2.11/lttng-docs-2.11.txt

   1 The LTTng Documentation
   2 =======================
   3 Philippe Proulx <pproulx@efficios.com>
   4 v2.11, 19 November 2018
   5
   6
   7 include::../common/copyright.txt[]
   8
   9
  10 include::../common/welcome.txt[]
  11
  12
  13 include::../common/audience.txt[]
  14
  15
  16 [[chapters]]
  17 === What's in this documentation?
  18
  19 The LTTng Documentation is divided into the following sections:
  20
  21 * **<<nuts-and-bolts,Nuts and bolts>>** explains the
  22   rudiments of software tracing and the rationale behind the
  23   LTTng project.
  24 +
  25 You can skip this section if you’re familiar with software tracing and
  26 with the LTTng project.
  27
  28 * **<<installing-lttng,Installation>>** describes the steps to
  29   install the LTTng packages on common Linux distributions and from
  30   their sources.
  31 +
  32 You can skip this section if you already properly installed LTTng on
  33 your target system.
  34
  35 * **<<getting-started,Quick start>>** is a concise guide to
  36   getting started quickly with LTTng kernel and user space tracing.
  37 +
  38 We recommend this section if you're new to LTTng or to software tracing
  39 in general.
  40 +
  41 You can skip this section if you're not new to LTTng.
  42
  43 * **<<core-concepts,Core concepts>>** explains the concepts at
  44   the heart of LTTng.
  45 +
  46 It's a good idea to become familiar with the core concepts
  47 before attempting to use the toolkit.
  48
  49 * **<<plumbing,Components of LTTng>>** describes the various components
  50   of the LTTng machinery, like the daemons, the libraries, and the
  51   command-line interface.
  52 * **<<instrumenting,Instrumentation>>** shows different ways to
  53   instrument user applications and the Linux kernel.
  54 +
  55 Instrumenting source code is essential to provide a meaningful
  56 source of events.
  57 +
  58 You can skip this section if you do not have a programming background.
  59
  60 * **<<controlling-tracing,Tracing control>>** is divided into topics
  61   which demonstrate how to use the vast array of features that
  62   LTTng{nbsp}{revision} offers.
  63 * **<<reference,Reference>>** contains reference tables.
  64 * **<<glossary,Glossary>>** is a specialized dictionary of terms related
  65   to LTTng or to the field of software tracing.
  66
  67
  68 include::../common/convention.txt[]
  69
  70
  71 include::../common/acknowledgements.txt[]
  72
  73
  74 [[whats-new]]
  75 == What's new in LTTng{nbsp}{revision}?
  76
  77 LTTng{nbsp}{revision} bears the name _Lafontaine_. This modern
  78 https://en.wikipedia.org/wiki/Saison[saison] from the
  79 https://oshlag.com/[Oshlag] microbrewery is a refreshing--zesty--rice
  80 beer with hints of fruit and spices. Some even say it makes for a great
  81 https://en.wikipedia.org/wiki/Somaek[Somaek] when mixed with
  82 Chamisul Soju, not that we've tried!
  83
  84 New features and changes in LTTng{nbsp}{revision}:
  85
  86 * Just like you can typically perform
  87   https://en.wikipedia.org/wiki/Log_rotation[log rotation], you can
  88   now <<session-rotation,_rotate_ a tracing session>>, that
  89   is, according to man:lttng-rotate(1), archive the current trace
  90   chunk (all the tracing session's trace data since the last rotation
  91   or since its inception) so that LTTng does not manage it anymore.
  92 +
  93 Once LTTng archives a trace chunk, you are free to read it, modify it,
  94 move it, or remove it.
  95 +
  96 You can rotate a tracing session immediately or set a rotation schedule
  97 to automate rotations.
  98
  99 * When you <<enabling-disabling-events,create an event rule>>, the
 100   filter expression syntax now supports the following new operators:
 101 +
 102 --
 103 ** `~` (bitwise not)
 104 ** `<<` (bitwise left shift)
 105 ** `>>` (bitwise right shift)
 106 ** `&` (bitwise AND)
 107 ** `^` (bitwise XOR)
 108 ** `|` (bitwise OR)
 109 --
 110 +
 111 The syntax also supports array indexing with the usual square brackets:
 112 +
 113 ----
 114 regs[3][1] & 0xff7 == 0x240
 115 ----
 116 +
 117 There are peculiarities for both the new operators and the array
 118 indexing brackets, like a custom precedence table and implicit casting.
 119 See man:lttng-enable-event(1) to get all the details about the filter
 120 expression syntax.
 121
 122 * You can now dynamically instrument any application's or library's
 123   function entry by symbol name thanks to the new
 124   opt:lttng-enable-event(1):--userspace-probe option of
 125   the `lttng enable-event` command:
 126 +
 127 [role="term"]
 128 ----
 129 $ lttng enable-event --kernel \
 130       --userspace-probe=/usr/lib/libc.so.6:malloc libc_malloc
 131 ----
 132 +
 133 The option also supports tracing existing
 134 https://www.sourceware.org/systemtap/wiki/AddingUserSpaceProbingToApps[SystemTap
 135 Statically Defined Tracing] (SDT) probe (DTrace-style marker). For example,
 136 given the following probe:
 137 +
 138 [source,c]
 139 ----
 140 DTRACE_PROBE2("server", "accept-request", request_id, ip_addr);
 141 ----
 142 +
 143 You can trace this probe with:
 144 +
 145 ----
 146 $ lttng enable-event --kernel \
 147       --userspace-probe=sdt:/path/to/server:server:accept-request \
 148       server_accept_request
 149 ----
 150 +
 151 This feature makes use of Linux's
 152 https://www.kernel.org/doc/Documentation/trace/uprobetracer.txt[uprobe]
 153 mechanism, therefore you must use the `--userspace-probe`
 154 instrumentation option with the opt:lttng-enable-event(1):--kernel
 155 domain option.
 156 +
 157 NOTE: As of LTTng{nbsp}{revision}, LTTng does not record function
 158 parameters with the opt:lttng-enable-event(1):--userspace-probe option.
 159
 160 * Two new <<adding-context,context>> fields are available for Linux
 161   kernel <<channel,channels>>:
 162 +
 163 --
 164 ** `callstack-kernel`
 165 ** `callstack-user`
 166 --
 167 +
 168 Thanks to those, you can record the Linux kernel and user call stacks
 169 when a kernel event occurs. For example:
 170 +
 171 [role="term"]
 172 ----
 173 $ lttng enable-event --kernel --syscall read
 174 $ lttng add-context --kernel --type=callstack-kernel --type=callstack-user
 175 ----
 176 +
 177 When an man:open(2) system call occurs, LTTng attaches the kernel and
 178 user call stacks to the recorded event.
 179 +
 180 NOTE: LTTng cannot always sample the user space call stack reliably.
 181 For instance, LTTng cannot sample the call stack of user applications
 182 and libraries compiled with the
 183 https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html[`-fomit-frame pointer`]
 184 option. In such a case, the tracing is not affected, but the sampled
 185 user space call stack may only contain the user call stack's topmost
 186 address.
 187
 188 * The <<lttng-relayd,relay daemon>> is more efficient and presents fewer
 189   connectivity issues, especially when a large number of targets send
 190   trace data to a given relay daemon.
 191
 192 * User applications and libraries instrumented with
 193   <<lttng-ust,LTTng-UST>> can now safely unload (man:dlclose(3)) a
 194   dynamically loaded
 195   <<building-tracepoint-providers-and-user-application,tracepoint
 196   provider package>>.
 197
 198
 199 [[nuts-and-bolts]]
 200 == Nuts and bolts
 201
 202 What is LTTng? As its name suggests, the _Linux Trace Toolkit: next
 203 generation_ is a modern toolkit for tracing Linux systems and
 204 applications. So your first question might be:
 205 **what is tracing?**
 206
 207
 208 [[what-is-tracing]]
 209 === What is tracing?
 210
 211 As the history of software engineering progressed and led to what
 212 we now take for granted--complex, numerous and
 213 interdependent software applications running in parallel on
 214 sophisticated operating systems like Linux--the authors of such
 215 components, software developers, began feeling a natural
 216 urge to have tools that would ensure the robustness and good performance
 217 of their masterpieces.
 218
 219 One major achievement in this field is, inarguably, the
 220 https://www.gnu.org/software/gdb/[GNU debugger (GDB)],
 221 an essential tool for developers to find and fix bugs. But even the best
 222 debugger won't help make your software run faster, and nowadays, faster
 223 software means either more work done by the same hardware, or cheaper
 224 hardware for the same work.
 225
 226 A _profiler_ is often the tool of choice to identify performance
 227 bottlenecks. Profiling is suitable to identify _where_ performance is
 228 lost in a given software. The profiler outputs a profile, a statistical
 229 summary of observed events, which you may use to discover which
 230 functions took the most time to execute. However, a profiler won't
 231 report _why_ some identified functions are the bottleneck. Bottlenecks
 232 might only occur when specific conditions are met, conditions that are
 233 sometimes impossible to capture by a statistical profiler, or impossible
 234 to reproduce with an application altered by the overhead of an
 235 event-based profiler. For a thorough investigation of software
 236 performance issues, a history of execution is essential, with the
 237 recorded values of variables and context fields you choose, and
 238 with as little influence as possible on the instrumented software. This
 239 is where tracing comes in handy.
 240
 241 _Tracing_ is a technique used to understand what goes on in a running
 242 software system. The software used for tracing is called a _tracer_,
 243 which is conceptually similar to a tape recorder. When recording,
 244 specific instrumentation points placed in the software source code
 245 generate events that are saved on a giant tape: a _trace_ file. You
 246 can trace user applications and the operating system at the same time,
 247 opening the possibility of resolving a wide range of problems that would
 248 otherwise be extremely challenging.
 249
 250 Tracing is often compared to _logging_. However, tracers and loggers are
 251 two different tools, serving two different purposes. Tracers are
 252 designed to record much lower-level events that occur much more
 253 frequently than log messages, often in the range of thousands per
 254 second, with very little execution overhead. Logging is more appropriate
 255 for a very high-level analysis of less frequent events: user accesses,
 256 exceptional conditions (errors and warnings, for example), database
 257 transactions, instant messaging communications, and such. Simply put,
 258 logging is one of the many use cases that can be satisfied with tracing.
 259
 260 The list of recorded events inside a trace file can be read manually
 261 like a log file for the maximum level of detail, but it is generally
 262 much more interesting to perform application-specific analyses to
 263 produce reduced statistics and graphs that are useful to resolve a
 264 given problem. Trace viewers and analyzers are specialized tools
 265 designed to do this.
 266
 267 In the end, this is what LTTng is: a powerful, open source set of
 268 tools to trace the Linux kernel and user applications at the same time.
 269 LTTng is composed of several components actively maintained and
 270 developed by its link:/community/#where[community].
 271
 272
 273 [[lttng-alternatives]]
 274 === Alternatives to noch:{LTTng}
 275
 276 Excluding proprietary solutions, a few competing software tracers
 277 exist for Linux:
 278
 279 * https://github.com/dtrace4linux/linux[dtrace4linux] is a port of
 280   Sun Microsystems's DTrace to Linux. The cmd:dtrace tool interprets
 281   user scripts and is responsible for loading code into the
 282   Linux kernel for further execution and collecting the outputted data.
 283 * https://en.wikipedia.org/wiki/Berkeley_Packet_Filter[eBPF] is a
 284   subsystem in the Linux kernel in which a virtual machine can execute
 285   programs passed from the user space to the kernel. You can attach
 286   such programs to tracepoints and kprobes thanks to a system call, and
 287   they can output data to the user space when executed thanks to
 288   different mechanisms (pipe, VM register values, and eBPF maps, to name
 289   a few).
 290 * https://www.kernel.org/doc/Documentation/trace/ftrace.txt[ftrace]
 291   is the de facto function tracer of the Linux kernel. Its user
 292   interface is a set of special files in sysfs.
 293 * https://perf.wiki.kernel.org/[perf] is
 294   a performance analyzing tool for Linux which supports hardware
 295   performance counters, tracepoints, as well as other counters and
 296   types of probes. perf's controlling utility is the cmd:perf command
 297   line/curses tool.
 298 * http://linux.die.net/man/1/strace[strace]
 299   is a command-line utility which records system calls made by a
 300   user process, as well as signal deliveries and changes of process
 301   state. strace makes use of https://en.wikipedia.org/wiki/Ptrace[ptrace]
 302   to fulfill its function.
 303 * http://www.sysdig.org/[sysdig], like SystemTap, uses scripts to
 304   analyze Linux kernel events. You write scripts, or _chisels_ in
 305   sysdig's jargon, in Lua and sysdig executes them while it traces the
 306   system or afterwards. sysdig's interface is the cmd:sysdig
 307   command-line tool as well as the curses-based cmd:csysdig tool.
 308 * https://sourceware.org/systemtap/[SystemTap] is a Linux kernel and
 309   user space tracer which uses custom user scripts to produce plain text
 310   traces. SystemTap converts the scripts to the C language, and then
 311   compiles them as Linux kernel modules which are loaded to produce
 312   trace data. SystemTap's primary user interface is the cmd:stap
 313   command-line tool.
 314
 315 The main distinctive features of LTTng is that it produces correlated
 316 kernel and user space traces, as well as doing so with the lowest
 317 overhead amongst other solutions. It produces trace files in the
 318 http://diamon.org/ctf[CTF] format, a file format optimized
 319 for the production and analyses of multi-gigabyte data.
 320
 321 LTTng is the result of more than 10{nbsp}years of active open source
 322 development by a community of passionate developers.
 323 LTTng{nbsp}{revision} is currently available on major desktop and server
 324 Linux distributions.
 325
 326 The main interface for tracing control is a single command-line tool
 327 named cmd:lttng. The latter can create several tracing sessions, enable
 328 and disable events on the fly, filter events efficiently with custom
 329 user expressions, start and stop tracing, and much more. LTTng can
 330 record the traces on the file system or send them over the network, and
 331 keep them totally or partially. You can view the traces once tracing
 332 becomes inactive or in real-time.
 333
 334 <<installing-lttng,Install LTTng now>> and
 335 <<getting-started,start tracing>>!
 336
 337
 338 [[installing-lttng]]
 339 == Installation
 340
 341 **LTTng** is a set of software <<plumbing,components>> which interact to
 342 <<instrumenting,instrument>> the Linux kernel and user applications, and
 343 to <<controlling-tracing,control tracing>> (start and stop
 344 tracing, enable and disable event rules, and the rest). Those
 345 components are bundled into the following packages:
 346
 347 * **LTTng-tools**: Libraries and command-line interface to
 348   control tracing.
 349 * **LTTng-modules**: Linux kernel modules to instrument and
 350   trace the kernel.
 351 * **LTTng-UST**: Libraries and Java/Python packages to instrument and
 352   trace user applications.
 353
 354 Most distributions mark the LTTng-modules and LTTng-UST packages as
 355 optional when installing LTTng-tools (which is always required). In the
 356 following sections, we always provide the steps to install all three,
 357 but note that:
 358
 359 * You only need to install LTTng-modules if you intend to trace the
 360   Linux kernel.
 361 * You only need to install LTTng-UST if you intend to trace user
 362   applications.
 363
 364 [role="growable"]
 365 .Availability of LTTng{nbsp}{revision} for major Linux distributions as of 12 November 2018.
 366 |====
 367 |Distribution |Available in releases |Alternatives
 368
 369 |https://www.ubuntu.com/[Ubuntu]
 370 |Ubuntu{nbsp}14.04 _Trusty Tahr_, Ubuntu{nbsp}16.04 _Xenial Xerus_,
 371 and Ubuntu{nbsp}18.04 _Bionic Beaver_:
 372 <<ubuntu-ppa,use the LTTng Stable{nbsp}{revision} PPA>>.
 373 |<<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 374
 375 |https://getfedora.org/[Fedora]
 376 |_Not available_
 377 |
 378 link:/docs/v2.10#doc-fedora[LTTng{nbsp}2.10 for Fedora{nbsp}27,
 379 Fedora{nbsp}28, and Fedora{nbsp}29].
 380
 381 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 382
 383 |https://www.debian.org/[Debian]
 384 |_Not available_
 385 |link:/docs/v2.10#doc-debian[LTTng{nbsp}2.10 for Debian "buster" (testing)].
 386
 387 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 388
 389 |https://www.archlinux.org/[Arch Linux]
 390 |_Not available_
 391 |<<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 392
 393 |https://alpinelinux.org/[Alpine Linux]
 394 |_Not available_
 395 |
 396 link:/docs/v2.10#doc-alpine-linux[LTTng{nbsp}2.10 for
 397 Alpine Linux{nbsp}3.7 and Alpine Linux{nbsp}3.8].
 398
 399 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 400
 401 |https://www.opensuse.org/[openSUSE]
 402 |_Not available_
 403 |link:/docs/v2.10#doc-opensuse[LTTng{nbsp}2.10 for openSUSE Leap{nbsp}15.0].
 404
 405 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 406
 407 |https://www.redhat.com/[RHEL] and https://www.suse.com/[SLES]
 408 |See http://packages.efficios.com/[EfficiOS Enterprise Packages].
 409 |
 410
 411 |https://buildroot.org/[Buildroot]
 412 |_Not available_
 413 |
 414 link:/docs/v2.10#doc-buildroot[LTTng{nbsp}2.10 for Buildroot{nbsp}2018.02,
 415 Buildroot{nbsp}2018.05, Buildroot{nbsp}2018.08, and Buildroot{nbsp}2018.11].
 416
 417 <<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 418
 419 |http://www.openembedded.org/wiki/Main_Page[OpenEmbedded] and
 420 https://www.yoctoproject.org/[Yocto]
 421 |_Not available_
 422 |<<building-from-source,Build LTTng{nbsp}{revision} from source>>.
 423 |====
 424
 425
 426 [[ubuntu-ppa]]
 427 === Ubuntu: noch:{LTTng} Stable {revision} PPA
 428
 429 The https://launchpad.net/~lttng/+archive/ubuntu/stable-{revision}[LTTng
 430 Stable{nbsp}{revision} PPA] offers the latest stable
 431 LTTng{nbsp}{revision} packages for:
 432
 433 * Ubuntu{nbsp}14.04 _Trusty Tahr_
 434 * Ubuntu{nbsp}16.04 _Xenial Xerus_
 435 * Ubuntu{nbsp}18.04 _Bionic Beaver_
 436
 437 To install LTTng{nbsp}{revision} from the LTTng Stable{nbsp}{revision} PPA:
 438
 439 . Add the LTTng Stable{nbsp}{revision} PPA repository and update the
 440   list of packages:
 441 +
 442 --
 443 [role="term"]
 444 ----
 445 # apt-add-repository ppa:lttng/stable-2.11
 446 # apt-get update
 447 ----
 448 --
 449
 450 . Install the main LTTng{nbsp}{revision} packages:
 451 +
 452 --
 453 [role="term"]
 454 ----
 455 # apt-get install lttng-tools
 456 # apt-get install lttng-modules-dkms
 457 # apt-get install liblttng-ust-dev
 458 ----
 459 --
 460
 461 . **If you need to instrument and trace
 462   <<java-application,Java applications>>**, install the LTTng-UST
 463   Java agent:
 464 +
 465 --
 466 [role="term"]
 467 ----
 468 # apt-get install liblttng-ust-agent-java
 469 ----
 470 --
 471
 472 . **If you need to instrument and trace
 473   <<python-application,Python{nbsp}3 applications>>**, install the
 474   LTTng-UST Python agent:
 475 +
 476 --
 477 [role="term"]
 478 ----
 479 # apt-get install python3-lttngust
 480 ----
 481 --
 482
 483
 484 [[enterprise-distributions]]
 485 === RHEL, SUSE, and other enterprise distributions
 486
 487 To install LTTng on enterprise Linux distributions, such as Red Hat
 488 Enterprise Linux (RHEL) and SUSE Linux Enterprise Server (SUSE), please
 489 see http://packages.efficios.com/[EfficiOS Enterprise Packages].
 490
 491
 492 [[building-from-source]]
 493 === Build from source
 494
 495 To build and install LTTng{nbsp}{revision} from source:
 496
 497 . Using your distribution's package manager, or from source, install
 498   the following dependencies of LTTng-tools and LTTng-UST:
 499 +
 500 --
 501 * https://sourceforge.net/projects/libuuid/[libuuid]
 502 * http://directory.fsf.org/wiki/Popt[popt]
 503 * http://liburcu.org/[Userspace RCU]
 504 * http://www.xmlsoft.org/[libxml2]
 505 * **Optional**: https://github.com/numactl/numactl[numactl]
 506 --
 507
 508 . Download, build, and install the latest LTTng-modules{nbsp}{revision}:
 509 +
 510 --
 511 [role="term"]
 512 ----
 513 $ cd $(mktemp -d) &&
 514 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.11.tar.bz2 &&
 515 tar -xf lttng-modules-latest-2.11.tar.bz2 &&
 516 cd lttng-modules-2.11.* &&
 517 make &&
 518 sudo make modules_install &&
 519 sudo depmod -a
 520 ----
 521 --
 522
 523 . Download, build, and install the latest LTTng-UST{nbsp}{revision}:
 524 +
 525 --
 526 [role="term"]
 527 ----
 528 $ cd $(mktemp -d) &&
 529 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.11.tar.bz2 &&
 530 tar -xf lttng-ust-latest-2.11.tar.bz2 &&
 531 cd lttng-ust-2.11.* &&
 532 ./configure &&
 533 make &&
 534 sudo make install &&
 535 sudo ldconfig
 536 ----
 537 --
 538 +
 539 Add `--disable-numa` to `./configure` if you don't have
 540 https://github.com/numactl/numactl[numactl].
 541 +
 542 --
 543 [IMPORTANT]
 544 .Java and Python application tracing
 545 ====
 546 If you need to instrument and trace <<java-application,Java
 547 applications>>, pass the `--enable-java-agent-jul`,
 548 `--enable-java-agent-log4j`, or `--enable-java-agent-all` options to the
 549 `configure` script, depending on which Java logging framework you use.
 550
 551 If you need to instrument and trace <<python-application,Python
 552 applications>>, pass the `--enable-python-agent` option to the
 553 `configure` script. You can set the `PYTHON` environment variable to the
 554 path to the Python interpreter for which to install the LTTng-UST Python
 555 agent package.
 556 ====
 557 --
 558 +
 559 --
 560 [NOTE]
 561 ====
 562 By default, LTTng-UST libraries are installed to
 563 dir:{/usr/local/lib}, which is the de facto directory in which to
 564 keep self-compiled and third-party libraries.
 565
 566 When <<building-tracepoint-providers-and-user-application,linking an
 567 instrumented user application with `liblttng-ust`>>:
 568
 569 * Append `/usr/local/lib` to the env:LD_LIBRARY_PATH environment
 570   variable.
 571 * Pass the `-L/usr/local/lib` and `-Wl,-rpath,/usr/local/lib` options to
 572   man:gcc(1), man:g++(1), or man:clang(1).
 573 ====
 574 --
 575
 576 . Download, build, and install the latest LTTng-tools{nbsp}{revision}:
 577 +
 578 --
 579 [role="term"]
 580 ----
 581 $ cd $(mktemp -d) &&
 582 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.11.tar.bz2 &&
 583 tar -xf lttng-tools-latest-2.11.tar.bz2 &&
 584 cd lttng-tools-2.11.* &&
 585 ./configure &&
 586 make &&
 587 sudo make install &&
 588 sudo ldconfig
 589 ----
 590 --
 591
 592 TIP: The https://github.com/eepp/vlttng[vlttng tool] can do all the
 593 previous steps automatically for a given version of LTTng and confine
 594 the installed files in a specific directory. This can be useful to test
 595 LTTng without installing it on your system.
 596
 597
 598 [[getting-started]]
 599 == Quick start
 600
 601 This is a short guide to get started quickly with LTTng kernel and user
 602 space tracing.
 603
 604 Before you follow this guide, make sure to <<installing-lttng,install>>
 605 LTTng.
 606
 607 This tutorial walks you through the steps to:
 608
 609 . <<tracing-the-linux-kernel,Trace the Linux kernel>>.
 610 . <<tracing-your-own-user-application,Trace a user application>> written
 611   in C.
 612 . <<viewing-and-analyzing-your-traces,View and analyze the
 613   recorded events>>.
 614
 615
 616 [[tracing-the-linux-kernel]]
 617 === Trace the Linux kernel
 618
 619 The following command lines start with the `#` prompt because you need
 620 root privileges to trace the Linux kernel. You can also trace the kernel
 621 as a regular user if your Unix user is a member of the
 622 <<tracing-group,tracing group>>.
 623
 624 . Create a <<tracing-session,tracing session>> which writes its traces
 625   to dir:{/tmp/my-kernel-trace}:
 626 +
 627 --
 628 [role="term"]
 629 ----
 630 # lttng create my-kernel-session --output=/tmp/my-kernel-trace
 631 ----
 632 --
 633
 634 . List the available kernel tracepoints and system calls:
 635 +
 636 --
 637 [role="term"]
 638 ----
 639 # lttng list --kernel
 640 # lttng list --kernel --syscall
 641 ----
 642 --
 643
 644 . Create <<event,event rules>> which match the desired instrumentation
 645   point names, for example the `sched_switch` and `sched_process_fork`
 646   tracepoints, and the man:open(2) and man:close(2) system calls:
 647 +
 648 --
 649 [role="term"]
 650 ----
 651 # lttng enable-event --kernel sched_switch,sched_process_fork
 652 # lttng enable-event --kernel --syscall open,close
 653 ----
 654 --
 655 +
 656 You can also create an event rule which matches _all_ the Linux kernel
 657 tracepoints (this will generate a lot of data when tracing):
 658 +
 659 --
 660 [role="term"]
 661 ----
 662 # lttng enable-event --kernel --all
 663 ----
 664 --
 665
 666 . <<basic-tracing-session-control,Start tracing>>:
 667 +
 668 --
 669 [role="term"]
 670 ----
 671 # lttng start
 672 ----
 673 --
 674
 675 . Do some operation on your system for a few seconds. For example,
 676   load a website, or list the files of a directory.
 677 . <<basic-tracing-session-control,Stop tracing>> and destroy the
 678   tracing session:
 679 +
 680 --
 681 [role="term"]
 682 ----
 683 # lttng stop
 684 # lttng destroy
 685 ----
 686 --
 687 +
 688 You need to stop tracing to make LTTng flush the remaining trace data
 689 and make the trace readable.
 690 +
 691 The man:lttng-destroy(1) command does not destroy the trace data; it
 692 only destroys the state of the tracing session.
 693
 694 . For the sake of this example, make the recorded trace accessible to
 695   the non-root users:
 696 +
 697 --
 698 [role="term"]
 699 ----
 700 # chown -R $(whoami) /tmp/my-kernel-trace
 701 ----
 702 --
 703
 704 See <<viewing-and-analyzing-your-traces,View and analyze the
 705 recorded events>> to view the recorded events.
 706
 707
 708 [[tracing-your-own-user-application]]
 709 === Trace a user application
 710
 711 This section steps you through a simple example to trace a
 712 _Hello world_ program written in C.
 713
 714 To create the traceable user application:
 715
 716 . Create the tracepoint provider header file, which defines the
 717   tracepoints and the events they can generate:
 718 +
 719 --
 720 [source,c]
 721 .path:{hello-tp.h}
 722 ----
 723 #undef TRACEPOINT_PROVIDER
 724 #define TRACEPOINT_PROVIDER hello_world
 725
 726 #undef TRACEPOINT_INCLUDE
 727 #define TRACEPOINT_INCLUDE "./hello-tp.h"
 728
 729 #if !defined(_HELLO_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
 730 #define _HELLO_TP_H
 731
 732 #include <lttng/tracepoint.h>
 733
 734 TRACEPOINT_EVENT(
 735     hello_world,
 736     my_first_tracepoint,
 737     TP_ARGS(
 738         int, my_integer_arg,
 739         char*, my_string_arg
 740     ),
 741     TP_FIELDS(
 742         ctf_string(my_string_field, my_string_arg)
 743         ctf_integer(int, my_integer_field, my_integer_arg)
 744     )
 745 )
 746
 747 #endif /* _HELLO_TP_H */
 748
 749 #include <lttng/tracepoint-event.h>
 750 ----
 751 --
 752
 753 . Create the tracepoint provider package source file:
 754 +
 755 --
 756 [source,c]
 757 .path:{hello-tp.c}
 758 ----
 759 #define TRACEPOINT_CREATE_PROBES
 760 #define TRACEPOINT_DEFINE
 761
 762 #include "hello-tp.h"
 763 ----
 764 --
 765
 766 . Build the tracepoint provider package:
 767 +
 768 --
 769 [role="term"]
 770 ----
 771 $ gcc -c -I. hello-tp.c
 772 ----
 773 --
 774
 775 . Create the _Hello World_ application source file:
 776 +
 777 --
 778 [source,c]
 779 .path:{hello.c}
 780 ----
 781 #include <stdio.h>
 782 #include "hello-tp.h"
 783
 784 int main(int argc, char *argv[])
 785 {
 786     int x;
 787
 788     puts("Hello, World!\nPress Enter to continue...");
 789
 790     /*
 791      * The following getchar() call is only placed here for the purpose
 792      * of this demonstration, to pause the application in order for
 793      * you to have time to list its tracepoints. It is not
 794      * needed otherwise.
 795      */
 796     getchar();
 797
 798     /*
 799      * A tracepoint() call.
 800      *
 801      * Arguments, as defined in hello-tp.h:
 802      *
 803      * 1. Tracepoint provider name   (required)
 804      * 2. Tracepoint name            (required)
 805      * 3. my_integer_arg             (first user-defined argument)
 806      * 4. my_string_arg              (second user-defined argument)
 807      *
 808      * Notice the tracepoint provider and tracepoint names are
 809      * NOT strings: they are in fact parts of variables that the
 810      * macros in hello-tp.h create.
 811      */
 812     tracepoint(hello_world, my_first_tracepoint, 23, "hi there!");
 813
 814     for (x = 0; x < argc; ++x) {
 815         tracepoint(hello_world, my_first_tracepoint, x, argv[x]);
 816     }
 817
 818     puts("Quitting now!");
 819     tracepoint(hello_world, my_first_tracepoint, x * x, "x^2");
 820
 821     return 0;
 822 }
 823 ----
 824 --
 825
 826 . Build the application:
 827 +
 828 --
 829 [role="term"]
 830 ----
 831 $ gcc -c hello.c
 832 ----
 833 --
 834
 835 . Link the application with the tracepoint provider package,
 836   `liblttng-ust`, and `libdl`:
 837 +
 838 --
 839 [role="term"]
 840 ----
 841 $ gcc -o hello hello.o hello-tp.o -llttng-ust -ldl
 842 ----
 843 --
 844
 845 Here's the whole build process:
 846
 847 [role="img-100"]
 848 .User space tracing tutorial's build steps.
 849 image::ust-flow.png[]
 850
 851 To trace the user application:
 852
 853 . Run the application with a few arguments:
 854 +
 855 --
 856 [role="term"]
 857 ----
 858 $ ./hello world and beyond
 859 ----
 860 --
 861 +
 862 You see:
 863 +
 864 --
 865 ----
 866 Hello, World!
 867 Press Enter to continue...
 868 ----
 869 --
 870
 871 . Start an LTTng <<lttng-sessiond,session daemon>>:
 872 +
 873 --
 874 [role="term"]
 875 ----
 876 $ lttng-sessiond --daemonize
 877 ----
 878 --
 879 +
 880 Note that a session daemon might already be running, for example as
 881 a service that the distribution's service manager started.
 882
 883 . List the available user space tracepoints:
 884 +
 885 --
 886 [role="term"]
 887 ----
 888 $ lttng list --userspace
 889 ----
 890 --
 891 +
 892 You see the `hello_world:my_first_tracepoint` tracepoint listed
 893 under the `./hello` process.
