
549 tracepoints (this will generate a lot of data when tracing):
554 # lttng enable-event --kernel --all
558 . <<basic-tracing-session-control,Start tracing>>:
567 . Do some operation on your system for a few seconds. For example,
568 load a website, or list the files of a directory.
569 . <<basic-tracing-session-control,Stop tracing>> and destroy the
580 The man:lttng-destroy(1) command does not destroy the trace data; it
581 only destroys the state of the tracing session.
583 . For the sake of this example, make the recorded trace accessible to
589 sudo chown -R $(whoami) /tmp/my-kernel-trace
593 See <<viewing-and-analyzing-your-traces,View and analyze the
594 recorded events>> to view the recorded events.
597 [[tracing-your-own-user-application]]
598 === Trace a user application
600 This section steps you through a simple example to trace a
601 _Hello world_ program written in C.
603 To create the traceable user application:
605 . Create the tracepoint provider header file, which defines the
606 tracepoints and the events they can generate:
612 #undef TRACEPOINT_PROVIDER
613 #define TRACEPOINT_PROVIDER hello_world
615 #undef TRACEPOINT_INCLUDE
616 #define TRACEPOINT_INCLUDE "./hello-tp.h"
618 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
621 #include <lttng/tracepoint.h>
631 ctf_string(my_string_field, my_string_arg)
632 ctf_integer(int, my_integer_field, my_integer_arg)
636 #endif /* _HELLO_TP_H */
638 #include <lttng/tracepoint-event.h>
642 . Create the tracepoint provider package source file:
648 #define TRACEPOINT_CREATE_PROBES
649 #define TRACEPOINT_DEFINE
651 #include "hello-tp.h"
655 . Build the tracepoint provider package:
660 $ gcc -c -I. hello-tp.c
664 . Create the _Hello World_ application source file:
671 #include "hello-tp.h"
673 int main(int argc, char *argv[])
677 puts("Hello, World!\nPress Enter to continue...");
680 * The following getchar() call is only placed here for the purpose
681 * of this demonstration, to pause the application in order for
682 * you to have time to list its tracepoints. It is not
688 * A tracepoint() call.
690 * Arguments, as defined in hello-tp.h:
692 * 1. Tracepoint provider name (required)
693 * 2. Tracepoint name (required)
694 * 3. my_integer_arg (first user-defined argument)
695 * 4. my_string_arg (second user-defined argument)
697 * Notice the tracepoint provider and tracepoint names are
698 * NOT strings: they are in fact parts of variables that the
699 * macros in hello-tp.h create.
701 tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
703 for (x = 0; x < argc; ++x) {
704 tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
707 puts("Quitting now!");
708 tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
715 . Build the application:
724 . Link the application with the tracepoint provider package,
725 `liblttng-ust`, and `libdl`:
730 $ gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
734 Here's the whole build process:
737 .User space tracing tutorial's build steps.
738 image::ust-flow.png[]
740 To trace the user application:
742 . Run the application with a few arguments:
747 $ ./hello world and beyond
756 Press Enter to continue...
760 . Start an LTTng <<lttng-sessiond,session daemon>>:
765 $ lttng-sessiond --daemonize
769 Note that a session daemon might already be running, for example as
770 a service that the distribution's service manager started.
772 . List the available user space tracepoints:
777 $ lttng list --userspace
781 You see the `hello_world:my_first_tracepoint` tracepoint listed
782 under the `./hello` process.
784 . Create a <<tracing-session,tracing session>>:
789 $ lttng create my-user-space-session
793 . Create an <<event,event rule>> which matches the
794 `hello_world:my_first_tracepoint` event name:
799 $ lttng enable-event --userspace hello_world:my_first_tracepoint
803 . <<basic-tracing-session-control,Start tracing>>:
812 . Go back to the running `hello` application and press Enter. The
813 program executes all `tracepoint()` instrumentation points and exits.
814 . <<basic-tracing-session-control,Stop tracing>> and destroy the
825 The man:lttng-destroy(1) command does not destroy the trace data; it
826 only destroys the state of the tracing session.
828 By default, LTTng saves the traces in
829 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
830 where +__name__+ is the tracing session name. The
831 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
833 See <<viewing-and-analyzing-your-traces,View and analyze the
834 recorded events>> to view the recorded events.
837 [[viewing-and-analyzing-your-traces]]
838 === View and analyze the recorded events
840 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
841 kernel>> and <<tracing-your-own-user-application,Trace a user
842 application>> tutorials, you can inspect the recorded events.
844 Many tools are available to read LTTng traces:
846 * **cmd:babeltrace** is a command-line utility which converts trace
847 formats; it supports the format that LTTng produces, CTF, as well as a
848 basic text output which can be ++grep++ed. The cmd:babeltrace command
849 is part of the http://diamon.org/babeltrace[Babeltrace] project.
850 * Babeltrace also includes
851 **https://www.python.org/[Python] bindings** so
852 that you can easily open and read an LTTng trace with your own script,
853 benefiting from the power of Python.
854 * http://tracecompass.org/[**Trace Compass**]
855 is a graphical user interface for viewing and analyzing any type of
856 logs or traces, including LTTng's.
857 * https://github.com/lttng/lttng-analyses[**LTTng analyses**] is a
858 project which includes many high-level analyses of LTTng kernel
859 traces, like scheduling statistics, interrupt frequency distribution,
860 top CPU usage, and more.
862 NOTE: This section assumes that the traces recorded during the previous
863 tutorials were saved to their default location, in the
864 dir:{$LTTNG_HOME/lttng-traces} directory. The env:LTTNG_HOME
865 environment variable defaults to `$HOME` if not set.
868 [[viewing-and-analyzing-your-traces-bt]]
869 ==== Use the cmd:babeltrace command-line tool
871 The simplest way to list all the recorded events of a trace is to pass
872 its path to cmd:babeltrace with no options:
876 $ babeltrace ~/lttng-traces/my-user-space-session*
879 cmd:babeltrace finds all traces recursively within the given path and
880 prints all their events, merging them in chronological order.
882 You can pipe the output of cmd:babeltrace into a tool like man:grep(1) for
887 $ babeltrace /tmp/my-kernel-trace | grep _switch
890 You can pipe the output of cmd:babeltrace into a tool like man:wc(1) to
891 count the recorded events:
895 $ babeltrace /tmp/my-kernel-trace | grep _open | wc --lines
899 [[viewing-and-analyzing-your-traces-bt-python]]
900 ==== Use the Babeltrace Python bindings
902 The <<viewing-and-analyzing-your-traces-bt,text output of cmd:babeltrace>>
903 is useful to isolate events by simple matching using man:grep(1) and
904 similar utilities. However, more elaborate filters, such as keeping only
905 event records with a field value falling within a specific range, are
906 not trivial to write using a shell. Moreover, reductions and even the
907 most basic computations involving multiple event records are virtually
908 impossible to implement.
910 Fortunately, Babeltrace ships with Python 3 bindings which makes it easy
911 to read the event records of an LTTng trace sequentially and compute the
914 The following script accepts an LTTng Linux kernel trace path as its
915 first argument and prints the short names of the top 5 running processes
916 on CPU 0 during the whole trace:
921 from collections import Counter
927 if len(sys.argv) != 2:
928 msg = 'Usage: python3 {} TRACEPATH'.format(sys.argv[0])
929 print(msg, file=sys.stderr)
932 # A trace collection contains one or more traces
933 col = babeltrace.TraceCollection()
935 # Add the trace provided by the user (LTTng traces always have
937 if col.add_trace(sys.argv[1], 'ctf') is None:
938 raise RuntimeError('Cannot add trace')
940 # This counter dict contains execution times:
942 # task command name -> total execution time (ns)
943 exec_times = Counter()
945 # This contains the last `sched_switch` timestamp
949 for event in col.events:
950 # Keep only `sched_switch` events
951 if event.name != 'sched_switch':
954 # Keep only events which happened on CPU 0
955 if event['cpu_id'] != 0:
959 cur_ts = event.timestamp
965 # Previous task command (short) name
966 prev_comm = event['prev_comm']
968 # Initialize entry in our dict if not yet done
969 if prev_comm not in exec_times:
970 exec_times[prev_comm] = 0
972 # Compute previous command execution time
973 diff = cur_ts - last_ts
975 # Update execution time of this command
976 exec_times[prev_comm] += diff
978 # Update last timestamp
982 for name, ns in exec_times.most_common(5):
984 print('{:20}{} s'.format(name, s))
989 if __name__ == '__main__':
990 sys.exit(0 if top5proc() else 1)
997 $ python3 top5proc.py /tmp/my-kernel-trace/kernel
1003 swapper/0 48.607245889 s
1004 chromium 7.192738188 s
1005 pavucontrol 0.709894415 s
1006 Compositor 0.660867933 s
1007 Xorg.bin 0.616753786 s
1010 Note that `swapper/0` is the "idle" process of CPU 0 on Linux; since we
1011 weren't using the CPU that much when tracing, its first position in the
1016 == [[understanding-lttng]]Core concepts
1018 From a user's perspective, the LTTng system is built on a few concepts,
1019 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
1020 operates by sending commands to the <<lttng-sessiond,session daemon>>.
1021 Understanding how those objects relate to eachother is key in mastering
1024 The core concepts are:
1026 * <<tracing-session,Tracing session>>
1027 * <<domain,Tracing domain>>
1028 * <<channel,Channel and ring buffer>>
1029 * <<"event","Instrumentation point, event rule, event, and event record">>
1035 A _tracing session_ is a stateful dialogue between you and
1036 a <<lttng-sessiond,session daemon>>. You can
1037 <<creating-destroying-tracing-sessions,create a new tracing
1038 session>> with the `lttng create` command.
1040 Anything that you do when you control LTTng tracers happens within a
1041 tracing session. In particular, a tracing session:
1044 * Has its own set of trace files.
1045 * Has its own state of activity (started or stopped).
1046 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
1048 * Has its own <<channel,channels>> which have their own
1049 <<event,event rules>>.
1052 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
1053 image::concepts.png[]
1055 Those attributes and objects are completely isolated between different
1058 A tracing session is analogous to a cash machine session:
1059 the operations you do on the banking system through the cash machine do
1060 not alter the data of other users of the same system. In the case of
1061 the cash machine, a session lasts as long as your bank card is inside.
1062 In the case of LTTng, a tracing session lasts from the `lttng create`
1063 command to the `lttng destroy` command.
1066 .Each Unix user has its own set of tracing sessions.
1067 image::many-sessions.png[]
1070 [[tracing-session-mode]]
1071 ==== Tracing session mode
1073 LTTng can send the generated trace data to different locations. The
1074 _tracing session mode_ dictates where to send it. The following modes
1075 are available in LTTng{nbsp}{revision}:
1077 [[local-mode]]Local mode::
1078 LTTng writes the traces to the file system of the machine being traced
1081 [[net-streaming-mode]]Network streaming mode::
1082 LTTng sends the traces over the network to a
1083 <<lttng-relayd,relay daemon>> running on a remote system.
1086 LTTng does not write the traces by default. Instead, you can request
1087 LTTng to <<taking-a-snapshot,take a snapshot>>, that is, a copy of the
1088 current tracing buffers, and to write it to the target's file system
1089 or to send it over the network to a <<lttng-relayd,relay daemon>>
1090 running on a remote system.
1092 [[live-mode]]Live mode::
1093 This mode is similar to the network streaming mode, but a live
1094 trace viewer can connect to the distant relay daemon to
1095 <<lttng-live,view event records as LTTng generates them>> by
1102 A _tracing domain_ is a namespace for event sources. A tracing domain
1103 has its own properties and features.
1105 There are currently five available tracing domains:
1109 * `java.util.logging` (JUL)
1113 You must specify a tracing domain when using some commands to avoid
1114 ambiguity. For example, since all the domains support named tracepoints
1115 as event sources (instrumentation points that you manually insert in the
1116 source code), you need to specify a tracing domain when
1117 <<enabling-disabling-events,creating an event rule>> because all the
1118 tracing domains could have tracepoints with the same names.
1120 Some features are reserved to specific tracing domains. Dynamic function
1121 entry and return instrumentation points, for example, are currently only
1122 supported in the Linux kernel tracing domain, but support for other
1123 tracing domains could be added in the future.
1125 You can create <<channel,channels>> in the Linux kernel and user space
1126 tracing domains. The other tracing domains have a single default
1131 === Channel and ring buffer
1133 A _channel_ is an object which is responsible for a set of ring buffers.
1134 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1135 tracer emits an event, it can record it to one or more
1136 sub-buffers. The attributes of a channel determine what to do when
1137 there's no space left for a new event record because all sub-buffers
1138 are full, where to send a full sub-buffer, and other behaviours.
1140 A channel is always associated to a <<domain,tracing domain>>. The
1141 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1142 a default channel which you cannot configure.
1144 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1145 an event, it records it to the sub-buffers of all
1146 the enabled channels with a satisfied event rule, as long as those
1147 channels are part of active <<tracing-session,tracing sessions>>.
1150 [[channel-buffering-schemes]]
1151 ==== Per-user vs. per-process buffering schemes
1153 A channel has at least one ring buffer _per CPU_. LTTng always
1154 records an event to the ring buffer associated to the CPU on which it
1157 Two _buffering schemes_ are available when you
1158 <<enabling-disabling-channels,create a channel>> in the
1159 user space <<domain,tracing domain>>:
1161 Per-user buffering::
1162 Allocate one set of ring buffers--one per CPU--shared by all the
1163 instrumented processes of each Unix user.
1167 .Per-user buffering scheme.
1168 image::per-user-buffering.png[]
1171 Per-process buffering::
1172 Allocate one set of ring buffers--one per CPU--for each
1173 instrumented process.
1177 .Per-process buffering scheme.
1178 image::per-process-buffering.png[]
1181 The per-process buffering scheme tends to consume more memory than the
1182 per-user option because systems generally have more instrumented
1183 processes than Unix users running instrumented processes. However, the
1184 per-process buffering scheme ensures that one process having a high
1185 event throughput won't fill all the shared sub-buffers of the same
1188 The Linux kernel tracing domain has only one available buffering scheme
1189 which is to allocate a single set of ring buffers for the whole system.
1190 This scheme is similar to the per-user option, but with a single, global
1191 user "running" the kernel.
1194 [[channel-overwrite-mode-vs-discard-mode]]
1195 ==== Overwrite vs. discard event loss modes
1197 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1198 arc in the following animation) of a specific channel's ring buffer.
1199 When there's no space left in a sub-buffer, the tracer marks it as
1200 consumable (red) and another, empty sub-buffer starts receiving the
1201 following event records. A <<lttng-consumerd,consumer daemon>>
1202 eventually consumes the marked sub-buffer (returns to white).
1205 [role="docsvg-channel-subbuf-anim"]
1210 In an ideal world, sub-buffers are consumed faster than they are filled,
1211 as is the case in the previous animation. In the real world,
1212 however, all sub-buffers can be full at some point, leaving no space to
1213 record the following events.
1215 By design, LTTng is a _non-blocking_ tracer: when no empty sub-buffer is
1216 available, it is acceptable to lose event records when the alternative
1217 would be to cause substantial delays in the instrumented application's
1218 execution. LTTng privileges performance over integrity; it aims at
1219 perturbing the traced system as little as possible in order to make
1220 tracing of subtle race conditions and rare interrupt cascades possible.
1222 When it comes to losing event records because no empty sub-buffer is
1223 available, the channel's _event loss mode_ determines what to do. The
1224 available event loss modes are:
1227 Drop the newest event records until a the tracer
1228 releases a sub-buffer.
1231 Clear the sub-buffer containing the oldest event records and start
1232 writing the newest event records there.
1234 This mode is sometimes called _flight recorder mode_ because it's
1236 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1237 always keep a fixed amount of the latest data.
1239 Which mechanism you should choose depends on your context: prioritize
1240 the newest or the oldest event records in the ring buffer?
1242 Beware that, in overwrite mode, the tracer abandons a whole sub-buffer
1243 as soon as a there's no space left for a new event record, whereas in
1244 discard mode, the tracer only discards the event record that doesn't
1247 In discard mode, LTTng increments a count of lost event records when an
1248 event record is lost and saves this count to the trace. Since
1249 LTTng{nbsp}2.8, in overwrite mode, LTTng writes to a given sub-buffer
1250 its sequence number within its data stream. With a <<local-mode,local>>,
1251 <<net-streaming-mode,network streaming>>, or <<live-mode,live>>
1252 <<tracing-session,tracing session>>, a trace reader can use such
1253 sequence numbers to report lost packets. In overwrite mode, LTTng
1254 doesn't write to the trace the exact number of lost event records in
1255 those lost sub-buffers.
1257 Trace analyses can use saved discarded event record and sub-buffer
1258 (packet) counts of the trace to decide whether or not to perform the
1259 analyses even if trace data is known to be missing.
1261 There are a few ways to decrease your probability of losing event
1263 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1264 how you can fine-une the sub-buffer count and size of a channel to
1265 virtually stop losing event records, though at the cost of greater
1269 [[channel-subbuf-size-vs-subbuf-count]]
1270 ==== Sub-buffer count and size
1272 When you <<enabling-disabling-channels,create a channel>>, you can
1273 set its number of sub-buffers and their size.
1275 Note that there is noticeable CPU overhead introduced when
1276 switching sub-buffers (marking a full one as consumable and switching
1277 to an empty one for the following events to be recorded). Knowing this,
1278 the following list presents a few practical situations along with how
1279 to configure the sub-buffer count and size for them:
1281 * **High event throughput**: In general, prefer bigger sub-buffers to
1282 lower the risk of losing event records.
1284 Having bigger sub-buffers also ensures a lower
1285 <<channel-switch-timer,sub-buffer switching frequency>>.