 894
 895 . Create a <<tracing-session,tracing session>>:
 896 +
 897 --
 898 [role="term"]
 899 ----
 900 $ lttng create my-user-space-session
 901 ----
 902 --
 903
 904 . Create an <<event,event rule>> which matches the
 905   `hello_world:my_first_tracepoint` event name:
 906 +
 907 --
 908 [role="term"]
 909 ----
 910 $ lttng enable-event --userspace hello_world:my_first_tracepoint
 911 ----
 912 --
 913
 914 . <<basic-tracing-session-control,Start tracing>>:
 915 +
 916 --
 917 [role="term"]
 918 ----
 919 $ lttng start
 920 ----
 921 --
 922
 923 . Go back to the running `hello` application and press Enter. The
 924   program executes all `tracepoint()` instrumentation points and exits.
 925 . <<basic-tracing-session-control,Stop tracing>> and destroy the
 926   tracing session:
 927 +
 928 --
 929 [role="term"]
 930 ----
 931 $ lttng stop
 932 $ lttng destroy
 933 ----
 934 --
 935 +
 936 You need to stop tracing to make LTTng flush the remaining trace data
 937 and make the trace readable.
 938 +
 939 The man:lttng-destroy(1) command does not destroy the trace data; it
 940 only destroys the state of the tracing session.
 941
 942 By default, LTTng saves the traces in
 943 +$LTTNG_HOME/lttng-traces/__name__-__date__-__time__+,
 944 where +__name__+ is the tracing session name. The
 945 env:LTTNG_HOME environment variable defaults to `$HOME` if not set.
 946
 947 See <<viewing-and-analyzing-your-traces,View and analyze the
 948 recorded events>> to view the recorded events.
 949
 950
 951 [[viewing-and-analyzing-your-traces]]
 952 === View and analyze the recorded events
 953
 954 Once you have completed the <<tracing-the-linux-kernel,Trace the Linux
 955 kernel>> and <<tracing-your-own-user-application,Trace a user
 956 application>> tutorials, you can inspect the recorded events.
 957
 958 Many tools are available to read LTTng traces:
 959
 960 * **cmd:babeltrace** is a command-line utility which converts trace
 961   formats; it supports the format that LTTng produces, CTF, as well as a
 962   basic text output which can be ++grep++ed. The cmd:babeltrace command
 963   is part of the http://diamon.org/babeltrace[Babeltrace] project.
 964 * Babeltrace also includes
 965   **https://www.python.org/[Python] bindings** so
 966   that you can easily open and read an LTTng trace with your own script,
 967   benefiting from the power of Python.
 968 * http://tracecompass.org/[**Trace Compass**]
 969   is a graphical user interface for viewing and analyzing any type of
 970   logs or traces, including LTTng's.
 971 * https://github.com/lttng/lttng-analyses[**LTTng analyses**] is a
 972   project which includes many high-level analyses of LTTng kernel
 973   traces, like scheduling statistics, interrupt frequency distribution,
 974   top CPU usage, and more.
 975
 976 NOTE: This section assumes that LTTng saved the traces it recorded
 977 during the previous tutorials to their default location, in the
 978 dir:{$LTTNG_HOME/lttng-traces} directory. The env:LTTNG_HOME
 979 environment variable defaults to `$HOME` if not set.
 980
 981
 982 [[viewing-and-analyzing-your-traces-bt]]
 983 ==== Use the cmd:babeltrace command-line tool
 984
 985 The simplest way to list all the recorded events of a trace is to pass
 986 its path to cmd:babeltrace with no options:
 987
 988 [role="term"]
 989 ----
 990 $ babeltrace ~/lttng-traces/my-user-space-session*
 991 ----
 992
 993 cmd:babeltrace finds all traces recursively within the given path and
 994 prints all their events, merging them in chronological order.
 995
 996 You can pipe the output of cmd:babeltrace into a tool like man:grep(1) for
 997 further filtering:
 998
 999 [role="term"]
1000 ----
1001 $ babeltrace /tmp/my-kernel-trace | grep _switch
1002 ----
1003
1004 You can pipe the output of cmd:babeltrace into a tool like man:wc(1) to
1005 count the recorded events:
1006
1007 [role="term"]
1008 ----
1009 $ babeltrace /tmp/my-kernel-trace | grep _open | wc --lines
1010 ----
1011
1012
1013 [[viewing-and-analyzing-your-traces-bt-python]]
1014 ==== Use the Babeltrace Python bindings
1015
1016 The <<viewing-and-analyzing-your-traces-bt,text output of cmd:babeltrace>>
1017 is useful to isolate events by simple matching using man:grep(1) and
1018 similar utilities. However, more elaborate filters, such as keeping only
1019 event records with a field value falling within a specific range, are
1020 not trivial to write using a shell. Moreover, reductions and even the
1021 most basic computations involving multiple event records are virtually
1022 impossible to implement.
1023
1024 Fortunately, Babeltrace ships with Python{nbsp}3 bindings which makes it
1025 easy to read the event records of an LTTng trace sequentially and
1026 compute the desired information.
1027
1028 The following script accepts an LTTng Linux kernel trace path as its
1029 first argument and prints the short names of the top five running
1030 processes on CPU{nbsp}0 during the whole trace:
1031
1032 [source,python]
1033 .path:{top5proc.py}
1034 ----
1035 from collections import Counter
1036 import babeltrace
1037 import sys
1038
1039
1040 def top5proc():
1041     if len(sys.argv) != 2:
1042         msg = 'Usage: python3 {} TRACEPATH'.format(sys.argv[0])
1043         print(msg, file=sys.stderr)
1044         return False
1045
1046     # A trace collection contains one or more traces
1047     col = babeltrace.TraceCollection()
1048
1049     # Add the trace provided by the user (LTTng traces always have
1050     # the 'ctf' format)
1051     if col.add_trace(sys.argv[1], 'ctf') is None:
1052         raise RuntimeError('Cannot add trace')
1053
1054     # This counter dict contains execution times:
1055     #
1056     #   task command name -> total execution time (ns)
1057     exec_times = Counter()
1058
1059     # This contains the last `sched_switch` timestamp
1060     last_ts = None
1061
1062     # Iterate on events
1063     for event in col.events:
1064         # Keep only `sched_switch` events
1065         if event.name != 'sched_switch':
1066             continue
1067
1068         # Keep only events which happened on CPU 0
1069         if event['cpu_id'] != 0:
1070             continue
1071
1072         # Event timestamp
1073         cur_ts = event.timestamp
1074
1075         if last_ts is None:
1076             # We start here
1077             last_ts = cur_ts
1078
1079         # Previous task command (short) name
1080         prev_comm = event['prev_comm']
1081
1082         # Initialize entry in our dict if not yet done
1083         if prev_comm not in exec_times:
1084             exec_times[prev_comm] = 0
1085
1086         # Compute previous command execution time
1087         diff = cur_ts - last_ts
1088
1089         # Update execution time of this command
1090         exec_times[prev_comm] += diff
1091
1092         # Update last timestamp
1093         last_ts = cur_ts
1094
1095     # Display top 5
1096     for name, ns in exec_times.most_common(5):
1097         s = ns / 1000000000
1098         print('{:20}{} s'.format(name, s))
1099
1100     return True
1101
1102
1103 if __name__ == '__main__':
1104     sys.exit(0 if top5proc() else 1)
1105 ----
1106
1107 Run this script:
1108
1109 [role="term"]
1110 ----
1111 $ python3 top5proc.py /tmp/my-kernel-trace/kernel
1112 ----
1113
1114 Output example:
1115
1116 ----
1117 swapper/0           48.607245889 s
1118 chromium            7.192738188 s
1119 pavucontrol         0.709894415 s
1120 Compositor          0.660867933 s
1121 Xorg.bin            0.616753786 s
1122 ----
1123
1124 Note that `swapper/0` is the "idle" process of CPU{nbsp}0 on Linux;
1125 since we weren't using the CPU that much when tracing, its first
1126 position in the list makes sense.
1127
1128
1129 [[core-concepts]]
1130 == [[understanding-lttng]]Core concepts
1131
1132 From a user's perspective, the LTTng system is built on a few concepts,
1133 or objects, on which the <<lttng-cli,cmd:lttng command-line tool>>
1134 operates by sending commands to the <<lttng-sessiond,session daemon>>.
1135 Understanding how those objects relate to eachother is key in mastering
1136 the toolkit.
1137
1138 The core concepts are:
1139
1140 * <<tracing-session,Tracing session>>
1141 * <<domain,Tracing domain>>
1142 * <<channel,Channel and ring buffer>>
1143 * <<"event","Instrumentation point, event rule, event, and event record">>
1144
1145
1146 [[tracing-session]]
1147 === Tracing session
1148
1149 A _tracing session_ is a stateful dialogue between you and
1150 a <<lttng-sessiond,session daemon>>. You can
1151 <<creating-destroying-tracing-sessions,create a new tracing
1152 session>> with the `lttng create` command.
1153
1154 Anything that you do when you control LTTng tracers happens within a
1155 tracing session. In particular, a tracing session:
1156
1157 * Has its own name.
1158 * Has its own set of trace files.
1159 * Has its own state of activity (started or stopped).
1160 * Has its own <<tracing-session-mode,mode>> (local, network streaming,
1161   snapshot, or live).
1162 * Has its own <<channel,channels>> to which are associated their own
1163   <<event,event rules>>.
1164
1165 [role="img-100"]
1166 .A _tracing session_ contains <<channel,channels>> that are members of <<domain,tracing domains>> and contain <<event,event rules>>.
1167 image::concepts.png[]
1168
1169 Those attributes and objects are completely isolated between different
1170 tracing sessions.
1171
1172 A tracing session is analogous to a cash machine session:
1173 the operations you do on the banking system through the cash machine do
1174 not alter the data of other users of the same system. In the case of
1175 the cash machine, a session lasts as long as your bank card is inside.
1176 In the case of LTTng, a tracing session lasts from the `lttng create`
1177 command to the `lttng destroy` command.
1178
1179 [role="img-100"]
1180 .Each Unix user has its own set of tracing sessions.
1181 image::many-sessions.png[]
1182
1183
1184 [[tracing-session-mode]]
1185 ==== Tracing session mode
1186
1187 LTTng can send the generated trace data to different locations. The
1188 _tracing session mode_ dictates where to send it. The following modes
1189 are available in LTTng{nbsp}{revision}:
1190
1191 Local mode::
1192   LTTng writes the traces to the file system of the machine it traces
1193   (target system).
1194
1195 Network streaming mode::
1196   LTTng sends the traces over the network to a
1197   <<lttng-relayd,relay daemon>> running on a remote system.
1198
1199 Snapshot mode::
1200   LTTng does not write the traces by default. Instead, you can request
1201   LTTng to <<taking-a-snapshot,take a snapshot>>, that is, a copy of the
1202   tracing session's current sub-buffers, and to write it to the
1203   target's file system or to send it over the network to a
1204   <<lttng-relayd,relay daemon>> running on a remote system.
1205
1206 Live mode::
1207   This mode is similar to the network streaming mode, but a live
1208   trace viewer can connect to the distant relay daemon to
1209   <<lttng-live,view event records as LTTng generates them>>.
1210
1211
1212 [[domain]]
1213 === Tracing domain
1214
1215 A _tracing domain_ is a namespace for event sources. A tracing domain
1216 has its own properties and features.
1217
1218 There are currently five available tracing domains:
1219
1220 * Linux kernel
1221 * User space
1222 * `java.util.logging` (JUL)
1223 * log4j
1224 * Python
1225
1226 You must specify a tracing domain when using some commands to avoid
1227 ambiguity. For example, since all the domains support named tracepoints
1228 as event sources (instrumentation points that you manually insert in the
1229 source code), you need to specify a tracing domain when
1230 <<enabling-disabling-events,creating an event rule>> because all the
1231 tracing domains could have tracepoints with the same names.
1232
1233 Some features are reserved to specific tracing domains. Dynamic function
1234 entry and return instrumentation points, for example, are currently only
1235 supported in the Linux kernel tracing domain, but support for other
1236 tracing domains could be added in the future.
1237
1238 You can create <<channel,channels>> in the Linux kernel and user space
1239 tracing domains. The other tracing domains have a single default
1240 channel.
1241
1242
1243 [[channel]]
1244 === Channel and ring buffer
1245
1246 A _channel_ is an object which is responsible for a set of ring buffers.
1247 Each ring buffer is divided into multiple sub-buffers. When an LTTng
1248 tracer emits an event, it can record it to one or more
1249 sub-buffers. The attributes of a channel determine what to do when
1250 there's no space left for a new event record because all sub-buffers
1251 are full, where to send a full sub-buffer, and other behaviours.
1252
1253 A channel is always associated to a <<domain,tracing domain>>. The
1254 `java.util.logging` (JUL), log4j, and Python tracing domains each have
1255 a default channel which you cannot configure.
1256
1257 A channel also owns <<event,event rules>>. When an LTTng tracer emits
1258 an event, it records it to the sub-buffers of all
1259 the enabled channels with a satisfied event rule, as long as those
1260 channels are part of active <<tracing-session,tracing sessions>>.
1261
1262
1263 [[channel-buffering-schemes]]
1264 ==== Per-user vs. per-process buffering schemes
1265
1266 A channel has at least one ring buffer _per CPU_. LTTng always
1267 records an event to the ring buffer associated to the CPU on which it
1268 occurs.
1269
1270 Two _buffering schemes_ are available when you
1271 <<enabling-disabling-channels,create a channel>> in the
1272 user space <<domain,tracing domain>>:
1273
1274 Per-user buffering::
1275   Allocate one set of ring buffers--one per CPU--shared by all the
1276   instrumented processes of each Unix user.
1277 +
1278 --
1279 [role="img-100"]
1280 .Per-user buffering scheme.
1281 image::per-user-buffering.png[]
1282 --
1283
1284 Per-process buffering::
1285   Allocate one set of ring buffers--one per CPU--for each
1286   instrumented process.
1287 +
1288 --
1289 [role="img-100"]
1290 .Per-process buffering scheme.
1291 image::per-process-buffering.png[]
1292 --
1293 +
1294 The per-process buffering scheme tends to consume more memory than the
1295 per-user option because systems generally have more instrumented
1296 processes than Unix users running instrumented processes. However, the
1297 per-process buffering scheme ensures that one process having a high
1298 event throughput won't fill all the shared sub-buffers of the same
1299 user, only its own.
1300
1301 The Linux kernel tracing domain has only one available buffering scheme
1302 which is to allocate a single set of ring buffers for the whole system.
1303 This scheme is similar to the per-user option, but with a single, global
1304 user "running" the kernel.
1305
1306
1307 [[channel-overwrite-mode-vs-discard-mode]]
1308 ==== Overwrite vs. discard event record loss modes
1309
1310 When an event occurs, LTTng records it to a specific sub-buffer (yellow
1311 arc in the following animations) of a specific channel's ring buffer.
1312 When there's no space left in a sub-buffer, the tracer marks it as
1313 consumable (red) and another, empty sub-buffer starts receiving the
1314 following event records. A <<lttng-consumerd,consumer daemon>>
1315 eventually consumes the marked sub-buffer (returns to white).
1316
1317 [NOTE]
1318 [role="docsvg-channel-subbuf-anim"]
1319 ====
1320 {note-no-anim}
1321 ====
1322
1323 In an ideal world, sub-buffers are consumed faster than they are filled,
1324 as it is the case in the previous animation. In the real world,
1325 however, all sub-buffers can be full at some point, leaving no space to
1326 record the following events.
1327
1328 By default, LTTng-modules and LTTng-UST are _non-blocking_ tracers: when
1329 no empty sub-buffer is available, it is acceptable to lose event records
1330 when the alternative would be to cause substantial delays in the
1331 instrumented application's execution. LTTng privileges performance over
1332 integrity; it aims at perturbing the target system as little as possible
1333 in order to make tracing of subtle race conditions and rare interrupt
1334 cascades possible.
1335
1336 Since LTTng{nbsp}2.10, the LTTng user space tracer, LTTng-UST, supports
1337 a _blocking mode_. See the <<blocking-timeout-example,blocking timeout
1338 example>> to learn how to use the blocking mode.
1339
1340 When it comes to losing event records because no empty sub-buffer is
1341 available, or because the <<opt-blocking-timeout,blocking timeout>> is
1342 reached, the channel's _event record loss mode_ determines what to do.
1343 The available event record loss modes are:
1344
1345 Discard mode::
1346   Drop the newest event records until a the tracer releases a
1347   sub-buffer.
1348 +
1349 This is the only available mode when you specify a
1350 <<opt-blocking-timeout,blocking timeout>>.
1351
1352 Overwrite mode::
1353   Clear the sub-buffer containing the oldest event records and start
1354   writing the newest event records there.
1355 +
1356 This mode is sometimes called _flight recorder mode_ because it's
1357 similar to a
1358 https://en.wikipedia.org/wiki/Flight_recorder[flight recorder]:
1359 always keep a fixed amount of the latest data.
1360
1361 Which mechanism you should choose depends on your context: prioritize
1362 the newest or the oldest event records in the ring buffer?
1363
1364 Beware that, in overwrite mode, the tracer abandons a _whole sub-buffer_
1365 as soon as a there's no space left for a new event record, whereas in
1366 discard mode, the tracer only discards the event record that doesn't
1367 fit.
1368
1369 In discard mode, LTTng increments a count of lost event records when an
1370 event record is lost and saves this count to the trace. In overwrite
1371 mode, since LTTng{nbsp}2.8, LTTng increments a count of lost sub-buffers
1372 when a sub-buffer is lost and saves this count to the trace. In this
1373 mode, LTTng does not write to the trace the exact number of lost event
1374 records in those lost sub-buffers. Trace analyses can use the trace's
1375 saved discarded event record and sub-buffer counts to decide whether or
1376 not to perform the analyses even if trace data is known to be missing.
1377
1378 There are a few ways to decrease your probability of losing event
1379 records.
1380 <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>> shows
1381 how you can fine-tune the sub-buffer count and size of a channel to
1382 virtually stop losing event records, though at the cost of greater
1383 memory usage.
1384
1385
1386 [[channel-subbuf-size-vs-subbuf-count]]
1387 ==== Sub-buffer count and size
1388
1389 When you <<enabling-disabling-channels,create a channel>>, you can
1390 set its number of sub-buffers and their size.
1391
1392 Note that there is noticeable CPU overhead introduced when
1393 switching sub-buffers (marking a full one as consumable and switching
1394 to an empty one for the following events to be recorded). Knowing this,
1395 the following list presents a few practical situations along with how
1396 to configure the sub-buffer count and size for them:
1397
1398 * **High event throughput**: In general, prefer bigger sub-buffers to
1399   lower the risk of losing event records.
1400 +
1401 Having bigger sub-buffers also ensures a lower
1402 <<channel-switch-timer,sub-buffer switching frequency>>.
1403 +
1404 The number of sub-buffers is only meaningful if you create the channel
1405 in overwrite mode: in this case, if a sub-buffer overwrite happens, the
1406 other sub-buffers are left unaltered.
1407
1408 * **Low event throughput**: In general, prefer smaller sub-buffers
1409   since the risk of losing event records is low.
1410 +
1411 Because events occur less frequently, the sub-buffer switching frequency
1412 should remain low and thus the tracer's overhead should not be a
1413 problem.
1414
1415 * **Low memory system**: If your target system has a low memory
1416   limit, prefer fewer first, then smaller sub-buffers.
1417 +
1418 Even if the system is limited in memory, you want to keep the
1419 sub-buffers as big as possible to avoid a high sub-buffer switching
1420 frequency.
1421
1422 Note that LTTng uses http://diamon.org/ctf/[CTF] as its trace format,
1423 which means event data is very compact. For example, the average
1424 LTTng kernel event record weights about 32{nbsp}bytes. Thus, a
1425 sub-buffer size of 1{nbsp}MiB is considered big.
1426
1427 The previous situations highlight the major trade-off between a few big
1428 sub-buffers and more, smaller sub-buffers: sub-buffer switching
1429 frequency vs. how much data is lost in overwrite mode. Assuming a
1430 constant event throughput and using the overwrite mode, the two
1431 following configurations have the same ring buffer total size:
1432
1433 [NOTE]
1434 [role="docsvg-channel-subbuf-size-vs-count-anim"]
1435 ====
1436 {note-no-anim}
1437 ====
1438
1439 * **Two sub-buffers of 4{nbsp}MiB each**: Expect a very low sub-buffer
1440   switching frequency, but if a sub-buffer overwrite happens, half of
1441   the event records so far (4{nbsp}MiB) are definitely lost.
1442 * **Eight sub-buffers of 1{nbsp}MiB each**: Expect four times the tracer's
1443   overhead as the previous configuration, but if a sub-buffer
1444   overwrite happens, only the eighth of event records so far are
1445   definitely lost.
1446
1447 In discard mode, the sub-buffers count parameter is pointless: use two
1448 sub-buffers and set their size according to the requirements of your
1449 situation.
1450
1451
1452 [[channel-switch-timer]]
1453 ==== Switch timer period
1454
1455 The _switch timer period_ is an important configurable attribute of
1456 a channel to ensure periodic sub-buffer flushing.
1457
1458 When the _switch timer_ expires, a sub-buffer switch happens. You can
1459 set the switch timer period attribute when you
1460 <<enabling-disabling-channels,create a channel>> to ensure that LTTng
1461 consumes and commits trace data to trace files or to a distant relay
1462 daemon periodically in case of a low event throughput.
1463
1464 [NOTE]
1465 [role="docsvg-channel-switch-timer"]
1466 ====
1467 {note-no-anim}
1468 ====
1469
1470 This attribute is also convenient when you use big sub-buffers to cope
1471 with a sporadic high event throughput, even if the throughput is
1472 normally low.
1473
1474
1475 [[channel-read-timer]]
1476 ==== Read timer period
1477
1478 By default, the LTTng tracers use a notification mechanism to signal a
1479 full sub-buffer so that a consumer daemon can consume it. When such
1480 notifications must be avoided, for example in real-time applications,
1481 you can use the channel's _read timer_ instead. When the read timer
1482 fires, the <<lttng-consumerd,consumer daemon>> checks for full,
1483 consumable sub-buffers.
1484
1485
1486 [[tracefile-rotation]]
1487 ==== Trace file count and size
1488
1489 By default, trace files can grow as large as needed. You can set the
1490 maximum size of each trace file that a channel writes when you
1491 <<enabling-disabling-channels,create a channel>>. When the size of
1492 a trace file reaches the channel's fixed maximum size, LTTng creates
1493 another file to contain the next event records. LTTng appends a file
1494 count to each trace file name in this case.
1495
1496 If you set the trace file size attribute when you create a channel, the
1497 maximum number of trace files that LTTng creates is _unlimited_ by
1498 default. To limit them, you can also set a maximum number of trace
1499 files. When the number of trace files reaches the channel's fixed
1500 maximum count, the oldest trace file is overwritten. This mechanism is
1501 called _trace file rotation_.
1502
1503 Even if you don't limit the trace file count, you cannot assume that
1504 LTTng doesn't manage any trace file. In other words, there is no safe
1505 way to know if LTTng still holds a given trace file open with the trace
1506 file rotation feature. The only way to obtain an unmanaged,
1507 self-contained LTTng trace before you
1508 <<creating-destroying-tracing-sessions,destroy>> the tracing session is
1509 with the <<session-rotation,tracing session rotation>> feature
1510 (available since LTTng{nbsp}2.11).
1511
1512
1513 [[event]]
1514 === Instrumentation point, event rule, event, and event record
1515
1516 An _event rule_ is a set of conditions which must be **all** satisfied
1517 for LTTng to record an occuring event.
1518
1519 You set the conditions when you <<enabling-disabling-events,create
1520 an event rule>>.
1521
1522 You always attach an event rule to <<channel,channel>> when you create
1523 it.
1524
1525 When an event passes the conditions of an event rule, LTTng records it
1526 in one of the attached channel's sub-buffers.
1527
1528 The available conditions, as of LTTng{nbsp}{revision}, are:
1529
1530 * The event rule _is enabled_.
1531 * The instrumentation point's type _is{nbsp}T_.
1532 * The instrumentation point's name (sometimes called _event name_)
1533   _matches{nbsp}N_, but _is not{nbsp}E_.
1534 * The instrumentation point's log level _is as severe as{nbsp}L_, or
1535   _is exactly{nbsp}L_.
1536 * The fields of the event's payload _satisfy_ a filter
1537   expression{nbsp}__F__.
1538
1539 As you can see, all the conditions but the dynamic filter are related to
1540 the event rule's status or to the instrumentation point, not to the
1541 occurring events. This is why, without a filter, checking if an event
1542 passes an event rule is not a dynamic task: when you create or modify an
1543 event rule, all the tracers of its tracing domain enable or disable the
1544 instrumentation points themselves once. This is possible because the
1545 attributes of an instrumentation point (type, name, and log level) are
1546 defined statically. In other words, without a dynamic filter, the tracer
1547 _does not evaluate_ the arguments of an instrumentation point unless it
1548 matches an enabled event rule.
1549
1550 Note that, for LTTng to record an event, the <<channel,channel>> to
1551 which a matching event rule is attached must also be enabled, and the
1552 <<tracing-session,tracing session>> owning this channel must be active
1553 (started).
1554
1555 [role="img-100"]
1556 .Logical path from an instrumentation point to an event record.
1557 image::event-rule.png[]
1558
1559 .Event, event record, or event rule?
1560 ****
1561 With so many similar terms, it's easy to get confused.
1562
1563 An **event** is the consequence of the execution of an _instrumentation
1564 point_, like a tracepoint that you manually place in some source code,
1565 or a Linux kernel kprobe. An event is said to _occur_ at a specific
1566 time. Different actions can be taken upon the occurrence of an event,
1567 like record the event's payload to a buffer.
1568
1569 An **event record** is the representation of an event in a sub-buffer. A
1570 tracer is responsible for capturing the payload of an event, current
1571 context variables, the event's ID, and the event's timestamp. LTTng
1572 can append this sub-buffer to a trace file.
1573
1574 An **event rule** is a set of conditions which must _all_ be satisfied
1575 for LTTng to record an occuring event. Events still occur without
1576 satisfying event rules, but LTTng does not record them.
1577 ****
1578
1579
1580 [[plumbing]]
1581 == Components of noch:{LTTng}
1582
1583 The second _T_ in _LTTng_ stands for _toolkit_: it would be wrong
1584 to call LTTng a simple _tool_ since it is composed of multiple
1585 interacting components. This section describes those components,
1586 explains their respective roles, and shows how they connect together to
1587 form the LTTng ecosystem.
1588
1589 The following diagram shows how the most important components of LTTng
1590 interact with user applications, the Linux kernel, and you:
1591
1592 [role="img-100"]
1593 .Control and trace data paths between LTTng components.
1594 image::plumbing.png[]
1595
1596 The LTTng project incorporates:
1597
1598 * **LTTng-tools**: Libraries and command-line interface to
1599   control tracing sessions.
1600 ** <<lttng-sessiond,Session daemon>> (man:lttng-sessiond(8)).
1601 ** <<lttng-consumerd,Consumer daemon>> (cmd:lttng-consumerd).
1602 ** <<lttng-relayd,Relay daemon>> (man:lttng-relayd(8)).
1603 ** <<liblttng-ctl-lttng,Tracing control library>> (`liblttng-ctl`).
1604 ** <<lttng-cli,Tracing control command-line tool>> (man:lttng(1)).
1605 * **LTTng-UST**: Libraries and Java/Python packages to trace user
1606   applications.
1607 ** <<lttng-ust,User space tracing library>> (`liblttng-ust`) and its
1608    headers to instrument and trace any native user application.
1609 ** <<prebuilt-ust-helpers,Preloadable user space tracing helpers>>:
1610 *** `liblttng-ust-libc-wrapper`
1611 *** `liblttng-ust-pthread-wrapper`
1612 *** `liblttng-ust-cyg-profile`
1613 *** `liblttng-ust-cyg-profile-fast`
1614 *** `liblttng-ust-dl`
1615 ** User space tracepoint provider source files generator command-line
1616    tool (man:lttng-gen-tp(1)).
1617 ** <<lttng-ust-agents,LTTng-UST Java agent>> to instrument and trace
1618    Java applications using `java.util.logging` or
1619    Apache log4j{nbsp}1.2 logging.
1620 ** <<lttng-ust-agents,LTTng-UST Python agent>> to instrument
1621    Python applications using the standard `logging` package.
1622 * **LTTng-modules**: <<lttng-modules,Linux kernel modules>> to trace
1623   the kernel.
1624 ** LTTng kernel tracer module.
1625 ** Tracing ring buffer kernel modules.
1626 ** Probe kernel modules.
1627 ** LTTng logger kernel module.
1628
1629
1630 [[lttng-cli]]
1631 === Tracing control command-line interface
1632
1633 [role="img-100"]
1634 .The tracing control command-line interface.
1635 image::plumbing-lttng-cli.png[]
1636
1637 The _man:lttng(1) command-line tool_ is the standard user interface to
1638 control LTTng <<tracing-session,tracing sessions>>. The cmd:lttng tool
1639 is part of LTTng-tools.
1640
1641 The cmd:lttng tool is linked with
1642 <<liblttng-ctl-lttng,`liblttng-ctl`>> to communicate with
1643 one or more <<lttng-sessiond,session daemons>> behind the scenes.
1644
1645 The cmd:lttng tool has a Git-like interface:
1646
1647 [role="term"]
1648 ----
1649 $ lttng <GENERAL OPTIONS> <COMMAND> <COMMAND OPTIONS>
1650 ----
1651
1652 The <<controlling-tracing,Tracing control>> section explores the
1653 available features of LTTng using the cmd:lttng tool.
1654
1655
1656 [[liblttng-ctl-lttng]]
1657 === Tracing control library
1658
1659 [role="img-100"]
1660 .The tracing control library.
1661 image::plumbing-liblttng-ctl.png[]
1662
1663 The _LTTng control library_, `liblttng-ctl`, is used to communicate
1664 with a <<lttng-sessiond,session daemon>> using a C API that hides the
1665 underlying protocol's details. `liblttng-ctl` is part of LTTng-tools.
1666
1667 The <<lttng-cli,cmd:lttng command-line tool>>
1668 is linked with `liblttng-ctl`.
1669
1670 You can use `liblttng-ctl` in C or $$C++$$ source code by including its
1671 "master" header:
1672
1673 [source,c]
1674 ----
1675 #include <lttng/lttng.h>
1676 ----
1677
1678 Some objects are referenced by name (C string), such as tracing
1679 sessions, but most of them require to create a handle first using
1680 `lttng_create_handle()`.
1681
1682 The best available developer documentation for `liblttng-ctl` is, as of
1683 LTTng{nbsp}{revision}, its installed header files. Every function and
1684 structure is thoroughly documented.
1685
1686
1687 [[lttng-ust]]
1688 === User space tracing library
1689
1690 [role="img-100"]
1691 .The user space tracing library.
1692 image::plumbing-liblttng-ust.png[]
1693
1694 The _user space tracing library_, `liblttng-ust` (see man:lttng-ust(3)),
1695 is the LTTng user space tracer. It receives commands from a
1696 <<lttng-sessiond,session daemon>>, for example to
1697 enable and disable specific instrumentation points, and writes event
1698 records to ring buffers shared with a
1699 <<lttng-consumerd,consumer daemon>>.
1700 `liblttng-ust` is part of LTTng-UST.
1701
1702 Public C header files are installed beside `liblttng-ust` to
1703 instrument any <<c-application,C or $$C++$$ application>>.
1704
1705 <<lttng-ust-agents,LTTng-UST agents>>, which are regular Java and Python
1706 packages, use their own library providing tracepoints which is
1707 linked with `liblttng-ust`.
1708
1709 An application or library does not have to initialize `liblttng-ust`
1710 manually: its constructor does the necessary tasks to properly register
1711 to a session daemon. The initialization phase also enables the
1712 instrumentation points matching the <<event,event rules>> that you
1713 already created.
1714
1715
1716 [[lttng-ust-agents]]
1717 === User space tracing agents
1718
1719 [role="img-100"]
1720 .The user space tracing agents.
1721 image::plumbing-lttng-ust-agents.png[]
1722
1723 The _LTTng-UST Java and Python agents_ are regular Java and Python
1724 packages which add LTTng tracing capabilities to the
1725 native logging frameworks. The LTTng-UST agents are part of LTTng-UST.
1726
1727 In the case of Java, the
1728 https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[`java.util.logging`
1729 core logging facilities] and
1730 https://logging.apache.org/log4j/1.2/[Apache log4j{nbsp}1.2] are supported.
1731 Note that Apache Log4{nbsp}2 is not supported.
1732
1733 In the case of Python, the standard
1734 https://docs.python.org/3/library/logging.html[`logging`] package
1735 is supported. Both Python{nbsp}2 and Python{nbsp}3 modules can import the
1736 LTTng-UST Python agent package.
1737
1738 The applications using the LTTng-UST agents are in the
1739 `java.util.logging` (JUL),
1740 log4j, and Python <<domain,tracing domains>>.
1741
1742 Both agents use the same mechanism to trace the log statements. When an
1743 agent initializes, it creates a log handler that attaches to the root
1744 logger. The agent also registers to a <<lttng-sessiond,session daemon>>.
1745 When the application executes a log statement, the root logger passes it
1746 to the agent's log handler. The agent's log handler calls a native
1747 function in a tracepoint provider package shared library linked with
1748 <<lttng-ust,`liblttng-ust`>>, passing the formatted log message and
1749 other fields, like its logger name and its log level. This native
1750 function contains a user space instrumentation point, hence tracing the
1751 log statement.
1752
1753 The log level condition of an
1754 <<event,event rule>> is considered when tracing
1755 a Java or a Python application, and it's compatible with the standard
1756 JUL, log4j, and Python log levels.
1757
1758
1759 [[lttng-modules]]
1760 === LTTng kernel modules
1761
1762 [role="img-100"]
1763 .The LTTng kernel modules.
1764 image::plumbing-lttng-modules.png[]
1765
1766 The _LTTng kernel modules_ are a set of Linux kernel modules
1767 which implement the kernel tracer of the LTTng project. The LTTng
1768 kernel modules are part of LTTng-modules.
1769
1770 The LTTng kernel modules include:
1771
1772 * A set of _probe_ modules.
1773 +
1774 Each module attaches to a specific subsystem
1775 of the Linux kernel using its tracepoint instrument points. There are
1776 also modules to attach to the entry and return points of the Linux
1777 system call functions.
1778
1779 * _Ring buffer_ modules.
1780 +
1781 A ring buffer implementation is provided as kernel modules. The LTTng
1782 kernel tracer writes to the ring buffer; a
1783 <<lttng-consumerd,consumer daemon>> reads from the ring buffer.
1784
1785 * The _LTTng kernel tracer_ module.
1786 * The _LTTng logger_ module.
1787 +
1788 The LTTng logger module implements the special path:{/proc/lttng-logger}
1789 file so that any executable can generate LTTng events by opening and
1790 writing to this file.
1791 +
1792 See <<proc-lttng-logger-abi,LTTng logger>>.
1793
1794 Generally, you do not have to load the LTTng kernel modules manually
1795 (using man:modprobe(8), for example): a root <<lttng-sessiond,session
1796 daemon>> loads the necessary modules when starting. If you have extra
1797 probe modules, you can specify to load them to the session daemon on
1798 the command line.
1799
1800 The LTTng kernel modules are installed in
1801 +/usr/lib/modules/__release__/extra+ by default, where +__release__+ is
1802 the kernel release (see `uname --kernel-release`).
1803
1804
1805 [[lttng-sessiond]]
1806 === Session daemon
1807
1808 [role="img-100"]
1809 .The session daemon.
1810 image::plumbing-sessiond.png[]
1811
1812 The _session daemon_, man:lttng-sessiond(8), is a daemon responsible for
1813 managing tracing sessions and for controlling the various components of
1814 LTTng. The session daemon is part of LTTng-tools.
1815
1816 The session daemon sends control requests to and receives control
1817 responses from:
1818
1819 * The <<lttng-ust,user space tracing library>>.
1820 +
1821 Any instance of the user space tracing library first registers to
1822 a session daemon. Then, the session daemon can send requests to
1823 this instance, such as:
1824 +
1825 --
1826 ** Get the list of tracepoints.
1827 ** Share an <<event,event rule>> so that the user space tracing library
1828    can enable or disable tracepoints. Amongst the possible conditions
1829    of an event rule is a filter expression which `liblttng-ust` evalutes
1830    when an event occurs.
1831 ** Share <<channel,channel>> attributes and ring buffer locations.
1832 --
1833 +
1834 The session daemon and the user space tracing library use a Unix
1835 domain socket for their communication.
1836
1837 * The <<lttng-ust-agents,user space tracing agents>>.
1838 +
1839 Any instance of a user space tracing agent first registers to
1840 a session daemon. Then, the session daemon can send requests to
1841 this instance, such as:
1842 +
1843 --
1844 ** Get the list of loggers.
1845 ** Enable or disable a specific logger.
1846 --
1847 +
1848 The session daemon and the user space tracing agent use a TCP connection
1849 for their communication.
1850
1851 * The <<lttng-modules,LTTng kernel tracer>>.
1852 * The <<lttng-consumerd,consumer daemon>>.
1853 +
1854 The session daemon sends requests to the consumer daemon to instruct
1855 it where to send the trace data streams, amongst other information.
1856
1857 * The <<lttng-relayd,relay daemon>>.
1858
1859 The session daemon receives commands from the
1860 <<liblttng-ctl-lttng,tracing control library>>.
1861
1862 The root session daemon loads the appropriate
1863 <<lttng-modules,LTTng kernel modules>> on startup. It also spawns
1864 a <<lttng-consumerd,consumer daemon>> as soon as you create
1865 an <<event,event rule>>.
1866
1867 The session daemon does not send and receive trace data: this is the
1868 role of the <<lttng-consumerd,consumer daemon>> and
1869 <<lttng-relayd,relay daemon>>. It does, however, generate the
1870 http://diamon.org/ctf/[CTF] metadata stream.
1871
1872 Each Unix user can have its own session daemon instance. The
1873 tracing sessions which different session daemons manage are completely
1874 independent.
1875
1876 The root user's session daemon is the only one which is
1877 allowed to control the LTTng kernel tracer, and its spawned consumer
1878 daemon is the only one which is allowed to consume trace data from the
1879 LTTng kernel tracer. Note, however, that any Unix user which is a member
1880 of the <<tracing-group,tracing group>> is allowed
1881 to create <<channel,channels>> in the
1882 Linux kernel <<domain,tracing domain>>, and thus to trace the Linux
1883 kernel.
1884
1885 The <<lttng-cli,cmd:lttng command-line tool>> automatically starts a
1886 session daemon when using its `create` command if none is currently
1887 running. You can also start the session daemon manually.
1888
1889
1890 [[lttng-consumerd]]
1891 === Consumer daemon
1892
1893 [role="img-100"]
1894 .The consumer daemon.
1895 image::plumbing-consumerd.png[]
1896
1897 The _consumer daemon_, cmd:lttng-consumerd, is a daemon which shares
1898 ring buffers with user applications or with the LTTng kernel modules to
1899 collect trace data and send it to some location (on disk or to a
1900 <<lttng-relayd,relay daemon>> over the network). The consumer daemon
1901 is part of LTTng-tools.
1902
1903 You do not start a consumer daemon manually: a consumer daemon is always
1904 spawned by a <<lttng-sessiond,session daemon>> as soon as you create an
1905 <<event,event rule>>, that is, before you start tracing. When you kill
1906 its owner session daemon, the consumer daemon also exits because it is
1907 the session daemon's child process. Command-line options of
1908 man:lttng-sessiond(8) target the consumer daemon process.
1909
1910 There are up to two running consumer daemons per Unix user, whereas only
1911 one session daemon can run per user. This is because each process can be
1912 either 32-bit or 64-bit: if the target system runs a mixture of 32-bit
1913 and 64-bit processes, it is more efficient to have separate
1914 corresponding 32-bit and 64-bit consumer daemons. The root user is an
1915 exception: it can have up to _three_ running consumer daemons: 32-bit
1916 and 64-bit instances for its user applications, and one more
1917 reserved for collecting kernel trace data.
1918
1919
1920 [[lttng-relayd]]
1921 === Relay daemon
1922
1923 [role="img-100"]
1924 .The relay daemon.
1925 image::plumbing-relayd.png[]
1926
1927 The _relay daemon_, man:lttng-relayd(8), is a daemon acting as a bridge
1928 between remote session and consumer daemons, local trace files, and a
1929 remote live trace viewer. The relay daemon is part of LTTng-tools.
1930
1931 The main purpose of the relay daemon is to implement a receiver of
1932 <<sending-trace-data-over-the-network,trace data over the network>>.
1933 This is useful when the target system does not have much file system
1934 space to record trace files locally.
1935
1936 The relay daemon is also a server to which a
1937 <<lttng-live,live trace viewer>> can
1938 connect. The live trace viewer sends requests to the relay daemon to
1939 receive trace data as the target system emits events. The
1940 communication protocol is named _LTTng live_; it is used over TCP
1941 connections.
1942
1943 Note that you can start the relay daemon on the target system directly.
1944 This is the setup of choice when the use case is to view events as
1945 the target system emits them without the need of a remote system.
1946
1947
1948 [[instrumenting]]
1949 == [[using-lttng]]Instrumentation
1950
1951 There are many examples of tracing and monitoring in our everyday life:
1952
1953 * You have access to real-time and historical weather reports and
1954   forecasts thanks to weather stations installed around the country.
1955 * You know your heart is safe thanks to an electrocardiogram.
1956 * You make sure not to drive your car too fast and to have enough fuel
1957   to reach your destination thanks to gauges visible on your dashboard.
1958
1959 All the previous examples have something in common: they rely on
1960 **instruments**. Without the electrodes attached to the surface of your
1961 body's skin, cardiac monitoring is futile.
1962
1963 LTTng, as a tracer, is no different from those real life examples. If
1964 you're about to trace a software system or, in other words, record its
1965 history of execution, you better have **instrumentation points** in the
1966 subject you're tracing, that is, the actual software.
1967
1968 Various ways were developed to instrument a piece of software for LTTng
1969 tracing. The most straightforward one is to manually place
1970 instrumentation points, called _tracepoints_, in the software's source
1971 code. It is also possible to add instrumentation points dynamically in
1972 the Linux kernel <<domain,tracing domain>>.
1973
1974 If you're only interested in tracing the Linux kernel, your
1975 instrumentation needs are probably already covered by LTTng's built-in
1976 <<lttng-modules,Linux kernel tracepoints>>. You may also wish to trace a
1977 user application which is already instrumented for LTTng tracing.
1978 In such cases, you can skip this whole section and read the topics of
1979 the <<controlling-tracing,Tracing control>> section.
1980
1981 Many methods are available to instrument a piece of software for LTTng
1982 tracing. They are:
1983
1984 * <<c-application,User space instrumentation for C and $$C++$$
1985   applications>>.
1986 * <<prebuilt-ust-helpers,Prebuilt user space tracing helpers>>.
1987 * <<java-application,User space Java agent>>.
1988 * <<python-application,User space Python agent>>.
1989 * <<proc-lttng-logger-abi,LTTng logger>>.
1990 * <<instrumenting-linux-kernel,LTTng kernel tracepoints>>.
1991
1992
1993 [[c-application]]
1994 === [[cxx-application]]User space instrumentation for C and $$C++$$ applications
1995
1996 The procedure to instrument a C or $$C++$$ user application with
1997 the <<lttng-ust,LTTng user space tracing library>>, `liblttng-ust`, is:
1998
1999 . <<tracepoint-provider,Create the source files of a tracepoint provider
2000   package>>.
2001 . <<probing-the-application-source-code,Add tracepoints to
2002   the application's source code>>.
2003 . <<building-tracepoint-providers-and-user-application,Build and link
2004   a tracepoint provider package and the user application>>.
2005
2006 If you need quick, man:printf(3)-like instrumentation, you can skip
2007 those steps and use <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>>
2008 instead.
2009
2010 IMPORTANT: You need to <<installing-lttng,install>> LTTng-UST to
2011 instrument a user application with `liblttng-ust`.
2012
2013
2014 [[tracepoint-provider]]
2015 ==== Create the source files of a tracepoint provider package
2016
2017 A _tracepoint provider_ is a set of compiled functions which provide
2018 **tracepoints** to an application, the type of instrumentation point
2019 supported by LTTng-UST. Those functions can emit events with
2020 user-defined fields and serialize those events as event records to one
2021 or more LTTng-UST <<channel,channel>> sub-buffers. The `tracepoint()`
2022 macro, which you <<probing-the-application-source-code,insert in a user
2023 application's source code>>, calls those functions.
2024
2025 A _tracepoint provider package_ is an object file (`.o`) or a shared
2026 library (`.so`) which contains one or more tracepoint providers.
2027 Its source files are:
2028
2029 * One or more <<tpp-header,tracepoint provider header>> (`.h`).
2030 * A <<tpp-source,tracepoint provider package source>> (`.c`).
2031
2032 A tracepoint provider package is dynamically linked with `liblttng-ust`,
2033 the LTTng user space tracer, at run time.
2034
2035 [role="img-100"]
2036 .User application linked with `liblttng-ust` and containing a tracepoint provider.
2037 image::ust-app.png[]
2038
2039 NOTE: If you need quick, man:printf(3)-like instrumentation, you can
2040 skip creating and using a tracepoint provider and use
2041 <<tracef,`tracef()`>> or <<tracelog,`tracelog()`>> instead.
2042
2043
2044 [[tpp-header]]
2045 ===== Create a tracepoint provider header file template
2046
2047 A _tracepoint provider header file_ contains the tracepoint
2048 definitions of a tracepoint provider.
2049
2050 To create a tracepoint provider header file:
2051
2052 . Start from this template:
2053 +
2054 --
2055 [source,c]
2056 .Tracepoint provider header file template (`.h` file extension).
2057 ----
2058 #undef TRACEPOINT_PROVIDER
2059 #define TRACEPOINT_PROVIDER provider_name
2060
2061 #undef TRACEPOINT_INCLUDE
2062 #define TRACEPOINT_INCLUDE "./tp.h"
2063
2064 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
2065 #define _TP_H
2066
2067 #include <lttng/tracepoint.h>
2068
2069 /*
2070  * Use TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
2071  * TRACEPOINT_EVENT_INSTANCE(), and TRACEPOINT_LOGLEVEL() here.
2072  */
2073
2074 #endif /* _TP_H */
2075
2076 #include <lttng/tracepoint-event.h>
2077 ----
2078 --
2079
2080 . Replace:
2081 +
2082 * `provider_name` with the name of your tracepoint provider.
2083 * `"tp.h"` with the name of your tracepoint provider header file.
2084
2085 . Below the `#include <lttng/tracepoint.h>` line, put your
2086   <<defining-tracepoints,tracepoint definitions>>.
2087
2088 Your tracepoint provider name must be unique amongst all the possible
2089 tracepoint provider names used on the same target system. We
2090 suggest to include the name of your project or company in the name,
2091 for example, `org_lttng_my_project_tpp`.
2092
2093 TIP: [[lttng-gen-tp]]You can use the man:lttng-gen-tp(1) tool to create
2094 this boilerplate for you. When using cmd:lttng-gen-tp, all you need to
2095 write are the <<defining-tracepoints,tracepoint definitions>>.
2096
2097
2098 [[defining-tracepoints]]
2099 ===== Create a tracepoint definition
2100
2101 A _tracepoint definition_ defines, for a given tracepoint:
2102
2103 * Its **input arguments**. They are the macro parameters that the
2104   `tracepoint()` macro accepts for this particular tracepoint
2105   in the user application's source code.
2106 * Its **output event fields**. They are the sources of event fields
2107   that form the payload of any event that the execution of the
2108   `tracepoint()` macro emits for this particular tracepoint.
2109
2110 You can create a tracepoint definition by using the
2111 `TRACEPOINT_EVENT()` macro below the `#include <lttng/tracepoint.h>`
2112 line in the
2113 <<tpp-header,tracepoint provider header file template>>.
2114
2115 The syntax of the `TRACEPOINT_EVENT()` macro is:
2116
2117 [source,c]
2118 .`TRACEPOINT_EVENT()` macro syntax.
2119 ----
2120 TRACEPOINT_EVENT(
2121     /* Tracepoint provider name */
2122     provider_name,
2123
2124     /* Tracepoint name */
2125     tracepoint_name,
2126
2127     /* Input arguments */
2128     TP_ARGS(
2129         arguments
2130     ),
2131
2132     /* Output event fields */
2133     TP_FIELDS(
2134         fields
2135     )
2136 )
2137 ----
2138
2139 Replace:
2140
2141 * `provider_name` with your tracepoint provider name.
2142 * `tracepoint_name` with your tracepoint name.
2143 * `arguments` with the <<tpp-def-input-args,input arguments>>.
2144 * `fields` with the <<tpp-def-output-fields,output event field>>
2145   definitions.
2146
2147 This tracepoint emits events named `provider_name:tracepoint_name`.
2148
2149 [IMPORTANT]
2150 .Event name's length limitation
2151 ====
2152 The concatenation of the tracepoint provider name and the
2153 tracepoint name must not exceed **254{nbsp}characters**. If it does, the
2154 instrumented application compiles and runs, but LTTng throws multiple
2155 warnings and you could experience serious issues.
2156 ====
2157
2158 [[tpp-def-input-args]]The syntax of the `TP_ARGS()` macro is:
2159
2160 [source,c]
2161 .`TP_ARGS()` macro syntax.
2162 ----
2163 TP_ARGS(
2164     type, arg_name
2165 )
2166 ----
2167
2168 Replace:
2169
2170 * `type` with the C type of the argument.
2171 * `arg_name` with the argument name.
2172
2173 You can repeat `type` and `arg_name` up to 10{nbsp}times to have
2174 more than one argument.
2175
2176 .`TP_ARGS()` usage with three arguments.
2177 ====
2178 [source,c]
2179 ----
2180 TP_ARGS(
2181     int, count,
2182     float, ratio,
2183     const char*, query
2184 )
2185 ----
2186 ====
2187
2188 The `TP_ARGS()` and `TP_ARGS(void)` forms are valid to create a
2189 tracepoint definition with no input arguments.
2190
2191 [[tpp-def-output-fields]]The `TP_FIELDS()` macro contains a list of
2192 `ctf_*()` macros. Each `ctf_*()` macro defines one event field. See
2193 man:lttng-ust(3) for a complete description of the available `ctf_*()`
2194 macros. A `ctf_*()` macro specifies the type, size, and byte order of
2195 one event field.
2196
2197 Each `ctf_*()` macro takes an _argument expression_ parameter. This is a
2198 C expression that the tracer evalutes at the `tracepoint()` macro site
2199 in the application's source code. This expression provides a field's
2200 source of data. The argument expression can include input argument names
2201 listed in the `TP_ARGS()` macro.
2202
2203 Each `ctf_*()` macro also takes a _field name_ parameter. Field names
2204 must be unique within a given tracepoint definition.
2205
2206 Here's a complete tracepoint definition example:
2207
2208 .Tracepoint definition.
2209 ====
2210 The following tracepoint definition defines a tracepoint which takes
2211 three input arguments and has four output event fields.
2212
2213 [source,c]
2214 ----
2215 #include "my-custom-structure.h"
2216
2217 TRACEPOINT_EVENT(
2218     my_provider,
2219     my_tracepoint,
2220     TP_ARGS(
2221         const struct my_custom_structure*, my_custom_structure,
2222         float, ratio,
2223         const char*, query
2224     ),
2225     TP_FIELDS(
2226         ctf_string(query_field, query)
2227         ctf_float(double, ratio_field, ratio)
2228         ctf_integer(int, recv_size, my_custom_structure->recv_size)
2229         ctf_integer(int, send_size, my_custom_structure->send_size)
2230     )
2231 )
2232 ----
2233
2234 You can refer to this tracepoint definition with the `tracepoint()`
2235 macro in your application's source code like this:
2236
2237 [source,c]
2238 ----
2239 tracepoint(my_provider, my_tracepoint,
2240            my_structure, some_ratio, the_query);
2241 ----
2242 ====
2243
2244 NOTE: The LTTng tracer only evaluates tracepoint arguments at run time
2245 if they satisfy an enabled <<event,event rule>>.
2246
2247
2248 [[using-tracepoint-classes]]
2249 ===== Use a tracepoint class
2250
2251 A _tracepoint class_ is a class of tracepoints which share the same
2252 output event field definitions. A _tracepoint instance_ is one
2253 instance of such a defined tracepoint class, with its own tracepoint
2254 name.
2255
2256 The <<defining-tracepoints,`TRACEPOINT_EVENT()` macro>> is actually a
2257 shorthand which defines both a tracepoint class and a tracepoint
2258 instance at the same time.
2259
2260 When you build a tracepoint provider package, the C or $$C++$$ compiler
2261 creates one serialization function for each **tracepoint class**. A
2262 serialization function is responsible for serializing the event fields
2263 of a tracepoint to a sub-buffer when tracing.
2264
2265 For various performance reasons, when your situation requires multiple
2266 tracepoint definitions with different names, but with the same event
2267 fields, we recommend that you manually create a tracepoint class
2268 and instantiate as many tracepoint instances as needed. One positive
2269 effect of such a design, amongst other advantages, is that all
2270 tracepoint instances of the same tracepoint class reuse the same
2271 serialization function, thus reducing
2272 https://en.wikipedia.org/wiki/Cache_pollution[cache pollution].
2273
2274 .Use a tracepoint class and tracepoint instances.
2275 ====
2276 Consider the following three tracepoint definitions:
2277
2278 [source,c]
2279 ----
2280 TRACEPOINT_EVENT(
2281     my_app,
2282     get_account,
2283     TP_ARGS(
2284         int, userid,
2285         size_t, len
2286     ),
2287     TP_FIELDS(
2288         ctf_integer(int, userid, userid)
2289         ctf_integer(size_t, len, len)
2290     )
2291 )
2292
2293 TRACEPOINT_EVENT(
2294     my_app,
2295     get_settings,
2296     TP_ARGS(
2297         int, userid,
2298         size_t, len
2299     ),
2300     TP_FIELDS(
2301         ctf_integer(int, userid, userid)
2302         ctf_integer(size_t, len, len)
2303     )
2304 )
2305
2306 TRACEPOINT_EVENT(
2307     my_app,
2308     get_transaction,
2309     TP_ARGS(
2310         int, userid,
2311         size_t, len
2312     ),
2313     TP_FIELDS(
2314         ctf_integer(int, userid, userid)
2315         ctf_integer(size_t, len, len)
2316     )
2317 )
2318 ----
2319
2320 In this case, we create three tracepoint classes, with one implicit
2321 tracepoint instance for each of them: `get_account`, `get_settings`, and
2322 `get_transaction`. However, they all share the same event field names
2323 and types. Hence three identical, yet independent serialization
2324 functions are created when you build the tracepoint provider package.
2325
2326 A better design choice is to define a single tracepoint class and three
2327 tracepoint instances:
2328
2329 [source,c]
2330 ----
2331 /* The tracepoint class */
2332 TRACEPOINT_EVENT_CLASS(
2333     /* Tracepoint provider name */
2334     my_app,
2335
2336     /* Tracepoint class name */
2337     my_class,
2338
2339     /* Input arguments */
2340     TP_ARGS(
2341         int, userid,
2342         size_t, len
2343     ),
2344
2345     /* Output event fields */
2346     TP_FIELDS(
2347         ctf_integer(int, userid, userid)
2348         ctf_integer(size_t, len, len)
2349     )
2350 )
2351
2352 /* The tracepoint instances */
2353 TRACEPOINT_EVENT_INSTANCE(
2354     /* Tracepoint provider name */
2355     my_app,
2356
2357     /* Tracepoint class name */
2358     my_class,
2359
2360     /* Tracepoint name */
2361     get_account,
2362
2363     /* Input arguments */
2364     TP_ARGS(
2365         int, userid,
2366         size_t, len
2367     )
2368 )
2369 TRACEPOINT_EVENT_INSTANCE(
2370     my_app,
2371     my_class,
2372     get_settings,
2373     TP_ARGS(
2374         int, userid,
2375         size_t, len
2376     )
2377 )
2378 TRACEPOINT_EVENT_INSTANCE(
2379     my_app,
2380     my_class,
2381     get_transaction,
2382     TP_ARGS(
2383         int, userid,
2384         size_t, len
2385     )
2386 )
2387 ----
2388 ====
2389
2390
2391 [[assigning-log-levels]]
2392 ===== Assign a log level to a tracepoint definition
2393
2394 You can assign an optional _log level_ to a
2395 <<defining-tracepoints,tracepoint definition>>.
2396
2397 Assigning different levels of severity to tracepoint definitions can
2398 be useful: when you <<enabling-disabling-events,create an event rule>>,
2399 you can target tracepoints having a log level as severe as a specific
2400 value.
2401
2402 The concept of LTTng-UST log levels is similar to the levels found
2403 in typical logging frameworks:
2404
2405 * In a logging framework, the log level is given by the function
2406   or method name you use at the log statement site: `debug()`,
2407   `info()`, `warn()`, `error()`, and so on.
2408 * In LTTng-UST, you statically assign the log level to a tracepoint
2409   definition; any `tracepoint()` macro invocation which refers to
2410   this definition has this log level.
2411
2412 You can assign a log level to a tracepoint definition with the
2413 `TRACEPOINT_LOGLEVEL()` macro. You must use this macro _after_ the
2414 <<defining-tracepoints,`TRACEPOINT_EVENT()`>> or
2415 <<using-tracepoint-classes,`TRACEPOINT_INSTANCE()`>> macro for a given
2416 tracepoint.
2417
2418 The syntax of the `TRACEPOINT_LOGLEVEL()` macro is:
2419
2420 [source,c]
2421 .`TRACEPOINT_LOGLEVEL()` macro syntax.
2422 ----
2423 TRACEPOINT_LOGLEVEL(provider_name, tracepoint_name, log_level)
2424 ----
2425
2426 Replace:
2427
2428 * `provider_name` with the tracepoint provider name.
2429 * `tracepoint_name` with the tracepoint name.
2430 * `log_level` with the log level to assign to the tracepoint
2431   definition named `tracepoint_name` in the `provider_name`
2432   tracepoint provider.
2433 +
2434 See man:lttng-ust(3) for a list of available log level names.
2435
2436 .Assign the `TRACE_DEBUG_UNIT` log level to a tracepoint definition.
2437 ====
2438 [source,c]
2439 ----
2440 /* Tracepoint definition */
2441 TRACEPOINT_EVENT(
2442     my_app,
2443     get_transaction,
2444     TP_ARGS(
2445         int, userid,
2446         size_t, len
2447     ),
2448     TP_FIELDS(
2449         ctf_integer(int, userid, userid)
2450         ctf_integer(size_t, len, len)
2451     )
2452 )
2453
2454 /* Log level assignment */
2455 TRACEPOINT_LOGLEVEL(my_app, get_transaction, TRACE_DEBUG_UNIT)
2456 ----
2457 ====
2458
2459
2460 [[tpp-source]]
2461 ===== Create a tracepoint provider package source file
2462
2463 A _tracepoint provider package source file_ is a C source file which
2464 includes a <<tpp-header,tracepoint provider header file>> to expand its
2465 macros into event serialization and other functions.
2466
2467 You can always use the following tracepoint provider package source
2468 file template:
2469
2470 [source,c]
2471 .Tracepoint provider package source file template.
2472 ----
2473 #define TRACEPOINT_CREATE_PROBES
2474
2475 #include "tp.h"
2476 ----
2477
2478 Replace `tp.h` with the name of your <<tpp-header,tracepoint provider
2479 header file>> name. You may also include more than one tracepoint
2480 provider header file here to create a tracepoint provider package
2481 holding more than one tracepoint providers.
2482
2483
2484 [[probing-the-application-source-code]]
2485 ==== Add tracepoints to an application's source code
2486
2487 Once you <<tpp-header,create a tracepoint provider header file>>, you
2488 can use the `tracepoint()` macro in your application's
2489 source code to insert the tracepoints that this header
2490 <<defining-tracepoints,defines>>.
2491
2492 The `tracepoint()` macro takes at least two parameters: the tracepoint
2493 provider name and the tracepoint name. The corresponding tracepoint
2494 definition defines the other parameters.
2495
2496 .`tracepoint()` usage.
2497 ====
2498 The following <<defining-tracepoints,tracepoint definition>> defines a
2499 tracepoint which takes two input arguments and has two output event
2500 fields.
2501
2502 [source,c]
2503 .Tracepoint provider header file.
2504 ----
2505 #include "my-custom-structure.h"
2506
2507 TRACEPOINT_EVENT(
2508     my_provider,
2509     my_tracepoint,
2510     TP_ARGS(
2511         int, argc,
2512         const char*, cmd_name
2513     ),
2514     TP_FIELDS(
2515         ctf_string(cmd_name, cmd_name)
2516         ctf_integer(int, number_of_args, argc)
2517     )
2518 )
2519 ----
2520
2521 You can refer to this tracepoint definition with the `tracepoint()`
2522 macro in your application's source code like this:
2523
2524 [source,c]
2525 .Application's source file.
2526 ----
2527 #include "tp.h"
2528
2529 int main(int argc, char* argv[])
2530 {
2531     tracepoint(my_provider, my_tracepoint, argc, argv[0]);
2532
2533     return 0;
2534 }
2535 ----
2536
2537 Note how the application's source code includes
2538 the tracepoint provider header file containing the tracepoint
2539 definitions to use, path:{tp.h}.
2540 ====
2541
2542 .`tracepoint()` usage with a complex tracepoint definition.
2543 ====
2544 Consider this complex tracepoint definition, where multiple event
2545 fields refer to the same input arguments in their argument expression
2546 parameter:
2547
2548 [source,c]
2549 .Tracepoint provider header file.
2550 ----
2551 /* For `struct stat` */
2552 #include <sys/types.h>
2553 #include <sys/stat.h>
2554 #include <unistd.h>
2555
2556 TRACEPOINT_EVENT(
2557     my_provider,
2558     my_tracepoint,
2559     TP_ARGS(
2560         int, my_int_arg,
2561         char*, my_str_arg,
2562         struct stat*, st
2563     ),
2564     TP_FIELDS(
2565         ctf_integer(int, my_constant_field, 23 + 17)
2566         ctf_integer(int, my_int_arg_field, my_int_arg)
2567         ctf_integer(int, my_int_arg_field2, my_int_arg * my_int_arg)
2568         ctf_integer(int, sum4_field, my_str_arg[0] + my_str_arg[1] +
2569                                      my_str_arg[2] + my_str_arg[3])
2570         ctf_string(my_str_arg_field, my_str_arg)
2571         ctf_integer_hex(off_t, size_field, st->st_size)
2572         ctf_float(double, size_dbl_field, (double) st->st_size)
2573         ctf_sequence_text(char, half_my_str_arg_field, my_str_arg,
2574                           size_t, strlen(my_str_arg) / 2)
2575     )
2576 )
2577 ----
2578
2579 You can refer to this tracepoint definition with the `tracepoint()`
2580 macro in your application's source code like this:
2581
2582 [source,c]
2583 .Application's source file.
2584 ----
2585 #define TRACEPOINT_DEFINE
2586 #include "tp.h"
2587
2588 int main(void)
2589 {
2590     struct stat s;
2591
2592     stat("/etc/fstab", &s);
2593     tracepoint(my_provider, my_tracepoint, 23, "Hello, World!", &s);
2594
2595     return 0;
2596 }
2597 ----
2598
2599 If you look at the event record that LTTng writes when tracing this
2600 program, assuming the file size of path:{/etc/fstab} is 301{nbsp}bytes,
2601 it should look like this:
2602
2603 .Event record fields
2604 |====
2605 |Field's name            |Field's value
2606 |`my_constant_field`     |40
2607 |`my_int_arg_field`      |23
2608 |`my_int_arg_field2`     |529
2609 |`sum4_field`            |389
2610 |`my_str_arg_field`      |`Hello, World!`
2611 |`size_field`            |0x12d
2612 |`size_dbl_field`        |301.0
2613 |`half_my_str_arg_field` |`Hello,`
2614 |====
2615 ====
2616
2617 Sometimes, the arguments you pass to `tracepoint()` are expensive to
2618 compute--they use the call stack, for example. To avoid this
2619 computation when the tracepoint is disabled, you can use the
2620 `tracepoint_enabled()` and `do_tracepoint()` macros.
2621
2622 The syntax of the `tracepoint_enabled()` and `do_tracepoint()` macros
2623 is:
2624
2625 [source,c]
2626 .`tracepoint_enabled()` and `do_tracepoint()` macros syntax.