1287 The number of sub-buffers is only meaningful if you create the channel
1288 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1289 other sub-buffers are left unaltered.
1291 * **Low event throughput**: In general, prefer smaller sub-buffers
1292 since the risk of losing event records is low.
1294 Because events occur less frequently, the sub-buffer switching frequency
1295 should remain low and thus the tracer's overhead should not be a
1298 * **Low memory system**: If your target system has a low memory
1299 limit, prefer fewer first, then smaller sub-buffers.
1301 Even if the system is limited in memory, you want to keep the
1302 sub-buffers as big as possible to avoid a high sub-buffer switching
1305 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1306 which means event data is very compact. For example, the average
1307 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1308 sub-buffer size of 1{nbsp}MiB is considered big.
1310 The previous situations highlight the major trade-off between a few big
1311 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1312 frequency vs. how much data is lost in overwrite mode. Assuming a
1313 constant event throughput and using the overwrite mode, the two
1314 following configurations have the same ring buffer total size:
1317 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1322 * **2 sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1323 switching frequency, but if a sub-buffer overwrite happens, half of
1324 the event records so far (4{nbsp}MiB) are definitely lost.
1325 * **8 sub-buffers of 1{nbsp}MiB each**: Expect 4{nbsp}times the tracer's
1326 overhead as the previous configuration, but if a sub-buffer
1327 overwrite happens, only the eighth of event records so far are
1330 In discard mode, the sub-buffers count parameter is pointless: use two
1331 sub-buffers and set their size according to the requirements of your
1335 [[channel-switch-timer]]
1336 ==== Switch timer period
1338 The _switch timer period_ is an important configurable attribute of
1339 a channel to ensure periodic sub-buffer flushing.
1341 When the _switch timer_ expires, a sub-buffer switch happens. You can
1342 set the switch timer period attribute when you
1343 <<enabling-disabling-channels,create a channel>> to ensure that event
1344 data is consumed and committed to trace files or to a distant relay
1345 daemon periodically in case of a low event throughput.
1348 [role="docsvg-channel-switch-timer"]
1353 This attribute is also convenient when you use big sub-buffers to cope
1354 with a sporadic high event throughput, even if the throughput is
1358 [[channel-read-timer]]
1359 ==== Read timer period
1361 By default, the LTTng tracers use a notification mechanism to signal a
1362 full sub-buffer so that a consumer daemon can consume it. When such
1363 notifications must be avoided, for example in real-time applications,
1364 you can use the channel's _read timer_ instead. When the read timer
1365 fires, the <<lttng-consumerd,consumer daemon>> checks for full,
1366 consumable sub-buffers.
1369 [[tracefile-rotation]]
1370 ==== Trace file count and size
1372 By default, trace files can grow as large as needed. You can set the
1373 maximum size of each trace file that a channel writes when you
1374 <<enabling-disabling-channels,create a channel>>. When the size of
1375 a trace file reaches the channel's fixed maximum size, LTTng creates
1376 another file to contain the next event records. LTTng appends a file
1377 count to each trace file name in this case.
1379 If you set the trace file size attribute when you create a channel, the
1380 maximum number of trace files that LTTng creates is _unlimited_ by
1381 default. To limit them, you can also set a maximum number of trace
1382 files. When the number of trace files reaches the channel's fixed
1383 maximum count, the oldest trace file is overwritten. This mechanism is
1384 called _trace file rotation_.
1388 === Instrumentation point, event rule, event, and event record
1390 An _event rule_ is a set of conditions which must be **all** satisfied
1391 for LTTng to record an occuring event.
1393 You set the conditions when you <<enabling-disabling-events,create
1396 You always attach an event rule to <<channel,channel>> when you create
1399 When an event passes the conditions of an event rule, LTTng records it
1400 in one of the attached channel's sub-buffers.
1402 The available conditions, as of LTTng{nbsp}{revision}, are:
1404 * The event rule _is enabled_.
1405 * The instrumentation point's type _is{nbsp}T_.
1406 * The instrumentation point's name (sometimes called _event name_)
1407 _matches{nbsp}N_, but _is not{nbsp}E_.
1408 * The instrumentation point's log level _is as severe as{nbsp}L_, or
1409 _is exactly{nbsp}L_.
1410 * The fields of the event's payload _satisfy_ a filter
1411 expression{nbsp}__F__.
1413 As you can see, all the conditions but the dynamic filter are related to
1414 the event rule's status or to the instrumentation point, not to the
1415 occurring events. This is why, without a filter, checking if an event
1416 passes an event rule is not a dynamic task: when you create or modify an
1417 event rule, all the tracers of its tracing domain enable or disable the
1418 instrumentation points themselves once. This is possible because the
1419 attributes of an instrumentation point (type, name, and log level) are
1420 defined statically. In other words, without a dynamic filter, the tracer
1421 _does not evaluate_ the arguments of an instrumentation point unless it
1422 matches an enabled event rule.
1424 Note that, for LTTng to record an event, the <<channel,channel>> to
1425 which a matching event rule is attached must also be enabled, and the
1426 tracing session owning this channel must be active.
1429 .Logical path from an instrumentation point to an event record.
1430 image::event-rule.png[]
1432 .Event, event record, or event rule?
1434 With so many similar terms, it's easy to get confused.
1436 An **event** is the consequence of the execution of an _instrumentation
1437 point_, like a tracepoint that you manually place in some source code,
1438 or a Linux kernel KProbe. An event is said to _occur_ at a specific
1439 time. Different actions can be taken upon the occurrence of an event,
1440 like record the event's payload to a buffer.
1442 An **event record** is the representation of an event in a sub-buffer. A
1443 tracer is responsible for capturing the payload of an event, current
1444 context variables, the event's ID, and the event's timestamp. LTTng
1445 can append this sub-buffer to a trace file.
1447 An **event rule** is a set of conditions which must all be satisfied for
1448 LTTng to record an occuring event. Events still occur without
1449 satisfying event rules, but LTTng does not record them.
1454 == Components of noch:{LTTng}
1456 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1457 to call LTTng a simple _tool_ since it is composed of multiple
1458 interacting components. This section describes those components,
1459 explains their respective roles, and shows how they connect together to
1460 form the LTTng ecosystem.
1462 The following diagram shows how the most important components of LTTng
1463 interact with user applications, the Linux kernel, and you:
1466 .Control and trace data paths between LTTng components.
1467 image::plumbing.png[]
1469 The LTTng project incorporates:
1471 * **LTTng-tools**: Libraries and command-line interface to
1472 control tracing sessions.
1473 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1474 ** <<lttng-consumerd,Consumer daemon>> (cmd:lttng-consumerd).
1475 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1476 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1477 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1478 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1480 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1481 headers to instrument and trace any native user application.
1482 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1483 *** `liblttng-ust-libc-wrapper`
1484 *** `liblttng-ust-pthread-wrapper`
1485 *** `liblttng-ust-cyg-profile`
1486 *** `liblttng-ust-cyg-profile-fast`
1487 *** `liblttng-ust-dl`
1488 ** User space tracepoint provider source files generator command-line
1489 tool (man:lttng-gen-tp(1)).
1490 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1491 Java applications using `java.util.logging` or
1492 Apache log4j 1.2 logging.
1493 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1494 Python applications using the standard `logging` package.
1495 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1497 ** LTTng kernel tracer module.
1498 ** Tracing ring buffer kernel modules.
1499 ** Probe kernel modules.
1500 ** LTTng logger kernel module.
1504 === Tracing control command-line interface
1507 .The tracing control command-line interface.
1508 image::plumbing-lttng-cli.png[]
1510 The _man:lttng(1) command-line tool_ is the standard user interface to
1511 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1512 is part of LTTng-tools.
1514 The cmd:lttng tool is linked with
1515 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1516 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1518 The cmd:lttng tool has a Git-like interface:
1522 $ lttng <GENERAL OPTIONS> <COMMAND> <COMMAND OPTIONS>
1525 The <<controlling-tracing,Tracing control>> section explores the
1526 available features of LTTng using the cmd:lttng tool.
1529 [[liblttng-ctl-lttng]]
1530 === Tracing control library
1533 .The tracing control library.
1534 image::plumbing-liblttng-ctl.png[]
1536 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1537 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1538 underlying protocol's details. `liblttng-ctl` is part of LTTng-tools.
1540 The <<lttng-cli,cmd:lttng command-line tool>>
1541 is linked with `liblttng-ctl`.
1543 You can use `liblttng-ctl` in C or $$C++$$ source code by including its
1548 #include <lttng/lttng.h>
1551 Some objects are referenced by name (C string), such as tracing
1552 sessions, but most of them require to create a handle first using
1553 `lttng_create_handle()`.
1555 The best available developer documentation for `liblttng-ctl` is, as of
1556 LTTng{nbsp}{revision}, its installed header files. Every function and
1557 structure is thoroughly documented.
1561 === User space tracing library
1564 .The user space tracing library.
1565 image::plumbing-liblttng-ust.png[]
1567 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1568 is the LTTng user space tracer. It receives commands from a
1569 <<lttng-sessiond,session daemon>>, for example to
1570 enable and disable specific instrumentation points, and writes event
1571 records to ring buffers shared with a
1572 <<lttng-consumerd,consumer daemon>>.
1573 `liblttng-ust` is part of LTTng-UST.
1575 Public C header files are installed beside `liblttng-ust` to
1576 instrument any <<c-application,C or $$C++$$ application>>.
1578 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1579 packages, use their own library providing tracepoints which is
1580 linked with `liblttng-ust`.
1582 An application or library does not have to initialize `liblttng-ust`
1583 manually: its constructor does the necessary tasks to properly register
1584 to a session daemon. The initialization phase also enables the
1585 instrumentation points matching the <<event,event rules>> that you
1589 [[lttng-ust-agents]]
1590 === User space tracing agents
1593 .The user space tracing agents.
1594 image::plumbing-lttng-ust-agents.png[]
1596 The _LTTng-UST Java and Python agents_ are regular Java and Python
1597 packages which add LTTng tracing capabilities to the
1598 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1600 In the case of Java, the
1601 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1602 core logging facilities] and
1603 https://logging.apache.org/log4j/1.2/[Apache log4j 1.2] are supported.
1604 Note that Apache Log4{nbsp}2 is not supported.
1606 In the case of Python, the standard
1607 https://docs.python.org/3/library/logging.html[`logging`] package
1608 is supported. Both Python 2 and Python 3 modules can import the
1609 LTTng-UST Python agent package.
1611 The applications using the LTTng-UST agents are in the
1612 `java.util.logging` (JUL),
1613 log4j, and Python <<domain,tracing domains>>.
1615 Both agents use the same mechanism to trace the log statements. When an
1616 agent is initialized, it creates a log handler that attaches to the root
1617 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
1618 When the application executes a log statement, it is passed to the
1619 agent's log handler by the root logger. The agent's log handler calls a
1620 native function in a tracepoint provider package shared library linked
1621 with <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
1622 other fields, like its logger name and its log level. This native
1623 function contains a user space instrumentation point, hence tracing the
1626 The log level condition of an
1627 <<event,event rule>> is considered when tracing
1628 a Java or a Python application, and it's compatible with the standard
1629 JUL, log4j, and Python log levels.
1633 === LTTng kernel modules
1636 .The LTTng kernel modules.
1637 image::plumbing-lttng-modules.png[]
1639 The _LTTng kernel modules_ are a set of Linux kernel modules
1640 which implement the kernel tracer of the LTTng project. The LTTng
1641 kernel modules are part of LTTng-modules.
1643 The LTTng kernel modules include:
1645 * A set of _probe_ modules.
1647 Each module attaches to a specific subsystem
1648 of the Linux kernel using its tracepoint instrument points. There are
1649 also modules to attach to the entry and return points of the Linux
1650 system call functions.
1652 * _Ring buffer_ modules.
1654 A ring buffer implementation is provided as kernel modules. The LTTng
1655 kernel tracer writes to the ring buffer; a
1656 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
1658 * The _LTTng kernel tracer_ module.
1659 * The _LTTng logger_ module.
1661 The LTTng logger module implements the special path:{/proc/lttng-logger}
1662 file so that any executable can generate LTTng events by opening and
1663 writing to this file.
1665 See <<proc-lttng-logger-abi,LTTng logger>>.
1667 Generally, you do not have to load the LTTng kernel modules manually
1668 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
1669 daemon>> loads the necessary modules when starting. If you have extra
1670 probe modules, you can specify to load them to the session daemon on
1673 The LTTng kernel modules are installed in
1674 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
1675 the kernel release (see `uname --kernel-release`).
1682 .The session daemon.
1683 image::plumbing-sessiond.png[]
1685 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
1686 managing tracing sessions and for controlling the various components of
1687 LTTng. The session daemon is part of LTTng-tools.
1689 The session daemon sends control requests to and receives control
1692 * The <<lttng-ust,user space tracing library>>.
1694 Any instance of the user space tracing library first registers to
1695 a session daemon. Then, the session daemon can send requests to
1696 this instance, such as:
1699 ** Get the list of tracepoints.
1700 ** Share an <<event,event rule>> so that the user space tracing library
1701 can enable or disable tracepoints. Amongst the possible conditions
1702 of an event rule is a filter expression which `liblttng-ust` evalutes
1703 when an event occurs.
1704 ** Share <<channel,channel>> attributes and ring buffer locations.
1707 The session daemon and the user space tracing library use a Unix
1708 domain socket for their communication.
1710 * The <<lttng-ust-agents,user space tracing agents>>.
1712 Any instance of a user space tracing agent first registers to
1713 a session daemon. Then, the session daemon can send requests to
1714 this instance, such as:
1717 ** Get the list of loggers.
1718 ** Enable or disable a specific logger.
1721 The session daemon and the user space tracing agent use a TCP connection
1722 for their communication.
1724 * The <<lttng-modules,LTTng kernel tracer>>.
1725 * The <<lttng-consumerd,consumer daemon>>.
1727 The session daemon sends requests to the consumer daemon to instruct
1728 it where to send the trace data streams, amongst other information.
1730 * The <<lttng-relayd,relay daemon>>.
1732 The session daemon receives commands from the
1733 <<liblttng-ctl-lttng,tracing control library>>.
1735 The root session daemon loads the appropriate
1736 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
1737 a <<lttng-consumerd,consumer daemon>> as soon as you create
1738 an <<event,event rule>>.
1740 The session daemon does not send and receive trace data: this is the
1741 role of the <<lttng-consumerd,consumer daemon>> and
1742 <<lttng-relayd,relay daemon>>. It does, however, generate the
1743 http://diamon.org/ctf/[CTF] metadata stream.
1745 Each Unix user can have its own session daemon instance. The
1746 tracing sessions managed by different session daemons are completely
1749 The root user's session daemon is the only one which is
1750 allowed to control the LTTng kernel tracer, and its spawned consumer
1751 daemon is the only one which is allowed to consume trace data from the
1752 LTTng kernel tracer. Note, however, that any Unix user which is a member
1753 of the <<tracing-group,tracing group>> is allowed
1754 to create <<channel,channels>> in the
1755 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
1758 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
1759 session daemon when using its `create` command if none is currently
1760 running. You can also start the session daemon manually.
1767 .The consumer daemon.
1768 image::plumbing-consumerd.png[]
1770 The _consumer daemon_, cmd:lttng-consumerd, is a daemon which shares
1771 ring buffers with user applications or with the LTTng kernel modules to
1772 collect trace data and send it to some location (on disk or to a
1773 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
1774 is part of LTTng-tools.
1776 You do not start a consumer daemon manually: a consumer daemon is always
1777 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
1778 <<event,event rule>>, that is, before you start tracing. When you kill
1779 its owner session daemon, the consumer daemon also exits because it is
1780 the session daemon's child process. Command-line options of
1781 man:lttng-sessiond(8) target the consumer daemon process.
1783 There are up to two running consumer daemons per Unix user, whereas only
1784 one session daemon can run per user. This is because each process can be
1785 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
1786 and 64-bit processes, it is more efficient to have separate
1787 corresponding 32-bit and 64-bit consumer daemons. The root user is an
1788 exception: it can have up to _three_ running consumer daemons: 32-bit
1789 and 64-bit instances for its user applications, and one more
1790 reserved for collecting kernel trace data.
1798 image::plumbing-relayd.png[]
1800 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
1801 between remote session and consumer daemons, local trace files, and a
1802 remote live trace viewer. The relay daemon is part of LTTng-tools.
1804 The main purpose of the relay daemon is to implement a receiver of
1805 <<sending-trace-data-over-the-network,trace data over the network>>.
1806 This is useful when the target system does not have much file system
1807 space to record trace files locally.
1809 The relay daemon is also a server to which a
1810 <<lttng-live,live trace viewer>> can
1811 connect. The live trace viewer sends requests to the relay daemon to
1812 receive trace data as the target system emits events. The
1813 communication protocol is named _LTTng live_; it is used over TCP
1816 Note that you can start the relay daemon on the target system directly.
1817 This is the setup of choice when the use case is to view events as
1818 the target system emits them without the need of a remote system.
1822 == [[using-lttng]]Instrumentation
1824 There are many examples of tracing and monitoring in our everyday life:
1826 * You have access to real-time and historical weather reports and
1827 forecasts thanks to weather stations installed around the country.
1828 * You know your heart is safe thanks to an electrocardiogram.
1829 * You make sure not to drive your car too fast and to have enough fuel
1830 to reach your destination thanks to gauges visible on your dashboard.
1832 All the previous examples have something in common: they rely on
1833 **instruments**. Without the electrodes attached to the surface of your
1834 body's skin, cardiac monitoring is futile.
1836 LTTng, as a tracer, is no different from those real life examples. If
1837 you're about to trace a software system or, in other words, record its
1838 history of execution, you better have **instrumentation points** in the
1839 subject you're tracing, that is, the actual software.
1841 Various ways were developed to instrument a piece of software for LTTng
1842 tracing. The most straightforward one is to manually place
1843 instrumentation points, called _tracepoints_, in the software's source
1844 code. It is also possible to add instrumentation points dynamically in
1845 the Linux kernel <<domain,tracing domain>>.
1847 If you're only interested in tracing the Linux kernel, your
1848 instrumentation needs are probably already covered by LTTng's built-in
1849 <<lttng-modules,Linux kernel tracepoints>>. You may also wish to trace a
1850 user application which is already instrumented for LTTng tracing.
1851 In such cases, you can skip this whole section and read the topics of
1852 the <<controlling-tracing,Tracing control>> section.
1854 Many methods are available to instrument a piece of software for LTTng
1857 * <<c-application,User space instrumentation for C and $$C++$$
1859 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
1860 * <<java-application,User space Java agent>>.
1861 * <<python-application,User space Python agent>>.
1862 * <<proc-lttng-logger-abi,LTTng logger>>.
1863 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
1867 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
1869 The procedure to instrument a C or $$C++$$ user application with
1870 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
1872 . <<tracepoint-provider,Create the source files of a tracepoint provider
1874 . <<probing-the-application-source-code,Add tracepoints to
1875 the application's source code>>.
1876 . <<building-tracepoint-providers-and-user-application,Build and link
1877 a tracepoint provider package and the user application>>.
1879 If you need quick, man:printf(3)-like instrumentation, you can skip
1880 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
1883 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
1884 instrument a user application with `liblttng-ust`.
1887 [[tracepoint-provider]]
1888 ==== Create the source files of a tracepoint provider package
1890 A _tracepoint provider_ is a set of compiled functions which provide
1891 **tracepoints** to an application, the type of instrumentation point
1892 supported by LTTng-UST. Those functions can emit events with
1893 user-defined fields and serialize those events as event records to one
1894 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
1895 macro, which you <<probing-the-application-source-code,insert in a user
1896 application's source code>>, calls those functions.
1898 A _tracepoint provider package_ is an object file (`.o`) or a shared
1899 library (`.so`) which contains one or more tracepoint providers.
1900 Its source files are:
1902 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
1903 * A <<tpp-source,tracepoint provider package source>> (`.c`).
1905 A tracepoint provider package is dynamically linked with `liblttng-ust`,
1906 the LTTng user space tracer, at run time.
1909 .User application linked with `liblttng-ust` and containing a tracepoint provider.
1910 image::ust-app.png[]
1912 NOTE: If you need quick, man:printf(3)-like instrumentation, you can
1913 skip creating and using a tracepoint provider and use
1914 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
1918 ===== Create a tracepoint provider header file template
1920 A _tracepoint provider header file_ contains the tracepoint
1921 definitions of a tracepoint provider.
1923 To create a tracepoint provider header file:
1925 . Start from this template:
1929 .Tracepoint provider header file template (`.h` file extension).
1931 #undef TRACEPOINT_PROVIDER
1932 #define TRACEPOINT_PROVIDER provider_name
1934 #undef TRACEPOINT_INCLUDE
1935 #define TRACEPOINT_INCLUDE "./tp.h"
1937 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
1940 #include <lttng/tracepoint.h>
1943 * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
1944 * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
1949 #include <lttng/tracepoint-event.h>
1955 * `provider_name` with the name of your tracepoint provider.
1956 * `"tp.h"` with the name of your tracepoint provider header file.
1958 . Below the `#include <lttng/tracepoint.h>` line, put your
1959 <<defining-tracepoints,tracepoint definitions>>.
1961 Your tracepoint provider name must be unique amongst all the possible
1962 tracepoint provider names used on the same target system. We
1963 suggest to include the name of your project or company in the name,
1964 for example, `org_lttng_my_project_tpp`.
1966 TIP: [[lttng-gen-tp]]You can use the man:lttng-gen-tp(1) tool to create
1967 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
1968 write are the <<defining-tracepoints,tracepoint definitions>>.
1971 [[defining-tracepoints]]
1972 ===== Create a tracepoint definition
1974 A _tracepoint definition_ defines, for a given tracepoint:
1976 * Its **input arguments**. They are the macro parameters that the
1977 `tracepoint()` macro accepts for this particular tracepoint
1978 in the user application's source code.
1979 * Its **output event fields**. They are the sources of event fields
1980 that form the payload of any event that the execution of the
1981 `tracepoint()` macro emits for this particular tracepoint.
1983 You can create a tracepoint definition by using the
1984 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
1986 <<tpp-header,tracepoint provider header file template>>.
1988 The syntax of the `TRACEPOINT_EVENT()` macro is:
1991 .`TRACEPOINT_EVENT()` macro syntax.
1994 /* Tracepoint provider name */
1997 /* Tracepoint name */
2000 /* Input arguments */
2005 /* Output event fields */
2014 * `provider_name` with your tracepoint provider name.
2015 * `tracepoint_name` with your tracepoint name.
2016 * `arguments` with the <<tpp-def-input-args,input arguments>>.
2017 * `fields` with the <<tpp-def-output-fields,output event field>>
2020 This tracepoint emits events named `provider_name:tracepoint_name`.
2023 .Event name's length limitation
2025 The concatenation of the tracepoint provider name and the
2026 tracepoint name must not exceed **254 characters**. If it does, the
2027 instrumented application compiles and runs, but LTTng throws multiple
2028 warnings and you could experience serious issues.
2031 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
2034 .`TP_ARGS()` macro syntax.
2043 * `type` with the C type of the argument.
2044 * `arg_name` with the argument name.
2046 You can repeat `type` and `arg_name` up to 10 times to have
2047 more than one argument.
2049 .`TP_ARGS()` usage with three arguments.
2061 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
2062 tracepoint definition with no input arguments.
2064 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
2065 `ctf_*()` macros. Each `ctf_*()` macro defines one event field. See
2066 man:lttng-ust(3) for a complete description of the available `ctf_*()`
2067 macros. A `ctf_*()` macro specifies the type, size, and byte order of
2070 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
2071 C expression that the tracer evalutes at the `tracepoint()` macro site
2072 in the application's source code. This expression provides a field's
2073 source of data. The argument expression can include input argument names
2074 listed in the `TP_ARGS()` macro.
2076 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
2077 must be unique within a given tracepoint definition.
2079 Here's a complete tracepoint definition example:
2081 .Tracepoint definition.
2083 The following tracepoint definition defines a tracepoint which takes
2084 three input arguments and has four output event fields.
2088 #include "my-custom-structure.h"
2094 const struct my_custom_structure*, my_custom_structure,
2099 ctf_string(query_field, query)
2100 ctf_float(double, ratio_field, ratio)
2101 ctf_integer(int, recv_size, my_custom_structure->recv_size)
2102 ctf_integer(int, send_size, my_custom_structure->send_size)
2107 You can refer to this tracepoint definition with the `tracepoint()`
2108 macro in your application's source code like this:
2112 tracepoint(my_provider, my_tracepoint,
2113 my_structure, some_ratio, the_query);
2117 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2118 if they satisfy an enabled <<event,event rule>>.
2121 [[using-tracepoint-classes]]
2122 ===== Use a tracepoint class
2124 A _tracepoint class_ is a class of tracepoints which share the same
2125 output event field definitions. A _tracepoint instance_ is one
2126 instance of such a defined tracepoint class, with its own tracepoint
2129 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2130 shorthand which defines both a tracepoint class and a tracepoint
2131 instance at the same time.
2133 When you build a tracepoint provider package, the C or $$C++$$ compiler
2134 creates one serialization function for each **tracepoint class**. A
2135 serialization function is responsible for serializing the event fields
2136 of a tracepoint to a sub-buffer when tracing.
2138 For various performance reasons, when your situation requires multiple
2139 tracepoint definitions with different names, but with the same event
2140 fields, we recommend that you manually create a tracepoint class
2141 and instantiate as many tracepoint instances as needed. One positive
2142 effect of such a design, amongst other advantages, is that all
2143 tracepoint instances of the same tracepoint class reuse the same
2144 serialization function, thus reducing
2145 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2147 .Use a tracepoint class and tracepoint instances.
2149 Consider the following three tracepoint definitions:
2161 ctf_integer(int, userid, userid)
2162 ctf_integer(size_t, len, len)
2174 ctf_integer(int, userid, userid)
2175 ctf_integer(size_t, len, len)
2187 ctf_integer(int, userid, userid)
2188 ctf_integer(size_t, len, len)
2193 In this case, we create three tracepoint classes, with one implicit
2194 tracepoint instance for each of them: `get_account`, `get_settings`, and
2195 `get_transaction`. However, they all share the same event field names
2196 and types. Hence three identical, yet independent serialization
2197 functions are created when you build the tracepoint provider package.
2199 A better design choice is to define a single tracepoint class and three
2200 tracepoint instances:
2204 /* The tracepoint class */
2205 TRACEPOINT_EVENT_CLASS(
2206 /* Tracepoint provider name */
2209 /* Tracepoint class name */
2212 /* Input arguments */
2218 /* Output event fields */
2220 ctf_integer(int, userid, userid)
2221 ctf_integer(size_t, len, len)
2225 /* The tracepoint instances */
2226 TRACEPOINT_EVENT_INSTANCE(
2227 /* Tracepoint provider name */
2230 /* Tracepoint class name */
2233 /* Tracepoint name */
2236 /* Input arguments */
2242 TRACEPOINT_EVENT_INSTANCE(
2251 TRACEPOINT_EVENT_INSTANCE(
2264 [[assigning-log-levels]]
2265 ===== Assign a log level to a tracepoint definition
2267 You can assign an optional _log level_ to a
2268 <<defining-tracepoints,tracepoint definition>>.
2270 Assigning different levels of severity to tracepoint definitions can
2271 be useful: when you <<enabling-disabling-events,create an event rule>>,
2272 you can target tracepoints having a log level as severe as a specific
2275 The concept of LTTng-UST log levels is similar to the levels found
2276 in typical logging frameworks:
2278 * In a logging framework, the log level is given by the function
2279 or method name you use at the log statement site: `debug()`,
2280 `info()`, `warn()`, `error()`, and so on.
2281 * In LTTng-UST, you statically assign the log level to a tracepoint
2282 definition; any `tracepoint()` macro invocation which refers to
2283 this definition has this log level.
2285 You can assign a log level to a tracepoint definition with the
2286 `TRACEPOINT_LOGLEVEL()` macro. You must use this macro _after_ the
2287 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2288 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2291 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2294 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2296 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2301 * `provider_name` with the tracepoint provider name.
2302 * `tracepoint_name` with the tracepoint name.
2303 * `log_level` with the log level to assign to the tracepoint
2304 definition named `tracepoint_name` in the `provider_name`
2305 tracepoint provider.
2307 See man:lttng-ust(3) for a list of available log level names.
2309 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2313 /* Tracepoint definition */
2322 ctf_integer(int, userid, userid)
2323 ctf_integer(size_t, len, len)
2327 /* Log level assignment */
2328 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2334 ===== Create a tracepoint provider package source file
2336 A _tracepoint provider package source file_ is a C source file which
2337 includes a <<tpp-header,tracepoint provider header file>> to expand its
2338 macros into event serialization and other functions.
2340 You can always use the following tracepoint provider package source
2344 .Tracepoint provider package source file template.
2346 #define TRACEPOINT_CREATE_PROBES
2351 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2352 header file>> name. You may also include more than one tracepoint
2353 provider header file here to create a tracepoint provider package
2354 holding more than one tracepoint providers.
2357 [[probing-the-application-source-code]]
2358 ==== Add tracepoints to an application's source code
2360 Once you <<tpp-header,create a tracepoint provider header file>>, you
2361 can use the `tracepoint()` macro in your application's
2362 source code to insert the tracepoints that this header
2363 <<defining-tracepoints,defines>>.
2365 The `tracepoint()` macro takes at least two parameters: the tracepoint
2366 provider name and the tracepoint name. The corresponding tracepoint
2367 definition defines the other parameters.
2369 .`tracepoint()` usage.
2371 The following <<defining-tracepoints,tracepoint definition>> defines a
2372 tracepoint which takes two input arguments and has two output event
2376 .Tracepoint provider header file.
2378 #include "my-custom-structure.h"
2385 const char*, cmd_name
2388 ctf_string(cmd_name, cmd_name)
2389 ctf_integer(int, number_of_args, argc)
2394 You can refer to this tracepoint definition with the `tracepoint()`
2395 macro in your application's source code like this:
2398 .Application's source file.
2402 int main(int argc, char* argv[])
2404 tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2410 Note how the application's source code includes
2411 the tracepoint provider header file containing the tracepoint
2412 definitions to use, path:{tp.h}.
2415 .`tracepoint()` usage with a complex tracepoint definition.
2417 Consider this complex tracepoint definition, where multiple event
2418 fields refer to the same input arguments in their argument expression
2422 .Tracepoint provider header file.
2424 /* For `struct stat` */
2425 #include <sys/types.h>
2426 #include <sys/stat.h>
2438 ctf_integer(int, my_constant_field, 23 + 17)
2439 ctf_integer(int, my_int_arg_field, my_int_arg)
2440 ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2441 ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2442 my_str_arg[2] + my_str_arg[3])
2443 ctf_string(my_str_arg_field, my_str_arg)
2444 ctf_integer_hex(off_t, size_field, st->st_size)
2445 ctf_float(double, size_dbl_field, (double) st->st_size)
2446 ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2447 size_t, strlen(my_str_arg) / 2)
2452 You can refer to this tracepoint definition with the `tracepoint()`
2453 macro in your application's source code like this:
2456 .Application's source file.
2458 #define TRACEPOINT_DEFINE
2465 stat("/etc/fstab", &s);
2466 tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2472 If you look at the event record that LTTng writes when tracing this
2473 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2474 it should look like this:
2476 .Event record fields
2478 |Field's name |Field's value
2479 |`my_constant_field` |40
2480 |`my_int_arg_field` |23
2481 |`my_int_arg_field2` |529
2483 |`my_str_arg_field` |`Hello, World!`
2484 |`size_field` |0x12d
2485 |`size_dbl_field` |301.0
2486 |`half_my_str_arg_field` |`Hello,`
2490 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2491 compute--they use the call stack, for example. To avoid this
2492 computation when the tracepoint is disabled, you can use the
2493 `tracepoint_enabled()` and `do_tracepoint()` macros.
2495 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2499 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2501 tracepoint_enabled(provider_name, tracepoint_name)
2502 do_tracepoint(provider_name, tracepoint_name, ...)
2507 * `provider_name` with the tracepoint provider name.
2508 * `tracepoint_name` with the tracepoint name.
2510 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2511 `tracepoint_name` from the provider named `provider_name` is enabled
2514 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2515 if the tracepoint is enabled. Using `tracepoint()` with
2516 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2517 the `tracepoint_enabled()` check, thus a race condition is
2518 possible in this situation:
2521 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2523 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2524 stuff = prepare_stuff();
2527 tracepoint(my_provider, my_tracepoint, stuff);
2530 If the tracepoint is enabled after the condition, then `stuff` is not
2531 prepared: the emitted event will either contain wrong data, or the whole
2532 application could crash (segmentation fault, for example).
2534 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2535 `STAP_PROBEV()` call. If you need it, you must emit
2539 [[building-tracepoint-providers-and-user-application]]
2540 ==== Build and link a tracepoint provider package and an application
2542 Once you have one or more <<tpp-header,tracepoint provider header
2543 files>> and a <<tpp-source,tracepoint provider package source file>>,
2544 you can create the tracepoint provider package by compiling its source
2545 file. From here, multiple build and run scenarios are possible. The
2546 following table shows common application and library configurations
2547 along with the required command lines to achieve them.
2549 In the following diagrams, we use the following file names:
2552 Executable application.
2555 Application's object file.
2558 Tracepoint provider package object file.
2561 Tracepoint provider package archive file.
2564 Tracepoint provider package shared object file.
2567 User library object file.
2570 User library shared object file.
2572 We use the following symbols in the diagrams of table below:
2575 .Symbols used in the build scenario diagrams.
2576 image::ust-sit-symbols.png[]
2578 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2579 variable in the following instructions.
2581 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2582 .Common tracepoint provider package scenarios.
2584 |Scenario |Instructions
2587 The instrumented application is statically linked with
2588 the tracepoint provider package object.
2590 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2593 include::../common/ust-sit-step-tp-o.txt[]
2595 To build the instrumented application:
2597 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2602 #define TRACEPOINT_DEFINE
2606 . Compile the application source file:
2615 . Build the application:
2620 $ gcc -o app app.o tpp.o -llttng-ust -ldl
2624 To run the instrumented application:
2626 * Start the application:
2636 The instrumented application is statically linked with the
2637 tracepoint provider package archive file.
2639 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
2642 To create the tracepoint provider package archive file:
2644 . Compile the <<tpp-source,tracepoint provider package source file>>:
2653 . Create the tracepoint provider package archive file:
2658 $ ar rcs tpp.a tpp.o
2662 To build the instrumented application:
2664 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2669 #define TRACEPOINT_DEFINE
2673 . Compile the application source file:
2682 . Build the application:
2687 $ gcc -o app app.o tpp.a -llttng-ust -ldl
2691 To run the instrumented application:
2693 * Start the application:
2703 The instrumented application is linked with the tracepoint provider
2704 package shared object.
2706 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
2709 include::../common/ust-sit-step-tp-so.txt[]
2711 To build the instrumented application:
2713 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2718 #define TRACEPOINT_DEFINE
2722 . Compile the application source file:
2731 . Build the application:
2736 $ gcc -o app app.o -ldl -L. -ltpp
2740 To run the instrumented application:
2742 * Start the application:
2752 The tracepoint provider package shared object is preloaded before the
2753 instrumented application starts.