2627 ----
2628 tracepoint_enabled(provider_name, tracepoint_name)
2629 do_tracepoint(provider_name, tracepoint_name, ...)
2630 ----
2631
2632 Replace:
2633
2634 * `provider_name` with the tracepoint provider name.
2635 * `tracepoint_name` with the tracepoint name.
2636
2637 `tracepoint_enabled()` returns a non-zero value if the tracepoint named
2638 `tracepoint_name` from the provider named `provider_name` is enabled
2639 **at run time**.
2640
2641 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
2642 if the tracepoint is enabled. Using `tracepoint()` with
2643 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
2644 the `tracepoint_enabled()` check, thus a race condition is
2645 possible in this situation:
2646
2647 [source,c]
2648 .Possible race condition when using `tracepoint_enabled()` with `tracepoint()`.
2649 ----
2650 if (tracepoint_enabled(my_provider, my_tracepoint)) {
2651     stuff = prepare_stuff();
2652 }
2653
2654 tracepoint(my_provider, my_tracepoint, stuff);
2655 ----
2656
2657 If the tracepoint is enabled after the condition, then `stuff` is not
2658 prepared: the emitted event will either contain wrong data, or the whole
2659 application could crash (segmentation fault, for example).
2660
2661 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have an
2662 `STAP_PROBEV()` call. If you need it, you must emit
2663 this call yourself.
2664
2665
2666 [[building-tracepoint-providers-and-user-application]]
2667 ==== Build and link a tracepoint provider package and an application
2668
2669 Once you have one or more <<tpp-header,tracepoint provider header
2670 files>> and a <<tpp-source,tracepoint provider package source file>>,
2671 you can create the tracepoint provider package by compiling its source
2672 file. From here, multiple build and run scenarios are possible. The
2673 following table shows common application and library configurations
2674 along with the required command lines to achieve them.
2675
2676 In the following diagrams, we use the following file names:
2677
2678 `app`::
2679   Executable application.
2680
2681 `app.o`::
2682   Application's object file.
2683
2684 `tpp.o`::
2685   Tracepoint provider package object file.
2686
2687 `tpp.a`::
2688   Tracepoint provider package archive file.
2689
2690 `libtpp.so`::
2691   Tracepoint provider package shared object file.
2692
2693 `emon.o`::
2694   User library object file.
2695
2696 `libemon.so`::
2697   User library shared object file.
2698
2699 We use the following symbols in the diagrams of table below:
2700
2701 [role="img-100"]
2702 .Symbols used in the build scenario diagrams.
2703 image::ust-sit-symbols.png[]
2704
2705 We assume that path:{.} is part of the env:LD_LIBRARY_PATH environment
2706 variable in the following instructions.
2707
2708 [role="growable ust-scenarios",cols="asciidoc,asciidoc"]
2709 .Common tracepoint provider package scenarios.
2710 |====
2711 |Scenario |Instructions
2712
2713 |
2714 The instrumented application is statically linked with
2715 the tracepoint provider package object.
2716
2717 image::ust-sit+app-linked-with-tp-o+app-instrumented.png[]
2718
2719 |
2720 include::../common/ust-sit-step-tp-o.txt[]
2721
2722 To build the instrumented application:
2723
2724 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2725 +
2726 --
2727 [source,c]
2728 ----
2729 #define TRACEPOINT_DEFINE
2730 ----
2731 --
2732
2733 . Compile the application source file:
2734 +
2735 --
2736 [role="term"]
2737 ----
2738 $ gcc -c app.c
2739 ----
2740 --
2741
2742 . Build the application:
2743 +
2744 --
2745 [role="term"]
2746 ----
2747 $ gcc -o app app.o tpp.o -llttng-ust -ldl
2748 ----
2749 --
2750
2751 To run the instrumented application:
2752
2753 * Start the application:
2754 +
2755 --
2756 [role="term"]
2757 ----
2758 $ ./app
2759 ----
2760 --
2761
2762 |
2763 The instrumented application is statically linked with the
2764 tracepoint provider package archive file.
2765
2766 image::ust-sit+app-linked-with-tp-a+app-instrumented.png[]
2767
2768 |
2769 To create the tracepoint provider package archive file:
2770
2771 . Compile the <<tpp-source,tracepoint provider package source file>>:
2772 +
2773 --
2774 [role="term"]
2775 ----
2776 $ gcc -I. -c tpp.c
2777 ----
2778 --
2779
2780 . Create the tracepoint provider package archive file:
2781 +
2782 --
2783 [role="term"]
2784 ----
2785 $ ar rcs tpp.a tpp.o
2786 ----
2787 --
2788
2789 To build the instrumented application:
2790
2791 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2792 +
2793 --
2794 [source,c]
2795 ----
2796 #define TRACEPOINT_DEFINE
2797 ----
2798 --
2799
2800 . Compile the application source file:
2801 +
2802 --
2803 [role="term"]
2804 ----
2805 $ gcc -c app.c
2806 ----
2807 --
2808
2809 . Build the application:
2810 +
2811 --
2812 [role="term"]
2813 ----
2814 $ gcc -o app app.o tpp.a -llttng-ust -ldl
2815 ----
2816 --
2817
2818 To run the instrumented application:
2819
2820 * Start the application:
2821 +
2822 --
2823 [role="term"]
2824 ----
2825 $ ./app
2826 ----
2827 --
2828
2829 |
2830 The instrumented application is linked with the tracepoint provider
2831 package shared object.
2832
2833 image::ust-sit+app-linked-with-tp-so+app-instrumented.png[]
2834
2835 |
2836 include::../common/ust-sit-step-tp-so.txt[]
2837
2838 To build the instrumented application:
2839
2840 . In path:{app.c}, before including path:{tpp.h}, add the following line:
2841 +
2842 --
2843 [source,c]
2844 ----
2845 #define TRACEPOINT_DEFINE
2846 ----
2847 --
2848
2849 . Compile the application source file:
2850 +
2851 --
2852 [role="term"]
2853 ----
2854 $ gcc -c app.c
2855 ----
2856 --
2857
2858 . Build the application:
2859 +
2860 --
2861 [role="term"]
2862 ----
2863 $ gcc -o app app.o -ldl -L. -ltpp
2864 ----
2865 --
2866
2867 To run the instrumented application:
2868
2869 * Start the application:
2870 +
2871 --
2872 [role="term"]
2873 ----
2874 $ ./app
2875 ----
2876 --
2877
2878 |
2879 The tracepoint provider package shared object is preloaded before the
2880 instrumented application starts.
2881
2882 image::ust-sit+tp-so-preloaded+app-instrumented.png[]
2883
2884 |
2885 include::../common/ust-sit-step-tp-so.txt[]
2886
2887 To build the instrumented application:
2888
2889 . In path:{app.c}, before including path:{tpp.h}, add the
2890   following lines:
2891 +
2892 --
2893 [source,c]
2894 ----
2895 #define TRACEPOINT_DEFINE
2896 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2897 ----
2898 --
2899
2900 . Compile the application source file:
2901 +
2902 --
2903 [role="term"]
2904 ----
2905 $ gcc -c app.c
2906 ----
2907 --
2908
2909 . Build the application:
2910 +
2911 --
2912 [role="term"]
2913 ----
2914 $ gcc -o app app.o -ldl
2915 ----
2916 --
2917
2918 To run the instrumented application with tracing support:
2919
2920 * Preload the tracepoint provider package shared object and
2921   start the application:
2922 +
2923 --
2924 [role="term"]
2925 ----
2926 $ LD_PRELOAD=./libtpp.so ./app
2927 ----
2928 --
2929
2930 To run the instrumented application without tracing support:
2931
2932 * Start the application:
2933 +
2934 --
2935 [role="term"]
2936 ----
2937 $ ./app
2938 ----
2939 --
2940
2941 |
2942 The instrumented application dynamically loads the tracepoint provider
2943 package shared object.
2944
2945 image::ust-sit+app-dlopens-tp-so+app-instrumented.png[]
2946
2947 |
2948 include::../common/ust-sit-step-tp-so.txt[]
2949
2950 To build the instrumented application:
2951
2952 . In path:{app.c}, before including path:{tpp.h}, add the
2953   following lines:
2954 +
2955 --
2956 [source,c]
2957 ----
2958 #define TRACEPOINT_DEFINE
2959 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
2960 ----
2961 --
2962
2963 . Compile the application source file:
2964 +
2965 --
2966 [role="term"]
2967 ----
2968 $ gcc -c app.c
2969 ----
2970 --
2971
2972 . Build the application:
2973 +
2974 --
2975 [role="term"]
2976 ----
2977 $ gcc -o app app.o -ldl
2978 ----
2979 --
2980
2981 To run the instrumented application:
2982
2983 * Start the application:
2984 +
2985 --
2986 [role="term"]
2987 ----
2988 $ ./app
2989 ----
2990 --
2991
2992 |
2993 The application is linked with the instrumented user library.
2994
2995 The instrumented user library is statically linked with the tracepoint
2996 provider package object file.
2997
2998 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-o+lib-instrumented.png[]
2999
3000 |
3001 include::../common/ust-sit-step-tp-o-fpic.txt[]
3002
3003 To build the instrumented user library:
3004
3005 . In path:{emon.c}, before including path:{tpp.h}, add the
3006   following line:
3007 +
3008 --
3009 [source,c]
3010 ----
3011 #define TRACEPOINT_DEFINE
3012 ----
3013 --
3014
3015 . Compile the user library source file:
3016 +
3017 --
3018 [role="term"]
3019 ----
3020 $ gcc -I. -fpic -c emon.c
3021 ----
3022 --
3023
3024 . Build the user library shared object:
3025 +
3026 --
3027 [role="term"]
3028 ----
3029 $ gcc -shared -o libemon.so emon.o tpp.o -llttng-ust -ldl
3030 ----
3031 --
3032
3033 To build the application:
3034
3035 . Compile the application source file:
3036 +
3037 --
3038 [role="term"]
3039 ----
3040 $ gcc -c app.c
3041 ----
3042 --
3043
3044 . Build the application:
3045 +
3046 --
3047 [role="term"]
3048 ----
3049 $ gcc -o app app.o -L. -lemon
3050 ----
3051 --
3052
3053 To run the application:
3054
3055 * Start the application:
3056 +
3057 --
3058 [role="term"]
3059 ----
3060 $ ./app
3061 ----
3062 --
3063
3064 |
3065 The application is linked with the instrumented user library.
3066
3067 The instrumented user library is linked with the tracepoint provider
3068 package shared object.
3069
3070 image::ust-sit+app-linked-with-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3071
3072 |
3073 include::../common/ust-sit-step-tp-so.txt[]
3074
3075 To build the instrumented user library:
3076
3077 . In path:{emon.c}, before including path:{tpp.h}, add the
3078   following line:
3079 +
3080 --
3081 [source,c]
3082 ----
3083 #define TRACEPOINT_DEFINE
3084 ----
3085 --
3086
3087 . Compile the user library source file:
3088 +
3089 --
3090 [role="term"]
3091 ----
3092 $ gcc -I. -fpic -c emon.c
3093 ----
3094 --
3095
3096 . Build the user library shared object:
3097 +
3098 --
3099 [role="term"]
3100 ----
3101 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3102 ----
3103 --
3104
3105 To build the application:
3106
3107 . Compile the application source file:
3108 +
3109 --
3110 [role="term"]
3111 ----
3112 $ gcc -c app.c
3113 ----
3114 --
3115
3116 . Build the application:
3117 +
3118 --
3119 [role="term"]
3120 ----
3121 $ gcc -o app app.o -L. -lemon
3122 ----
3123 --
3124
3125 To run the application:
3126
3127 * Start the application:
3128 +
3129 --
3130 [role="term"]
3131 ----
3132 $ ./app
3133 ----
3134 --
3135
3136 |
3137 The tracepoint provider package shared object is preloaded before the
3138 application starts.
3139
3140 The application is linked with the instrumented user library.
3141
3142 image::ust-sit+tp-so-preloaded+app-linked-with-lib+lib-instrumented.png[]
3143
3144 |
3145 include::../common/ust-sit-step-tp-so.txt[]
3146
3147 To build the instrumented user library:
3148
3149 . In path:{emon.c}, before including path:{tpp.h}, add the
3150   following lines:
3151 +
3152 --
3153 [source,c]
3154 ----
3155 #define TRACEPOINT_DEFINE
3156 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3157 ----
3158 --
3159
3160 . Compile the user library source file:
3161 +
3162 --
3163 [role="term"]
3164 ----
3165 $ gcc -I. -fpic -c emon.c
3166 ----
3167 --
3168
3169 . Build the user library shared object:
3170 +
3171 --
3172 [role="term"]
3173 ----
3174 $ gcc -shared -o libemon.so emon.o -ldl
3175 ----
3176 --
3177
3178 To build the application:
3179
3180 . Compile the application source file:
3181 +
3182 --
3183 [role="term"]
3184 ----
3185 $ gcc -c app.c
3186 ----
3187 --
3188
3189 . Build the application:
3190 +
3191 --
3192 [role="term"]
3193 ----
3194 $ gcc -o app app.o -L. -lemon
3195 ----
3196 --
3197
3198 To run the application with tracing support:
3199
3200 * Preload the tracepoint provider package shared object and
3201   start the application:
3202 +
3203 --
3204 [role="term"]
3205 ----
3206 $ LD_PRELOAD=./libtpp.so ./app
3207 ----
3208 --
3209
3210 To run the application without tracing support:
3211
3212 * Start the application:
3213 +
3214 --
3215 [role="term"]
3216 ----
3217 $ ./app
3218 ----
3219 --
3220
3221 |
3222 The application is linked with the instrumented user library.
3223
3224 The instrumented user library dynamically loads the tracepoint provider
3225 package shared object.
3226
3227 image::ust-sit+app-linked-with-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3228
3229 |
3230 include::../common/ust-sit-step-tp-so.txt[]
3231
3232 To build the instrumented user library:
3233
3234 . In path:{emon.c}, before including path:{tpp.h}, add the
3235   following lines:
3236 +
3237 --
3238 [source,c]
3239 ----
3240 #define TRACEPOINT_DEFINE
3241 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3242 ----
3243 --
3244
3245 . Compile the user library source file:
3246 +
3247 --
3248 [role="term"]
3249 ----
3250 $ gcc -I. -fpic -c emon.c
3251 ----
3252 --
3253
3254 . Build the user library shared object:
3255 +
3256 --
3257 [role="term"]
3258 ----
3259 $ gcc -shared -o libemon.so emon.o -ldl
3260 ----
3261 --
3262
3263 To build the application:
3264
3265 . Compile the application source file:
3266 +
3267 --
3268 [role="term"]
3269 ----
3270 $ gcc -c app.c
3271 ----
3272 --
3273
3274 . Build the application:
3275 +
3276 --
3277 [role="term"]
3278 ----
3279 $ gcc -o app app.o -L. -lemon
3280 ----
3281 --
3282
3283 To run the application:
3284
3285 * Start the application:
3286 +
3287 --
3288 [role="term"]
3289 ----
3290 $ ./app
3291 ----
3292 --
3293
3294 |
3295 The application dynamically loads the instrumented user library.
3296
3297 The instrumented user library is linked with the tracepoint provider
3298 package shared object.
3299
3300 image::ust-sit+app-dlopens-lib+lib-linked-with-tp-so+lib-instrumented.png[]
3301
3302 |
3303 include::../common/ust-sit-step-tp-so.txt[]
3304
3305 To build the instrumented user library:
3306
3307 . In path:{emon.c}, before including path:{tpp.h}, add the
3308   following line:
3309 +
3310 --
3311 [source,c]
3312 ----
3313 #define TRACEPOINT_DEFINE
3314 ----
3315 --
3316
3317 . Compile the user library source file:
3318 +
3319 --
3320 [role="term"]
3321 ----
3322 $ gcc -I. -fpic -c emon.c
3323 ----
3324 --
3325
3326 . Build the user library shared object:
3327 +
3328 --
3329 [role="term"]
3330 ----
3331 $ gcc -shared -o libemon.so emon.o -ldl -L. -ltpp
3332 ----
3333 --
3334
3335 To build the application:
3336
3337 . Compile the application source file:
3338 +
3339 --
3340 [role="term"]
3341 ----
3342 $ gcc -c app.c
3343 ----
3344 --
3345
3346 . Build the application:
3347 +
3348 --
3349 [role="term"]
3350 ----
3351 $ gcc -o app app.o -ldl -L. -lemon
3352 ----
3353 --
3354
3355 To run the application:
3356
3357 * Start the application:
3358 +
3359 --
3360 [role="term"]
3361 ----
3362 $ ./app
3363 ----
3364 --
3365
3366 |
3367 The application dynamically loads the instrumented user library.
3368
3369 The instrumented user library dynamically loads the tracepoint provider
3370 package shared object.
3371
3372 image::ust-sit+app-dlopens-lib+lib-dlopens-tp-so+lib-instrumented.png[]
3373
3374 |
3375 include::../common/ust-sit-step-tp-so.txt[]
3376
3377 To build the instrumented user library:
3378
3379 . In path:{emon.c}, before including path:{tpp.h}, add the
3380   following lines:
3381 +
3382 --
3383 [source,c]
3384 ----
3385 #define TRACEPOINT_DEFINE
3386 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3387 ----
3388 --
3389
3390 . Compile the user library source file:
3391 +
3392 --
3393 [role="term"]
3394 ----
3395 $ gcc -I. -fpic -c emon.c
3396 ----
3397 --
3398
3399 . Build the user library shared object:
3400 +
3401 --
3402 [role="term"]
3403 ----
3404 $ gcc -shared -o libemon.so emon.o -ldl
3405 ----
3406 --
3407
3408 To build the application:
3409
3410 . Compile the application source file:
3411 +
3412 --
3413 [role="term"]
3414 ----
3415 $ gcc -c app.c
3416 ----
3417 --
3418
3419 . Build the application:
3420 +
3421 --
3422 [role="term"]
3423 ----
3424 $ gcc -o app app.o -ldl -L. -lemon
3425 ----
3426 --
3427
3428 To run the application:
3429
3430 * Start the application:
3431 +
3432 --
3433 [role="term"]
3434 ----
3435 $ ./app
3436 ----
3437 --
3438
3439 |
3440 The tracepoint provider package shared object is preloaded before the
3441 application starts.
3442
3443 The application dynamically loads the instrumented user library.
3444
3445 image::ust-sit+tp-so-preloaded+app-dlopens-lib+lib-instrumented.png[]
3446
3447 |
3448 include::../common/ust-sit-step-tp-so.txt[]
3449
3450 To build the instrumented user library:
3451
3452 . In path:{emon.c}, before including path:{tpp.h}, add the
3453   following lines:
3454 +
3455 --
3456 [source,c]
3457 ----
3458 #define TRACEPOINT_DEFINE
3459 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
3460 ----
3461 --
3462
3463 . Compile the user library source file:
3464 +
3465 --
3466 [role="term"]
3467 ----
3468 $ gcc -I. -fpic -c emon.c
3469 ----
3470 --
3471
3472 . Build the user library shared object:
3473 +
3474 --
3475 [role="term"]
3476 ----
3477 $ gcc -shared -o libemon.so emon.o -ldl
3478 ----
3479 --
3480
3481 To build the application:
3482
3483 . Compile the application source file:
3484 +
3485 --
3486 [role="term"]
3487 ----
3488 $ gcc -c app.c
3489 ----
3490 --
3491
3492 . Build the application:
3493 +
3494 --
3495 [role="term"]
3496 ----
3497 $ gcc -o app app.o -L. -lemon
3498 ----
3499 --
3500
3501 To run the application with tracing support:
3502
3503 * Preload the tracepoint provider package shared object and
3504   start the application:
3505 +
3506 --
3507 [role="term"]
3508 ----
3509 $ LD_PRELOAD=./libtpp.so ./app
3510 ----
3511 --
3512
3513 To run the application without tracing support:
3514
3515 * Start the application:
3516 +
3517 --
3518 [role="term"]
3519 ----
3520 $ ./app
3521 ----
3522 --
3523
3524 |
3525 The application is statically linked with the tracepoint provider
3526 package object file.
3527
3528 The application is linked with the instrumented user library.
3529
3530 image::ust-sit+app-linked-with-tp-o+app-linked-with-lib+lib-instrumented.png[]
3531
3532 |
3533 include::../common/ust-sit-step-tp-o.txt[]
3534
3535 To build the instrumented user library:
3536
3537 . In path:{emon.c}, before including path:{tpp.h}, add the
3538   following line:
3539 +
3540 --
3541 [source,c]
3542 ----
3543 #define TRACEPOINT_DEFINE
3544 ----
3545 --
3546
3547 . Compile the user library source file:
3548 +
3549 --
3550 [role="term"]
3551 ----
3552 $ gcc -I. -fpic -c emon.c
3553 ----
3554 --
3555
3556 . Build the user library shared object:
3557 +
3558 --
3559 [role="term"]
3560 ----
3561 $ gcc -shared -o libemon.so emon.o
3562 ----
3563 --
3564
3565 To build the application:
3566
3567 . Compile the application source file:
3568 +
3569 --
3570 [role="term"]
3571 ----
3572 $ gcc -c app.c
3573 ----
3574 --
3575
3576 . Build the application:
3577 +
3578 --
3579 [role="term"]
3580 ----
3581 $ gcc -o app app.o tpp.o -llttng-ust -ldl -L. -lemon
3582 ----
3583 --
3584
3585 To run the instrumented application:
3586
3587 * Start the application:
3588 +
3589 --
3590 [role="term"]
3591 ----
3592 $ ./app
3593 ----
3594 --
3595
3596 |
3597 The application is statically linked with the tracepoint provider
3598 package object file.
3599
3600 The application dynamically loads the instrumented user library.
3601
3602 image::ust-sit+app-linked-with-tp-o+app-dlopens-lib+lib-instrumented.png[]
3603
3604 |
3605 include::../common/ust-sit-step-tp-o.txt[]
3606
3607 To build the application:
3608
3609 . In path:{app.c}, before including path:{tpp.h}, add the following line:
3610 +
3611 --
3612 [source,c]
3613 ----
3614 #define TRACEPOINT_DEFINE
3615 ----
3616 --
3617
3618 . Compile the application source file:
3619 +
3620 --
3621 [role="term"]
3622 ----
3623 $ gcc -c app.c
3624 ----
3625 --
3626
3627 . Build the application:
3628 +
3629 --
3630 [role="term"]
3631 ----
3632 $ gcc -Wl,--export-dynamic -o app app.o tpp.o \
3633     -llttng-ust -ldl
3634 ----
3635 --
3636 +
3637 The `--export-dynamic` option passed to the linker is necessary for the
3638 dynamically loaded library to ``see'' the tracepoint symbols defined in
3639 the application.
3640
3641 To build the instrumented user library:
3642
3643 . Compile the user library source file:
3644 +
3645 --
3646 [role="term"]
3647 ----
3648 $ gcc -I. -fpic -c emon.c
3649 ----
3650 --
3651
3652 . Build the user library shared object:
3653 +
3654 --
3655 [role="term"]
3656 ----
3657 $ gcc -shared -o libemon.so emon.o
3658 ----
3659 --
3660
3661 To run the application:
3662
3663 * Start the application:
3664 +
3665 --
3666 [role="term"]
3667 ----
3668 $ ./app
3669 ----
3670 --
3671 |====
3672
3673
3674 [[using-lttng-ust-with-daemons]]
3675 ===== Use noch:{LTTng-UST} with daemons
3676
3677 If your instrumented application calls man:fork(2), man:clone(2),
3678 or BSD's man:rfork(2), without a following man:exec(3)-family
3679 system call, you must preload the path:{liblttng-ust-fork.so} shared
3680 object when you start the application.
3681
3682 [role="term"]
3683 ----
3684 $ LD_PRELOAD=liblttng-ust-fork.so ./my-app
3685 ----
3686
3687 If your tracepoint provider package is
3688 a shared library which you also preload, you must put both
3689 shared objects in env:LD_PRELOAD:
3690
3691 [role="term"]
3692 ----
3693 $ LD_PRELOAD=liblttng-ust-fork.so:/path/to/tp.so ./my-app
3694 ----
3695
3696
3697 [role="since-2.9"]
3698 [[liblttng-ust-fd]]
3699 ===== Use noch:{LTTng-UST} with applications which close file descriptors that don't belong to them
3700
3701 If your instrumented application closes one or more file descriptors
3702 which it did not open itself, you must preload the
3703 path:{liblttng-ust-fd.so} shared object when you start the application:
3704
3705 [role="term"]
3706 ----
3707 $ LD_PRELOAD=liblttng-ust-fd.so ./my-app
3708 ----
3709
3710 Typical use cases include closing all the file descriptors after
3711 man:fork(2) or man:rfork(2) and buggy applications doing
3712 ``double closes''.
3713
3714
3715 [[lttng-ust-pkg-config]]
3716 ===== Use noch:{pkg-config}
3717
3718 On some distributions, LTTng-UST ships with a
3719 https://www.freedesktop.org/wiki/Software/pkg-config/[pkg-config]
3720 metadata file. If this is your case, then you can use cmd:pkg-config to
3721 build an application on the command line:
3722
3723 [role="term"]
3724 ----
3725 $ gcc -o my-app my-app.o tp.o $(pkg-config --cflags --libs lttng-ust)
3726 ----
3727
3728
3729 [[instrumenting-32-bit-app-on-64-bit-system]]
3730 ===== [[advanced-instrumenting-techniques]]Build a 32-bit instrumented application for a 64-bit target system
3731
3732 In order to trace a 32-bit application running on a 64-bit system,
3733 LTTng must use a dedicated 32-bit
3734 <<lttng-consumerd,consumer daemon>>.
3735
3736 The following steps show how to build and install a 32-bit consumer
3737 daemon, which is _not_ part of the default 64-bit LTTng build, how to
3738 build and install the 32-bit LTTng-UST libraries, and how to build and
3739 link an instrumented 32-bit application in that context.
3740
3741 To build a 32-bit instrumented application for a 64-bit target system,
3742 assuming you have a fresh target system with no installed Userspace RCU
3743 or LTTng packages:
3744
3745 . Download, build, and install a 32-bit version of Userspace RCU:
3746 +
3747 --
3748 [role="term"]
3749 ----
3750 $ cd $(mktemp -d) &&
3751 wget http://lttng.org/files/urcu/userspace-rcu-latest-0.9.tar.bz2 &&
3752 tar -xf userspace-rcu-latest-0.9.tar.bz2 &&
3753 cd userspace-rcu-0.9.* &&
3754 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 &&
3755 make &&
3756 sudo make install &&
3757 sudo ldconfig
3758 ----
3759 --
3760
3761 . Using your distribution's package manager, or from source, install
3762   the following 32-bit versions of the following dependencies of
3763   LTTng-tools and LTTng-UST:
3764 +
3765 --
3766 * https://sourceforge.net/projects/libuuid/[libuuid]
3767 * http://directory.fsf.org/wiki/Popt[popt]
3768 * http://www.xmlsoft.org/[libxml2]
3769 --
3770
3771 . Download, build, and install a 32-bit version of the latest
3772   LTTng-UST{nbsp}{revision}:
3773 +
3774 --
3775 [role="term"]
3776 ----
3777 $ cd $(mktemp -d) &&
3778 wget http://lttng.org/files/lttng-ust/lttng-ust-latest-2.11.tar.bz2 &&
3779 tar -xf lttng-ust-latest-2.11.tar.bz2 &&
3780 cd lttng-ust-2.11.* &&
3781 ./configure --libdir=/usr/local/lib32 \
3782             CFLAGS=-m32 CXXFLAGS=-m32 \
3783             LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' &&
3784 make &&
3785 sudo make install &&
3786 sudo ldconfig
3787 ----
3788 --
3789 +
3790 [NOTE]
3791 ====
3792 Depending on your distribution,
3793 32-bit libraries could be installed at a different location than
3794 `/usr/lib32`. For example, Debian is known to install
3795 some 32-bit libraries in `/usr/lib/i386-linux-gnu`.
3796
3797 In this case, make sure to set `LDFLAGS` to all the
3798 relevant 32-bit library paths, for example:
3799
3800 [role="term"]
3801 ----
3802 $ LDFLAGS='-L/usr/lib/i386-linux-gnu -L/usr/lib32'
3803 ----
3804 ====
3805
3806 . Download the latest LTTng-tools{nbsp}{revision}, build, and install
3807   the 32-bit consumer daemon:
3808 +
3809 --
3810 [role="term"]
3811 ----
3812 $ cd $(mktemp -d) &&
3813 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.11.tar.bz2 &&
3814 tar -xf lttng-tools-latest-2.11.tar.bz2 &&
3815 cd lttng-tools-2.11.* &&
3816 ./configure --libdir=/usr/local/lib32 CFLAGS=-m32 CXXFLAGS=-m32 \
3817             LDFLAGS='-L/usr/local/lib32 -L/usr/lib32' \
3818             --disable-bin-lttng --disable-bin-lttng-crash \
3819             --disable-bin-lttng-relayd --disable-bin-lttng-sessiond &&
3820 make &&
3821 cd src/bin/lttng-consumerd &&
3822 sudo make install &&
3823 sudo ldconfig
3824 ----
3825 --
3826
3827 . From your distribution or from source,
3828   <<installing-lttng,install>> the 64-bit versions of
3829   LTTng-UST and Userspace RCU.
3830 . Download, build, and install the 64-bit version of the
3831   latest LTTng-tools{nbsp}{revision}:
3832 +
3833 --
3834 [role="term"]
3835 ----
3836 $ cd $(mktemp -d) &&
3837 wget http://lttng.org/files/lttng-tools/lttng-tools-latest-2.11.tar.bz2 &&
3838 tar -xf lttng-tools-latest-2.11.tar.bz2 &&
3839 cd lttng-tools-2.11.* &&
3840 ./configure --with-consumerd32-libdir=/usr/local/lib32 \
3841             --with-consumerd32-bin=/usr/local/lib32/lttng/libexec/lttng-consumerd &&
3842 make &&
3843 sudo make install &&
3844 sudo ldconfig
3845 ----
3846 --
3847
3848 . Pass the following options to man:gcc(1), man:g++(1), or man:clang(1)
3849   when linking your 32-bit application:
3850 +
3851 ----
3852 -m32 -L/usr/lib32 -L/usr/local/lib32 \
3853 -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32
3854 ----
3855 +
3856 For example, let's rebuild the quick start example in
3857 <<tracing-your-own-user-application,Trace a user application>> as an
3858 instrumented 32-bit application:
3859 +
3860 --
3861 [role="term"]
3862 ----
3863 $ gcc -m32 -c -I. hello-tp.c
3864 $ gcc -m32 -c hello.c
3865 $ gcc -m32 -o hello hello.o hello-tp.o \
3866     -L/usr/lib32 -L/usr/local/lib32 \
3867     -Wl,-rpath,/usr/lib32,-rpath,/usr/local/lib32 \
3868     -llttng-ust -ldl
3869 ----
3870 --
3871
3872 No special action is required to execute the 32-bit application and
3873 to trace it: use the command-line man:lttng(1) tool as usual.
3874
3875
3876 [role="since-2.5"]
3877 [[tracef]]
3878 ==== Use `tracef()`
3879
3880 man:tracef(3) is a small LTTng-UST API designed for quick,
3881 man:printf(3)-like instrumentation without the burden of
3882 <<tracepoint-provider,creating>> and
3883 <<building-tracepoint-providers-and-user-application,building>>
3884 a tracepoint provider package.
3885
3886 To use `tracef()` in your application:
3887
3888 . In the C or C++ source files where you need to use `tracef()`,
3889   include `<lttng/tracef.h>`:
3890 +
3891 --
3892 [source,c]
3893 ----
3894 #include <lttng/tracef.h>
3895 ----
3896 --
3897
3898 . In the application's source code, use `tracef()` like you would use
3899   man:printf(3):
3900 +
3901 --
3902 [source,c]
3903 ----
3904     /* ... */
3905
3906     tracef("my message: %d (%s)", my_integer, my_string);
3907
3908     /* ... */
3909 ----
3910 --
3911
3912 . Link your application with `liblttng-ust`:
3913 +
3914 --
3915 [role="term"]
3916 ----
3917 $ gcc -o app app.c -llttng-ust
3918 ----
3919 --
3920
3921 To trace the events that `tracef()` calls emit:
3922
3923 * <<enabling-disabling-events,Create an event rule>> which matches the
3924   `lttng_ust_tracef:*` event name:
3925 +
3926 --
3927 [role="term"]
3928 ----
3929 $ lttng enable-event --userspace 'lttng_ust_tracef:*'
3930 ----
3931 --
3932
3933 [IMPORTANT]
3934 .Limitations of `tracef()`
3935 ====
3936 The `tracef()` utility function was developed to make user space tracing
3937 super simple, albeit with notable disadvantages compared to
3938 <<defining-tracepoints,user-defined tracepoints>>:
3939
3940 * All the emitted events have the same tracepoint provider and
3941   tracepoint names, respectively `lttng_ust_tracef` and `event`.
3942 * There is no static type checking.
3943 * The only event record field you actually get, named `msg`, is a string
3944   potentially containing the values you passed to `tracef()`
3945   using your own format string. This also means that you cannot filter
3946   events with a custom expression at run time because there are no
3947   isolated fields.
3948 * Since `tracef()` uses the C standard library's man:vasprintf(3)
3949   function behind the scenes to format the strings at run time, its
3950   expected performance is lower than with user-defined tracepoints,
3951   which do not require a conversion to a string.
3952
3953 Taking this into consideration, `tracef()` is useful for some quick
3954 prototyping and debugging, but you should not consider it for any
3955 permanent and serious applicative instrumentation.
3956 ====
3957
3958
3959 [role="since-2.7"]
3960 [[tracelog]]
3961 ==== Use `tracelog()`
3962
3963 The man:tracelog(3) API is very similar to <<tracef,`tracef()`>>, with
3964 the difference that it accepts an additional log level parameter.
3965
3966 The goal of `tracelog()` is to ease the migration from logging to
3967 tracing.
3968
3969 To use `tracelog()` in your application:
3970
3971 . In the C or C++ source files where you need to use `tracelog()`,
3972   include `<lttng/tracelog.h>`:
3973 +
3974 --
3975 [source,c]
3976 ----
3977 #include <lttng/tracelog.h>
3978 ----
3979 --
3980
3981 . In the application's source code, use `tracelog()` like you would use
3982   man:printf(3), except for the first parameter which is the log
3983   level:
3984 +
3985 --
3986 [source,c]
3987 ----
3988     /* ... */
3989
3990     tracelog(TRACE_WARNING, "my message: %d (%s)",
3991              my_integer, my_string);
3992
3993     /* ... */
3994 ----
3995 --
3996 +
3997 See man:lttng-ust(3) for a list of available log level names.
3998
3999 . Link your application with `liblttng-ust`:
4000 +
4001 --
4002 [role="term"]
4003 ----
4004 $ gcc -o app app.c -llttng-ust
4005 ----
4006 --
4007
4008 To trace the events that `tracelog()` calls emit with a log level
4009 _as severe as_ a specific log level:
4010
4011 * <<enabling-disabling-events,Create an event rule>> which matches the
4012   `lttng_ust_tracelog:*` event name and a minimum level
4013   of severity:
4014 +
4015 --
4016 [role="term"]
4017 ----
4018 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4019                    --loglevel=TRACE_WARNING
4020 ----
4021 --
4022
4023 To trace the events that `tracelog()` calls emit with a
4024 _specific log level_:
4025
4026 * Create an event rule which matches the `lttng_ust_tracelog:*`
4027   event name and a specific log level:
4028 +
4029 --
4030 [role="term"]
4031 ----
4032 $ lttng enable-event --userspace 'lttng_ust_tracelog:*'
4033                    --loglevel-only=TRACE_INFO
4034 ----
4035 --
4036
4037
4038 [[prebuilt-ust-helpers]]
4039 === Prebuilt user space tracing helpers
4040
4041 The LTTng-UST package provides a few helpers in the form or preloadable
4042 shared objects which automatically instrument system functions and
4043 calls.
4044
4045 The helper shared objects are normally found in dir:{/usr/lib}. If you
4046 built LTTng-UST <<building-from-source,from source>>, they are probably
4047 located in dir:{/usr/local/lib}.
4048
4049 The installed user space tracing helpers in LTTng-UST{nbsp}{revision}
4050 are:
4051
4052 path:{liblttng-ust-libc-wrapper.so}::
4053 path:{liblttng-ust-pthread-wrapper.so}::
4054   <<liblttng-ust-libc-pthread-wrapper,C{nbsp}standard library
4055   memory and POSIX threads function tracing>>.
4056
4057 path:{liblttng-ust-cyg-profile.so}::
4058 path:{liblttng-ust-cyg-profile-fast.so}::
4059   <<liblttng-ust-cyg-profile,Function entry and exit tracing>>.
4060
4061 path:{liblttng-ust-dl.so}::
4062   <<liblttng-ust-dl,Dynamic linker tracing>>.
4063
4064 To use a user space tracing helper with any user application:
4065
4066 * Preload the helper shared object when you start the application:
4067 +
4068 --
4069 [role="term"]
4070 ----
4071 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so my-app
4072 ----
4073 --
4074 +
4075 You can preload more than one helper:
4076 +
4077 --
4078 [role="term"]
4079 ----
4080 $ LD_PRELOAD=liblttng-ust-libc-wrapper.so:liblttng-ust-dl.so my-app
4081 ----
4082 --
4083
4084
4085 [role="since-2.3"]
4086 [[liblttng-ust-libc-pthread-wrapper]]
4087 ==== Instrument C standard library memory and POSIX threads functions
4088
4089 The path:{liblttng-ust-libc-wrapper.so} and
4090 path:{liblttng-ust-pthread-wrapper.so} helpers
4091 add instrumentation to some C standard library and POSIX
4092 threads functions.
4093
4094 [role="growable"]
4095 .Functions instrumented by preloading path:{liblttng-ust-libc-wrapper.so}.
4096 |====
4097 |TP provider name |TP name |Instrumented function
4098
4099 .6+|`lttng_ust_libc` |`malloc`         |man:malloc(3)
4100                      |`calloc`         |man:calloc(3)
4101                      |`realloc`        |man:realloc(3)
4102                      |`free`           |man:free(3)
4103                      |`memalign`       |man:memalign(3)
4104                      |`posix_memalign` |man:posix_memalign(3)
4105 |====
4106
4107 [role="growable"]
4108 .Functions instrumented by preloading path:{liblttng-ust-pthread-wrapper.so}.
4109 |====
4110 |TP provider name |TP name |Instrumented function
4111
4112 .4+|`lttng_ust_pthread` |`pthread_mutex_lock_req` |man:pthread_mutex_lock(3p) (request time)
4113                         |`pthread_mutex_lock_acq` |man:pthread_mutex_lock(3p) (acquire time)
4114                         |`pthread_mutex_trylock`  |man:pthread_mutex_trylock(3p)
4115                         |`pthread_mutex_unlock`   |man:pthread_mutex_unlock(3p)
4116 |====
4117
4118 When you preload the shared object, it replaces the functions listed
4119 in the previous tables by wrappers which contain tracepoints and call
4120 the replaced functions.
4121
4122
4123 [[liblttng-ust-cyg-profile]]
4124 ==== Instrument function entry and exit
4125
4126 The path:{liblttng-ust-cyg-profile*.so} helpers can add instrumentation
4127 to the entry and exit points of functions.
4128
4129 man:gcc(1) and man:clang(1) have an option named
4130 https://gcc.gnu.org/onlinedocs/gcc/Instrumentation-Options.html[`-finstrument-functions`]
4131 which generates instrumentation calls for entry and exit to functions.
4132 The LTTng-UST function tracing helpers,
4133 path:{liblttng-ust-cyg-profile.so} and
4134 path:{liblttng-ust-cyg-profile-fast.so}, take advantage of this feature
4135 to add tracepoints to the two generated functions (which contain
4136 `cyg_profile` in their names, hence the helper's name).
4137
4138 To use the LTTng-UST function tracing helper, the source files to
4139 instrument must be built using the `-finstrument-functions` compiler
4140 flag.
4141
4142 There are two versions of the LTTng-UST function tracing helper:
4143
4144 * **path:{liblttng-ust-cyg-profile-fast.so}** is a lightweight variant
4145   that you should only use when it can be _guaranteed_ that the
4146   complete event stream is recorded without any lost event record.
4147   Any kind of duplicate information is left out.
4148 +
4149 Assuming no event record is lost, having only the function addresses on
4150 entry is enough to create a call graph, since an event record always
4151 contains the ID of the CPU that generated it.
4152 +
4153 You can use a tool like man:addr2line(1) to convert function addresses
4154 back to source file names and line numbers.
4155
4156 * **path:{liblttng-ust-cyg-profile.so}** is a more robust variant
4157 which also works in use cases where event records might get discarded or
4158 not recorded from application startup.
4159 In these cases, the trace analyzer needs more information to be
4160 able to reconstruct the program flow.
4161
4162 See man:lttng-ust-cyg-profile(3) to learn more about the instrumentation
4163 points of this helper.
4164
4165 All the tracepoints that this helper provides have the
4166 log level `TRACE_DEBUG_FUNCTION` (see man:lttng-ust(3)).
4167
4168 TIP: It's sometimes a good idea to limit the number of source files that
4169 you compile with the `-finstrument-functions` option to prevent LTTng
4170 from writing an excessive amount of trace data at run time. When using
4171 man:gcc(1), you can use the
4172 `-finstrument-functions-exclude-function-list` option to avoid
4173 instrument entries and exits of specific function names.
4174
4175
4176 [role="since-2.4"]
4177 [[liblttng-ust-dl]]
4178 ==== Instrument the dynamic linker
4179
4180 The path:{liblttng-ust-dl.so} helper adds instrumentation to the
4181 man:dlopen(3) and man:dlclose(3) function calls.
4182
4183 See man:lttng-ust-dl(3) to learn more about the instrumentation points
4184 of this helper.
4185
4186
4187 [role="since-2.4"]
4188 [[java-application]]
4189 === User space Java agent
4190
4191 You can instrument any Java application which uses one of the following
4192 logging frameworks:
4193
4194 * The https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[**`java.util.logging`**]
4195   (JUL) core logging facilities.
4196 * http://logging.apache.org/log4j/1.2/[**Apache log4j{nbsp}1.2**], since
4197   LTTng{nbsp}2.6. Note that Apache Log4j{nbsp}2 is not supported.
4198
4199 [role="img-100"]
4200 .LTTng-UST Java agent imported by a Java application.
4201 image::java-app.png[]
4202
4203 Note that the methods described below are new in LTTng{nbsp}{revision}.
4204 Previous LTTng versions use another technique.
4205
4206 NOTE: We use http://openjdk.java.net/[OpenJDK]{nbsp}8 for development
4207 and https://ci.lttng.org/[continuous integration], thus this version is
4208 directly supported. However, the LTTng-UST Java agent is also tested
4209 with OpenJDK{nbsp}7.
4210
4211
4212 [role="since-2.8"]
4213 [[jul]]
4214 ==== Use the LTTng-UST Java agent for `java.util.logging`
4215
4216 To use the LTTng-UST Java agent in a Java application which uses
4217 `java.util.logging` (JUL):
4218
4219 . In the Java application's source code, import the LTTng-UST
4220   log handler package for `java.util.logging`:
4221 +
4222 --
4223 [source,java]
4224 ----
4225 import org.lttng.ust.agent.jul.LttngLogHandler;
4226 ----
4227 --
4228
4229 . Create an LTTng-UST JUL log handler:
4230 +
4231 --
4232 [source,java]
4233 ----
4234 Handler lttngUstLogHandler = new LttngLogHandler();
4235 ----
4236 --
4237
4238 . Add this handler to the JUL loggers which should emit LTTng events:
4239 +
4240 --
4241 [source,java]
4242 ----
4243 Logger myLogger = Logger.getLogger("some-logger");
4244
4245 myLogger.addHandler(lttngUstLogHandler);
4246 ----
4247 --
4248
4249 . Use `java.util.logging` log statements and configuration as usual.
4250   The loggers with an attached LTTng-UST log handler can emit
4251   LTTng events.
4252
4253 . Before exiting the application, remove the LTTng-UST log handler from
4254   the loggers attached to it and call its `close()` method:
4255 +
4256 --
4257 [source,java]
4258 ----
4259 myLogger.removeHandler(lttngUstLogHandler);
4260 lttngUstLogHandler.close();
4261 ----
4262 --
4263 +
4264 This is not strictly necessary, but it is recommended for a clean
4265 disposal of the handler's resources.
4266
4267 . Include the LTTng-UST Java agent's common and JUL-specific JAR files,
4268   path:{lttng-ust-agent-common.jar} and path:{lttng-ust-agent-jul.jar},
4269   in the
4270   https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4271   path] when you build the Java application.
4272 +
4273 The JAR files are typically located in dir:{/usr/share/java}.
4274 +
4275 IMPORTANT: The LTTng-UST Java agent must be
4276 <<installing-lttng,installed>> for the logging framework your
4277 application uses.
4278
4279 .Use the LTTng-UST Java agent for `java.util.logging`.
4280 ====
4281 [source,java]
4282 .path:{Test.java}
4283 ----
4284 import java.io.IOException;
4285 import java.util.logging.Handler;
4286 import java.util.logging.Logger;
4287 import org.lttng.ust.agent.jul.LttngLogHandler;
4288
4289 public class Test
4290 {
4291     private static final int answer = 42;
4292
4293     public static void main(String[] argv) throws Exception
4294     {
4295         // Create a logger
4296         Logger logger = Logger.getLogger("jello");
4297
4298         // Create an LTTng-UST log handler
4299         Handler lttngUstLogHandler = new LttngLogHandler();
4300
4301         // Add the LTTng-UST log handler to our logger
4302         logger.addHandler(lttngUstLogHandler);
4303
4304         // Log at will!
4305         logger.info("some info");
4306         logger.warning("some warning");
4307         Thread.sleep(500);
4308         logger.finer("finer information; the answer is " + answer);
4309         Thread.sleep(123);
4310         logger.severe("error!");
4311
4312         // Not mandatory, but cleaner
4313         logger.removeHandler(lttngUstLogHandler);
4314         lttngUstLogHandler.close();
4315     }
4316 }
4317 ----
4318
4319 Build this example:
4320
4321 [role="term"]
4322 ----
4323 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4324 ----
4325
4326 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4327 <<enabling-disabling-events,create an event rule>> matching the
4328 `jello` JUL logger, and <<basic-tracing-session-control,start tracing>>:
4329
4330 [role="term"]
4331 ----
4332 $ lttng create
4333 $ lttng enable-event --jul jello
4334 $ lttng start
4335 ----
4336
4337 Run the compiled class:
4338
4339 [role="term"]
4340 ----
4341 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4342 ----
4343
4344 <<basic-tracing-session-control,Stop tracing>> and inspect the
4345 recorded events:
4346
4347 [role="term"]
4348 ----
4349 $ lttng stop
4350 $ lttng view
4351 ----
4352 ====
4353
4354 In the resulting trace, an <<event,event record>> generated by a Java
4355 application using `java.util.logging` is named `lttng_jul:event` and
4356 has the following fields:
4357
4358 `msg`::
4359   Log record's message.
4360
4361 `logger_name`::
4362   Logger name.
4363
4364 `class_name`::
4365   Name of the class in which the log statement was executed.
4366
4367 `method_name`::
4368   Name of the method in which the log statement was executed.
4369
4370 `long_millis`::
4371   Logging time (timestamp in milliseconds).
4372
4373 `int_loglevel`::
4374   Log level integer value.
4375
4376 `int_threadid`::
4377   ID of the thread in which the log statement was executed.
4378
4379 You can use the opt:lttng-enable-event(1):--loglevel or
4380 opt:lttng-enable-event(1):--loglevel-only option of the
4381 man:lttng-enable-event(1) command to target a range of JUL log levels
4382 or a specific JUL log level.
4383
4384
4385 [role="since-2.8"]
4386 [[log4j]]
4387 ==== Use the LTTng-UST Java agent for Apache log4j
4388
4389 To use the LTTng-UST Java agent in a Java application which uses
4390 Apache log4j{nbsp}1.2:
4391
4392 . In the Java application's source code, import the LTTng-UST
4393   log appender package for Apache log4j:
4394 +
4395 --
4396 [source,java]
4397 ----
4398 import org.lttng.ust.agent.log4j.LttngLogAppender;
4399 ----
4400 --
4401
4402 . Create an LTTng-UST log4j log appender:
4403 +
4404 --
4405 [source,java]
4406 ----
4407 Appender lttngUstLogAppender = new LttngLogAppender();
4408 ----
4409 --
4410
4411 . Add this appender to the log4j loggers which should emit LTTng events:
4412 +
4413 --
4414 [source,java]
4415 ----
4416 Logger myLogger = Logger.getLogger("some-logger");
4417
4418 myLogger.addAppender(lttngUstLogAppender);
4419 ----
4420 --
4421
4422 . Use Apache log4j log statements and configuration as usual. The
4423   loggers with an attached LTTng-UST log appender can emit LTTng events.
4424
4425 . Before exiting the application, remove the LTTng-UST log appender from
4426   the loggers attached to it and call its `close()` method:
4427 +
4428 --
4429 [source,java]
4430 ----
4431 myLogger.removeAppender(lttngUstLogAppender);
4432 lttngUstLogAppender.close();
4433 ----
4434 --
4435 +
4436 This is not strictly necessary, but it is recommended for a clean
4437 disposal of the appender's resources.
4438
4439 . Include the LTTng-UST Java agent's common and log4j-specific JAR
4440   files, path:{lttng-ust-agent-common.jar} and
4441   path:{lttng-ust-agent-log4j.jar}, in the
4442   https://docs.oracle.com/javase/tutorial/essential/environment/paths.html[class
4443   path] when you build the Java application.
4444 +
4445 The JAR files are typically located in dir:{/usr/share/java}.
4446 +
4447 IMPORTANT: The LTTng-UST Java agent must be
4448 <<installing-lttng,installed>> for the logging framework your
4449 application uses.
4450
4451 .Use the LTTng-UST Java agent for Apache log4j.
4452 ====
4453 [source,java]
4454 .path:{Test.java}
4455 ----
4456 import org.apache.log4j.Appender;
4457 import org.apache.log4j.Logger;
4458 import org.lttng.ust.agent.log4j.LttngLogAppender;
4459
4460 public class Test
4461 {
4462     private static final int answer = 42;
4463
4464     public static void main(String[] argv) throws Exception
4465     {
4466         // Create a logger
4467         Logger logger = Logger.getLogger("jello");
4468
4469         // Create an LTTng-UST log appender
4470         Appender lttngUstLogAppender = new LttngLogAppender();
4471
4472         // Add the LTTng-UST log appender to our logger
4473         logger.addAppender(lttngUstLogAppender);
4474
4475         // Log at will!
4476         logger.info("some info");
4477         logger.warn("some warning");
4478         Thread.sleep(500);
4479         logger.debug("debug information; the answer is " + answer);
4480         Thread.sleep(123);
4481         logger.fatal("error!");
4482
4483         // Not mandatory, but cleaner
4484         logger.removeAppender(lttngUstLogAppender);
4485         lttngUstLogAppender.close();
4486     }
4487 }
4488
4489 ----
4490
4491 Build this example (`$LOG4JPATH` is the path to the Apache log4j JAR
4492 file):
4493
4494 [role="term"]
4495 ----
4496 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH Test.java
4497 ----
4498
4499 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4500 <<enabling-disabling-events,create an event rule>> matching the
4501 `jello` log4j logger, and <<basic-tracing-session-control,start tracing>>:
4502
4503 [role="term"]
4504 ----
4505 $ lttng create
4506 $ lttng enable-event --log4j jello
4507 $ lttng start
4508 ----
4509
4510 Run the compiled class:
4511
4512 [role="term"]
4513 ----
4514 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-log4j.jar:$LOG4JPATH:. Test
4515 ----
4516
4517 <<basic-tracing-session-control,Stop tracing>> and inspect the
4518 recorded events:
4519
4520 [role="term"]
4521 ----
4522 $ lttng stop
4523 $ lttng view
4524 ----
4525 ====
4526
4527 In the resulting trace, an <<event,event record>> generated by a Java
4528 application using log4j is named `lttng_log4j:event` and
4529 has the following fields:
4530
4531 `msg`::
4532   Log record's message.
4533
4534 `logger_name`::
4535   Logger name.
4536
4537 `class_name`::
4538   Name of the class in which the log statement was executed.
4539
4540 `method_name`::
4541   Name of the method in which the log statement was executed.
4542
4543 `filename`::
4544   Name of the file in which the executed log statement is located.
4545
4546 `line_number`::
4547   Line number at which the log statement was executed.
4548
4549 `timestamp`::
4550   Logging timestamp.
4551
4552 `int_loglevel`::
4553   Log level integer value.
4554
4555 `thread_name`::
4556   Name of the Java thread in which the log statement was executed.
4557
4558 You can use the opt:lttng-enable-event(1):--loglevel or
4559 opt:lttng-enable-event(1):--loglevel-only option of the
4560 man:lttng-enable-event(1) command to target a range of Apache log4j log levels
4561 or a specific log4j log level.
4562
4563
4564 [role="since-2.8"]
4565 [[java-application-context]]
4566 ==== Provide application-specific context fields in a Java application
4567
4568 A Java application-specific context field is a piece of state provided
4569 by the application which <<adding-context,you can add>>, using the
4570 man:lttng-add-context(1) command, to each <<event,event record>>
4571 produced by the log statements of this application.
4572
4573 For example, a given object might have a current request ID variable.
4574 You can create a context information retriever for this object and
4575 assign a name to this current request ID. You can then, using the
4576 man:lttng-add-context(1) command, add this context field by name to
4577 the JUL or log4j <<channel,channel>>.
4578
4579 To provide application-specific context fields in a Java application:
4580
4581 . In the Java application's source code, import the LTTng-UST
4582   Java agent context classes and interfaces:
4583 +
4584 --
4585 [source,java]
4586 ----
4587 import org.lttng.ust.agent.context.ContextInfoManager;
4588 import org.lttng.ust.agent.context.IContextInfoRetriever;
4589 ----
4590 --
4591
4592 . Create a context information retriever class, that is, a class which
4593   implements the `IContextInfoRetriever` interface:
4594 +
4595 --
4596 [source,java]
4597 ----
4598 class MyContextInfoRetriever implements IContextInfoRetriever
4599 {
4600     @Override
4601     public Object retrieveContextInfo(String key)
4602     {
4603         if (key.equals("intCtx")) {
4604             return (short) 17;
4605         } else if (key.equals("strContext")) {
4606             return "context value!";
4607         } else {
4608             return null;
4609         }
4610     }
4611 }
4612 ----
4613 --
4614 +
4615 This `retrieveContextInfo()` method is the only member of the
4616 `IContextInfoRetriever` interface. Its role is to return the current
4617 value of a state by name to create a context field. The names of the
4618 context fields and which state variables they return depends on your
4619 specific scenario.
4620 +
4621 All primitive types and objects are supported as context fields.
4622 When `retrieveContextInfo()` returns an object, the context field
4623 serializer calls its `toString()` method to add a string field to
4624 event records. The method can also return `null`, which means that
4625 no context field is available for the required name.
4626
4627 . Register an instance of your context information retriever class to
4628   the context information manager singleton:
4629 +
4630 --
4631 [source,java]
4632 ----
4633 IContextInfoRetriever cir = new MyContextInfoRetriever();
4634 ContextInfoManager cim = ContextInfoManager.getInstance();
4635 cim.registerContextInfoRetriever("retrieverName", cir);
4636 ----
4637 --
4638
4639 . Before exiting the application, remove your context information
4640   retriever from the context information manager singleton:
4641 +
4642 --
4643 [source,java]
4644 ----
4645 ContextInfoManager cim = ContextInfoManager.getInstance();
4646 cim.unregisterContextInfoRetriever("retrieverName");
4647 ----
4648 --
4649 +
4650 This is not strictly necessary, but it is recommended for a clean
4651 disposal of some manager's resources.
4652
4653 . Build your Java application with LTTng-UST Java agent support as
4654   usual, following the procedure for either the <<jul,JUL>> or
4655   <<log4j,Apache log4j>> framework.
4656
4657
4658 .Provide application-specific context fields in a Java application.
4659 ====
4660 [source,java]
4661 .path:{Test.java}
4662 ----
4663 import java.util.logging.Handler;
4664 import java.util.logging.Logger;
4665 import org.lttng.ust.agent.jul.LttngLogHandler;
4666 import org.lttng.ust.agent.context.ContextInfoManager;
4667 import org.lttng.ust.agent.context.IContextInfoRetriever;
4668
4669 public class Test
4670 {
4671     // Our context information retriever class
4672     private static class MyContextInfoRetriever
4673     implements IContextInfoRetriever
4674     {
4675         @Override
4676         public Object retrieveContextInfo(String key) {
4677             if (key.equals("intCtx")) {
4678                 return (short) 17;
4679             } else if (key.equals("strContext")) {
4680                 return "context value!";
4681             } else {
4682                 return null;
4683             }
4684         }
4685     }
4686
4687     private static final int answer = 42;
4688
4689     public static void main(String args[]) throws Exception
4690     {
4691         // Get the context information manager instance
4692         ContextInfoManager cim = ContextInfoManager.getInstance();
4693
4694         // Create and register our context information retriever
4695         IContextInfoRetriever cir = new MyContextInfoRetriever();
4696         cim.registerContextInfoRetriever("myRetriever", cir);
4697
4698         // Create a logger
4699         Logger logger = Logger.getLogger("jello");
4700
4701         // Create an LTTng-UST log handler
4702         Handler lttngUstLogHandler = new LttngLogHandler();
4703
4704         // Add the LTTng-UST log handler to our logger
4705         logger.addHandler(lttngUstLogHandler);
4706
4707         // Log at will!
4708         logger.info("some info");
4709         logger.warning("some warning");
4710         Thread.sleep(500);
4711         logger.finer("finer information; the answer is " + answer);
4712         Thread.sleep(123);
4713         logger.severe("error!");
4714
4715         // Not mandatory, but cleaner
4716         logger.removeHandler(lttngUstLogHandler);
4717         lttngUstLogHandler.close();
4718         cim.unregisterContextInfoRetriever("myRetriever");
4719     }
4720 }
4721 ----
4722
4723 Build this example:
4724
4725 [role="term"]
4726 ----
4727 $ javac -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar Test.java
4728 ----
4729
4730 <<creating-destroying-tracing-sessions,Create a tracing session>>
4731 and <<enabling-disabling-events,create an event rule>> matching the
4732 `jello` JUL logger:
4733
4734 [role="term"]
4735 ----
4736 $ lttng create
4737 $ lttng enable-event --jul jello
4738 ----
4739
4740 <<adding-context,Add the application-specific context fields>> to the
4741 JUL channel:
4742
4743 [role="term"]
4744 ----
4745 $ lttng add-context --jul --type='$app.myRetriever:intCtx'
4746 $ lttng add-context --jul --type='$app.myRetriever:strContext'
4747 ----
4748
4749 <<basic-tracing-session-control,Start tracing>>:
4750
4751 [role="term"]
4752 ----
4753 $ lttng start
4754 ----
4755
4756 Run the compiled class:
4757
4758 [role="term"]
4759 ----
4760 $ java -cp /usr/share/java/jarpath/lttng-ust-agent-common.jar:/usr/share/java/jarpath/lttng-ust-agent-jul.jar:. Test
4761 ----
4762
4763 <<basic-tracing-session-control,Stop tracing>> and inspect the
4764 recorded events:
4765
4766 [role="term"]
4767 ----
4768 $ lttng stop
4769 $ lttng view
4770 ----
4771 ====
4772
4773
4774 [role="since-2.7"]
4775 [[python-application]]
4776 === User space Python agent
4777
4778 You can instrument a Python{nbsp}2 or Python{nbsp}3 application which
4779 uses the standard
4780 https://docs.python.org/3/library/logging.html[`logging`] package.
4781
4782 Each log statement emits an LTTng event once the
4783 application module imports the
4784 <<lttng-ust-agents,LTTng-UST Python agent>> package.
4785
4786 [role="img-100"]
4787 .A Python application importing the LTTng-UST Python agent.
4788 image::python-app.png[]
4789
4790 To use the LTTng-UST Python agent:
4791
4792 . In the Python application's source code, import the LTTng-UST Python
4793   agent:
4794 +
4795 --
4796 [source,python]
4797 ----
4798 import lttngust
4799 ----
4800 --
4801 +
4802 The LTTng-UST Python agent automatically adds its logging handler to the
4803 root logger at import time.
4804 +
4805 Any log statement that the application executes before this import does
4806 not emit an LTTng event.
4807 +
4808 IMPORTANT: The LTTng-UST Python agent must be
4809 <<installing-lttng,installed>>.
4810
4811 . Use log statements and logging configuration as usual.
4812   Since the LTTng-UST Python agent adds a handler to the _root_
4813   logger, you can trace any log statement from any logger.
4814
4815 .Use the LTTng-UST Python agent.
4816 ====
4817 [source,python]
4818 .path:{test.py}
4819 ----
4820 import lttngust
4821 import logging
4822 import time
4823
4824
4825 def example():
4826     logging.basicConfig()
4827     logger = logging.getLogger('my-logger')
4828
4829     while True:
4830         logger.debug('debug message')
4831         logger.info('info message')
4832         logger.warn('warn message')
4833         logger.error('error message')
4834         logger.critical('critical message')
4835         time.sleep(1)
4836
4837
4838 if __name__ == '__main__':
4839     example()
4840 ----
4841
4842 NOTE: `logging.basicConfig()`, which adds to the root logger a basic
4843 logging handler which prints to the standard error stream, is not
4844 strictly required for LTTng-UST tracing to work, but in versions of
4845 Python preceding{nbsp}3.2, you could see a warning message which indicates
4846 that no handler exists for the logger `my-logger`.
4847
4848 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4849 <<enabling-disabling-events,create an event rule>> matching the
4850 `my-logger` Python logger, and <<basic-tracing-session-control,start
4851 tracing>>:
4852
4853 [role="term"]
4854 ----
4855 $ lttng create
4856 $ lttng enable-event --python my-logger
4857 $ lttng start
4858 ----
4859
4860 Run the Python script:
4861
4862 [role="term"]
4863 ----
4864 $ python test.py
4865 ----
4866
4867 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
4868 events:
4869
4870 [role="term"]
4871 ----
4872 $ lttng stop
4873 $ lttng view
4874 ----
4875 ====
4876
4877 In the resulting trace, an <<event,event record>> generated by a Python
4878 application is named `lttng_python:event` and has the following fields:
4879
4880 `asctime`::
4881   Logging time (string).
4882
4883 `msg`::
4884   Log record's message.
4885
4886 `logger_name`::
4887   Logger name.
4888
4889 `funcName`::
4890   Name of the function in which the log statement was executed.
4891
4892 `lineno`::
4893   Line number at which the log statement was executed.
4894
4895 `int_loglevel`::
4896   Log level integer value.
4897
4898 `thread`::
4899   ID of the Python thread in which the log statement was executed.
4900
4901 `threadName`::
4902   Name of the Python thread in which the log statement was executed.
4903
4904 You can use the opt:lttng-enable-event(1):--loglevel or
4905 opt:lttng-enable-event(1):--loglevel-only option of the
4906 man:lttng-enable-event(1) command to target a range of Python log levels
4907 or a specific Python log level.
4908
4909 When an application imports the LTTng-UST Python agent, the agent tries
4910 to register to a <<lttng-sessiond,session daemon>>. Note that you must
4911 <<start-sessiond,start the session daemon>> _before_ you run the Python
4912 application. If a session daemon is found, the agent tries to register
4913 to it during five seconds, after which the application continues
4914 without LTTng tracing support. You can override this timeout value with
4915 the env:LTTNG_UST_PYTHON_REGISTER_TIMEOUT environment variable
4916 (milliseconds).
4917
4918 If the session daemon stops while a Python application with an imported
4919 LTTng-UST Python agent runs, the agent retries to connect and to
4920 register to a session daemon every three seconds. You can override this
4921 delay with the env:LTTNG_UST_PYTHON_REGISTER_RETRY_DELAY environment
4922 variable.