2755 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
2758 include::../common/ust-sit-step-tp-so.txt[]
2760 To build the instrumented application:
2762 . In path:{app.c}, before including path:{tpp.h}, add the
2768 #define TRACEPOINT_DEFINE
2769 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2773 . Compile the application source file:
2782 . Build the application:
2787 $ gcc -o app app.o -ldl
2791 To run the instrumented application with tracing support:
2793 * Preload the tracepoint provider package shared object and
2794 start the application:
2799 $ LD_PRELOAD=./libtpp.so ./app
2803 To run the instrumented application without tracing support:
2805 * Start the application:
2815 The instrumented application dynamically loads the tracepoint provider
2816 package shared object.
2818 See the <<dlclose-warning,warning about `dlclose()`>>.
2820 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
2823 include::../common/ust-sit-step-tp-so.txt[]
2825 To build the instrumented application:
2827 . In path:{app.c}, before including path:{tpp.h}, add the
2833 #define TRACEPOINT_DEFINE
2834 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2838 . Compile the application source file:
2847 . Build the application:
2852 $ gcc -o app app.o -ldl
2856 To run the instrumented application:
2858 * Start the application:
2868 The application is linked with the instrumented user library.
2870 The instrumented user library is statically linked with the tracepoint
2871 provider package object file.
2873 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
2876 include::../common/ust-sit-step-tp-o-fpic.txt[]
2878 To build the instrumented user library:
2880 . In path:{emon.c}, before including path:{tpp.h}, add the
2886 #define TRACEPOINT_DEFINE
2890 . Compile the user library source file:
2895 $ gcc -I. -fpic -c emon.c
2899 . Build the user library shared object:
2904 $ gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
2908 To build the application:
2910 . Compile the application source file:
2919 . Build the application:
2924 $ gcc -o app app.o -L. -lemon
2928 To run the application:
2930 * Start the application:
2940 The application is linked with the instrumented user library.
2942 The instrumented user library is linked with the tracepoint provider
2943 package shared object.
2945 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
2948 include::../common/ust-sit-step-tp-so.txt[]
2950 To build the instrumented user library:
2952 . In path:{emon.c}, before including path:{tpp.h}, add the
2958 #define TRACEPOINT_DEFINE
2962 . Compile the user library source file:
2967 $ gcc -I. -fpic -c emon.c
2971 . Build the user library shared object:
2976 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
2980 To build the application:
2982 . Compile the application source file:
2991 . Build the application:
2996 $ gcc -o app app.o -L. -lemon
3000 To run the application:
3002 * Start the application:
3012 The tracepoint provider package shared object is preloaded before the
3015 The application is linked with the instrumented user library.
3017 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
3020 include::../common/ust-sit-step-tp-so.txt[]
3022 To build the instrumented user library:
3024 . In path:{emon.c}, before including path:{tpp.h}, add the
3030 #define TRACEPOINT_DEFINE
3031 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3035 . Compile the user library source file:
3040 $ gcc -I. -fpic -c emon.c
3044 . Build the user library shared object:
3049 $ gcc -shared -o libemon.so emon.o -ldl
3053 To build the application:
3055 . Compile the application source file:
3064 . Build the application:
3069 $ gcc -o app app.o -L. -lemon
3073 To run the application with tracing support:
3075 * Preload the tracepoint provider package shared object and
3076 start the application:
3081 $ LD_PRELOAD=./libtpp.so ./app
3085 To run the application without tracing support:
3087 * Start the application:
3097 The application is linked with the instrumented user library.
3099 The instrumented user library dynamically loads the tracepoint provider
3100 package shared object.
3102 See the <<dlclose-warning,warning about `dlclose()`>>.
3104 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3107 include::../common/ust-sit-step-tp-so.txt[]
3109 To build the instrumented user library:
3111 . In path:{emon.c}, before including path:{tpp.h}, add the
3117 #define TRACEPOINT_DEFINE
3118 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3122 . Compile the user library source file:
3127 $ gcc -I. -fpic -c emon.c
3131 . Build the user library shared object:
3136 $ gcc -shared -o libemon.so emon.o -ldl
3140 To build the application:
3142 . Compile the application source file:
3151 . Build the application:
3156 $ gcc -o app app.o -L. -lemon
3160 To run the application:
3162 * Start the application:
3172 The application dynamically loads the instrumented user library.
3174 The instrumented user library is linked with the tracepoint provider
3175 package shared object.
3177 See the <<dlclose-warning,warning about `dlclose()`>>.
3179 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3182 include::../common/ust-sit-step-tp-so.txt[]
3184 To build the instrumented user library:
3186 . In path:{emon.c}, before including path:{tpp.h}, add the
3192 #define TRACEPOINT_DEFINE
3196 . Compile the user library source file:
3201 $ gcc -I. -fpic -c emon.c
3205 . Build the user library shared object:
3210 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3214 To build the application:
3216 . Compile the application source file:
3225 . Build the application:
3230 $ gcc -o app app.o -ldl -L. -lemon
3234 To run the application:
3236 * Start the application:
3246 The application dynamically loads the instrumented user library.
3248 The instrumented user library dynamically loads the tracepoint provider
3249 package shared object.
3251 See the <<dlclose-warning,warning about `dlclose()`>>.
3253 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3256 include::../common/ust-sit-step-tp-so.txt[]
3258 To build the instrumented user library:
3260 . In path:{emon.c}, before including path:{tpp.h}, add the
3266 #define TRACEPOINT_DEFINE
3267 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3271 . Compile the user library source file:
3276 $ gcc -I. -fpic -c emon.c
3280 . Build the user library shared object:
3285 $ gcc -shared -o libemon.so emon.o -ldl
3289 To build the application:
3291 . Compile the application source file:
3300 . Build the application:
3305 $ gcc -o app app.o -ldl -L. -lemon
3309 To run the application:
3311 * Start the application:
3321 The tracepoint provider package shared object is preloaded before the
3324 The application dynamically loads the instrumented user library.
3326 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3329 include::../common/ust-sit-step-tp-so.txt[]
3331 To build the instrumented user library:
3333 . In path:{emon.c}, before including path:{tpp.h}, add the
3339 #define TRACEPOINT_DEFINE
3340 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3344 . Compile the user library source file:
3349 $ gcc -I. -fpic -c emon.c
3353 . Build the user library shared object:
3358 $ gcc -shared -o libemon.so emon.o -ldl
3362 To build the application:
3364 . Compile the application source file:
3373 . Build the application:
3378 $ gcc -o app app.o -L. -lemon
3382 To run the application with tracing support:
3384 * Preload the tracepoint provider package shared object and
3385 start the application:
3390 $ LD_PRELOAD=./libtpp.so ./app
3394 To run the application without tracing support:
3396 * Start the application:
3406 The application is statically linked with the tracepoint provider
3407 package object file.
3409 The application is linked with the instrumented user library.
3411 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3414 include::../common/ust-sit-step-tp-o.txt[]
3416 To build the instrumented user library:
3418 . In path:{emon.c}, before including path:{tpp.h}, add the
3424 #define TRACEPOINT_DEFINE
3428 . Compile the user library source file:
3433 $ gcc -I. -fpic -c emon.c
3437 . Build the user library shared object:
3442 $ gcc -shared -o libemon.so emon.o
3446 To build the application:
3448 . Compile the application source file:
3457 . Build the application:
3462 $ gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3466 To run the instrumented application:
3468 * Start the application:
3478 The application is statically linked with the tracepoint provider
3479 package object file.
3481 The application dynamically loads the instrumented user library.
3483 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3486 include::../common/ust-sit-step-tp-o.txt[]
3488 To build the application:
3490 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3495 #define TRACEPOINT_DEFINE
3499 . Compile the application source file:
3508 . Build the application:
3513 $ gcc -Wl,--export-dynamic -o app app.o tpp.o \
3518 The `--export-dynamic` option passed to the linker is necessary for the
3519 dynamically loaded library to ``see'' the tracepoint symbols defined in
3522 To build the instrumented user library:
3524 . Compile the user library source file:
3529 $ gcc -I. -fpic -c emon.c
3533 . Build the user library shared object:
3538 $ gcc -shared -o libemon.so emon.o
3542 To run the application:
3544 * Start the application:
3556 .Do not use man:dlclose(3) on a tracepoint provider package
3558 Never use man:dlclose(3) on any shared object which:
3560 * Is linked with, statically or dynamically, a tracepoint provider
3562 * Calls man:dlopen(3) itself to dynamically open a tracepoint provider
3563 package shared object.
3565 This is currently considered **unsafe** due to a lack of reference
3566 counting from LTTng-UST to the shared object.
3568 A known workaround (available since glibc 2.2) is to use the
3569 `RTLD_NODELETE` flag when calling man:dlopen(3) initially. This has the
3570 effect of not unloading the loaded shared object, even if man:dlclose(3)
3573 You can also preload the tracepoint provider package shared object with
3574 the env:LD_PRELOAD environment variable to overcome this limitation.
3578 [[using-lttng-ust-with-daemons]]
3579 ===== Use noch:{LTTng-UST} with daemons
3581 If your instrumented application calls man:fork(2), man:clone(2),
3582 or BSD's man:rfork(2), without a following man:exec(3)-family
3583 system call, you must preload the path:{liblttng-ust-fork.so} shared
3584 object when starting the application.
3588 $ LD_PRELOAD=liblttng-ust-fork.so ./my-app
3591 If your tracepoint provider package is
3592 a shared library which you also preload, you must put both
3593 shared objects in env:LD_PRELOAD:
3597 $ LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3601 [[lttng-ust-pkg-config]]
3602 ===== Use noch:{pkg-config}
3604 On some distributions, LTTng-UST ships with a
3605 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3606 metadata file. If this is your case, then you can use cmd:pkg-config to
3607 build an application on the command line:
3611 $ gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3615 [[instrumenting-32-bit-app-on-64-bit-system]]
3616 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3618 In order to trace a 32-bit application running on a 64-bit system,
3619 LTTng must use a dedicated 32-bit
3620 <<lttng-consumerd,consumer daemon>>.
3622 The following steps show how to build and install a 32-bit consumer
3623 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3624 build and install the 32-bit LTTng-UST libraries, and how to build and
3625 link an instrumented 32-bit application in that context.
3627 To build a 32-bit instrumented application for a 64-bit target system,
3628 assuming you have a fresh target system with no installed Userspace RCU
3631 . Download, build, and install a 32-bit version of Userspace RCU:
3636 $ cd $(mktemp -d) &&
3637 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
3638 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
3639 cd userspace-rcu-0.9.* &&
3640 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
3642 sudo make install &&
3647 . Using your distribution's package manager, or from source, install
3648 the following 32-bit versions of the following dependencies of
3649 LTTng-tools and LTTng-UST:
3652 * https://sourceforge.net/projects/libuuid/[libuuid]
3653 * http://directory.fsf.org/wiki/Popt[popt]
3654 * http://www.xmlsoft.org/[libxml2]
3657 . Download, build, and install a 32-bit version of the latest
3658 LTTng-UST{nbsp}{revision}:
3663 $ cd $(mktemp -d) &&
3664 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.8.tar.bz2 &&
3665 tar -xf lttng-ust-latest-2.8.tar.bz2 &&
3666 cd lttng-ust-2.8.* &&
3667 ./configure --libdir=/usr/local/lib32 \
3668 CFLAGS=-m32 CXXFLAGS=-m32 \
3669 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3671 sudo make install &&
3678 Depending on your distribution,
3679 32-bit libraries could be installed at a different location than
3680 `/usr/lib32`. For example, Debian is known to install
3681 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
3683 In this case, make sure to set `LDFLAGS` to all the
3684 relevant 32-bit library paths, for example:
3688 $ LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
3692 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
3693 the 32-bit consumer daemon:
3698 $ cd $(mktemp -d) &&
3699 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.8.tar.bz2 &&
3700 tar -xf lttng-tools-latest-2.8.tar.bz2 &&
3701 cd lttng-tools-2.8.* &&
3702 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
3703 LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' \
3704 --disable-bin-lttng --disable-bin-lttng-crash \
3705 --disable-bin-lttng-relayd --disable-bin-lttng-sessiond &&
3707 cd src/bin/lttng-consumerd &&
3708 sudo make install &&
3713 . From your distribution or from source,
3714 <<installing-lttng,install>> the 64-bit versions of
3715 LTTng-UST and Userspace RCU.
3716 . Download, build, and install the 64-bit version of the
3717 latest LTTng-tools{nbsp}{revision}:
3722 $ cd $(mktemp -d) &&
3723 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.8.tar.bz2 &&
3724 tar -xf lttng-tools-latest-2.8.tar.bz2 &&
3725 cd lttng-tools-2.8.* &&
3726 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
3727 --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
3729 sudo make install &&
3734 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
3735 when linking your 32-bit application:
3738 -m32 -L/usr/lib32 -L/usr/local/lib32 \
3739 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
3742 For example, let's rebuild the quick start example in
3743 <<tracing-your-own-user-application,Trace a user application>> as an
3744 instrumented 32-bit application:
3749 $ gcc -m32 -c -I. hello-tp.c
3750 $ gcc -m32 -c hello.c
3751 $ gcc -m32 -o hello hello.o hello-tp.o \
3752 -L/usr/lib32 -L/usr/local/lib32 \
3753 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
3758 No special action is required to execute the 32-bit application and
3759 to trace it: use the command-line man:lttng(1) tool as usual.
3766 man:tracef(3) is a small LTTng-UST API designed for quick,
3767 man:printf(3)-like instrumentation without the burden of
3768 <<tracepoint-provider,creating>> and
3769 <<building-tracepoint-providers-and-user-application,building>>
3770 a tracepoint provider package.
3772 To use `tracef()` in your application:
3774 . In the C or C++ source files where you need to use `tracef()`,
3775 include `<lttng/tracef.h>`:
3780 #include <lttng/tracef.h>
3784 . In the application's source code, use `tracef()` like you would use
3792 tracef("my message: %d (%s)", my_integer, my_string);
3798 . Link your application with `liblttng-ust`:
3803 $ gcc -o app app.c -llttng-ust
3807 To trace the events that `tracef()` calls emit:
3809 * <<enabling-disabling-events,Create an event rule>> which matches the
3810 `lttng_ust_tracef:*` event name:
3815 $ lttng enable-event --userspace 'lttng_ust_tracef:*'
3820 .Limitations of `tracef()`
3822 The `tracef()` utility function was developed to make user space tracing
3823 super simple, albeit with notable disadvantages compared to
3824 <<defining-tracepoints,user-defined tracepoints>>:
3826 * All the emitted events have the same tracepoint provider and
3827 tracepoint names, respectively `lttng_ust_tracef` and `event`.
3828 * There is no static type checking.
3829 * The only event record field you actually get, named `msg`, is a string
3830 potentially containing the values you passed to `tracef()`
3831 using your own format string. This also means that you cannot filter
3832 events with a custom expression at run time because there are no
3834 * Since `tracef()` uses the C standard library's man:vasprintf(3)
3835 function behind the scenes to format the strings at run time, its
3836 expected performance is lower than with user-defined tracepoints,
3837 which do not require a conversion to a string.
3839 Taking this into consideration, `tracef()` is useful for some quick
3840 prototyping and debugging, but you should not consider it for any
3841 permanent and serious applicative instrumentation.
3847 ==== Use `tracelog()`
3849 The man:tracelog(3) API is very similar to <<tracef,`tracef()`>>, with
3850 the difference that it accepts an additional log level parameter.
3852 The goal of `tracelog()` is to ease the migration from logging to
3855 To use `tracelog()` in your application:
3857 . In the C or C++ source files where you need to use `tracelog()`,
3858 include `<lttng/tracelog.h>`:
3863 #include <lttng/tracelog.h>
3867 . In the application's source code, use `tracelog()` like you would use
3868 man:printf(3), except for the first parameter which is the log
3876 tracelog(TRACE_WARNING, "my message: %d (%s)",
3877 my_integer, my_string);
3883 See man:lttng-ust(3) for a list of available log level names.
3885 . Link your application with `liblttng-ust`:
3890 $ gcc -o app app.c -llttng-ust
3894 To trace the events that `tracelog()` calls emit with a log level
3895 _as severe as_ a specific log level:
3897 * <<enabling-disabling-events,Create an event rule>> which matches the
3898 `lttng_ust_tracelog:*` event name and a minimum level
3904 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
3905 --loglevel=TRACE_WARNING
3909 To trace the events that `tracelog()` calls emit with a
3910 _specific log level_:
3912 * Create an event rule which matches the `lttng_ust_tracelog:*`
3913 event name and a specific log level:
3918 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
3919 --loglevel-only=TRACE_INFO
3924 [[prebuilt-ust-helpers]]
3925 === Prebuilt user space tracing helpers
3927 The LTTng-UST package provides a few helpers in the form of preloadable
3928 shared objects which automatically instrument system functions and
3931 The helper shared objects are normally found in dir:{/usr/lib}. If you
3932 built LTTng-UST <<building-from-source,from source>>, they are probably
3933 located in dir:{/usr/local/lib}.
3935 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
3938 path:{liblttng-ust-libc-wrapper.so}::
3939 path:{liblttng-ust-pthread-wrapper.so}::
3940 <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
3941 memory and POSIX threads function tracing>>.
3943 path:{liblttng-ust-cyg-profile.so}::
3944 path:{liblttng-ust-cyg-profile-fast.so}::
3945 <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
3947 path:{liblttng-ust-dl.so}::
3948 <<liblttng-ust-dl,Dynamic linker tracing>>.
3950 To use a user space tracing helper with any user application:
3952 * Preload the helper shared object when you start the application:
3957 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
3961 You can preload more than one helper:
3966 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
3972 [[liblttng-ust-libc-pthread-wrapper]]
3973 ==== Instrument C standard library memory and POSIX threads functions
3975 The path:{liblttng-ust-libc-wrapper.so} and
3976 path:{liblttng-ust-pthread-wrapper.so} helpers
3977 add instrumentation to some C standard library and POSIX
3981 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
3983 |TP provider name |TP name |Instrumented function
3985 .6+|`lttng_ust_libc` |`malloc` |man:malloc(3)
3986 |`calloc` |man:calloc(3)
3987 |`realloc` |man:realloc(3)
3988 |`free` |man:free(3)
3989 |`memalign` |man:memalign(3)
3990 |`posix_memalign` |man:posix_memalign(3)
3994 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
3996 |TP provider name |TP name |Instrumented function
3998 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
3999 |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
4000 |`pthread_mutex_trylock` |man:pthread_mutex_trylock(3p)
4001 |`pthread_mutex_unlock` |man:pthread_mutex_unlock(3p)
4004 When you preload the shared object, it replaces the functions listed
4005 in the previous tables by wrappers which contain tracepoints and call
4006 the replaced functions.
4009 [[liblttng-ust-cyg-profile]]
4010 ==== Instrument function entry and exit
4012 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
4013 to the entry and exit points of functions.
4015 man:gcc(1) and man:clang(1) have an option named
4016 https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html[`-finstrument-functions`]
4017 which generates instrumentation calls for entry and exit to functions.
4018 The LTTng-UST function tracing helpers,
4019 path:{liblttng-ust-cyg-profile.so} and
4020 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
4021 to add tracepoints to the two generated functions (which contain
4022 `cyg_profile` in their names, hence the helper's name).
4024 To use the LTTng-UST function tracing helper, the source files to
4025 instrument must be built using the `-finstrument-functions` compiler
4028 There are two versions of the LTTng-UST function tracing helper:
4030 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
4031 that you should only use when it can be _guaranteed_ that the
4032 complete event stream is recorded without any lost event record.
4033 Any kind of duplicate information is left out.
4035 Assuming no event record is lost, having only the function addresses on
4036 entry is enough to create a call graph, since an event record always
4037 contains the ID of the CPU that generated it.
4039 You can use a tool like man:addr2line(1) to convert function addresses
4040 back to source file names and line numbers.
4042 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
4043 which also works in use cases where event records might get discarded or
4044 not recorded from application startup.
4045 In these cases, the trace analyzer needs more information to be
4046 able to reconstruct the program flow.
4048 See man:lttng-ust-cyg-profile(3) to learn more about the instrumentation
4049 points of this helper.
4051 All the tracepoints that this helper provides have the
4052 log level `TRACE_DEBUG_FUNCTION` (see man:lttng-ust(3)).
4054 TIP: It's sometimes a good idea to limit the number of source files that
4055 you compile with the `-finstrument-functions` option to prevent LTTng
4056 from writing an excessive amount of trace data at run time. When using
4057 man:gcc(1), you can use the
4058 `-finstrument-functions-exclude-function-list` option to avoid
4059 instrument entries and exits of specific function names.
4064 ==== Instrument the dynamic linker
4066 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4067 man:dlopen(3) and man:dlclose(3) function calls.
4069 See man:lttng-ust-dl(3) to learn more about the instrumentation points
4074 [[java-application]]
4075 === User space Java agent
4077 You can instrument any Java application which uses one of the following
4080 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4081 (JUL) core logging facilities.
4082 * http://logging.apache.org/log4j/1.2/[**Apache log4j 1.2**], since
4083 LTTng 2.6. Note that Apache Log4j{nbsp}2 is not supported.
4086 .LTTng-UST Java agent imported by a Java application.
4087 image::java-app.png[]
4089 Note that the methods described below are new in LTTng{nbsp}2.8.
4090 Previous LTTng versions use another technique.
4092 NOTE: We use http://openjdk.java.net/[OpenJDK]{nbsp}8 for development
4093 and https://ci.lttng.org/[continuous integration], thus this version is
4094 directly supported. However, the LTTng-UST Java agent is also tested
4095 with OpenJDK{nbsp}7.
4100 ==== Use the LTTng-UST Java agent for `java.util.logging`
4102 To use the LTTng-UST Java agent in a Java application which uses
4103 `java.util.logging` (JUL):
4105 . In the Java application's source code, import the LTTng-UST
4106 log handler package for `java.util.logging`:
4111 import org.lttng.ust.agent.jul.LttngLogHandler;
4115 . Create an LTTng-UST JUL log handler:
4120 Handler lttngUstLogHandler = new LttngLogHandler();
4124 . Add this handler to the JUL loggers which should emit LTTng events:
4129 Logger myLogger = Logger.getLogger("some-logger");
4131 myLogger.addHandler(lttngUstLogHandler);
4135 . Use `java.util.logging` log statements and configuration as usual.
4136 The loggers with an attached LTTng-UST log handler can emit
4139 . Before exiting the application, remove the LTTng-UST log handler from
4140 the loggers attached to it and call its `close()` method:
4145 myLogger.removeHandler(lttngUstLogHandler);
4146 lttngUstLogHandler.close();
4150 This is not strictly necessary, but it is recommended for a clean
4151 disposal of the handler's resources.
4153 . Include the LTTng-UST Java agent's common and JUL-specific JAR files,
4154 path:{lttng-ust-agent-common.jar} and path:{lttng-ust-agent-jul.jar},
4156 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4157 path] when you build the Java application.
4159 The JAR files are typically located in dir:{/usr/share/java}.
4161 IMPORTANT: The LTTng-UST Java agent must be
4162 <<installing-lttng,installed>> for the logging framework your
4165 .Use the LTTng-UST Java agent for `java.util.logging`.
4170 import java.io.IOException;
4171 import java.util.logging.Handler;
4172 import java.util.logging.Logger;
4173 import org.lttng.ust.agent.jul.LttngLogHandler;
4177 private static final int answer = 42;
4179 public static void main(String[] argv) throws Exception
4182 Logger logger = Logger.getLogger("jello");
4184 // Create an LTTng-UST log handler
4185 Handler lttngUstLogHandler = new LttngLogHandler();
4187 // Add the LTTng-UST log handler to our logger
4188 logger.addHandler(lttngUstLogHandler);
4191 logger.info("some info");
4192 logger.warning("some warning");
4194 logger.finer("finer information; the answer is " + answer);
4196 logger.severe("error!");
4198 // Not mandatory, but cleaner
4199 logger.removeHandler(lttngUstLogHandler);
4200 lttngUstLogHandler.close();
4209 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4212 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4213 <<enabling-disabling-events,create an event rule>> matching the
4214 `jello` JUL logger, and <<basic-tracing-session-control,start tracing>>:
4219 $ lttng enable-event --jul jello
4223 Run the compiled class:
4227 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4230 <<basic-tracing-session-control,Stop tracing>> and inspect the
4240 In the resulting trace, an <<event,event record>> generated by a Java
4241 application using `java.util.logging` is named `lttng_jul:event` and
4242 has the following fields:
4245 Log record's message.
4251 Name of the class in which the log statement was executed.
4254 Name of the method in which the log statement was executed.
4257 Logging time (timestamp in milliseconds).
4260 Log level integer value.
4263 ID of the thread in which the log statement was executed.
4265 You can use the opt:lttng-enable-event(1):--loglevel or
4266 opt:lttng-enable-event(1):--loglevel-only option of the
4267 man:lttng-enable-event(1) command to target a range of JUL log levels
4268 or a specific JUL log level.
4273 ==== Use the LTTng-UST Java agent for Apache log4j
4275 To use the LTTng-UST Java agent in a Java application which uses
4278 . In the Java application's source code, import the LTTng-UST
4279 log appender package for Apache log4j:
4284 import org.lttng.ust.agent.log4j.LttngLogAppender;
4288 . Create an LTTng-UST log4j log appender:
4293 Appender lttngUstLogAppender = new LttngLogAppender();
4297 . Add this appender to the log4j loggers which should emit LTTng events:
4302 Logger myLogger = Logger.getLogger("some-logger");
4304 myLogger.addAppender(lttngUstLogAppender);
4308 . Use Apache log4j log statements and configuration as usual. The
4309 loggers with an attached LTTng-UST log appender can emit LTTng events.
4311 . Before exiting the application, remove the LTTng-UST log appender from
4312 the loggers attached to it and call its `close()` method:
4317 myLogger.removeAppender(lttngUstLogAppender);
4318 lttngUstLogAppender.close();
4322 This is not strictly necessary, but it is recommended for a clean
4323 disposal of the appender's resources.
4325 . Include the LTTng-UST Java agent's common and log4j-specific JAR
4326 files, path:{lttng-ust-agent-common.jar} and
4327 path:{lttng-ust-agent-log4j.jar}, in the
4328 https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4329 path] when you build the Java application.
4331 The JAR files are typically located in dir:{/usr/share/java}.
4333 IMPORTANT: The LTTng-UST Java agent must be
4334 <<installing-lttng,installed>> for the logging framework your
4337 .Use the LTTng-UST Java agent for Apache log4j.
4342 import org.apache.log4j.Appender;
4343 import org.apache.log4j.Logger;
4344 import org.lttng.ust.agent.log4j.LttngLogAppender;
4348 private static final int answer = 42;
4350 public static void main(String[] argv) throws Exception
4353 Logger logger = Logger.getLogger("jello");
4355 // Create an LTTng-UST log appender
4356 Appender lttngUstLogAppender = new LttngLogAppender();
4358 // Add the LTTng-UST log appender to our logger
4359 logger.addAppender(lttngUstLogAppender);
4362 logger.info("some info");
4363 logger.warn("some warning");
4365 logger.debug("debug information; the answer is " + answer);
4367 logger.fatal("error!");
4369 // Not mandatory, but cleaner
4370 logger.removeAppender(lttngUstLogAppender);
4371 lttngUstLogAppender.close();
4377 Build this example (`$LOG4JPATH` is the path to the Apache log4j JAR
4382 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH Test.java
4385 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4386 <<enabling-disabling-events,create an event rule>> matching the
4387 `jello` log4j logger, and <<basic-tracing-session-control,start tracing>>:
4392 $ lttng enable-event --log4j jello
4396 Run the compiled class:
4400 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH:. Test
4403 <<basic-tracing-session-control,Stop tracing>> and inspect the
4413 In the resulting trace, an <<event,event record>> generated by a Java
4414 application using log4j is named `lttng_log4j:event` and
4415 has the following fields:
4418 Log record's message.
4424 Name of the class in which the log statement was executed.
4427 Name of the method in which the log statement was executed.
4430 Name of the file in which the executed log statement is located.
4433 Line number at which the log statement was executed.
4439 Log level integer value.
4442 Name of the Java thread in which the log statement was executed.
4444 You can use the opt:lttng-enable-event(1):--loglevel or
4445 opt:lttng-enable-event(1):--loglevel-only option of the
4446 man:lttng-enable-event(1) command to target a range of Apache log4j log levels
4447 or a specific log4j log level.
4451 [[java-application-context]]
4452 ==== Provide application-specific context fields in a Java application
4454 A Java application-specific context field is a piece of state provided
4455 by the application which <<adding-context,you can add>>, using the
4456 man:lttng-add-context(1) command, to each <<event,event record>>
4457 produced by the log statements of this application.
4459 For example, a given object might have a current request ID variable.
4460 You can create a context information retriever for this object and
4461 assign a name to this current request ID. You can then, using the
4462 man:lttng-add-context(1) command, add this context field by name to
4463 the JUL or log4j <<channel,channel>>.
4465 To provide application-specific context fields in a Java application:
4467 . In the Java application's source code, import the LTTng-UST
4468 Java agent context classes and interfaces:
4473 import org.lttng.ust.agent.context.ContextInfoManager;
4474 import org.lttng.ust.agent.context.IContextInfoRetriever;
4478 . Create a context information retriever class, that is, a class which
4479 implements the `IContextInfoRetriever` interface:
4484 class MyContextInfoRetriever implements IContextInfoRetriever
4487 public Object retrieveContextInfo(String key)
4489 if (key.equals("intCtx")) {
4491 } else if (key.equals("strContext")) {
4492 return "context value!";
4501 This `retrieveContextInfo()` method is the only member of the
4502 `IContextInfoRetriever` interface. Its role is to return the current
4503 value of a state by name to create a context field. The names of the
4504 context fields and which state variables they return depends on your
4507 All primitive types and objects are supported as context fields.
4508 When `retrieveContextInfo()` returns an object, the context field
4509 serializer calls its `toString()` method to add a string field to
4510 event records. The method can also return `null`, which means that
4511 no context field is available for the required name.
4513 . Register an instance of your context information retriever class to
4514 the context information manager singleton:
4519 IContextInfoRetriever cir = new MyContextInfoRetriever();
4520 ContextInfoManager cim = ContextInfoManager.getInstance();
4521 cim.registerContextInfoRetriever("retrieverName", cir);
4525 . Before exiting the application, remove your context information
4526 retriever from the context information manager singleton:
4531 ContextInfoManager cim = ContextInfoManager.getInstance();
4532 cim.unregisterContextInfoRetriever("retrieverName");
4536 This is not strictly necessary, but it is recommended for a clean
4537 disposal of some manager's resources.
4539 . Build your Java application with LTTng-UST Java agent support as
4540 usual, following the procedure for either the <<jul,JUL>> or
4541 <<log4j,Apache log4j>> framework.
4544 .Provide application-specific context fields in a Java application.
4549 import java.util.logging.Handler;
4550 import java.util.logging.Logger;
4551 import org.lttng.ust.agent.jul.LttngLogHandler;
4552 import org.lttng.ust.agent.context.ContextInfoManager;
4553 import org.lttng.ust.agent.context.IContextInfoRetriever;
4557 // Our context information retriever class
4558 private static class MyContextInfoRetriever
4559 implements IContextInfoRetriever
4562 public Object retrieveContextInfo(String key) {
4563 if (key.equals("intCtx")) {
4565 } else if (key.equals("strContext")) {
4566 return "context value!";
4573 private static final int answer = 42;
4575 public static void main(String args[]) throws Exception
4577 // Get the context information manager instance
4578 ContextInfoManager cim = ContextInfoManager.getInstance();
4580 // Create and register our context information retriever
4581 IContextInfoRetriever cir = new MyContextInfoRetriever();
4582 cim.registerContextInfoRetriever("myRetriever", cir);
4585 Logger logger = Logger.getLogger("jello");
4587 // Create an LTTng-UST log handler
4588 Handler lttngUstLogHandler = new LttngLogHandler();
4590 // Add the LTTng-UST log handler to our logger
4591 logger.addHandler(lttngUstLogHandler);
4594 logger.info("some info");
4595 logger.warning("some warning");
4597 logger.finer("finer information; the answer is " + answer);
4599 logger.severe("error!");
4601 // Not mandatory, but cleaner
4602 logger.removeHandler(lttngUstLogHandler);
4603 lttngUstLogHandler.close();
4604 cim.unregisterContextInfoRetriever("myRetriever");
4613 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4616 <<creating-destroying-tracing-sessions,Create a tracing session>>
4617 and <<enabling-disabling-events,create an event rule>> matching the
4623 $ lttng enable-event --jul jello
4626 <<adding-context,Add the application-specific context fields>> to the
4631 $ lttng add-context --jul --type='$app.myRetriever:intCtx'
4632 $ lttng add-context --jul --type='$app.myRetriever:strContext'
4635 <<basic-tracing-session-control,Start tracing>>:
4642 Run the compiled class:
4646 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4649 <<basic-tracing-session-control,Stop tracing>> and inspect the
4661 [[python-application]]
4662 === User space Python agent
4664 You can instrument a Python 2 or Python 3 application which uses the
4665 standard https://docs.python.org/3/library/logging.html[`logging`]
4668 Each log statement emits an LTTng event once the
4669 application module imports the
4670 <<lttng-ust-agents,LTTng-UST Python agent>> package.
4673 .A Python application importing the LTTng-UST Python agent.
4674 image::python-app.png[]
4676 To use the LTTng-UST Python agent:
4678 . In the Python application's source code, import the LTTng-UST Python
4688 The LTTng-UST Python agent automatically adds its logging handler to the
4689 root logger at import time.
4691 Any log statement that the application executes before this import does
4692 not emit an LTTng event.
4694 IMPORTANT: The LTTng-UST Python agent must be
4695 <<installing-lttng,installed>>.
4697 . Use log statements and logging configuration as usual.
4698 Since the LTTng-UST Python agent adds a handler to the _root_
4699 logger, you can trace any log statement from any logger.
4701 .Use the LTTng-UST Python agent.
4712 logging.basicConfig()
4713 logger = logging.getLogger('my-logger')
4716 logger.debug('debug message')
4717 logger.info('info message')
4718 logger.warn('warn message')
4719 logger.error('error message')
4720 logger.critical('critical message')
4724 if __name__ == '__main__':
4728 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
4729 logging handler which prints to the standard error stream, is not
4730 strictly required for LTTng-UST tracing to work, but in versions of
4731 Python preceding 3.2, you could see a warning message which indicates
4732 that no handler exists for the logger `my-logger`.
4734 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4735 <<enabling-disabling-events,create an event rule>> matching the
4736 `my-logger` Python logger, and <<basic-tracing-session-control,start
4742 $ lttng enable-event --python my-logger
4746 Run the Python script:
4753 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
4763 In the resulting trace, an <<event,event record>> generated by a Python
4764 application is named `lttng_python:event` and has the following fields:
4767 Logging time (string).
4770 Log record's message.
4776 Name of the function in which the log statement was executed.
4779 Line number at which the log statement was executed.
4782 Log level integer value.
4785 ID of the Python thread in which the log statement was executed.
4788 Name of the Python thread in which the log statement was executed.
4790 You can use the opt:lttng-enable-event(1):--loglevel or
4791 opt:lttng-enable-event(1):--loglevel-only option of the
4792 man:lttng-enable-event(1) command to target a range of Python log levels
4793 or a specific Python log level.