4923
4924
4925 [role="since-2.5"]
4926 [[proc-lttng-logger-abi]]
4927 === LTTng logger
4928
4929 The `lttng-tracer` Linux kernel module, part of
4930 <<lttng-modules,LTTng-modules>>, creates the special LTTng logger file
4931 path:{/proc/lttng-logger} when it's loaded. Any application can write
4932 text data to this file to emit an LTTng event.
4933
4934 [role="img-100"]
4935 .An application writes to the LTTng logger file to emit an LTTng event.
4936 image::lttng-logger.png[]
4937
4938 The LTTng logger is the quickest method--not the most efficient,
4939 however--to add instrumentation to an application. It is designed
4940 mostly to instrument shell scripts:
4941
4942 [role="term"]
4943 ----
4944 $ echo "Some message, some $variable" > /proc/lttng-logger
4945 ----
4946
4947 Any event that the LTTng logger emits is named `lttng_logger` and
4948 belongs to the Linux kernel <<domain,tracing domain>>. However, unlike
4949 other instrumentation points in the kernel tracing domain, **any Unix
4950 user** can <<enabling-disabling-events,create an event rule>> which
4951 matches its event name, not only the root user or users in the
4952 <<tracing-group,tracing group>>.
4953
4954 To use the LTTng logger:
4955
4956 * From any application, write text data to the path:{/proc/lttng-logger}
4957   file.
4958
4959 The `msg` field of `lttng_logger` event records contains the
4960 recorded message.
4961
4962 NOTE: The maximum message length of an LTTng logger event is
4963 1024{nbsp}bytes. Writing more than this makes the LTTng logger emit more
4964 than one event to contain the remaining data.
4965
4966 You should not use the LTTng logger to trace a user application which
4967 can be instrumented in a more efficient way, namely:
4968
4969 * <<c-application,C and $$C++$$ applications>>.
4970 * <<java-application,Java applications>>.
4971 * <<python-application,Python applications>>.
4972
4973 .Use the LTTng logger.
4974 ====
4975 [source,bash]
4976 .path:{test.bash}
4977 ----
4978 echo 'Hello, World!' > /proc/lttng-logger
4979 sleep 2
4980 df --human-readable --print-type / > /proc/lttng-logger
4981 ----
4982
4983 <<creating-destroying-tracing-sessions,Create a tracing session>>,
4984 <<enabling-disabling-events,create an event rule>> matching the
4985 `lttng_logger` Linux kernel tracepoint, and
4986 <<basic-tracing-session-control,start tracing>>:
4987
4988 [role="term"]
4989 ----
4990 $ lttng create
4991 $ lttng enable-event --kernel lttng_logger
4992 $ lttng start
4993 ----
4994
4995 Run the Bash script:
4996
4997 [role="term"]
4998 ----
4999 $ bash test.bash
5000 ----
5001
5002 <<basic-tracing-session-control,Stop tracing>> and inspect the recorded
5003 events:
5004
5005 [role="term"]
5006 ----
5007 $ lttng stop
5008 $ lttng view
5009 ----
5010 ====
5011
5012
5013 [[instrumenting-linux-kernel]]
5014 === LTTng kernel tracepoints
5015
5016 NOTE: This section shows how to _add_ instrumentation points to the
5017 Linux kernel. The kernel's subsystems are already thoroughly
5018 instrumented at strategic places for LTTng when you
5019 <<installing-lttng,install>> the <<lttng-modules,LTTng-modules>>
5020 package.
5021
5022 ////
5023 There are two methods to instrument the Linux kernel:
5024
5025 . <<linux-add-lttng-layer,Add an LTTng layer>> over an existing ftrace
5026   tracepoint which uses the `TRACE_EVENT()` API.
5027 +
5028 Choose this if you want to instrumentation a Linux kernel tree with an
5029 instrumentation point compatible with ftrace, perf, and SystemTap.
5030
5031 . Use an <<linux-lttng-tracepoint-event,LTTng-only approach>> to
5032   instrument an out-of-tree kernel module.
5033 +
5034 Choose this if you don't need ftrace, perf, or SystemTap support.
5035 ////
5036
5037
5038 [[linux-add-lttng-layer]]
5039 ==== [[instrumenting-linux-kernel-itself]][[mainline-trace-event]][[lttng-adaptation-layer]]Add an LTTng layer to an existing ftrace tracepoint
5040
5041 This section shows how to add an LTTng layer to existing ftrace
5042 instrumentation using the `TRACE_EVENT()` API.
5043
5044 This section does not document the `TRACE_EVENT()` macro. You can
5045 read the following articles to learn more about this API:
5046
5047 * http://lwn.net/Articles/379903/[Using the TRACE_EVENT() macro (Part{nbsp}1)]
5048 * http://lwn.net/Articles/381064/[Using the TRACE_EVENT() macro (Part{nbsp}2)]
5049 * http://lwn.net/Articles/383362/[Using the TRACE_EVENT() macro (Part{nbsp}3)]
5050
5051 The following procedure assumes that your ftrace tracepoints are
5052 correctly defined in their own header and that they are created in
5053 one source file using the `CREATE_TRACE_POINTS` definition.
5054
5055 To add an LTTng layer over an existing ftrace tracepoint:
5056
5057 . Make sure the following kernel configuration options are
5058   enabled:
5059 +
5060 --
5061 * `CONFIG_MODULES`
5062 * `CONFIG_KALLSYMS`
5063 * `CONFIG_HIGH_RES_TIMERS`
5064 * `CONFIG_TRACEPOINTS`
5065 --
5066
5067 . Build the Linux source tree with your custom ftrace tracepoints.
5068 . Boot the resulting Linux image on your target system.
5069 +
5070 Confirm that the tracepoints exist by looking for their names in the
5071 dir:{/sys/kernel/debug/tracing/events/subsys} directory, where `subsys`
5072 is your subsystem's name.
5073
5074 . Get a copy of the latest LTTng-modules{nbsp}{revision}:
5075 +
5076 --
5077 [role="term"]
5078 ----
5079 $ cd $(mktemp -d) &&
5080 wget http://lttng.org/files/lttng-modules/lttng-modules-latest-2.11.tar.bz2 &&
5081 tar -xf lttng-modules-latest-2.11.tar.bz2 &&
5082 cd lttng-modules-2.11.*
5083 ----
5084 --
5085
5086 . In dir:{instrumentation/events/lttng-module}, relative to the root
5087   of the LTTng-modules source tree, create a header file named
5088   +__subsys__.h+ for your custom subsystem +__subsys__+ and write your
5089   LTTng-modules tracepoint definitions using the LTTng-modules
5090   macros in it.
5091 +
5092 Start with this template:
5093 +
5094 --
5095 [source,c]
5096 .path:{instrumentation/events/lttng-module/my_subsys.h}
5097 ----
5098 #undef TRACE_SYSTEM
5099 #define TRACE_SYSTEM my_subsys
5100
5101 #if !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ)
5102 #define _LTTNG_MY_SUBSYS_H
5103
5104 #include "../../../probes/lttng-tracepoint-event.h"
5105 #include <linux/tracepoint.h>
5106
5107 LTTNG_TRACEPOINT_EVENT(
5108     /*
5109      * Format is identical to TRACE_EVENT()'s version for the three
5110      * following macro parameters:
5111      */
5112     my_subsys_my_event,
5113     TP_PROTO(int my_int, const char *my_string),
5114     TP_ARGS(my_int, my_string),
5115
5116     /* LTTng-modules specific macros */
5117     TP_FIELDS(
5118         ctf_integer(int, my_int_field, my_int)
5119         ctf_string(my_bar_field, my_bar)
5120     )
5121 )
5122
5123 #endif /* !defined(_LTTNG_MY_SUBSYS_H) || defined(TRACE_HEADER_MULTI_READ) */
5124
5125 #include "../../../probes/define_trace.h"
5126 ----
5127 --
5128 +
5129 The entries in the `TP_FIELDS()` section are the list of fields for the
5130 LTTng tracepoint. This is similar to the `TP_STRUCT__entry()` part of
5131 ftrace's `TRACE_EVENT()` macro.
5132 +
5133 See <<lttng-modules-tp-fields,Tracepoint fields macros>> for a
5134 complete description of the available `ctf_*()` macros.
5135
5136 . Create the LTTng-modules probe's kernel module C source file,
5137   +probes/lttng-probe-__subsys__.c+, where +__subsys__+ is your
5138   subsystem name:
5139 +
5140 --
5141 [source,c]
5142 .path:{probes/lttng-probe-my-subsys.c}
5143 ----
5144 #include <linux/module.h>
5145 #include "../lttng-tracer.h"
5146
5147 /*
5148  * Build-time verification of mismatch between mainline
5149  * TRACE_EVENT() arguments and the LTTng-modules adaptation
5150  * layer LTTNG_TRACEPOINT_EVENT() arguments.
5151  */
5152 #include <trace/events/my_subsys.h>
5153
5154 /* Create LTTng tracepoint probes */
5155 #define LTTNG_PACKAGE_BUILD
5156 #define CREATE_TRACE_POINTS
5157 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
5158
5159 #include "../instrumentation/events/lttng-module/my_subsys.h"
5160
5161 MODULE_LICENSE("GPL and additional rights");
5162 MODULE_AUTHOR("Your name <your-email>");
5163 MODULE_DESCRIPTION("LTTng my_subsys probes");
5164 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION) "."
5165     __stringify(LTTNG_MODULES_MINOR_VERSION) "."
5166     __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION)
5167     LTTNG_MODULES_EXTRAVERSION);
5168 ----
5169 --
5170
5171 . Edit path:{probes/KBuild} and add your new kernel module object
5172   next to the existing ones:
5173 +
5174 --
5175 [source,make]
5176 .path:{probes/KBuild}
5177 ----
5178 # ...
5179
5180 obj-m += lttng-probe-module.o
5181 obj-m += lttng-probe-power.o
5182
5183 obj-m += lttng-probe-my-subsys.o
5184
5185 # ...
5186 ----
5187 --
5188
5189 . Build and install the LTTng kernel modules:
5190 +
5191 --
5192 [role="term"]
5193 ----
5194 $ make KERNELDIR=/path/to/linux
5195 # make modules_install && depmod -a
5196 ----
5197 --
5198 +
5199 Replace `/path/to/linux` with the path to the Linux source tree where
5200 you defined and used tracepoints with ftrace's `TRACE_EVENT()` macro.
5201
5202 Note that you can also use the
5203 <<lttng-tracepoint-event-code,`LTTNG_TRACEPOINT_EVENT_CODE()` macro>>
5204 instead of `LTTNG_TRACEPOINT_EVENT()` to use custom local variables and
5205 C code that need to be executed before the event fields are recorded.
5206
5207 The best way to learn how to use the previous LTTng-modules macros is to
5208 inspect the existing LTTng-modules tracepoint definitions in the
5209 dir:{instrumentation/events/lttng-module} header files. Compare them
5210 with the Linux kernel mainline versions in the
5211 dir:{include/trace/events} directory of the Linux source tree.
5212
5213
5214 [role="since-2.7"]
5215 [[lttng-tracepoint-event-code]]
5216 ===== Use custom C code to access the data for tracepoint fields
5217
5218 Although we recommended to always use the
5219 <<lttng-adaptation-layer,`LTTNG_TRACEPOINT_EVENT()`>> macro to describe
5220 the arguments and fields of an LTTng-modules tracepoint when possible,
5221 sometimes you need a more complex process to access the data that the
5222 tracer records as event record fields. In other words, you need local
5223 variables and multiple C{nbsp}statements instead of simple
5224 argument-based expressions that you pass to the
5225 <<lttng-modules-tp-fields,`ctf_*()` macros of `TP_FIELDS()`>>.
5226
5227 You can use the `LTTNG_TRACEPOINT_EVENT_CODE()` macro instead of
5228 `LTTNG_TRACEPOINT_EVENT()` to declare custom local variables and define
5229 a block of C{nbsp}code to be executed before LTTng records the fields.
5230 The structure of this macro is:
5231
5232 [source,c]
5233 .`LTTNG_TRACEPOINT_EVENT_CODE()` macro syntax.
5234 ----
5235 LTTNG_TRACEPOINT_EVENT_CODE(
5236     /*
5237      * Format identical to the LTTNG_TRACEPOINT_EVENT()
5238      * version for the following three macro parameters:
5239      */
5240     my_subsys_my_event,
5241     TP_PROTO(int my_int, const char *my_string),
5242     TP_ARGS(my_int, my_string),
5243
5244     /* Declarations of custom local variables */
5245     TP_locvar(
5246         int a = 0;
5247         unsigned long b = 0;
5248         const char *name = "(undefined)";
5249         struct my_struct *my_struct;
5250     ),
5251
5252     /*
5253      * Custom code which uses both tracepoint arguments
5254      * (in TP_ARGS()) and local variables (in TP_locvar()).
5255      *
5256      * Local variables are actually members of a structure pointed
5257      * to by the special variable tp_locvar.
5258      */
5259     TP_code(
5260         if (my_int) {
5261             tp_locvar->a = my_int + 17;
5262             tp_locvar->my_struct = get_my_struct_at(tp_locvar->a);
5263             tp_locvar->b = my_struct_compute_b(tp_locvar->my_struct);
5264             tp_locvar->name = my_struct_get_name(tp_locvar->my_struct);
5265             put_my_struct(tp_locvar->my_struct);
5266
5267             if (tp_locvar->b) {
5268                 tp_locvar->a = 1;
5269             }
5270         }
5271     ),
5272
5273     /*
5274      * Format identical to the LTTNG_TRACEPOINT_EVENT()
5275      * version for this, except that tp_locvar members can be
5276      * used in the argument expression parameters of
5277      * the ctf_*() macros.
5278      */
5279     TP_FIELDS(
5280         ctf_integer(unsigned long, my_struct_b, tp_locvar->b)
5281         ctf_integer(int, my_struct_a, tp_locvar->a)
5282         ctf_string(my_string_field, my_string)
5283         ctf_string(my_struct_name, tp_locvar->name)
5284     )
5285 )
5286 ----
5287
5288 IMPORTANT: The C code defined in `TP_code()` must not have any side
5289 effects when executed. In particular, the code must not allocate
5290 memory or get resources without deallocating this memory or putting
5291 those resources afterwards.
5292
5293
5294 [[instrumenting-linux-kernel-tracing]]
5295 ==== Load and unload a custom probe kernel module
5296
5297 You must load a <<lttng-adaptation-layer,created LTTng-modules probe
5298 kernel module>> in the kernel before it can emit LTTng events.
5299
5300 To load the default probe kernel modules and a custom probe kernel
5301 module:
5302
5303 * Use the opt:lttng-sessiond(8):--extra-kmod-probes option to give extra
5304   probe modules to load when starting a root <<lttng-sessiond,session
5305   daemon>>:
5306 +
5307 --
5308 .Load the `my_subsys`, `usb`, and the default probe modules.
5309 ====
5310 [role="term"]
5311 ----
5312 # lttng-sessiond --extra-kmod-probes=my_subsys,usb
5313 ----
5314 ====
5315 --
5316 +
5317 You only need to pass the subsystem name, not the whole kernel module
5318 name.
5319
5320 To load _only_ a given custom probe kernel module:
5321
5322 * Use the opt:lttng-sessiond(8):--kmod-probes option to give the probe
5323   modules to load when starting a root session daemon:
5324 +
5325 --
5326 .Load only the `my_subsys` and `usb` probe modules.
5327 ====
5328 [role="term"]
5329 ----
5330 # lttng-sessiond --kmod-probes=my_subsys,usb
5331 ----
5332 ====
5333 --
5334
5335 To confirm that a probe module is loaded:
5336
5337 * Use man:lsmod(8):
5338 +
5339 --
5340 [role="term"]
5341 ----
5342 $ lsmod | grep lttng_probe_usb
5343 ----
5344 --
5345
5346 To unload the loaded probe modules:
5347
5348 * Kill the session daemon with `SIGTERM`:
5349 +
5350 --
5351 [role="term"]
5352 ----
5353 # pkill lttng-sessiond
5354 ----
5355 --
5356 +
5357 You can also use man:modprobe(8)'s `--remove` option if the session
5358 daemon terminates abnormally.
5359
5360
5361 [[controlling-tracing]]
5362 == Tracing control
5363
5364 Once an application or a Linux kernel is
5365 <<instrumenting,instrumented>> for LTTng tracing,
5366 you can _trace_ it.
5367
5368 This section is divided in topics on how to use the various
5369 <<plumbing,components of LTTng>>, in particular the <<lttng-cli,cmd:lttng
5370 command-line tool>>, to _control_ the LTTng daemons and tracers.
5371
5372 NOTE: In the following subsections, we refer to an man:lttng(1) command
5373 using its man page name. For example, instead of _Run the `create`
5374 command to..._, we use _Run the man:lttng-create(1) command to..._.
5375
5376
5377 [[start-sessiond]]
5378 === Start a session daemon
5379
5380 In some situations, you need to run a <<lttng-sessiond,session daemon>>
5381 (man:lttng-sessiond(8)) _before_ you can use the man:lttng(1)
5382 command-line tool.
5383
5384 You will see the following error when you run a command while no session
5385 daemon is running:
5386
5387 ----
5388 Error: No session daemon is available
5389 ----
5390
5391 The only command that automatically runs a session daemon is
5392 man:lttng-create(1), which you use to
5393 <<creating-destroying-tracing-sessions,create a tracing session>>. While
5394 this is most of the time the first operation that you do, sometimes it's
5395 not. Some examples are:
5396
5397 * <<list-instrumentation-points,List the available instrumentation points>>.
5398 * <<saving-loading-tracing-session,Load a tracing session configuration>>.
5399
5400 [[tracing-group]] Each Unix user must have its own running session
5401 daemon to trace user applications. The session daemon that the root user
5402 starts is the only one allowed to control the LTTng kernel tracer. Users
5403 that are part of the _tracing group_ can control the root session
5404 daemon. The default tracing group name is `tracing`; you can set it to
5405 something else with the opt:lttng-sessiond(8):--group option when you
5406 start the root session daemon.
5407
5408 To start a user session daemon:
5409
5410 * Run man:lttng-sessiond(8):
5411 +
5412 --
5413 [role="term"]
5414 ----
5415 $ lttng-sessiond --daemonize
5416 ----
5417 --
5418
5419 To start the root session daemon:
5420
5421 * Run man:lttng-sessiond(8) as the root user:
5422 +
5423 --
5424 [role="term"]
5425 ----
5426 # lttng-sessiond --daemonize
5427 ----
5428 --
5429
5430 In both cases, remove the opt:lttng-sessiond(8):--daemonize option to
5431 start the session daemon in foreground.
5432
5433 To stop a session daemon, use man:kill(1) on its process ID (standard
5434 `TERM` signal).
5435
5436 Note that some Linux distributions could manage the LTTng session daemon
5437 as a service. In this case, you should use the service manager to
5438 start, restart, and stop session daemons.
5439
5440
5441 [[creating-destroying-tracing-sessions]]
5442 === Create and destroy a tracing session
5443
5444 Almost all the LTTng control operations happen in the scope of
5445 a <<tracing-session,tracing session>>, which is the dialogue between the
5446 <<lttng-sessiond,session daemon>> and you.
5447
5448 To create a tracing session with a generated name:
5449
5450 * Use the man:lttng-create(1) command:
5451 +
5452 --
5453 [role="term"]
5454 ----
5455 $ lttng create
5456 ----
5457 --
5458
5459 The created tracing session's name is `auto` followed by the
5460 creation date.
5461
5462 To create a tracing session with a specific name:
5463
5464 * Use the optional argument of the man:lttng-create(1) command:
5465 +
5466 --
5467 [role="term"]
5468 ----
5469 $ lttng create my-session
5470 ----
5471 --
5472 +
5473 Replace `my-session` with the specific tracing session name.
5474
5475 LTTng appends the creation date to the created tracing session's name.
5476
5477 LTTng writes the traces of a tracing session in
5478 +$LTTNG_HOME/lttng-trace/__name__+ by default, where +__name__+ is the
5479 name of the tracing session. Note that the env:LTTNG_HOME environment
5480 variable defaults to `$HOME` if not set.
5481
5482 To output LTTng traces to a non-default location:
5483
5484 * Use the opt:lttng-create(1):--output option of the man:lttng-create(1) command:
5485 +
5486 --
5487 [role="term"]
5488 ----
5489 $ lttng create my-session --output=/tmp/some-directory
5490 ----
5491 --
5492
5493 You may create as many tracing sessions as you wish.
5494
5495 To list all the existing tracing sessions for your Unix user:
5496
5497 * Use the man:lttng-list(1) command:
5498 +
5499 --
5500 [role="term"]
5501 ----
5502 $ lttng list
5503 ----
5504 --
5505
5506 When you create a tracing session, it is set as the _current tracing
5507 session_. The following man:lttng(1) commands operate on the current
5508 tracing session when you don't specify one:
5509
5510 [role="list-3-cols"]
5511 * man:lttng-add-context(1)
5512 * man:lttng-destroy(1)
5513 * man:lttng-disable-channel(1)
5514 * man:lttng-disable-event(1)
5515 * man:lttng-disable-rotation(1)
5516 * man:lttng-enable-channel(1)
5517 * man:lttng-enable-event(1)
5518 * man:lttng-enable-rotation(1)
5519 * man:lttng-load(1)
5520 * man:lttng-regenerate(1)
5521 * man:lttng-rotate(1)
5522 * man:lttng-save(1)
5523 * man:lttng-snapshot(1)
5524 * man:lttng-start(1)
5525 * man:lttng-status(1)
5526 * man:lttng-stop(1)
5527 * man:lttng-track(1)
5528 * man:lttng-untrack(1)
5529 * man:lttng-view(1)
5530
5531 To change the current tracing session:
5532
5533 * Use the man:lttng-set-session(1) command:
5534 +
5535 --
5536 [role="term"]
5537 ----
5538 $ lttng set-session new-session
5539 ----
5540 --
5541 +
5542 Replace `new-session` by the name of the new current tracing session.
5543
5544 When you are done tracing in a given tracing session, you can destroy
5545 it. This operation frees the resources taken by the tracing session
5546 to destroy; it does not destroy the trace data that LTTng wrote for
5547 this tracing session.
5548
5549 To destroy the current tracing session:
5550
5551 * Use the man:lttng-destroy(1) command:
5552 +
5553 --
5554 [role="term"]
5555 ----
5556 $ lttng destroy
5557 ----
5558 --
5559
5560
5561 [[list-instrumentation-points]]
5562 === List the available instrumentation points
5563
5564 The <<lttng-sessiond,session daemon>> can query the running instrumented
5565 user applications and the Linux kernel to get a list of available
5566 instrumentation points. For the Linux kernel <<domain,tracing domain>>,
5567 they are tracepoints and system calls. For the user space tracing
5568 domain, they are tracepoints. For the other tracing domains, they are
5569 logger names.
5570
5571 To list the available instrumentation points:
5572
5573 * Use the man:lttng-list(1) command with the requested tracing domain's
5574   option amongst:
5575 +
5576 --
5577 * opt:lttng-list(1):--kernel: Linux kernel tracepoints (your Unix user
5578   must be a root user, or it must be a member of the
5579   <<tracing-group,tracing group>>).
5580 * opt:lttng-list(1):--kernel with opt:lttng-list(1):--syscall: Linux
5581   kernel system calls (your Unix user must be a root user, or it must be
5582   a member of the tracing group).
5583 * opt:lttng-list(1):--userspace: user space tracepoints.
5584 * opt:lttng-list(1):--jul: `java.util.logging` loggers.
5585 * opt:lttng-list(1):--log4j: Apache log4j loggers.
5586 * opt:lttng-list(1):--python: Python loggers.
5587 --
5588
5589 .List the available user space tracepoints.
5590 ====
5591 [role="term"]
5592 ----
5593 $ lttng list --userspace
5594 ----
5595 ====
5596
5597 .List the available Linux kernel system call tracepoints.
5598 ====
5599 [role="term"]
5600 ----
5601 $ lttng list --kernel --syscall
5602 ----
5603 ====
5604
5605
5606 [[enabling-disabling-events]]
5607 === Create and enable an event rule
5608
5609 Once you <<creating-destroying-tracing-sessions,create a tracing
5610 session>>, you can create <<event,event rules>> with the
5611 man:lttng-enable-event(1) command.
5612
5613 You specify each condition with a command-line option. The available
5614 condition arguments are shown in the following table.
5615
5616 [role="growable",cols="asciidoc,asciidoc,default"]
5617 .Condition command-line arguments for the man:lttng-enable-event(1) command.
5618 |====
5619 |Argument |Description |Applicable tracing domains
5620
5621 |
5622 One of:
5623
5624 . `--syscall`
5625 . +--probe=__ADDR__+
5626 . +--function=__ADDR__+
5627 . +--userspace-probe=__PATH__:__SYMBOL__+
5628 . +--userspace-probe=sdt:__PATH__:__PROVIDER__:__NAME__+
5629
5630 |
5631 Instead of using the default _tracepoint_ instrumentation type, use:
5632
5633 . A Linux system call (entry and exit).
5634 . A Linux https://lwn.net/Articles/132196/[kprobe] (symbol or address).
5635 . The entry and return points of a Linux function (symbol or address).
5636 . The entry point of a user application or library function (path to
5637   application/library and symbol).
5638 . A https://www.sourceware.org/systemtap/wiki/AddingUserSpaceProbingToApps[SystemTap
5639   Statically Defined Tracing] (SDT) probe (path to application/library,
5640   provider and probe names).
5641
5642 |Linux kernel.
5643
5644 |First positional argument.
5645
5646 |
5647 Tracepoint or system call name.
5648
5649 With the opt:lttng-enable-event(1):--probe,
5650 opt:lttng-enable-event(1):--function, and
5651 opt:lttng-enable-event(1):--userspace-probe options, this is a custom
5652 name given to the event rule. With the JUL, log4j, and Python domains,
5653 this is a logger name.
5654
5655 With a tracepoint, logger, or system call name, you can use the special
5656 `*` globbing character to match anything (for example, `sched_*`,
5657 `my_comp*:*msg_*`).
5658
5659 |All.
5660
5661 |
5662 One of:
5663
5664 . +--loglevel=__LEVEL__+
5665 . +--loglevel-only=__LEVEL__+
5666
5667 |
5668 . Match only tracepoints or log statements with a logging level at
5669   least as severe as +__LEVEL__+.
5670 . Match only tracepoints or log statements with a logging level
5671   equal to +__LEVEL__+.
5672
5673 See man:lttng-enable-event(1) for the list of available logging level
5674 names.
5675
5676 |User space, JUL, log4j, and Python.
5677
5678 |+--exclude=__EXCLUSIONS__+
5679
5680 |
5681 When you use a `*` character at the end of the tracepoint or logger
5682 name (first positional argument), exclude the specific names in the
5683 comma-delimited list +__EXCLUSIONS__+.
5684
5685 |
5686 User space, JUL, log4j, and Python.
5687
5688 |+--filter=__EXPR__+
5689
5690 |
5691 Match only events which satisfy the expression +__EXPR__+.
5692
5693 See man:lttng-enable-event(1) to learn more about the syntax of a
5694 filter expression.
5695
5696 |All.
5697
5698 |====
5699
5700 You attach an event rule to a <<channel,channel>> on creation. If you do
5701 not specify the channel with the opt:lttng-enable-event(1):--channel
5702 option, and if the event rule to create is the first in its
5703 <<domain,tracing domain>> for a given tracing session, then LTTng
5704 creates a _default channel_ for you. This default channel is reused in
5705 subsequent invocations of the man:lttng-enable-event(1) command for the
5706 same tracing domain.
5707
5708 An event rule is always enabled at creation time.
5709
5710 The following examples show how you can combine the previous
5711 command-line options to create simple to more complex event rules.
5712
5713 .Create an event rule targetting a Linux kernel tracepoint (default channel).
5714 ====
5715 [role="term"]
5716 ----
5717 $ lttng enable-event --kernel sched_switch
5718 ----
5719 ====
5720
5721 .Create an event rule matching four Linux kernel system calls (default channel).
5722 ====
5723 [role="term"]
5724 ----
5725 $ lttng enable-event --kernel --syscall open,write,read,close
5726 ----
5727 ====
5728
5729 .Create event rules matching tracepoints with filter expressions (default channel).
5730 ====
5731 [role="term"]
5732 ----
5733 $ lttng enable-event --kernel sched_switch --filter='prev_comm == "bash"'
5734 ----
5735
5736 [role="term"]
5737 ----
5738 $ lttng enable-event --kernel --all \
5739                    --filter='$ctx.tid == 1988 || $ctx.tid == 1534'
5740 ----
5741
5742 [role="term"]
5743 ----
5744 $ lttng enable-event --jul my_logger \
5745                    --filter='$app.retriever:cur_msg_id > 3'
5746 ----
5747
5748 IMPORTANT: Make sure to always quote the filter string when you
5749 use man:lttng(1) from a shell.
5750 ====
5751
5752 .Create an event rule matching any user space tracepoint of a given tracepoint provider with a log level range (default channel).
5753 ====
5754 [role="term"]
5755 ----
5756 $ lttng enable-event --userspace my_app:'*' --loglevel=TRACE_INFO
5757 ----
5758
5759 IMPORTANT: Make sure to always quote the wildcard character when you
5760 use man:lttng(1) from a shell.
5761 ====
5762
5763 .Create an event rule matching multiple Python loggers with a wildcard and with exclusions (default channel).
5764 ====
5765 [role="term"]
5766 ----
5767 $ lttng enable-event --python my-app.'*' \
5768                    --exclude='my-app.module,my-app.hello'
5769 ----
5770 ====
5771
5772 .Create an event rule matching any Apache log4j logger with a specific log level (default channel).
5773 ====
5774 [role="term"]
5775 ----
5776 $ lttng enable-event --log4j --all --loglevel-only=LOG4J_WARN
5777 ----
5778 ====
5779
5780 .Create an event rule attached to a specific channel matching a specific user space tracepoint provider and tracepoint.
5781 ====
5782 [role="term"]
5783 ----
5784 $ lttng enable-event --userspace my_app:my_tracepoint --channel=my-channel
5785 ----
5786 ====
5787
5788 .Create an event rule matching the `malloc` function entry in path:{/usr/lib/libc.so.6}:
5789 ====
5790 [role="term"]
5791 ----
5792 $ lttng enable-event --kernel --userspace-probe=/usr/lib/libc.so.6:malloc \
5793                    libc_malloc
5794 ----
5795 ====
5796
5797 .Create an event rule matching the `server`/`accept_request` https://www.sourceware.org/systemtap/wiki/AddingUserSpaceProbingToApps[SDT probe] in path:{/usr/bin/serv}:
5798 ====
5799 [role="term"]
5800 ----
5801 $ lttng enable-event --kernel --userspace-probe=sdt:serv:server:accept_request \
5802                   server_accept_request
5803 ----
5804 ====
5805
5806 The event rules of a given channel form a whitelist: as soon as an
5807 emitted event passes one of them, LTTng can record the event. For
5808 example, an event named `my_app:my_tracepoint` emitted from a user space
5809 tracepoint with a `TRACE_ERROR` log level passes both of the following
5810 rules:
5811
5812 [role="term"]
5813 ----
5814 $ lttng enable-event --userspace my_app:my_tracepoint
5815 $ lttng enable-event --userspace my_app:my_tracepoint \
5816                    --loglevel=TRACE_INFO
5817 ----
5818
5819 The second event rule is redundant: the first one includes
5820 the second one.
5821
5822
5823 [[disable-event-rule]]
5824 === Disable an event rule
5825
5826 To disable an event rule that you <<enabling-disabling-events,created>>
5827 previously, use the man:lttng-disable-event(1) command. This command
5828 disables _all_ the event rules (of a given tracing domain and channel)
5829 which match an instrumentation point. The other conditions are not
5830 supported as of LTTng{nbsp}{revision}.
5831
5832 The LTTng tracer does not record an emitted event which passes
5833 a _disabled_ event rule.
5834
5835 .Disable an event rule matching a Python logger (default channel).
5836 ====
5837 [role="term"]
5838 ----
5839 $ lttng disable-event --python my-logger
5840 ----
5841 ====
5842
5843 .Disable an event rule matching all `java.util.logging` loggers (default channel).
5844 ====
5845 [role="term"]
5846 ----
5847 $ lttng disable-event --jul '*'
5848 ----
5849 ====
5850
5851 .Disable _all_ the event rules of the default channel.
5852 ====
5853 The opt:lttng-disable-event(1):--all-events option is not, like the
5854 opt:lttng-enable-event(1):--all option of man:lttng-enable-event(1), the
5855 equivalent of the event name `*` (wildcard): it disables _all_ the event
5856 rules of a given channel.
5857
5858 [role="term"]
5859 ----
5860 $ lttng disable-event --jul --all-events
5861 ----
5862 ====
5863
5864 NOTE: You cannot delete an event rule once you create it.
5865
5866
5867 [[status]]
5868 === Get the status of a tracing session
5869
5870 To get the status of the current tracing session, that is, its
5871 parameters, its channels, event rules, and their attributes:
5872
5873 * Use the man:lttng-status(1) command:
5874 +
5875 --
5876 [role="term"]
5877 ----
5878 $ lttng status
5879 ----
5880 --
5881 +
5882
5883 To get the status of any tracing session:
5884
5885 * Use the man:lttng-list(1) command with the tracing session's name:
5886 +
5887 --
5888 [role="term"]
5889 ----
5890 $ lttng list my-session
5891 ----
5892 --
5893 +
5894 Replace `my-session` with the desired tracing session's name.
5895
5896
5897 [[basic-tracing-session-control]]
5898 === Start and stop a tracing session
5899
5900 Once you <<creating-destroying-tracing-sessions,create a tracing
5901 session>> and
5902 <<enabling-disabling-events,create one or more event rules>>,
5903 you can start and stop the tracers for this tracing session.
5904
5905 To start tracing in the current tracing session:
5906
5907 * Use the man:lttng-start(1) command:
5908 +
5909 --
5910 [role="term"]
5911 ----
5912 $ lttng start
5913 ----
5914 --
5915
5916 LTTng is very flexible: you can launch user applications before
5917 or after the you start the tracers. The tracers only record the events
5918 if they pass enabled event rules and if they occur while the tracers are
5919 started.
5920
5921 To stop tracing in the current tracing session:
5922
5923 * Use the man:lttng-stop(1) command:
5924 +
5925 --
5926 [role="term"]
5927 ----
5928 $ lttng stop
5929 ----
5930 --
5931 +
5932 If there were <<channel-overwrite-mode-vs-discard-mode,lost event
5933 records>> or lost sub-buffers since the last time you ran
5934 man:lttng-start(1), warnings are printed when you run the
5935 man:lttng-stop(1) command.
5936
5937 IMPORTANT: You need to stop tracing to make LTTng flush the remaining
5938 trace data and make the trace readable.
5939
5940
5941 [[enabling-disabling-channels]]
5942 === Create a channel
5943
5944 Once you create a tracing session, you can create a <<channel,channel>>
5945 with the man:lttng-enable-channel(1) command.
5946
5947 Note that LTTng automatically creates a default channel when, for a
5948 given <<domain,tracing domain>>, no channels exist and you
5949 <<enabling-disabling-events,create>> the first event rule. This default
5950 channel is named `channel0` and its attributes are set to reasonable
5951 values. Therefore, you only need to create a channel when you need
5952 non-default attributes.
5953
5954 You specify each non-default channel attribute with a command-line
5955 option when you use the man:lttng-enable-channel(1) command. The
5956 available command-line options are:
5957
5958 [role="growable",cols="asciidoc,asciidoc"]
5959 .Command-line options for the man:lttng-enable-channel(1) command.
5960 |====
5961 |Option |Description
5962
5963 |`--overwrite`
5964
5965 |
5966 Use the _overwrite_
5967 <<channel-overwrite-mode-vs-discard-mode,event record loss mode>> instead
5968 of the default _discard_ mode.
5969
5970 |`--buffers-pid` (user space tracing domain only)
5971
5972 |
5973 Use the per-process <<channel-buffering-schemes,buffering scheme>>
5974 instead of the default per-user buffering scheme.
5975
5976 |+--subbuf-size=__SIZE__+
5977
5978 |
5979 Allocate sub-buffers of +__SIZE__+ bytes (power of two), for each CPU,
5980 either for each Unix user (default), or for each instrumented process.
5981
5982 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5983
5984 |+--num-subbuf=__COUNT__+
5985
5986 |
5987 Allocate +__COUNT__+ sub-buffers (power of two), for each CPU, either
5988 for each Unix user (default), or for each instrumented process.
5989
5990 See <<channel-subbuf-size-vs-subbuf-count,Sub-buffer count and size>>.
5991
5992 |+--tracefile-size=__SIZE__+
5993
5994 |
5995 Set the maximum size of each trace file that this channel writes within
5996 a stream to +__SIZE__+ bytes instead of no maximum.
5997
5998 See <<tracefile-rotation,Trace file count and size>>.
5999
6000 |+--tracefile-count=__COUNT__+
6001
6002 |
6003 Limit the number of trace files that this channel creates to
6004 +__COUNT__+ channels instead of no limit.
6005
6006 See <<tracefile-rotation,Trace file count and size>>.
6007
6008 |+--switch-timer=__PERIODUS__+
6009
6010 |
6011 Set the <<channel-switch-timer,switch timer period>>
6012 to +__PERIODUS__+{nbsp}µs.
6013
6014 |+--read-timer=__PERIODUS__+
6015
6016 |
6017 Set the <<channel-read-timer,read timer period>>
6018 to +__PERIODUS__+{nbsp}µs.
6019
6020 |[[opt-blocking-timeout]]+--blocking-timeout=__TIMEOUTUS__+
6021
6022 |
6023 Set the timeout of user space applications which load LTTng-UST
6024 in blocking mode to +__TIMEOUTUS__+:
6025
6026 0 (default)::
6027     Never block (non-blocking mode).
6028
6029 `inf`::
6030     Block forever until space is available in a sub-buffer to record
6031     the event.
6032
6033 __n__, a positive value::
6034     Wait for at most __n__ µs when trying to write into a sub-buffer.
6035
6036 Note that, for this option to have any effect on an instrumented
6037 user space application, you need to run the application with a set
6038 env:LTTNG_UST_ALLOW_BLOCKING environment variable.
6039
6040 |+--output=__TYPE__+ (Linux kernel tracing domain only)
6041
6042 |
6043 Set the channel's output type to +__TYPE__+, either `mmap` or `splice`.
6044
6045 |====
6046
6047 You can only create a channel in the Linux kernel and user space
6048 <<domain,tracing domains>>: other tracing domains have their own channel
6049 created on the fly when <<enabling-disabling-events,creating event
6050 rules>>.
6051
6052 [IMPORTANT]
6053 ====
6054 Because of a current LTTng limitation, you must create all channels
6055 _before_ you <<basic-tracing-session-control,start tracing>> in a given
6056 tracing session, that is, before the first time you run
6057 man:lttng-start(1).
6058
6059 Since LTTng automatically creates a default channel when you use the
6060 man:lttng-enable-event(1) command with a specific tracing domain, you
6061 cannot, for example, create a Linux kernel event rule, start tracing,
6062 and then create a user space event rule, because no user space channel
6063 exists yet and it's too late to create one.
6064
6065 For this reason, make sure to configure your channels properly
6066 before starting the tracers for the first time!
6067 ====
6068
6069 The following examples show how you can combine the previous
6070 command-line options to create simple to more complex channels.
6071
6072 .Create a Linux kernel channel with default attributes.
6073 ====
6074 [role="term"]
6075 ----
6076 $ lttng enable-channel --kernel my-channel
6077 ----
6078 ====
6079
6080 .Create a user space channel with four sub-buffers or 1{nbsp}MiB each, per CPU, per instrumented process.
6081 ====
6082 [role="term"]
6083 ----
6084 $ lttng enable-channel --userspace --num-subbuf=4 --subbuf-size=1M \
6085                      --buffers-pid my-channel
6086 ----
6087 ====
6088
6089 .[[blocking-timeout-example]]Create a default user space channel with an infinite blocking timeout.
6090 ====
6091 <<creating-destroying-tracing-sessions,Create a tracing-session>>,
6092 create the channel, <<enabling-disabling-events,create an event rule>>,
6093 and <<basic-tracing-session-control,start tracing>>:
6094
6095 [role="term"]
6096 ----
6097 $ lttng create
6098 $ lttng enable-channel --userspace --blocking-timeout=inf blocking-channel
6099 $ lttng enable-event --userspace --channel=blocking-channel --all
6100 $ lttng start
6101 ----
6102
6103 Run an application instrumented with LTTng-UST and allow it to block:
6104
6105 [role="term"]
6106 ----
6107 $ LTTNG_UST_ALLOW_BLOCKING=1 my-app
6108 ----
6109 ====
6110
6111 .Create a Linux kernel channel which rotates eight trace files of 4{nbsp}MiB each for each stream
6112 ====
6113 [role="term"]
6114 ----
6115 $ lttng enable-channel --kernel --tracefile-count=8 \
6116                      --tracefile-size=4194304 my-channel
6117 ----
6118 ====
6119
6120 .Create a user space channel in overwrite (or _flight recorder_) mode.
6121 ====
6122 [role="term"]
6123 ----
6124 $ lttng enable-channel --userspace --overwrite my-channel
6125 ----
6126 ====
6127
6128 You can <<enabling-disabling-events,create>> the same event rule in
6129 two different channels:
6130
6131 [role="term"]
6132 ----
6133 $ lttng enable-event --userspace --channel=my-channel app:tp
6134 $ lttng enable-event --userspace --channel=other-channel app:tp
6135 ----
6136
6137 If both channels are enabled, when a tracepoint named `app:tp` is
6138 reached, LTTng records two events, one for each channel.
6139
6140
6141 [[disable-channel]]
6142 === Disable a channel
6143
6144 To disable a specific channel that you <<enabling-disabling-channels,created>>
6145 previously, use the man:lttng-disable-channel(1) command.
6146
6147 .Disable a specific Linux kernel channel.
6148 ====
6149 [role="term"]
6150 ----
6151 $ lttng disable-channel --kernel my-channel
6152 ----
6153 ====
6154
6155 The state of a channel precedes the individual states of event rules
6156 attached to it: event rules which belong to a disabled channel, even if
6157 they are enabled, are also considered disabled.
6158
6159
6160 [[adding-context]]
6161 === Add context fields to a channel
6162
6163 Event record fields in trace files provide important information about
6164 events that occured previously, but sometimes some external context may
6165 help you solve a problem faster. Examples of context fields are:
6166
6167 * The **process ID**, **thread ID**, **process name**, and
6168   **process priority** of the thread in which the event occurs.
6169 * The **hostname** of the system on which the event occurs.
6170 * The Linux kernel and user call stacks (since
6171   LTTng{nbsp}{revision}).
6172 * The current values of many possible **performance counters** using
6173   perf, for example:
6174 ** CPU cycles, stalled cycles, idle cycles, and the other cycle types.
6175 ** Cache misses.
6176 ** Branch instructions, misses, and loads.
6177 ** CPU faults.
6178 * Any context defined at the application level (supported for the
6179   JUL and log4j <<domain,tracing domains>>).
6180
6181 To get the full list of available context fields, see
6182 `lttng add-context --list`. Some context fields are reserved for a
6183 specific <<domain,tracing domain>> (Linux kernel or user space).
6184
6185 You add context fields to <<channel,channels>>. All the events
6186 that a channel with added context fields records contain those fields.
6187
6188 To add context fields to one or all the channels of a given tracing
6189 session:
6190
6191 * Use the man:lttng-add-context(1) command.
6192
6193 .Add context fields to all the channels of the current tracing session.
6194 ====
6195 The following command line adds the virtual process identifier and
6196 the per-thread CPU cycles count fields to all the user space channels
6197 of the current tracing session.
6198
6199 [role="term"]
6200 ----
6201 $ lttng add-context --userspace --type=vpid --type=perf:thread:cpu-cycles
6202 ----
6203 ====
6204
6205 .Add performance counter context fields by raw ID
6206 ====
6207 See man:lttng-add-context(1) for the exact format of the context field
6208 type, which is partly compatible with the format used in
6209 man:perf-record(1).
6210
6211 [role="term"]
6212 ----
6213 $ lttng add-context --userspace --type=perf:thread:raw:r0110:test
6214 $ lttng add-context --kernel --type=perf:cpu:raw:r0013c:x86unhalted
6215 ----
6216 ====
6217
6218 .Add context fields to a specific channel.
6219 ====
6220 The following command line adds the thread identifier and user call
6221 stack context fields to the Linux kernel channel named `my-channel` in
6222 the current tracing session.
6223
6224 [role="term"]
6225 ----
6226 $ lttng add-context --kernel --channel=my-channel \
6227      --type=tid --type=callstack-user
6228 ----
6229 ====
6230
6231 .Add an application-specific context field to a specific channel.
6232 ====
6233 The following command line adds the `cur_msg_id` context field of the
6234 `retriever` context retriever for all the instrumented
6235 <<java-application,Java applications>> recording <<event,event records>>
6236 in the channel named `my-channel`:
6237
6238 [role="term"]
6239 ----
6240 $ lttng add-context --kernel --channel=my-channel \
6241                   --type='$app:retriever:cur_msg_id'
6242 ----
6243
6244 IMPORTANT: Make sure to always quote the `$` character when you
6245 use man:lttng-add-context(1) from a shell.
6246 ====
6247
6248 NOTE: You cannot remove context fields from a channel once you add it.
6249
6250
6251 [role="since-2.7"]
6252 [[pid-tracking]]
6253 === Track process IDs
6254
6255 It's often useful to allow only specific process IDs (PIDs) to emit
6256 events. For example, you may wish to record all the system calls made by
6257 a given process (à la http://linux.die.net/man/1/strace[strace]).
6258
6259 The man:lttng-track(1) and man:lttng-untrack(1) commands serve this
6260 purpose. Both commands operate on a whitelist of process IDs. You _add_
6261 entries to this whitelist with the man:lttng-track(1) command and remove
6262 entries with the man:lttng-untrack(1) command. Any process which has one
6263 of the PIDs in the whitelist is allowed to emit LTTng events which pass
6264 an enabled <<event,event rule>>.
6265
6266 NOTE: The PID tracker tracks the _numeric process IDs_. Should a
6267 process with a given tracked ID exit and another process be given this
6268 ID, then the latter would also be allowed to emit events.
6269
6270 .Track and untrack process IDs.
6271 ====
6272 For the sake of the following example, assume the target system has
6273 16{nbsp}possible PIDs.
6274
6275 When you
6276 <<creating-destroying-tracing-sessions,create a tracing session>>,
6277 the whitelist contains all the possible PIDs:
6278
6279 [role="img-100"]
6280 .All PIDs are tracked.
6281 image::track-all.png[]
6282
6283 When the whitelist is full and you use the man:lttng-track(1) command to
6284 specify some PIDs to track, LTTng first clears the whitelist, then it
6285 tracks the specific PIDs. After:
6286
6287 [role="term"]
6288 ----
6289 $ lttng track --pid=3,4,7,10,13
6290 ----
6291
6292 the whitelist is:
6293
6294 [role="img-100"]
6295 .PIDs 3, 4, 7, 10, and 13 are tracked.
6296 image::track-3-4-7-10-13.png[]
6297
6298 You can add more PIDs to the whitelist afterwards:
6299
6300 [role="term"]
6301 ----
6302 $ lttng track --pid=1,15,16
6303 ----
6304
6305 The result is:
6306
6307 [role="img-100"]
6308 .PIDs 1, 15, and 16 are added to the whitelist.
6309 image::track-1-3-4-7-10-13-15-16.png[]
6310
6311 The man:lttng-untrack(1) command removes entries from the PID tracker's
6312 whitelist. Given the previous example, the following command:
6313
6314 [role="term"]
6315 ----
6316 $ lttng untrack --pid=3,7,10,13
6317 ----
6318
6319 leads to this whitelist:
6320
6321 [role="img-100"]
6322 .PIDs 3, 7, 10, and 13 are removed from the whitelist.
6323 image::track-1-4-15-16.png[]
6324
6325 LTTng can track all possible PIDs again using the
6326 opt:lttng-track(1):--all option:
6327
6328 [role="term"]
6329 ----
6330 $ lttng track --pid --all
6331 ----
6332
6333 The result is, again:
6334
6335 [role="img-100"]
6336 .All PIDs are tracked.
6337 image::track-all.png[]
6338 ====
6339
6340 .Track only specific PIDs
6341 ====
6342 A very typical use case with PID tracking is to start with an empty
6343 whitelist, then <<basic-tracing-session-control,start the tracers>>, and
6344 then add PIDs manually while tracers are active. You can accomplish this
6345 by using the opt:lttng-untrack(1):--all option of the
6346 man:lttng-untrack(1) command to clear the whitelist after you
6347 <<creating-destroying-tracing-sessions,create a tracing session>>:
6348
6349 [role="term"]
6350 ----
6351 $ lttng untrack --pid --all
6352 ----
6353
6354 gives:
6355
6356 [role="img-100"]
6357 .No PIDs are tracked.
6358 image::untrack-all.png[]
6359
6360 If you trace with this whitelist configuration, the tracer records no
6361 events for this <<domain,tracing domain>> because no processes are
6362 tracked. You can use the man:lttng-track(1) command as usual to track
6363 specific PIDs, for example:
6364
6365 [role="term"]
6366 ----
6367 $ lttng track --pid=6,11
6368 ----
6369
6370 Result:
6371
6372 [role="img-100"]
6373 .PIDs 6 and 11 are tracked.
6374 image::track-6-11.png[]
6375 ====
6376
6377
6378 [role="since-2.5"]
6379 [[saving-loading-tracing-session]]
6380 === Save and load tracing session configurations
6381
6382 Configuring a <<tracing-session,tracing session>> can be long. Some of
6383 the tasks involved are:
6384
6385 * <<enabling-disabling-channels,Create channels>> with
6386   specific attributes.
6387 * <<adding-context,Add context fields>> to specific channels.
6388 * <<enabling-disabling-events,Create event rules>> with specific log
6389   level and filter conditions.
6390
6391 If you use LTTng to solve real world problems, chances are you have to
6392 record events using the same tracing session setup over and over,
6393 modifying a few variables each time in your instrumented program
6394 or environment. To avoid constant tracing session reconfiguration,
6395 the man:lttng(1) command-line tool can save and load tracing session
6396 configurations to/from XML files.
6397
6398 To save a given tracing session configuration:
6399
6400 * Use the man:lttng-save(1) command:
6401 +
6402 --
6403 [role="term"]
6404 ----
6405 $ lttng save my-session
6406 ----
6407 --
6408 +
6409 Replace `my-session` with the name of the tracing session to save.
6410
6411 LTTng saves tracing session configurations to
6412 dir:{$LTTNG_HOME/.lttng/sessions} by default. Note that the
6413 env:LTTNG_HOME environment variable defaults to `$HOME` if not set. Use
6414 the opt:lttng-save(1):--output-path option to change this destination
6415 directory.
6416
6417 LTTng saves all configuration parameters, for example:
6418
6419 * The tracing session name.
6420 * The trace data output path.
6421 * The channels with their state and all their attributes.
6422 * The context fields you added to channels.
6423 * The event rules with their state, log level and filter conditions.
6424
6425 To load a tracing session:
6426
6427 * Use the man:lttng-load(1) command:
6428 +
6429 --
6430 [role="term"]
6431 ----
6432 $ lttng load my-session
6433 ----
6434 --
6435 +
6436 Replace `my-session` with the name of the tracing session to load.
6437
6438 When LTTng loads a configuration, it restores your saved tracing session
6439 as if you just configured it manually.
6440
6441 See man:lttng(1) for the complete list of command-line options. You
6442 can also save and load all many sessions at a time, and decide in which
6443 directory to output the XML files.
6444
6445
6446 [[sending-trace-data-over-the-network]]
6447 === Send trace data over the network
6448
6449 LTTng can send the recorded trace data to a remote system over the
6450 network instead of writing it to the local file system.
6451
6452 To send the trace data over the network:
6453
6454 . On the _remote_ system (which can also be the target system),
6455   start an LTTng <<lttng-relayd,relay daemon>> (man:lttng-relayd(8)):
6456 +
6457 --
6458 [role="term"]
6459 ----
6460 $ lttng-relayd
6461 ----
6462 --
6463
6464 . On the _target_ system, create a tracing session configured to
6465   send trace data over the network:
6466 +
6467 --
6468 [role="term"]
6469 ----
6470 $ lttng create my-session --set-url=net://remote-system
6471 ----
6472 --
6473 +
6474 Replace `remote-system` by the host name or IP address of the
6475 remote system. See man:lttng-create(1) for the exact URL format.
6476
6477 . On the target system, use the man:lttng(1) command-line tool as usual.
6478   When tracing is active, the target's consumer daemon sends sub-buffers
6479   to the relay daemon running on the remote system instead of flushing
6480   them to the local file system. The relay daemon writes the received
6481   packets to the local file system.
6482
6483 The relay daemon writes trace files to
6484 +$LTTNG_HOME/lttng-traces/__hostname__/__session__+ by default, where
6485 +__hostname__+ is the host name of the target system and +__session__+
6486 is the tracing session name. Note that the env:LTTNG_HOME environment
6487 variable defaults to `$HOME` if not set. Use the
6488 opt:lttng-relayd(8):--output option of man:lttng-relayd(8) to write
6489 trace files to another base directory.
6490
6491
6492 [role="since-2.4"]
6493 [[lttng-live]]
6494 === View events as LTTng emits them (noch:{LTTng} live)
6495
6496 LTTng live is a network protocol implemented by the <<lttng-relayd,relay
6497 daemon>> (man:lttng-relayd(8)) to allow compatible trace viewers to
6498 display events as LTTng emits them on the target system while tracing is
6499 active.
6500
6501 The relay daemon creates a _tee_: it forwards the trace data to both
6502 the local file system and to connected live viewers:
6503
6504 [role="img-90"]
6505 .The relay daemon creates a _tee_, forwarding the trace data to both trace files and a connected live viewer.
6506 image::live.png[]
6507
6508 To use LTTng live:
6509
6510 . On the _target system_, create a <<tracing-session,tracing session>>
6511   in _live mode_:
6512 +
6513 --
6514 [role="term"]
6515 ----
6516 $ lttng create my-session --live
6517 ----
6518 --
6519 +
6520 This spawns a local relay daemon.
6521
6522 . Start the live viewer and configure it to connect to the relay
6523   daemon. For example, with http://diamon.org/babeltrace[Babeltrace]:
6524 +
6525 --
6526 [role="term"]
6527 ----
6528 $ babeltrace --input-format=lttng-live \
6529            net://localhost/host/hostname/my-session
6530 ----
6531 --
6532 +
6533 Replace:
6534 +
6535 --
6536 * `hostname` with the host name of the target system.
6537 * `my-session` with the name of the tracing session to view.
6538 --
6539
6540 . Configure the tracing session as usual with the man:lttng(1)
6541   command-line tool, and <<basic-tracing-session-control,start tracing>>.
6542
6543 You can list the available live tracing sessions with Babeltrace:
6544
6545 [role="term"]
6546 ----
6547 $ babeltrace --input-format=lttng-live net://localhost
6548 ----
6549
6550 You can start the relay daemon on another system. In this case, you need
6551 to specify the relay daemon's URL when you create the tracing session
6552 with the opt:lttng-create(1):--set-url option. You also need to replace
6553 `localhost` in the procedure above with the host name of the system on
6554 which the relay daemon is running.
6555
6556 See man:lttng-create(1) and man:lttng-relayd(8) for the complete list of
6557 command-line options.
6558
6559
6560 [role="since-2.3"]
6561 [[taking-a-snapshot]]
6562 === Take a snapshot of the current sub-buffers of a tracing session
6563
6564 The normal behavior of LTTng is to append full sub-buffers to growing
6565 trace data files. This is ideal to keep a full history of the events
6566 that occurred on the target system, but it can
6567 represent too much data in some situations. For example, you may wish
6568 to trace your application continuously until some critical situation
6569 happens, in which case you only need the latest few recorded
6570 events to perform the desired analysis, not multi-gigabyte trace files.
6571
6572 With the man:lttng-snapshot(1) command, you can take a snapshot of the
6573 current sub-buffers of a given <<tracing-session,tracing session>>.
6574 LTTng can write the snapshot to the local file system or send it over
6575 the network.
6576
6577 [role="img-100"]
6578 .A snapshot is a copy of the current sub-buffers, which are not cleared after the operation.
6579 image::snapshot.png[]
6580
6581 If you wish to create unmanaged, self-contained, non-overlapping
6582 trace chunk archives instead of a simple copy of the current
6583 sub-buffers, see the <<session-rotation,tracing session rotation>>
6584 feature (available since LTTng{nbsp}2.11).
6585
6586 To take a snapshot:
6587
6588 . Create a tracing session in _snapshot mode_:
6589 +
6590 --
6591 [role="term"]
6592 ----
6593 $ lttng create my-session --snapshot
6594 ----
6595 --
6596 +
6597 The <<channel-overwrite-mode-vs-discard-mode,event record loss mode>> of
6598 <<channel,channels>> created in this mode is automatically set to
6599 _overwrite_ (flight recorder mode).
6600
6601 . Configure the tracing session as usual with the man:lttng(1)
6602   command-line tool, and <<basic-tracing-session-control,start tracing>>.
6603
6604 . **Optional**: When you need to take a snapshot,
6605   <<basic-tracing-session-control,stop tracing>>.
6606 +
6607 You can take a snapshot when the tracers are active, but if you stop
6608 them first, you are sure that the data in the sub-buffers does not
6609 change before you actually take the snapshot.
6610
6611 . Take a snapshot:
6612 +
6613 --
6614 [role="term"]
6615 ----
6616 $ lttng snapshot record --name=my-first-snapshot
6617 ----
6618 --
6619 +
6620 LTTng writes the current sub-buffers of all the current tracing
6621 session's channels to trace files on the local file system. Those trace
6622 files have `my-first-snapshot` in their name.
6623
6624 There is no difference between the format of a normal trace file and the
6625 format of a snapshot: viewers of LTTng traces also support LTTng
6626 snapshots.
6627
6628 By default, LTTng writes snapshot files to the path shown by
6629 `lttng snapshot list-output`. You can change this path or decide to send
6630 snapshots over the network using either:
6631
6632 . An output path or URL that you specify when you
6633   <<creating-destroying-tracing-sessions,create the tracing session>>.
6634 . A snapshot output path or URL that you add using
6635   `lttng snapshot add-output`.
6636 . An output path or URL that you provide directly to the
6637   `lttng snapshot record` command.
6638
6639 Method{nbsp}3 overrides method{nbsp}2, which overrides method 1. When
6640 you specify a URL, a relay daemon must listen on a remote system (see
6641 <<sending-trace-data-over-the-network,Send trace data over the
6642 network>>).
6643
6644
6645 [role="since-2.11"]
6646 [[session-rotation]]
6647 === Archive the current trace chunk (rotate a tracing session)
6648
6649 The <<taking-a-snapshot,snapshot user guide>> shows how you can dump
6650 a tracing session's current sub-buffers to the file system or send them
6651 over the network. When you take a snapshot, LTTng does not clear the
6652 tracing session's ring buffers: if you take another snapshot immediately
6653 after, both snapshots could contain overlapping trace data.
6654
6655 Inspired by https://en.wikipedia.org/wiki/Log_rotation[log rotation],
6656 _tracing session rotation_ is a feature which appends the content of the
6657 ring buffers to what's already on the file system or sent over the
6658 network since the tracing session's creation or since the last
6659 rotation, and then clears those ring buffers to avoid trace data
6660 overlaps.
6661
6662 What LTTng is about to write when performing a tracing session rotation
6663 is called the _current trace chunk_. When this current trace chunk is
6664 written to the file system or sent over the network, it is called a
6665 _trace chunk archive_. Therefore, a tracing session rotation _archives_
6666 the current trace chunk.
6667
6668 [role="img-100"]
6669 .A tracing session rotation operation _archives_ the current trace chunk.
6670 image::rotation.png[]
6671
6672 A trace chunk archive is a self-contained LTTng trace and is not managed
6673 anymore by LTTng: you can read it, modify it, move it, or remove it.
6674
6675 There are two methods to perform a tracing session rotation:
6676 immediately or automatically.
6677
6678 To perform an immediate tracing session rotation:
6679
6680 . <<creating-destroying-tracing-sessions,Create a tracing session>>
6681   in _normal mode_ or _network streaming mode_
6682   (only those two creation modes support tracing session rotation):
6683 +
6684 --
6685 [role="term"]
6686 ----
6687 $ lttng create my-session
6688 ----
6689 --
6690
6691 . <<enabling-disabling-events,Create one or more event rules>>
6692 and <<basic-tracing-session-control,start tracing>>:
6693 +
6694 --
6695 [role="term"]
6696 ----
6697 $ lttng enable-event --kernel sched_'*'
6698 $ lttng start
6699 ----
6700 --
6701
6702 . When needed, immediately rotate the current tracing session:
6703 +
6704 --
6705 [role="term"]
6706 ----
6707 $ lttng rotate
6708 ----
6709 --
6710 +
6711 The cmd:lttng-rotate command prints the path to the created trace
6712 chunk archive. See man:lttng-rotate(1) to learn about the format
6713 of trace chunk archive directory names.
6714 +
6715 You can perform other immediate rotations while the tracing session is
6716 active. It is guaranteed that all the trace chunk archives do not
6717 contain overlapping trace data. You can also perform an immediate
6718 rotation once the tracing session is
6719 <<basic-tracing-session-control,stopped>>.
6720
6721 . When you are done tracing,
6722   <<creating-destroying-tracing-sessions,destroy the current tracing
6723   session>>:
6724 +
6725 --
6726 [role="term"]
6727 ----
6728 $ lttng destroy
6729 ----
6730 --
6731 +
6732 The tracing session destruction operation creates one last trace
6733 chunk archive from the current trace chunk.
6734
6735 An automatic tracing session rotation is a rotation which LTTng
6736 performs automatically based on one of the following conditions:
6737
6738 * A timer with a configured period times out.
6739 * The total size of the flushed part of the current trace chunk
6740   becomes greater than or equal to a configured value.
6741
6742 To configure a future, automatic tracing session rotation, you need
6743 to set a _rotation schedule_.
6744
6745 To set a rotation schedule:
6746
6747 . <<creating-destroying-tracing-sessions,Create a tracing session>>
6748   in _normal mode_ or _network streaming mode_
6749   (only those two creation modes support tracing session rotation):
6750 +
6751 --
6752 [role="term"]
6753 ----
6754 $ lttng create my-session
6755 ----
6756 --
6757
6758 . <<enabling-disabling-events,Create one or more event rules>>:
6759 +
6760 --
6761 [role="term"]
6762 ----
6763 $ lttng enable-event --kernel sched_'*'
6764 ----
6765 --
6766
6767 . Set a tracing session rotation schedule:
6768 +
6769 --
6770 [role="term"]
6771 ----
6772 $ lttng enable-rotation --timer=12s
6773 ----
6774 --
6775 +
6776 In this example, we set a rotation schedule so that LTTng performs a
6777 tracing session rotation every 12{nbsp}seconds.
6778 +
6779 See man:lttng-enable-rotation(1) to learn more about other ways to set a
6780 rotation schedule.
6781
6782 . <<basic-tracing-session-control,Start tracing>>:
6783 +
6784 --
6785 [role="term"]
6786 ----
6787 $ lttng start
6788 ----
6789 --
6790 +
6791 LTTng performs tracing session rotations automatically while the tracing
6792 session is active thanks to the rotation schedule.
6793
6794 . When you are done tracing,
6795   <<creating-destroying-tracing-sessions,destroy the current tracing
6796   session>>:
6797 +
6798 --
6799 [role="term"]
6800 ----
6801 $ lttng destroy
6802 ----
6803 --
6804 +
6805 The tracing session destruction operation creates one last trace chunk
6806 archive from the current trace chunk.
6807
6808 You can use man:lttng-disable-rotation(1) to unset an a tracing session
6809 rotation schedule.
6810
6811 NOTE: man:lttng-rotate(1) and man:lttng-enable-rotation(1) list
6812 limitations regarding those two commands.