4795 When an application imports the LTTng-UST Python agent, the agent tries
4796 to register to a <<lttng-sessiond,session daemon>>. Note that you must
4797 <<start-sessiond,start the session daemon>> _before_ you run the Python
4798 application. If a session daemon is found, the agent tries to register
4799 to it during 5{nbsp}seconds, after which the application continues
4800 without LTTng tracing support. You can override this timeout value with
4801 the env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
4804 If the session daemon stops while a Python application with an imported
4805 LTTng-UST Python agent runs, the agent retries to connect and to
4806 register to a session daemon every 3{nbsp}seconds. You can override this
4807 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
4812 [[proc-lttng-logger-abi]]
4815 The `lttng-tracer` Linux kernel module, part of
4816 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger file
4817 path:{/proc/lttng-logger} when it's loaded. Any application can write
4818 text data to this file to emit an LTTng event.
4821 .An application writes to the LTTng logger file to emit an LTTng event.
4822 image::lttng-logger.png[]
4824 The LTTng logger is the quickest method--not the most efficient,
4825 however--to add instrumentation to an application. It is designed
4826 mostly to instrument shell scripts:
4830 $ echo "Some message, some $variable" > /proc/lttng-logger
4833 Any event that the LTTng logger emits is named `lttng_logger` and
4834 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
4835 other instrumentation points in the kernel tracing domain, **any Unix
4836 user** can <<enabling-disabling-events,create an event rule>> which
4837 matches its event name, not only the root user or users in the
4838 <<tracing-group,tracing group>>.
4840 To use the LTTng logger:
4842 * From any application, write text data to the path:{/proc/lttng-logger}
4845 The `msg` field of `lttng_logger` event records contains the
4848 NOTE: The maximum message length of an LTTng logger event is
4849 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
4850 than one event to contain the remaining data.
4852 You should not use the LTTng logger to trace a user application which
4853 can be instrumented in a more efficient way, namely:
4855 * <<c-application,C and $$C++$$ applications>>.
4856 * <<java-application,Java applications>>.
4857 * <<python-application,Python applications>>.
4859 .Use the LTTng logger.
4864 echo 'Hello, World!' > /proc/lttng-logger
4866 df --human-readable --print-type / > /proc/lttng-logger
4869 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4870 <<enabling-disabling-events,create an event rule>> matching the
4871 `lttng_logger` Linux kernel tracepoint, and
4872 <<basic-tracing-session-control,start tracing>>:
4877 $ lttng enable-event --kernel lttng_logger
4881 Run the Bash script:
4888 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
4899 [[instrumenting-linux-kernel]]
4900 === LTTng kernel tracepoints
4902 NOTE: This section shows how to _add_ instrumentation points to the
4903 Linux kernel. The kernel's subsystems are already thoroughly
4904 instrumented at strategic places for LTTng when you
4905 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
4909 There are two methods to instrument the Linux kernel:
4911 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
4912 tracepoint which uses the `TRACE_EVENT()` API.
4914 Choose this if you want to instrumentation a Linux kernel tree with an
4915 instrumentation point compatible with ftrace, perf, and SystemTap.
4917 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
4918 instrument an out-of-tree kernel module.
4920 Choose this if you don't need ftrace, perf, or SystemTap support.
4924 [[linux-add-lttng-layer]]
4925 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
4927 This section shows how to add an LTTng layer to existing ftrace
4928 instrumentation using the `TRACE_EVENT()` API.
4930 This section does not document the `TRACE_EVENT()` macro. You can
4931 read the following articles to learn more about this API:
4933 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part 1)]
4934 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part 2)]
4935 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part 3)]
4937 The following procedure assumes that your ftrace tracepoints are
4938 correctly defined in their own header and that they are created in
4939 one source file using the `CREATE_TRACE_POINTS` definition.
4941 To add an LTTng layer over an existing ftrace tracepoint:
4943 . Make sure the following kernel configuration options are
4949 * `CONFIG_HIGH_RES_TIMERS`
4950 * `CONFIG_TRACEPOINTS`
4953 . Build the Linux source tree with your custom ftrace tracepoints.
4954 . Boot the resulting Linux image on your target system.
4956 Confirm that the tracepoints exist by looking for their names in the
4957 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
4958 is your subsystem's name.
4960 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
4965 $ cd $(mktemp -d) &&
4966 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.8.tar.bz2 &&
4967 tar -xf lttng-modules-latest-2.8.tar.bz2 &&
4968 cd lttng-modules-2.8.*
4972 . In dir:{instrumentation/events/lttng-module}, relative to the root
4973 of the LTTng-modules source tree, create a header file named
4974 +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
4975 LTTng-modules tracepoint definitions using the LTTng-modules
4978 Start with this template:
4982 .path:{instrumentation/events/lttng-module/my_subsys.h}
4985 #define TRACE_SYSTEM my_subsys
4987 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
4988 #define _LTTNG_MY_SUBSYS_H
4990 #include "../../../probes/lttng-tracepoint-event.h"
4991 #include <linux/tracepoint.h>
4993 LTTNG_TRACEPOINT_EVENT(
4995 * Format is identical to TRACE_EVENT()'s version for the three
4996 * following macro parameters:
4999 TP_PROTO(int my_int, const char *my_string),
5000 TP_ARGS(my_int, my_string),
5002 /* LTTng-modules specific macros */
5004 ctf_integer(int, my_int_field, my_int)
5005 ctf_string(my_bar_field, my_bar)
5009 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
5011 #include "../../../probes/define_trace.h"
5015 The entries in the `TP_FIELDS()` section are the list of fields for the
5016 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
5017 ftrace's `TRACE_EVENT()` macro.
5019 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
5020 complete description of the available `ctf_*()` macros.
5022 . Create the LTTng-modules probe's kernel module C source file,
5023 +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
5028 .path:{probes/lttng-probe-my-subsys.c}
5030 #include <linux/module.h>
5031 #include "../lttng-tracer.h"
5034 * Build-time verification of mismatch between mainline
5035 * TRACE_EVENT() arguments and the LTTng-modules adaptation
5036 * layer LTTNG_TRACEPOINT_EVENT() arguments.
5038 #include <trace/events/my_subsys.h>
5040 /* Create LTTng tracepoint probes */
5041 #define LTTNG_PACKAGE_BUILD
5042 #define CREATE_TRACE_POINTS
5043 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
5045 #include "../instrumentation/events/lttng-module/my_subsys.h"
5047 MODULE_LICENSE("GPL and additional rights");
5048 MODULE_AUTHOR("Your name <your-email>");
5049 MODULE_DESCRIPTION("LTTng my_subsys probes");
5050 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
5051 __stringify(LTTNG_MODULES_MINOR_VERSION) "."
5052 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
5053 LTTNG_MODULES_EXTRAVERSION);
5057 . Edit path:{probes/KBuild} and add your new kernel module object
5058 next to the existing ones:
5062 .path:{probes/KBuild}
5066 obj-m += lttng-probe-module.o
5067 obj-m += lttng-probe-power.o
5069 obj-m += lttng-probe-my-subsys.o
5075 . Build and install the LTTng kernel modules:
5080 $ make KERNELDIR=/path/to/linux
5081 # make modules_install && depmod -a
5085 Replace `/path/to/linux` with the path to the Linux source tree where
5086 you defined and used tracepoints with ftrace's `TRACE_EVENT()` macro.
5088 Note that you can also use the
5089 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
5090 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
5091 C code that need to be executed before the event fields are recorded.
5093 The best way to learn how to use the previous LTTng-modules macros is to
5094 inspect the existing LTTng-modules tracepoint definitions in the
5095 dir:{instrumentation/events/lttng-module} header files. Compare them
5096 with the Linux kernel mainline versions in the
5097 dir:{include/trace/events} directory of the Linux source tree.
5101 [[lttng-tracepoint-event-code]]
5102 ===== Use custom C code to access the data for tracepoint fields
5104 Although we recommended to always use the
5105 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
5106 the arguments and fields of an LTTng-modules tracepoint when possible,
5107 sometimes you need a more complex process to access the data that the
5108 tracer records as event record fields. In other words, you need local
5109 variables and multiple C{nbsp}statements instead of simple
5110 argument-based expressions that you pass to the
5111 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
5113 You can use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
5114 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
5115 a block of C{nbsp}code to be executed before LTTng records the fields.
5116 The structure of this macro is:
5119 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
5121 LTTNG_TRACEPOINT_EVENT_CODE(
5123 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5124 * version for the following three macro parameters:
5127 TP_PROTO(int my_int, const char *my_string),
5128 TP_ARGS(my_int, my_string),
5130 /* Declarations of custom local variables */
5133 unsigned long b = 0;
5134 const char *name = "(undefined)";
5135 struct my_struct *my_struct;
5139 * Custom code which uses both tracepoint arguments
5140 * (in TP_ARGS()) and local variables (in TP_locvar()).
5142 * Local variables are actually members of a structure pointed
5143 * to by the special variable tp_locvar.
5147 tp_locvar->a = my_int + 17;
5148 tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
5149 tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
5150 tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
5151 put_my_struct(tp_locvar->my_struct);
5160 * Format identical to the LTTNG_TRACEPOINT_EVENT()
5161 * version for this, except that tp_locvar members can be
5162 * used in the argument expression parameters of
5163 * the ctf_*() macros.
5166 ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
5167 ctf_integer(int, my_struct_a, tp_locvar->a)
5168 ctf_string(my_string_field, my_string)
5169 ctf_string(my_struct_name, tp_locvar->name)
5174 IMPORTANT: The C code defined in `TP_code()` must not have any side
5175 effects when executed. In particular, the code must not allocate
5176 memory or get resources without deallocating this memory or putting
5177 those resources afterwards.
5180 [[instrumenting-linux-kernel-tracing]]
5181 ==== Load and unload a custom probe kernel module
5183 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
5184 kernel module>> in the kernel before it can emit LTTng events.
5186 To load the default probe kernel modules and a custom probe kernel
5189 * Use the opt:lttng-sessiond(8):--extra-kmod-probes option to give extra
5190 probe modules to load when starting a root <<lttng-sessiond,session
5194 .Load the `my_subsys`, `usb`, and the default probe modules.
5198 # lttng-sessiond --extra-kmod-probes=my_subsys,usb
5203 You only need to pass the subsystem name, not the whole kernel module
5206 To load _only_ a given custom probe kernel module:
5208 * Use the opt:lttng-sessiond(8):--kmod-probes option to give the probe
5209 modules to load when starting a root session daemon:
5212 .Load only the `my_subsys` and `usb` probe modules.
5216 # lttng-sessiond --kmod-probes=my_subsys,usb
5221 To confirm that a probe module is loaded:
5228 $ lsmod | grep lttng_probe_usb
5232 To unload the loaded probe modules:
5234 * Kill the session daemon with `SIGTERM`:
5239 # pkill lttng-sessiond
5243 You can also use man:modprobe(8)'s `--remove` option if the session
5244 daemon terminates abnormally.
5247 [[controlling-tracing]]
5250 Once an application or a Linux kernel is
5251 <<instrumenting,instrumented>> for LTTng tracing,
5254 This section is divided in topics on how to use the various
5255 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
5256 command-line tool>>, to _control_ the LTTng daemons and tracers.
5258 NOTE: In the following subsections, we refer to an man:lttng(1) command
5259 using its man page name. For example, instead of _Run the `create`
5260 command to..._, we use _Run the man:lttng-create(1) command to..._.
5264 === Start a session daemon
5266 In some situations, you need to run a <<lttng-sessiond,session daemon>>
5267 (man:lttng-sessiond(8)) _before_ you can use the man:lttng(1)
5270 You will see the following error when you run a command while no session
5274 Error: No session daemon is available
5277 The only command that automatically runs a session daemon is
5278 man:lttng-create(1), which you use to
5279 <<creating-destroying-tracing-sessions,create a tracing session>>. While
5280 this is most of the time the first operation that you do, sometimes it's
5281 not. Some examples are:
5283 * <<list-instrumentation-points,List the available instrumentation points>>.
5284 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
5286 [[tracing-group]] Each Unix user must have its own running session
5287 daemon to trace user applications. The session daemon that the root user
5288 starts is the only one allowed to control the LTTng kernel tracer. Users
5289 that are part of the _tracing group_ can control the root session
5290 daemon. The default tracing group name is `tracing`; you can set it to
5291 something else with the opt:lttng-sessiond(8):--group option when you
5292 start the root session daemon.
5294 To start a user session daemon:
5296 * Run man:lttng-sessiond(8):
5301 $ lttng-sessiond --daemonize
5305 To start the root session daemon:
5307 * Run man:lttng-sessiond(8) as the root user:
5312 # lttng-sessiond --daemonize
5316 In both cases, remove the opt:lttng-sessiond(8):--daemonize option to
5317 start the session daemon in foreground.
5319 To stop a session daemon, use man:kill(1) on its process ID (standard
5322 Note that some Linux distributions could manage the LTTng session daemon
5323 as a service. In this case, you should use the service manager to
5324 start, restart, and stop session daemons.
5327 [[creating-destroying-tracing-sessions]]
5328 === Create and destroy a tracing session
5330 Almost all the LTTng control operations happen in the scope of
5331 a <<tracing-session,tracing session>>, which is the dialogue between the
5332 <<lttng-sessiond,session daemon>> and you.
5334 To create a tracing session with a generated name:
5336 * Use the man:lttng-create(1) command:
5345 The created tracing session's name is `auto` followed by the
5348 To create a tracing session with a specific name:
5350 * Use the optional argument of the man:lttng-create(1) command:
5355 $ lttng create my-session
5359 Replace `my-session` with the specific tracing session name.
5361 LTTng appends the creation date to the created tracing session's name.
5363 LTTng writes the traces of a tracing session in
5364 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
5365 name of the tracing session. Note that the env:LTTNG_HOME environment
5366 variable defaults to `$HOME` if not set.
5368 To output LTTng traces to a non-default location:
5370 * Use the opt:lttng-create(1):--output option of the man:lttng-create(1) command:
5375 $ lttng create my-session --output=/tmp/some-directory
5379 You may create as many tracing sessions as you wish.
5381 To list all the existing tracing sessions for your Unix user:
5383 * Use the man:lttng-list(1) command:
5392 When you create a tracing session, it is set as the _current tracing
5393 session_. The following man:lttng(1) commands operate on the current
5394 tracing session when you don't specify one:
5396 [role="list-3-cols"]
5412 To change the current tracing session:
5414 * Use the man:lttng-set-session(1) command:
5419 $ lttng set-session new-session
5423 Replace `new-session` by the name of the new current tracing session.
5425 When you are done tracing in a given tracing session, you can destroy
5426 it. This operation frees the resources taken by the tracing session
5427 to destroy; it does not destroy the trace data that LTTng wrote for
5428 this tracing session.
5430 To destroy the current tracing session:
5432 * Use the man:lttng-destroy(1) command:
5442 [[list-instrumentation-points]]
5443 === List the available instrumentation points
5445 The <<lttng-sessiond,session daemon>> can query the running instrumented
5446 user applications and the Linux kernel to get a list of available
5447 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
5448 they are tracepoints and system calls. For the user space tracing
5449 domain, they are tracepoints. For the other tracing domains, they are
5452 To list the available instrumentation points:
5454 * Use the man:lttng-list(1) command with the requested tracing domain's
5458 * opt:lttng-list(1):--kernel: Linux kernel tracepoints (your Unix user
5459 must be a root user, or it must be a member of the
5460 <<tracing-group,tracing group>>).
5461 * opt:lttng-list(1):--kernel with opt:lttng-list(1):--syscall: Linux
5462 kernel system calls (your Unix user must be a root user, or it must be
5463 a member of the tracing group).
5464 * opt:lttng-list(1):--userspace: user space tracepoints.
5465 * opt:lttng-list(1):--jul: `java.util.logging` loggers.
5466 * opt:lttng-list(1):--log4j: Apache log4j loggers.
5467 * opt:lttng-list(1):--python: Python loggers.
5470 .List the available user space tracepoints.
5474 $ lttng list --userspace
5478 .List the available Linux kernel system call tracepoints.
5482 $ lttng list --kernel --syscall
5487 [[enabling-disabling-events]]
5488 === Create and enable an event rule
5490 Once you <<creating-destroying-tracing-sessions,create a tracing
5491 session>>, you can create <<event,event rules>> with the
5492 man:lttng-enable-event(1) command.
5494 You specify each condition with a command-line option. The available
5495 condition options are shown in the following table.
5497 [role="growable",cols="asciidoc,asciidoc,default"]
5498 .Condition command-line options for the man:lttng-enable-event(1) command.
5500 |Option |Description |Applicable tracing domains
5506 . +--probe=__ADDR__+
5507 . +--function=__ADDR__+
5510 Instead of using the default _tracepoint_ instrumentation type, use:
5512 . A Linux system call.
5513 . A Linux https://lwn.net/Articles/132196/[KProbe] (symbol or address).
5514 . The entry and return points of a Linux function (symbol or address).
5518 |First positional argument.
5521 Tracepoint or system call name. In the case of a Linux KProbe or
5522 function, this is a custom name given to the event rule. With the
5523 JUL, log4j, and Python domains, this is a logger name.
5525 With a tracepoint, logger, or system call name, the last character
5526 can be `*` to match anything that remains.
5533 . +--loglevel=__LEVEL__+
5534 . +--loglevel-only=__LEVEL__+
5537 . Match only tracepoints or log statements with a logging level at
5538 least as severe as +__LEVEL__+.
5539 . Match only tracepoints or log statements with a logging level
5540 equal to +__LEVEL__+.
5542 See man:lttng-enable-event(1) for the list of available logging level
5545 |User space, JUL, log4j, and Python.
5547 |+--exclude=__EXCLUSIONS__+
5550 When you use a `*` character at the end of the tracepoint or logger
5551 name (first positional argument), exclude the specific names in the
5552 comma-delimited list +__EXCLUSIONS__+.
5555 User space, JUL, log4j, and Python.
5557 |+--filter=__EXPR__+
5560 Match only events which satisfy the expression +__EXPR__+.
5562 See man:lttng-enable-event(1) to learn more about the syntax of a
5569 You attach an event rule to a <<channel,channel>> on creation. If you do
5570 not specify the channel with the opt:lttng-enable-event(1):--channel
5571 option, and if the event rule to create is the first in its
5572 <<domain,tracing domain>> for a given tracing session, then LTTng
5573 creates a _default channel_ for you. This default channel is reused in
5574 subsequent invocations of the man:lttng-enable-event(1) command for the
5575 same tracing domain.
5577 An event rule is always enabled at creation time.
5579 The following examples show how you can combine the previous
5580 command-line options to create simple to more complex event rules.
5582 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5586 $ lttng enable-event --kernel sched_switch
5590 .Create an event rule matching four Linux kernel system calls (default channel).
5594 $ lttng enable-event --kernel --syscall open,write,read,close
5598 .Create event rules matching tracepoints with filter expressions (default channel).
5602 $ lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
5607 $ lttng enable-event --kernel --all \
5608 --filter='$ctx.tid == 1988 || $ctx.tid == 1534'
5613 $ lttng enable-event --jul my_logger \
5614 --filter='$app.retriever:cur_msg_id > 3'
5617 IMPORTANT: Make sure to always quote the filter string when you
5618 use man:lttng(1) from a shell.
5621 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
5625 $ lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
5628 IMPORTANT: Make sure to always quote the wildcard character when you
5629 use man:lttng(1) from a shell.
5632 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
5636 $ lttng enable-event --python my-app.'*' \
5637 --exclude='my-app.module,my-app.hello'
5641 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
5645 $ lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
5649 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
5653 $ lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
5657 The event rules of a given channel form a whitelist: as soon as an
5658 emitted event passes one of them, LTTng can record the event. For
5659 example, an event named `my_app:my_tracepoint` emitted from a user space
5660 tracepoint with a `TRACE_ERROR` log level passes both of the following
5665 $ lttng enable-event --userspace my_app:my_tracepoint
5666 $ lttng enable-event --userspace my_app:my_tracepoint \
5667 --loglevel=TRACE_INFO
5670 The second event rule is redundant: the first one includes
5674 [[disable-event-rule]]
5675 === Disable an event rule
5677 To disable an event rule that you <<enabling-disabling-events,created>>
5678 previously, use the man:lttng-disable-event(1) command. This command
5679 disables _all_ the event rules (of a given tracing domain and channel)
5680 which match an instrumentation point. The other conditions are not
5681 supported as of LTTng{nbsp}{revision}.
5683 The LTTng tracer does not record an emitted event which passes
5684 a _disabled_ event rule.
5686 .Disable an event rule matching a Python logger (default channel).
5690 $ lttng disable-event --python my-logger
5694 .Disable an event rule matching all `java.util.logging` loggers (default channel).
5698 $ lttng disable-event --jul '*'
5702 .Disable _all_ the event rules of the default channel.
5704 The opt:lttng-disable-event(1):--all-events option is not, like the
5705 opt:lttng-enable-event(1):--all option of man:lttng-enable-event(1), the
5706 equivalent of the event name `*` (wildcard): it disables _all_ the event
5707 rules of a given channel.
5711 $ lttng disable-event --jul --all-events
5715 NOTE: You cannot delete an event rule once you create it.
5719 === Get the status of a tracing session
5721 To get the status of the current tracing session, that is, its
5722 parameters, its channels, event rules, and their attributes:
5724 * Use the man:lttng-status(1) command:
5734 To get the status of any tracing session:
5736 * Use the man:lttng-list(1) command with the tracing session's name:
5741 $ lttng list my-session
5745 Replace `my-session` with the desired tracing session's name.
5748 [[basic-tracing-session-control]]
5749 === Start and stop a tracing session
5751 Once you <<creating-destroying-tracing-sessions,create a tracing
5753 <<enabling-disabling-events,create one or more event rules>>,
5754 you can start and stop the tracers for this tracing session.
5756 To start tracing in the current tracing session:
5758 * Use the man:lttng-start(1) command:
5767 LTTng is very flexible: you can launch user applications before
5768 or after the you start the tracers. The tracers only record the events
5769 if they pass enabled event rules and if they occur while the tracers are
5772 To stop tracing in the current tracing session:
5774 * Use the man:lttng-stop(1) command:
5783 If there were <<channel-overwrite-mode-vs-discard-mode,lost event
5784 records>> or lost sub-buffers since the last time you ran
5785 man:lttng-start(1), warnings are printed when you run the
5786 man:lttng-stop(1) command.
5789 [[enabling-disabling-channels]]
5790 === Create a channel
5792 Once you create a tracing session, you can create a <<channel,channel>>
5793 with the man:lttng-enable-channel(1) command.
5795 Note that LTTng automatically creates a default channel when, for a
5796 given <<domain,tracing domain>>, no channels exist and you
5797 <<enabling-disabling-events,create>> the first event rule. This default
5798 channel is named `channel0` and its attributes are set to reasonable
5799 values. Therefore, you only need to create a channel when you need
5800 non-default attributes.
5802 You specify each non-default channel attribute with a command-line
5803 option when you use the man:lttng-enable-channel(1) command. The
5804 available command-line options are:
5806 [role="growable",cols="asciidoc,asciidoc"]
5807 .Command-line options for the man:lttng-enable-channel(1) command.
5809 |Option |Description
5815 <<channel-overwrite-mode-vs-discard-mode,event loss mode>> instead of
5816 the default _discard_ mode.
5818 |`--buffers-pid` (user space tracing domain only)
5821 Use the per-process <<channel-buffering-schemes,buffering scheme>>
5822 instead of the default per-user buffering scheme.
5824 |+--subbuf-size=__SIZE__+
5827 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
5828 either for each Unix user (default), or for each instrumented process.
5830 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5832 |+--num-subbuf=__COUNT__+
5835 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
5836 for each Unix user (default), or for each instrumented process.
5838 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5840 |+--tracefile-size=__SIZE__+
5843 Set the maximum size of each trace file that this channel writes within
5844 a stream to +__SIZE__+ bytes instead of no maximum.
5846 See <<tracefile-rotation,Trace file count and size>>.
5848 |+--tracefile-count=__COUNT__+
5851 Limit the number of trace files that this channel creates to
5852 +__COUNT__+ channels instead of no limit.
5854 See <<tracefile-rotation,Trace file count and size>>.
5856 |+--switch-timer=__PERIODUS__+
5859 Set the <<channel-switch-timer,switch timer period>>
5860 to +__PERIODUS__+{nbsp}µs.
5862 |+--read-timer=__PERIODUS__+
5865 Set the <<channel-read-timer,read timer period>>
5866 to +__PERIODUS__+{nbsp}µs.
5868 |+--output=__TYPE__+ (Linux kernel tracing domain only)
5871 Set the channel's output type to +__TYPE__+, either `mmap` or `splice`.
5875 You can only create a channel in the Linux kernel and user space
5876 <<domain,tracing domains>>: other tracing domains have their own channel
5877 created on the fly when <<enabling-disabling-events,creating event
5882 Because of a current LTTng limitation, you must create all channels
5883 _before_ you <<basic-tracing-session-control,start tracing>> in a given
5884 tracing session, that is, before the first time you run
5887 Since LTTng automatically creates a default channel when you use the
5888 man:lttng-enable-event(1) command with a specific tracing domain, you
5889 cannot, for example, create a Linux kernel event rule, start tracing,
5890 and then create a user space event rule, because no user space channel
5891 exists yet and it's too late to create one.
5893 For this reason, make sure to configure your channels properly
5894 before starting the tracers for the first time!
5897 The following examples show how you can combine the previous
5898 command-line options to create simple to more complex channels.
5900 .Create a Linux kernel channel with default attributes.
5904 $ lttng enable-channel --kernel my-channel
5908 .Create a user space channel with 4 sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
5912 $ lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
5913 --buffers-pid my-channel
5917 .Create a Linux kernel channel which rotates 8 trace files of 4{nbsp}MiB each for each stream
5921 $ lttng enable-channel --kernel --tracefile-count=8 \
5922 --tracefile-size=4194304 my-channel
5926 .Create a user space channel in overwrite (or _flight recorder_) mode.
5930 $ lttng enable-channel --userspace --overwrite my-channel
5934 You can <<enabling-disabling-events,create>> the same event rule in
5935 two different channels:
5939 $ lttng enable-event --userspace --channel=my-channel app:tp
5940 $ lttng enable-event --userspace --channel=other-channel app:tp
5943 If both channels are enabled, when a tracepoint named `app:tp` is
5944 reached, LTTng records two events, one for each channel.
5948 === Disable a channel
5950 To disable a specific channel that you <<enabling-disabling-channels,created>>
5951 previously, use the man:lttng-disable-channel(1) command.
5953 .Disable a specific Linux kernel channel.
5957 $ lttng disable-channel --kernel my-channel
5961 The state of a channel precedes the individual states of event rules
5962 attached to it: event rules which belong to a disabled channel, even if
5963 they are enabled, are also considered disabled.
5967 === Add context fields to a channel
5969 Event record fields in trace files provide important information about
5970 events that occured previously, but sometimes some external context may
5971 help you solve a problem faster. Examples of context fields are:
5973 * The **process ID**, **thread ID**, **process name**, and
5974 **process priority** of the thread in which the event occurs.
5975 * The **hostname** of the system on which the event occurs.
5976 * The current values of many possible **performance counters** using
5978 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
5980 ** Branch instructions, misses, and loads.
5982 * Any context defined at the application level (supported for the
5983 JUL and log4j <<domain,tracing domains>>).
5985 To get the full list of available context fields, see
5986 `lttng add-context --list`. Some context fields are reserved for a
5987 specific <<domain,tracing domain>> (Linux kernel or user space).
5989 You add context fields to <<channel,channels>>. All the events
5990 that a channel with added context fields records contain those fields.
5992 To add context fields to one or all the channels of a given tracing
5995 * Use the man:lttng-add-context(1) command.
5997 .Add context fields to all the channels of the current tracing session.
5999 The following command line adds the virtual process identifier and
6000 the per-thread CPU cycles count fields to all the user space channels
6001 of the current tracing session.
6005 $ lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
6009 .Add a context field to a specific channel.
6011 The following command line adds the thread identifier context field
6012 to the Linux kernel channel named `my-channel` in the current
6017 $ lttng add-context --kernel --channel=my-channel --type=tid
6021 .Add an application-specific context field to a specific channel.
6023 The following command line adds the `cur_msg_id` context field of the
6024 `retriever` context retriever for all the instrumented
6025 <<java-application,Java applications>> recording <<event,event records>>
6026 in the channel named `my-channel`:
6030 $ lttng add-context --kernel --channel=my-channel \
6031 --type='$app:retriever:cur_msg_id'
6034 IMPORTANT: Make sure to always quote the `$` character when you
6035 use man:lttng-add-context(1) from a shell.
6038 NOTE: You cannot remove context fields from a channel once you add it.
6043 === Track process IDs
6045 It's often useful to allow only specific process IDs (PIDs) to emit
6046 events. For example, you may wish to record all the system calls made by
6047 a given process (à la http://linux.die.net/man/1/strace[strace]).
6049 The man:lttng-track(1) and man:lttng-untrack(1) commands serve this
6050 purpose. Both commands operate on a whitelist of process IDs. You _add_
6051 entries to this whitelist with the man:lttng-track(1) command and remove
6052 entries with the man:lttng-untrack(1) command. Any process which has one
6053 of the PIDs in the whitelist is allowed to emit LTTng events which pass
6054 an enabled <<event,event rule>>.
6056 NOTE: The PID tracker tracks the _numeric process IDs_. Should a
6057 process with a given tracked ID exit and another process be given this
6058 ID, then the latter would also be allowed to emit events.
6060 .Track and untrack process IDs.
6062 For the sake of the following example, assume the target system has 16
6066 <<creating-destroying-tracing-sessions,create a tracing session>>,
6067 the whitelist contains all the possible PIDs:
6070 .All PIDs are tracked.
6071 image::track-all.png[]
6073 When the whitelist is full and you use the man:lttng-track(1) command to
6074 specify some PIDs to track, LTTng first clears the whitelist, then it
6075 tracks the specific PIDs. After:
6079 $ lttng track --pid=3,4,7,10,13
6085 .PIDs 3, 4, 7, 10, and 13 are tracked.
6086 image::track-3-4-7-10-13.png[]
6088 You can add more PIDs to the whitelist afterwards:
6092 $ lttng track --pid=1,15,16
6098 .PIDs 1, 15, and 16 are added to the whitelist.
6099 image::track-1-3-4-7-10-13-15-16.png[]
6101 The man:lttng-untrack(1) command removes entries from the PID tracker's
6102 whitelist. Given the previous example, the following command:
6106 $ lttng untrack --pid=3,7,10,13
6109 leads to this whitelist:
6112 .PIDs 3, 7, 10, and 13 are removed from the whitelist.
6113 image::track-1-4-15-16.png[]
6115 LTTng can track all possible PIDs again using the
6116 opt:lttng-track(1):--all option:
6120 $ lttng track --pid --all
6123 The result is, again:
6126 .All PIDs are tracked.
6127 image::track-all.png[]
6130 .Track only specific PIDs
6132 A very typical use case with PID tracking is to start with an empty
6133 whitelist, then <<basic-tracing-session-control,start the tracers>>, and
6134 then add PIDs manually while tracers are active. You can accomplish this
6135 by using the opt:lttng-untrack(1):--all option of the
6136 man:lttng-untrack(1) command to clear the whitelist after you
6137 <<creating-destroying-tracing-sessions,create a tracing session>>:
6141 $ lttng untrack --pid --all
6147 .No PIDs are tracked.
6148 image::untrack-all.png[]
6150 If you trace with this whitelist configuration, the tracer records no
6151 events for this <<domain,tracing domain>> because no processes are
6152 tracked. You can use the man:lttng-track(1) command as usual to track
6153 specific PIDs, for example:
6157 $ lttng track --pid=6,11
6163 .PIDs 6 and 11 are tracked.
6164 image::track-6-11.png[]
6169 [[saving-loading-tracing-session]]
6170 === Save and load tracing session configurations
6172 Configuring a <<tracing-session,tracing session>> can be long. Some of
6173 the tasks involved are:
6175 * <<enabling-disabling-channels,Create channels>> with
6176 specific attributes.
6177 * <<adding-context,Add context fields>> to specific channels.
6178 * <<enabling-disabling-events,Create event rules>> with specific log
6179 level and filter conditions.
6181 If you use LTTng to solve real world problems, chances are you have to
6182 record events using the same tracing session setup over and over,
6183 modifying a few variables each time in your instrumented program
6184 or environment. To avoid constant tracing session reconfiguration,
6185 the man:lttng(1) command-line tool can save and load tracing session
6186 configurations to/from XML files.
6188 To save a given tracing session configuration:
6190 * Use the man:lttng-save(1) command:
6195 $ lttng save my-session
6199 Replace `my-session` with the name of the tracing session to save.
6201 LTTng saves tracing session configurations to
6202 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
6203 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
6204 the opt:lttng-save(1):--output-path option to change this destination
6207 LTTng saves all configuration parameters, for example:
6209 * The tracing session name.
6210 * The trace data output path.
6211 * The channels with their state and all their attributes.
6212 * The context fields you added to channels.
6213 * The event rules with their state, log level and filter conditions.
6215 To load a tracing session:
6217 * Use the man:lttng-load(1) command:
6222 $ lttng load my-session
6226 Replace `my-session` with the name of the tracing session to load.
6228 When LTTng loads a configuration, it restores your saved tracing session
6229 as if you just configured it manually.
6231 See man:lttng(1) for the complete list of command-line options. You
6232 can also save and load all many sessions at a time, and decide in which
6233 directory to output the XML files.
6236 [[sending-trace-data-over-the-network]]
6237 === Send trace data over the network
6239 LTTng can send the recorded trace data to a remote system over the
6240 network instead of writing it to the local file system.
6242 To send the trace data over the network:
6244 . On the _remote_ system (which can also be the target system),
6245 start an LTTng <<lttng-relayd,relay daemon>> (man:lttng-relayd(8)):
6254 . On the _target_ system, create a tracing session configured to
6255 send trace data over the network:
6260 $ lttng create my-session --set-url=net://remote-system
6264 Replace `remote-system` by the host name or IP address of the
6265 remote system. See man:lttng-create(1) for the exact URL format.
6267 . On the target system, use the man:lttng(1) command-line tool as usual.
6268 When tracing is active, the target's consumer daemon sends sub-buffers
6269 to the relay daemon running on the remote system instead of flushing
6270 them to the local file system. The relay daemon writes the received
6271 packets to the local file system.
6273 The relay daemon writes trace files to
6274 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
6275 +__hostname__+ is the host name of the target system and +__session__+
6276 is the tracing session name. Note that the env:LTTNG_HOME environment
6277 variable defaults to `$HOME` if not set. Use the
6278 opt:lttng-relayd(8):--output option of man:lttng-relayd(8) to write
6279 trace files to another base directory.
6284 === View events as LTTng emits them (noch:{LTTng} live)
6286 LTTng live is a network protocol implemented by the <<lttng-relayd,relay
6287 daemon>> (man:lttng-relayd(8)) to allow compatible trace viewers to
6288 display events as LTTng emits them on the target system while tracing is
6291 The relay daemon creates a _tee_: it forwards the trace data to both
6292 the local file system and to connected live viewers:
6295 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
6300 . On the _target system_, create a <<tracing-session,tracing session>>
6306 $ lttng create my-session --live
6310 This spawns a local relay daemon.
6312 . Start the live viewer and configure it to connect to the relay
6313 daemon. For example, with http://diamon.org/babeltrace[Babeltrace]:
6318 $ babeltrace --input-format=lttng-live
6319 net://localhost/host/hostname/my-session
6326 * `hostname` with the host name of the target system.
6327 * `my-session` with the name of the tracing session to view.
6330 . Configure the tracing session as usual with the man:lttng(1)
6331 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6333 You can list the available live tracing sessions with Babeltrace:
6337 $ babeltrace --input-format=lttng-live net://localhost
6340 You can start the relay daemon on another system. In this case, you need
6341 to specify the relay daemon's URL when you create the tracing session
6342 with the opt:lttng-create(1):--set-url option. You also need to replace
6343 `localhost` in the procedure above with the host name of the system on
6344 which the relay daemon is running.
6346 See man:lttng-create(1) and man:lttng-relayd(8) for the complete list of
6347 command-line options.
6351 [[taking-a-snapshot]]
6352 === Take a snapshot of the current sub-buffers of a tracing session
6354 The normal behavior of LTTng is to append full sub-buffers to growing
6355 trace data files. This is ideal to keep a full history of the events
6356 that occurred on the target system, but it can
6357 represent too much data in some situations. For example, you may wish
6358 to trace your application continuously until some critical situation
6359 happens, in which case you only need the latest few recorded
6360 events to perform the desired analysis, not multi-gigabyte trace files.
6362 With the man:lttng-snapshot(1) command, you can take a snapshot of the
6363 current sub-buffers of a given <<tracing-session,tracing session>>.
6364 LTTng can write the snapshot to the local file system or send it over
6369 . Create a tracing session in _snapshot mode_:
6374 $ lttng create my-session --snapshot
6378 The <<channel-overwrite-mode-vs-discard-mode,event loss mode>> of
6379 <<channel,channels>> created in this mode is automatically set to
6380 _overwrite_ (flight recorder mode).
6382 . Configure the tracing session as usual with the man:lttng(1)
6383 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6385 . **Optional**: When you need to take a snapshot,
6386 <<basic-tracing-session-control,stop tracing>>.
6388 You can take a snapshot when the tracers are active, but if you stop
6389 them first, you are sure that the data in the sub-buffers does not
6390 change before you actually take the snapshot.
6397 $ lttng snapshot record --name=my-first-snapshot
6401 LTTng writes the current sub-buffers of all the current tracing
6402 session's channels to trace files on the local file system. Those trace
6403 files have `my-first-snapshot` in their name.
6405 There is no difference between the format of a normal trace file and the
6406 format of a snapshot: viewers of LTTng traces also support LTTng
6409 By default, LTTng writes snapshot files to the path shown by
6410 `lttng snapshot list-output`. You can change this path or decide to send
6411 snapshots over the network using either:
6413 . An output path or URL that you specify when you create the
6415 . An snapshot output path or URL that you add using
6416 `lttng snapshot add-output`
6417 . An output path or URL that you provide directly to the
6418 `lttng snapshot record` command.
6420 Method 3 overrides method 2, which overrides method 1. When you
6421 specify a URL, a relay daemon must listen on a remote system (see
6422 <<sending-trace-data-over-the-network,Send trace data over the network>>).
6427 === Use the machine interface
6429 With any command of the man:lttng(1) command-line tool, you can set the
6430 opt:lttng(1):--mi option to `xml` (before the command name) to get an
6431 XML machine interface output, for example:
6435 $ lttng --mi=xml enable-event --kernel --syscall open
6438 A schema definition (XSD) is
6439 https://github.com/lttng/lttng-tools/blob/stable-2.8/src/common/mi-lttng-3.0.xsd[available]
6440 to ease the integration with external tools as much as possible.
6444 [[metadata-regenerate]]
6445 === Regenerate the metadata of an LTTng trace
6447 An LTTng trace, which is a http://diamon.org/ctf[CTF] trace, has both
6448 data stream files and a metadata file. This metadata file contains,
6449 amongst other things, information about the offset of the clock sources
6450 used to timestamp <<event,event records>> when tracing.
6452 If, once a <<tracing-session,tracing session>> is
6453 <<basic-tracing-session-control,started>>, a major
6454 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
6455 happens, the trace's clock offset also needs to be updated. You
6456 can use the man:lttng-metadata(1) command to do so.
6458 The main use case of this command is to allow a system to boot with
6459 an incorrect wall time and trace it with LTTng before its wall time
6460 is corrected. Once the system is known to be in a state where its
6461 wall time is correct, it can run `lttng metadata regenerate`.
6463 To regenerate the metadata of an LTTng trace:
6465 * Use the `regenerate` action of the man:lttng-metadata(1) command:
6470 $ lttng metadata regenerate
6476 `lttng metadata regenerate` has the following limitations:
6478 * Tracing session <<creating-destroying-tracing-sessions,created>>
6480 * User space <<channel,channels>>, if any, using
6481 <<channel-buffering-schemes,per-user buffering>>.
6486 [[persistent-memory-file-systems]]
6487 === Record trace data on persistent memory file systems
6489 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
6490 (NVRAM) is random-access memory that retains its information when power
6491 is turned off (non-volatile). Systems with such memory can store data
6492 structures in RAM and retrieve them after a reboot, without flushing
6493 to typical _storage_.
6495 Linux supports NVRAM file systems thanks to either
6496 http://pramfs.sourceforge.net/[PRAMFS] or
6497 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
6498 (requires Linux 4.1+).
6500 This section does not describe how to operate such file systems;
6501 we assume that you have a working persistent memory file system.
6503 When you create a <<tracing-session,tracing session>>, you can specify
6504 the path of the shared memory holding the sub-buffers. If you specify a
6505 location on an NVRAM file system, then you can retrieve the latest
6506 recorded trace data when the system reboots after a crash.
6508 To record trace data on a persistent memory file system and retrieve the
6509 trace data after a system crash:
6511 . Create a tracing session with a sub-buffer shared memory path located
6512 on an NVRAM file system:
6517 $ lttng create my-session --shm-path=/path/to/shm
6521 . Configure the tracing session as usual with the man:lttng(1)
6522 command-line tool, and <<basic-tracing-session-control,start tracing>>.
6524 . After a system crash, use the man:lttng-crash(1) command-line tool to
6525 view the trace data recorded on the NVRAM file system:
6530 $ lttng-crash /path/to/shm
6534 The binary layout of the ring buffer files is not exactly the same as
6535 the trace files layout. This is why you need to use man:lttng-crash(1)
6536 instead of your preferred trace viewer directly.
6538 To convert the ring buffer files to LTTng trace files:
6540 * Use the opt:lttng-crash(1):--extract option of man:lttng-crash(1):
6545 $ lttng-crash --extract=/path/to/trace /path/to/shm
6553 [[lttng-modules-ref]]
6554 === noch:{LTTng-modules}
6557 [[lttng-modules-tp-fields]]
6558 ==== Tracepoint fields macros (for `TP_FIELDS()`)
6560 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
6561 tracepoint fields, which must be listed within `TP_FIELDS()` in
6562 `LTTNG_TRACEPOINT_EVENT()`, are:
6564 [role="func-desc growable",cols="asciidoc,asciidoc"]
6565 .Available macros to define LTTng-modules tracepoint fields
6567 |Macro |Description and parameters
6570 +ctf_integer(__t__, __n__, __e__)+
6572 +ctf_integer_nowrite(__t__, __n__, __e__)+
6574 +ctf_user_integer(__t__, __n__, __e__)+
6576 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
6578 Standard integer, displayed in base 10.
6581 Integer C type (`int`, `long`, `size_t`, ...).
6587 Argument expression.
6590 +ctf_integer_hex(__t__, __n__, __e__)+
6592 +ctf_user_integer_hex(__t__, __n__, __e__)+
6594 Standard integer, displayed in base 16.
6603 Argument expression.
6605 |+ctf_integer_oct(__t__, __n__, __e__)+
6607 Standard integer, displayed in base 8.
6616 Argument expression.
6619 +ctf_integer_network(__t__, __n__, __e__)+
6621 +ctf_user_integer_network(__t__, __n__, __e__)+
6623 Integer in network byte order (big-endian), displayed in base 10.
6632 Argument expression.
6635 +ctf_integer_network_hex(__t__, __n__, __e__)+
6637 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
6639 Integer in network byte order, displayed in base 16.
6648 Argument expression.
6651 +ctf_string(__n__, __e__)+
6653 +ctf_string_nowrite(__n__, __e__)+
6655 +ctf_user_string(__n__, __e__)+
6657 +ctf_user_string_nowrite(__n__, __e__)+
6659 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
6665 Argument expression.
6668 +ctf_array(__t__, __n__, __e__, __s__)+
6670 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
6672 +ctf_user_array(__t__, __n__, __e__, __s__)+
6674 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
6676 Statically-sized array of integers.
6679 Array element C type.
6685 Argument expression.
6691 +ctf_array_bitfield(__t__, __n__, __e__, __s__)+
6693 +ctf_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
6695 +ctf_user_array_bitfield(__t__, __n__, __e__, __s__)+
6697 +ctf_user_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
6699 Statically-sized array of bits.
6701 The type of +__e__+ must be an integer type. +__s__+ is the number
6702 of elements of such type in +__e__+, not the number of bits.
6705 Array element C type.
6711 Argument expression.
6717 +ctf_array_text(__t__, __n__, __e__, __s__)+
6719 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
6721 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
6723 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
6725 Statically-sized array, printed as text.
6727 The string does not need to be null-terminated.
6730 Array element C type (always `char`).
6736 Argument expression.
6742 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
6744 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
6746 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
6748 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
6750 Dynamically-sized array of integers.
6752 The type of +__E__+ must be unsigned.
6755 Array element C type.
6761 Argument expression.
6764 Length expression C type.
6770 +ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
6772 +ctf_user_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
6774 Dynamically-sized array of integers, displayed in base 16.
6776 The type of +__E__+ must be unsigned.
6779 Array element C type.
6785 Argument expression.
6788 Length expression C type.
6793 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
6795 Dynamically-sized array of integers in network byte order (big-endian),
6796 displayed in base 10.
6798 The type of +__E__+ must be unsigned.
6801 Array element C type.
6807 Argument expression.
6810 Length expression C type.
6816 +ctf_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
6818 +ctf_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
6820 +ctf_user_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
6822 +ctf_user_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
6824 Dynamically-sized array of bits.
6826 The type of +__e__+ must be an integer type. +__s__+ is the number
6827 of elements of such type in +__e__+, not the number of bits.
6829 The type of +__E__+ must be unsigned.
6832 Array element C type.
6838 Argument expression.
6841 Length expression C type.
6847 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
6849 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
6851 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
6853 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
6855 Dynamically-sized array, displayed as text.
6857 The string does not need to be null-terminated.
6859 The type of +__E__+ must be unsigned.
6861 The behaviour is undefined if +__e__+ is `NULL`.
6864 Sequence element C type (always `char`).
6870 Argument expression.
6873 Length expression C type.
6879 Use the `_user` versions when the argument expression, `e`, is
6880 a user space address. In the cases of `ctf_user_integer*()` and
6881 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
6884 The `_nowrite` versions omit themselves from the session trace, but are
6885 otherwise identical. This means the `_nowrite` fields won't be written
6886 in the recorded trace. Their primary purpose is to make some
6887 of the event context available to the
6888 <<enabling-disabling-events,event filters>> without having to
6889 commit the data to sub-buffers.
6895 Terms related to LTTng and to tracing in general:
6898 The http://diamon.org/babeltrace[Babeltrace] project, which includes
6899 the cmd:babeltrace command, some libraries, and Python bindings.
6901 <<channel-buffering-schemes,buffering scheme>>::
6902 A layout of sub-buffers applied to a given channel.
6904 <<channel,channel>>::
6905 An entity which is responsible for a set of ring buffers.
6907 <<event,Event rules>> are always attached to a specific channel.
6910 A reference of time for a tracer.
6912 <<lttng-consumerd,consumer daemon>>::
6913 A process which is responsible for consuming the full sub-buffers
6914 and write them to a file system or send them over the network.
6916 <<channel-overwrite-mode-vs-discard-mode,discard mode>>:: The event loss
6917 mode in which the tracer _discards_ new event records when there's no
6918 sub-buffer space left to store them.
6921 The consequence of the execution of an instrumentation
6922 point, like a tracepoint that you manually place in some source code,
6923 or a Linux kernel KProbe.
6925 An event is said to _occur_ at a specific time. Different actions can
6926 be taken upon the occurrence of an event, like record the event's payload
6929 <<channel-overwrite-mode-vs-discard-mode,event loss mode>>::
6930 The mechanism by which event records of a given channel are lost
6931 (not recorded) when there is no sub-buffer space left to store them.
6933 [[def-event-name]]event name::
6934 The name of an event, which is also the name of the event record.
6935 This is also called the _instrumentation point name_.
6938 A record, in a trace, of the payload of an event which occured.
6940 <<event,event rule>>::
6941 Set of conditions which must be satisfied for one or more occuring
6942 events to be recorded.
6944 `java.util.logging`::
6946 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities].
6948 <<instrumenting,instrumentation>>::
6949 The use of LTTng probes to make a piece of software traceable.
6951 instrumentation point::
6952 A point in the execution path of a piece of software that, when
6953 reached by this execution, can emit an event.
6955 instrumentation point name::
6956 See _<<def-event-name,event name>>_.
6959 A http://logging.apache.org/log4j/1.2/[logging library] for Java
6960 developed by the Apache Software Foundation.
6963 Level of severity of a log statement or user space
6964 instrumentation point.
6967 The _Linux Trace Toolkit: next generation_ project.
6969 <<lttng-cli,cmd:lttng>>::
6970 A command-line tool provided by the LTTng-tools project which you
6971 can use to send and receive control messages to and from a
6975 The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
6976 which is a set of analyzing programs that are used to obtain a
6977 higher level view of an LTTng trace.
6979 cmd:lttng-consumerd::
6980 The name of the consumer daemon program.
6983 A utility provided by the LTTng-tools project which can convert
6984 ring buffer files (usually
6985 <<persistent-memory-file-systems,saved on a persistent memory file system>>)
6988 LTTng Documentation::
6991 <<lttng-live,LTTng live>>::
6992 A communication protocol between the relay daemon and live viewers
6993 which makes it possible to see events "live", as they are received by
6996 <<lttng-modules,LTTng-modules>>::
6997 The https://github.com/lttng/lttng-modules[LTTng-modules] project,
6998 which contains the Linux kernel modules to make the Linux kernel
6999 instrumentation points available for LTTng tracing.
7002 The name of the relay daemon program.
7004 cmd:lttng-sessiond::
7005 The name of the session daemon program.
7008 The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
7009 contains the various programs and libraries used to
7010 <<controlling-tracing,control tracing>>.
7012 <<lttng-ust,LTTng-UST>>::
7013 The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
7014 contains libraries to instrument user applications.
7016 <<lttng-ust-agents,LTTng-UST Java agent>>::
7017 A Java package provided by the LTTng-UST project to allow the
7018 LTTng instrumentation of `java.util.logging` and Apache log4j 1.2
7021 <<lttng-ust-agents,LTTng-UST Python agent>>::
7022 A Python package provided by the LTTng-UST project to allow the
7023 LTTng instrumentation of Python logging statements.
7025 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
7026 The event loss mode in which new event records overwrite older
7027 event records when there's no sub-buffer space left to store them.
7029 <<channel-buffering-schemes,per-process buffering>>::
7030 A buffering scheme in which each instrumented process has its own
7031 sub-buffers for a given user space channel.
7033 <<channel-buffering-schemes,per-user buffering>>::
7034 A buffering scheme in which all the processes of a Unix user share the
7035 same sub-buffer for a given user space channel.
7037 <<lttng-relayd,relay daemon>>::
7038 A process which is responsible for receiving the trace data sent by
7039 a distant consumer daemon.
7042 A set of sub-buffers.
7044 <<lttng-sessiond,session daemon>>::
7045 A process which receives control commands from you and orchestrates
7046 the tracers and various LTTng daemons.
7048 <<taking-a-snapshot,snapshot>>::
7049 A copy of the current data of all the sub-buffers of a given tracing
7050 session, saved as trace files.
7053 One part of an LTTng ring buffer which contains event records.
7056 The time information attached to an event when it is emitted.
7059 A set of files which are the concatenations of one or more
7060 flushed sub-buffers.
7063 The action of recording the events emitted by an application
7064 or by a system, or to initiate such recording by controlling
7068 The http://tracecompass.org[Trace Compass] project and application.
7071 An instrumentation point using the tracepoint mechanism of the Linux
7072 kernel or of LTTng-UST.
7074 tracepoint definition::
7075 The definition of a single tracepoint.
7078 The name of a tracepoint.
7080 tracepoint provider::
7081 A set of functions providing tracepoints to an instrumented user
7084 Not to be confused with a _tracepoint provider package_: many tracepoint
7085 providers can exist within a tracepoint provider package.
7087 tracepoint provider package::
7088 One or more tracepoint providers compiled as an object file or as
7092 A software which records emitted events.
7094 <<domain,tracing domain>>::
7095 A namespace for event sources.
7097 <<tracing-group,tracing group>>::
7098 The Unix group in which a Unix user can be to be allowed to trace the
7101 <<tracing-session,tracing session>>::
7102 A stateful dialogue between you and a <<lttng-sessiond,session
7106 An application running in user space, as opposed to a Linux kernel
7107 module, for example.