6813
6814
6815 [role="since-2.6"]
6816 [[mi]]
6817 === Use the machine interface
6818
6819 With any command of the man:lttng(1) command-line tool, you can set the
6820 opt:lttng(1):--mi option to `xml` (before the command name) to get an
6821 XML machine interface output, for example:
6822
6823 [role="term"]
6824 ----
6825 $ lttng --mi=xml enable-event --kernel --syscall open
6826 ----
6827
6828 A schema definition (XSD) is
6829 https://github.com/lttng/lttng-tools/blob/stable-2.11/src/common/mi-lttng-3.0.xsd[available]
6830 to ease the integration with external tools as much as possible.
6831
6832
6833 [role="since-2.8"]
6834 [[metadata-regenerate]]
6835 === Regenerate the metadata of an LTTng trace
6836
6837 An LTTng trace, which is a http://diamon.org/ctf[CTF] trace, has both
6838 data stream files and a metadata file. This metadata file contains,
6839 amongst other things, information about the offset of the clock sources
6840 used to timestamp <<event,event records>> when tracing.
6841
6842 If, once a <<tracing-session,tracing session>> is
6843 <<basic-tracing-session-control,started>>, a major
6844 https://en.wikipedia.org/wiki/Network_Time_Protocol[NTP] correction
6845 happens, the trace's clock offset also needs to be updated. You
6846 can use the `metadata` item of the man:lttng-regenerate(1) command
6847 to do so.
6848
6849 The main use case of this command is to allow a system to boot with
6850 an incorrect wall time and trace it with LTTng before its wall time
6851 is corrected. Once the system is known to be in a state where its
6852 wall time is correct, it can run `lttng regenerate metadata`.
6853
6854 To regenerate the metadata of an LTTng trace:
6855
6856 * Use the `metadata` item of the man:lttng-regenerate(1) command:
6857 +
6858 --
6859 [role="term"]
6860 ----
6861 $ lttng regenerate metadata
6862 ----
6863 --
6864
6865 [IMPORTANT]
6866 ====
6867 `lttng regenerate metadata` has the following limitations:
6868
6869 * Tracing session <<creating-destroying-tracing-sessions,created>>
6870   in non-live mode.
6871 * User space <<channel,channels>>, if any, are using
6872   <<channel-buffering-schemes,per-user buffering>>.
6873 ====
6874
6875
6876 [role="since-2.9"]
6877 [[regenerate-statedump]]
6878 === Regenerate the state dump of a tracing session
6879
6880 The LTTng kernel and user space tracers generate state dump
6881 <<event,event records>> when the application starts or when you
6882 <<basic-tracing-session-control,start a tracing session>>. An analysis
6883 can use the state dump event records to set an initial state before it
6884 builds the rest of the state from the following event records.
6885 http://tracecompass.org/[Trace Compass] is a notable example of an
6886 application which uses the state dump of an LTTng trace.
6887
6888 When you <<taking-a-snapshot,take a snapshot>>, it's possible that the
6889 state dump event records are not included in the snapshot because they
6890 were recorded to a sub-buffer that has been consumed or overwritten
6891 already.
6892
6893 You can use the `lttng regenerate statedump` command to emit the state
6894 dump event records again.
6895
6896 To regenerate the state dump of the current tracing session, provided
6897 create it in snapshot mode, before you take a snapshot:
6898
6899 . Use the `statedump` item of the man:lttng-regenerate(1) command:
6900 +
6901 --
6902 [role="term"]
6903 ----
6904 $ lttng regenerate statedump
6905 ----
6906 --
6907
6908 . <<basic-tracing-session-control,Stop the tracing session>>:
6909 +
6910 --
6911 [role="term"]
6912 ----
6913 $ lttng stop
6914 ----
6915 --
6916
6917 . <<taking-a-snapshot,Take a snapshot>>:
6918 +
6919 --
6920 [role="term"]
6921 ----
6922 $ lttng snapshot record --name=my-snapshot
6923 ----
6924 --
6925
6926 Depending on the event throughput, you should run steps 1 and 2
6927 as closely as possible.
6928
6929 NOTE: To record the state dump events, you need to
6930 <<enabling-disabling-events,create event rules>> which enable them.
6931 LTTng-UST state dump tracepoints start with `lttng_ust_statedump:`.
6932 LTTng-modules state dump tracepoints start with `lttng_statedump_`.
6933
6934
6935 [role="since-2.7"]
6936 [[persistent-memory-file-systems]]
6937 === Record trace data on persistent memory file systems
6938
6939 https://en.wikipedia.org/wiki/Non-volatile_random-access_memory[Non-volatile random-access memory]
6940 (NVRAM) is random-access memory that retains its information when power
6941 is turned off (non-volatile). Systems with such memory can store data
6942 structures in RAM and retrieve them after a reboot, without flushing
6943 to typical _storage_.
6944
6945 Linux supports NVRAM file systems thanks to either
6946 http://pramfs.sourceforge.net/[PRAMFS] or
6947 https://www.kernel.org/doc/Documentation/filesystems/dax.txt[DAX]{nbsp}+{nbsp}http://lkml.iu.edu/hypermail/linux/kernel/1504.1/03463.html[pmem]
6948 (requires Linux{nbsp}4.1+).
6949
6950 This section does not describe how to operate such file systems;
6951 we assume that you have a working persistent memory file system.
6952
6953 When you create a <<tracing-session,tracing session>>, you can specify
6954 the path of the shared memory holding the sub-buffers. If you specify a
6955 location on an NVRAM file system, then you can retrieve the latest
6956 recorded trace data when the system reboots after a crash.
6957
6958 To record trace data on a persistent memory file system and retrieve the
6959 trace data after a system crash:
6960
6961 . Create a tracing session with a sub-buffer shared memory path located
6962   on an NVRAM file system:
6963 +
6964 --
6965 [role="term"]
6966 ----
6967 $ lttng create my-session --shm-path=/path/to/shm
6968 ----
6969 --
6970
6971 . Configure the tracing session as usual with the man:lttng(1)
6972   command-line tool, and <<basic-tracing-session-control,start tracing>>.
6973
6974 . After a system crash, use the man:lttng-crash(1) command-line tool to
6975   view the trace data recorded on the NVRAM file system:
6976 +
6977 --
6978 [role="term"]
6979 ----
6980 $ lttng-crash /path/to/shm
6981 ----
6982 --
6983
6984 The binary layout of the ring buffer files is not exactly the same as
6985 the trace files layout. This is why you need to use man:lttng-crash(1)
6986 instead of your preferred trace viewer directly.
6987
6988 To convert the ring buffer files to LTTng trace files:
6989
6990 * Use the opt:lttng-crash(1):--extract option of man:lttng-crash(1):
6991 +
6992 --
6993 [role="term"]
6994 ----
6995 $ lttng-crash --extract=/path/to/trace /path/to/shm
6996 ----
6997 --
6998
6999
7000 [role="since-2.10"]
7001 [[notif-trigger-api]]
7002 === Get notified when a channel's buffer usage is too high or too low
7003
7004 With LTTng's $$C/C++$$ notification and trigger API, your user
7005 application can get notified when the buffer usage of one or more
7006 <<channel,channels>> becomes too low or too high. You can use this API
7007 and enable or disable <<event,event rules>> during tracing to avoid
7008 <<channel-overwrite-mode-vs-discard-mode,discarded event records>>.
7009
7010 .Have a user application get notified when an LTTng channel's buffer usage is too high.
7011 ====
7012 In this example, we create and build an application which gets notified
7013 when the buffer usage of a specific LTTng channel is higher than
7014 75{nbsp}%. We only print that it is the case in the example, but we
7015 could as well use the API of <<liblttng-ctl-lttng,`liblttng-ctl`>> to
7016 disable event rules when this happens.
7017
7018 . Create the application's C source file:
7019 +
7020 --
7021 [source,c]
7022 .path:{notif-app.c}
7023 ----
7024 #include <stdio.h>
7025 #include <assert.h>
7026 #include <lttng/domain.h>
7027 #include <lttng/action/action.h>
7028 #include <lttng/action/notify.h>
7029 #include <lttng/condition/condition.h>
7030 #include <lttng/condition/buffer-usage.h>
7031 #include <lttng/condition/evaluation.h>
7032 #include <lttng/notification/channel.h>
7033 #include <lttng/notification/notification.h>
7034 #include <lttng/trigger/trigger.h>
7035 #include <lttng/endpoint.h>
7036
7037 int main(int argc, char *argv[])
7038 {
7039     int exit_status = 0;
7040     struct lttng_notification_channel *notification_channel;
7041     struct lttng_condition *condition;
7042     struct lttng_action *action;
7043     struct lttng_trigger *trigger;
7044     const char *tracing_session_name;
7045     const char *channel_name;
7046
7047     assert(argc >= 3);
7048     tracing_session_name = argv[1];
7049     channel_name = argv[2];
7050
7051     /*
7052      * Create a notification channel. A notification channel
7053      * connects the user application to the LTTng session daemon.
7054      * This notification channel can be used to listen to various
7055      * types of notifications.
7056      */
7057     notification_channel = lttng_notification_channel_create(
7058         lttng_session_daemon_notification_endpoint);
7059
7060     /*
7061      * Create a "high buffer usage" condition. In this case, the
7062      * condition is reached when the buffer usage is greater than or
7063      * equal to 75 %. We create the condition for a specific tracing
7064      * session name, channel name, and for the user space tracing
7065      * domain.
7066      *
7067      * The "low buffer usage" condition type also exists.
7068      */
7069     condition = lttng_condition_buffer_usage_high_create();
7070     lttng_condition_buffer_usage_set_threshold_ratio(condition, .75);
7071     lttng_condition_buffer_usage_set_session_name(
7072         condition, tracing_session_name);
7073     lttng_condition_buffer_usage_set_channel_name(condition,
7074         channel_name);
7075     lttng_condition_buffer_usage_set_domain_type(condition,
7076         LTTNG_DOMAIN_UST);
7077
7078     /*
7079      * Create an action (get a notification) to take when the
7080      * condition created above is reached.
7081      */
7082     action = lttng_action_notify_create();
7083
7084     /*
7085      * Create a trigger. A trigger associates a condition to an
7086      * action: the action is executed when the condition is reached.
7087      */
7088     trigger = lttng_trigger_create(condition, action);
7089
7090     /* Register the trigger to LTTng. */
7091     lttng_register_trigger(trigger);
7092
7093     /*
7094      * Now that we have registered a trigger, a notification will be
7095      * emitted everytime its condition is met. To receive this
7096      * notification, we must subscribe to notifications that match
7097      * the same condition.
7098      */
7099     lttng_notification_channel_subscribe(notification_channel,
7100         condition);
7101
7102     /*
7103      * Notification loop. You can put this in a dedicated thread to
7104      * avoid blocking the main thread.
7105      */
7106     for (;;) {
7107         struct lttng_notification *notification;
7108         enum lttng_notification_channel_status status;
7109         const struct lttng_evaluation *notification_evaluation;
7110         const struct lttng_condition *notification_condition;
7111         double buffer_usage;
7112
7113         /* Receive the next notification. */
7114         status = lttng_notification_channel_get_next_notification(
7115             notification_channel, &notification);
7116
7117         switch (status) {
7118         case LTTNG_NOTIFICATION_CHANNEL_STATUS_OK:
7119             break;
7120         case LTTNG_NOTIFICATION_CHANNEL_STATUS_NOTIFICATIONS_DROPPED:
7121             /*
7122              * The session daemon can drop notifications if
7123              * a monitoring application is not consuming the
7124              * notifications fast enough.
7125              */
7126             continue;
7127         case LTTNG_NOTIFICATION_CHANNEL_STATUS_CLOSED:
7128             /*
7129              * The notification channel has been closed by the
7130              * session daemon. This is typically caused by a session
7131              * daemon shutting down.
7132              */
7133             goto end;
7134         default:
7135             /* Unhandled conditions or errors. */
7136             exit_status = 1;
7137             goto end;
7138         }
7139
7140         /*
7141          * A notification provides, amongst other things:
7142          *
7143          * * The condition that caused this notification to be
7144          *   emitted.
7145          * * The condition evaluation, which provides more
7146          *   specific information on the evaluation of the
7147          *   condition.
7148          *
7149          * The condition evaluation provides the buffer usage
7150          * value at the moment the condition was reached.
7151          */
7152         notification_condition = lttng_notification_get_condition(
7153             notification);
7154         notification_evaluation = lttng_notification_get_evaluation(
7155             notification);
7156
7157         /* We're subscribed to only one condition. */
7158         assert(lttng_condition_get_type(notification_condition) ==
7159             LTTNG_CONDITION_TYPE_BUFFER_USAGE_HIGH);
7160
7161         /*
7162          * Get the exact sampled buffer usage from the
7163          * condition evaluation.
7164          */
7165         lttng_evaluation_buffer_usage_get_usage_ratio(
7166             notification_evaluation, &buffer_usage);
7167
7168         /*
7169          * At this point, instead of printing a message, we
7170          * could do something to reduce the channel's buffer
7171          * usage, like disable specific events.
7172          */
7173         printf("Buffer usage is %f %% in tracing session \"%s\", "
7174             "user space channel \"%s\".\n", buffer_usage * 100,
7175             tracing_session_name, channel_name);
7176         lttng_notification_destroy(notification);
7177     }
7178
7179 end:
7180     lttng_action_destroy(action);
7181     lttng_condition_destroy(condition);
7182     lttng_trigger_destroy(trigger);
7183     lttng_notification_channel_destroy(notification_channel);
7184     return exit_status;
7185 }
7186 ----
7187 --
7188
7189 . Build the `notif-app` application, linking it to `liblttng-ctl`:
7190 +
7191 --
7192 [role="term"]
7193 ----
7194 $ gcc -o notif-app notif-app.c -llttng-ctl
7195 ----
7196 --
7197
7198 . <<creating-destroying-tracing-sessions,Create a tracing session>>,
7199   <<enabling-disabling-events,create an event rule>> matching all the
7200   user space tracepoints, and
7201   <<basic-tracing-session-control,start tracing>>:
7202 +
7203 --
7204 [role="term"]
7205 ----
7206 $ lttng create my-session
7207 $ lttng enable-event --userspace --all
7208 $ lttng start
7209 ----
7210 --
7211 +
7212 If you create the channel manually with the man:lttng-enable-channel(1)
7213 command, you can control how frequently are the current values of the
7214 channel's properties sampled to evaluate user conditions with the
7215 opt:lttng-enable-channel(1):--monitor-timer option.
7216
7217 . Run the `notif-app` application. This program accepts the
7218   <<tracing-session,tracing session>> name and the user space channel
7219   name as its two first arguments. The channel which LTTng automatically
7220   creates with the man:lttng-enable-event(1) command above is named
7221   `channel0`:
7222 +
7223 --
7224 [role="term"]
7225 ----
7226 $ ./notif-app my-session channel0
7227 ----
7228 --
7229
7230 . In another terminal, run an application with a very high event
7231   throughput so that the 75{nbsp}% buffer usage condition is reached.
7232 +
7233 In the first terminal, the application should print lines like this:
7234 +
7235 ----
7236 Buffer usage is 81.45197 % in tracing session "my-session", user space
7237 channel "channel0".
7238 ----
7239 +
7240 If you don't see anything, try modifying the condition in
7241 path:{notif-app.c} to a lower value (0.1, for example), rebuilding it
7242 (step{nbsp}2) and running it again (step{nbsp}4).
7243 ====
7244
7245
7246 [[reference]]
7247 == Reference
7248
7249 [[lttng-modules-ref]]
7250 === noch:{LTTng-modules}
7251
7252
7253 [role="since-2.9"]
7254 [[lttng-tracepoint-enum]]
7255 ==== `LTTNG_TRACEPOINT_ENUM()` usage
7256
7257 Use the `LTTNG_TRACEPOINT_ENUM()` macro to define an enumeration:
7258
7259 [source,c]
7260 ----
7261 LTTNG_TRACEPOINT_ENUM(name, TP_ENUM_VALUES(entries))
7262 ----
7263
7264 Replace:
7265
7266 * `name` with the name of the enumeration (C identifier, unique
7267   amongst all the defined enumerations).
7268 * `entries` with a list of enumeration entries.
7269
7270 The available enumeration entry macros are:
7271
7272 +ctf_enum_value(__name__, __value__)+::
7273   Entry named +__name__+ mapped to the integral value +__value__+.
7274
7275 +ctf_enum_range(__name__, __begin__, __end__)+::
7276   Entry named +__name__+ mapped to the range of integral values between
7277   +__begin__+ (included) and +__end__+ (included).
7278
7279 +ctf_enum_auto(__name__)+::
7280   Entry named +__name__+ mapped to the integral value following the
7281   last mapping's value.
7282 +
7283 The last value of a `ctf_enum_value()` entry is its +__value__+
7284 parameter.
7285 +
7286 The last value of a `ctf_enum_range()` entry is its +__end__+ parameter.
7287 +
7288 If `ctf_enum_auto()` is the first entry in the list, its integral
7289 value is 0.
7290
7291 Use the `ctf_enum()` <<lttng-modules-tp-fields,field definition macro>>
7292 to use a defined enumeration as a tracepoint field.
7293
7294 .Define an enumeration with `LTTNG_TRACEPOINT_ENUM()`.
7295 ====
7296 [source,c]
7297 ----
7298 LTTNG_TRACEPOINT_ENUM(
7299     my_enum,
7300     TP_ENUM_VALUES(
7301         ctf_enum_auto("AUTO: EXPECT 0")
7302         ctf_enum_value("VALUE: 23", 23)
7303         ctf_enum_value("VALUE: 27", 27)
7304         ctf_enum_auto("AUTO: EXPECT 28")
7305         ctf_enum_range("RANGE: 101 TO 303", 101, 303)
7306         ctf_enum_auto("AUTO: EXPECT 304")
7307     )
7308 )
7309 ----
7310 ====
7311
7312
7313 [role="since-2.7"]
7314 [[lttng-modules-tp-fields]]
7315 ==== Tracepoint fields macros (for `TP_FIELDS()`)
7316
7317 [[tp-fast-assign]][[tp-struct-entry]]The available macros to define
7318 tracepoint fields, which must be listed within `TP_FIELDS()` in
7319 `LTTNG_TRACEPOINT_EVENT()`, are:
7320
7321 [role="func-desc growable",cols="asciidoc,asciidoc"]
7322 .Available macros to define LTTng-modules tracepoint fields
7323 |====
7324 |Macro |Description and parameters
7325
7326 |
7327 +ctf_integer(__t__, __n__, __e__)+
7328
7329 +ctf_integer_nowrite(__t__, __n__, __e__)+
7330
7331 +ctf_user_integer(__t__, __n__, __e__)+
7332
7333 +ctf_user_integer_nowrite(__t__, __n__, __e__)+
7334 |
7335 Standard integer, displayed in base{nbsp}10.
7336
7337 +__t__+::
7338   Integer C type (`int`, `long`, `size_t`, ...).
7339
7340 +__n__+::
7341   Field name.
7342
7343 +__e__+::
7344   Argument expression.
7345
7346 |
7347 +ctf_integer_hex(__t__, __n__, __e__)+
7348
7349 +ctf_user_integer_hex(__t__, __n__, __e__)+
7350 |
7351 Standard integer, displayed in base{nbsp}16.
7352
7353 +__t__+::
7354   Integer C type.
7355
7356 +__n__+::
7357   Field name.
7358
7359 +__e__+::
7360   Argument expression.
7361
7362 |+ctf_integer_oct(__t__, __n__, __e__)+
7363 |
7364 Standard integer, displayed in base{nbsp}8.
7365
7366 +__t__+::
7367   Integer C type.
7368
7369 +__n__+::
7370   Field name.
7371
7372 +__e__+::
7373   Argument expression.
7374
7375 |
7376 +ctf_integer_network(__t__, __n__, __e__)+
7377
7378 +ctf_user_integer_network(__t__, __n__, __e__)+
7379 |
7380 Integer in network byte order (big-endian), displayed in base{nbsp}10.
7381
7382 +__t__+::
7383   Integer C type.
7384
7385 +__n__+::
7386   Field name.
7387
7388 +__e__+::
7389   Argument expression.
7390
7391 |
7392 +ctf_integer_network_hex(__t__, __n__, __e__)+
7393
7394 +ctf_user_integer_network_hex(__t__, __n__, __e__)+
7395 |
7396 Integer in network byte order, displayed in base{nbsp}16.
7397
7398 +__t__+::
7399   Integer C type.
7400
7401 +__n__+::
7402   Field name.
7403
7404 +__e__+::
7405   Argument expression.
7406
7407 |
7408 +ctf_enum(__N__, __t__, __n__, __e__)+
7409
7410 +ctf_enum_nowrite(__N__, __t__, __n__, __e__)+
7411
7412 +ctf_user_enum(__N__, __t__, __n__, __e__)+
7413
7414 +ctf_user_enum_nowrite(__N__, __t__, __n__, __e__)+
7415 |
7416 Enumeration.
7417
7418 +__N__+::
7419   Name of a <<lttng-tracepoint-enum,previously defined enumeration>>.
7420
7421 +__t__+::
7422   Integer C type (`int`, `long`, `size_t`, ...).
7423
7424 +__n__+::
7425   Field name.
7426
7427 +__e__+::
7428   Argument expression.
7429
7430 |
7431 +ctf_string(__n__, __e__)+
7432
7433 +ctf_string_nowrite(__n__, __e__)+
7434
7435 +ctf_user_string(__n__, __e__)+
7436
7437 +ctf_user_string_nowrite(__n__, __e__)+
7438 |
7439 Null-terminated string; undefined behavior if +__e__+ is `NULL`.
7440
7441 +__n__+::
7442   Field name.
7443
7444 +__e__+::
7445   Argument expression.
7446
7447 |
7448 +ctf_array(__t__, __n__, __e__, __s__)+
7449
7450 +ctf_array_nowrite(__t__, __n__, __e__, __s__)+
7451
7452 +ctf_user_array(__t__, __n__, __e__, __s__)+
7453
7454 +ctf_user_array_nowrite(__t__, __n__, __e__, __s__)+
7455 |
7456 Statically-sized array of integers.
7457
7458 +__t__+::
7459   Array element C type.
7460
7461 +__n__+::
7462   Field name.
7463
7464 +__e__+::
7465   Argument expression.
7466
7467 +__s__+::
7468   Number of elements.
7469
7470 |
7471 +ctf_array_bitfield(__t__, __n__, __e__, __s__)+
7472
7473 +ctf_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7474
7475 +ctf_user_array_bitfield(__t__, __n__, __e__, __s__)+
7476
7477 +ctf_user_array_bitfield_nowrite(__t__, __n__, __e__, __s__)+
7478 |
7479 Statically-sized array of bits.
7480
7481 The type of +__e__+ must be an integer type. +__s__+ is the number
7482 of elements of such type in +__e__+, not the number of bits.
7483
7484 +__t__+::
7485   Array element C type.
7486
7487 +__n__+::
7488   Field name.
7489
7490 +__e__+::
7491   Argument expression.
7492
7493 +__s__+::
7494   Number of elements.
7495
7496 |
7497 +ctf_array_text(__t__, __n__, __e__, __s__)+
7498
7499 +ctf_array_text_nowrite(__t__, __n__, __e__, __s__)+
7500
7501 +ctf_user_array_text(__t__, __n__, __e__, __s__)+
7502
7503 +ctf_user_array_text_nowrite(__t__, __n__, __e__, __s__)+
7504 |
7505 Statically-sized array, printed as text.
7506
7507 The string does not need to be null-terminated.
7508
7509 +__t__+::
7510   Array element C type (always `char`).
7511
7512 +__n__+::
7513   Field name.
7514
7515 +__e__+::
7516   Argument expression.
7517
7518 +__s__+::
7519   Number of elements.
7520
7521 |
7522 +ctf_sequence(__t__, __n__, __e__, __T__, __E__)+
7523
7524 +ctf_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7525
7526 +ctf_user_sequence(__t__, __n__, __e__, __T__, __E__)+
7527
7528 +ctf_user_sequence_nowrite(__t__, __n__, __e__, __T__, __E__)+
7529 |
7530 Dynamically-sized array of integers.
7531
7532 The type of +__E__+ must be unsigned.
7533
7534 +__t__+::
7535   Array element C type.
7536
7537 +__n__+::
7538   Field name.
7539
7540 +__e__+::
7541   Argument expression.
7542
7543 +__T__+::
7544   Length expression C type.
7545
7546 +__E__+::
7547   Length expression.
7548
7549 |
7550 +ctf_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7551
7552 +ctf_user_sequence_hex(__t__, __n__, __e__, __T__, __E__)+
7553 |
7554 Dynamically-sized array of integers, displayed in base{nbsp}16.
7555
7556 The type of +__E__+ must be unsigned.
7557
7558 +__t__+::
7559   Array element C type.
7560
7561 +__n__+::
7562   Field name.
7563
7564 +__e__+::
7565   Argument expression.
7566
7567 +__T__+::
7568   Length expression C type.
7569
7570 +__E__+::
7571   Length expression.
7572
7573 |+ctf_sequence_network(__t__, __n__, __e__, __T__, __E__)+
7574 |
7575 Dynamically-sized array of integers in network byte order (big-endian),
7576 displayed in base{nbsp}10.
7577
7578 The type of +__E__+ must be unsigned.
7579
7580 +__t__+::
7581   Array element C type.
7582
7583 +__n__+::
7584   Field name.
7585
7586 +__e__+::
7587   Argument expression.
7588
7589 +__T__+::
7590   Length expression C type.
7591
7592 +__E__+::
7593   Length expression.
7594
7595 |
7596 +ctf_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7597
7598 +ctf_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7599
7600 +ctf_user_sequence_bitfield(__t__, __n__, __e__, __T__, __E__)+
7601
7602 +ctf_user_sequence_bitfield_nowrite(__t__, __n__, __e__, __T__, __E__)+
7603 |
7604 Dynamically-sized array of bits.
7605
7606 The type of +__e__+ must be an integer type. +__s__+ is the number
7607 of elements of such type in +__e__+, not the number of bits.
7608
7609 The type of +__E__+ must be unsigned.
7610
7611 +__t__+::
7612   Array element C type.
7613
7614 +__n__+::
7615   Field name.
7616
7617 +__e__+::
7618   Argument expression.
7619
7620 +__T__+::
7621   Length expression C type.
7622
7623 +__E__+::
7624   Length expression.
7625
7626 |
7627 +ctf_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7628
7629 +ctf_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7630
7631 +ctf_user_sequence_text(__t__, __n__, __e__, __T__, __E__)+
7632
7633 +ctf_user_sequence_text_nowrite(__t__, __n__, __e__, __T__, __E__)+
7634 |
7635 Dynamically-sized array, displayed as text.
7636
7637 The string does not need to be null-terminated.
7638
7639 The type of +__E__+ must be unsigned.
7640
7641 The behaviour is undefined if +__e__+ is `NULL`.
7642
7643 +__t__+::
7644   Sequence element C type (always `char`).
7645
7646 +__n__+::
7647   Field name.
7648
7649 +__e__+::
7650   Argument expression.
7651
7652 +__T__+::
7653   Length expression C type.
7654
7655 +__E__+::
7656   Length expression.
7657 |====
7658
7659 Use the `_user` versions when the argument expression, `e`, is
7660 a user space address. In the cases of `ctf_user_integer*()` and
7661 `ctf_user_float*()`, `&e` must be a user space address, thus `e` must
7662 be addressable.
7663
7664 The `_nowrite` versions omit themselves from the session trace, but are
7665 otherwise identical. This means the `_nowrite` fields won't be written
7666 in the recorded trace. Their primary purpose is to make some
7667 of the event context available to the
7668 <<enabling-disabling-events,event filters>> without having to
7669 commit the data to sub-buffers.
7670
7671
7672 [[glossary]]
7673 == Glossary
7674
7675 Terms related to LTTng and to tracing in general:
7676
7677 Babeltrace::
7678   The http://diamon.org/babeltrace[Babeltrace] project, which includes
7679   the cmd:babeltrace command, libraries, and Python bindings.
7680
7681 [[def-buffering-scheme]]<<channel-buffering-schemes,buffering scheme>>::
7682   A layout of <<def-sub-buffer,sub-buffers>> applied to a given channel.
7683
7684 <<channel,channel>>::
7685   An entity which is responsible for a set of
7686   <<def-ring-buffer,ring buffers>>.
7687 +
7688 <<event,Event rules>> are always attached to a specific channel.
7689
7690 clock::
7691   A source of time for a <<def-tracer,tracer>>.
7692
7693 <<lttng-consumerd,consumer daemon>>::
7694   A process which is responsible for consuming the full
7695   <<def-sub-buffer,sub-buffers>> and write them to a file system or
7696   send them over the network.
7697
7698 [[def-current-trace-chunk]]current trace chunk::
7699   A <<def-trace-chunk,trace chunk>> which includes the current content
7700   of all the <<tracing-session,tracing session>>'s
7701   <<def-sub-buffer,sub-buffers>> and the stream files produced since the
7702   latest event amongst:
7703 +
7704 * The creation of the tracing session.
7705 * The last <<session-rotation,tracing session rotation>>, if any.
7706
7707 <<channel-overwrite-mode-vs-discard-mode,discard mode>>:: The event
7708   record loss mode in which the <<def-tracer,tracer>> _discards_ new
7709   event records when there's no
7710   <<def-sub-buffer,sub-buffer>> space left to store them.
7711
7712 [[def-event]]event::
7713   The consequence of the execution of an
7714   <<def-instrumentation-point,instrumentation point>>, like a
7715   <<def-tracepoint,tracepoint>> that you manually place in some source
7716   code, or a Linux kernel kprobe.
7717 +
7718 An event is said to _occur_ at a specific time. Different actions can
7719 be taken upon the occurrence of an event, like record the event's payload
7720 to a <<def-sub-buffer,sub-buffer>>.
7721
7722 [[def-event-name]]event name::
7723   The name of an event, which is also the name of the event record.
7724   This is also called the _instrumentation point name_.
7725
7726 [[def-event-record]]event record::
7727   A record, in a <<def-trace,trace>>, of the payload of an event
7728   which occured.
7729
7730 [[def-event-record-loss-mode]]<<channel-overwrite-mode-vs-discard-mode,event record loss mode>>::
7731   The mechanism by which event records of a given <<channel,channel>>
7732   are lost (not recorded) when there is no <<def-sub-buffer,sub-buffer>>
7733   space left to store them.
7734
7735 <<event,event rule>>::
7736   Set of conditions which must be satisfied for one or more occuring
7737   events to be recorded.
7738
7739 `java.util.logging`::
7740   Java platform's
7741   https://docs.oracle.com/javase/7/docs/api/java/util/logging/package-summary.html[core logging facilities].
7742
7743 <<instrumenting,instrumentation>>::
7744   The use of LTTng probes to make a piece of software traceable.
7745
7746 [[def-instrumentation-point]]instrumentation point::
7747   A point in the execution path of a piece of software that, when
7748   reached by this execution, can emit an <<def-event,event>>.
7749
7750 instrumentation point name::
7751   See _<<def-event-name,event name>>_.
7752
7753 log4j::
7754   A http://logging.apache.org/log4j/1.2/[logging library] for Java
7755   developed by the Apache Software Foundation.
7756
7757 log level::
7758   Level of severity of a log statement or user space
7759   instrumentation point.
7760
7761 LTTng::
7762   The _Linux Trace Toolkit: next generation_ project.
7763
7764 <<lttng-cli,cmd:lttng>>::
7765   A command-line tool provided by the LTTng-tools project which you
7766   can use to send and receive control messages to and from a
7767   session daemon.
7768
7769 LTTng analyses::
7770   The https://github.com/lttng/lttng-analyses[LTTng analyses] project,
7771   which is a set of analyzing programs that are used to obtain a
7772   higher level view of an LTTng <<def-trace,trace>>.
7773
7774 cmd:lttng-consumerd::
7775   The name of the consumer daemon program.
7776
7777 cmd:lttng-crash::
7778   A utility provided by the LTTng-tools project which can convert
7779   <<def-ring-buffer,ring buffer>> files (usually
7780   <<persistent-memory-file-systems,saved on a persistent memory file system>>)
7781   to <<def-trace,trace>> files.
7782
7783 LTTng Documentation::
7784   This document.
7785
7786 <<lttng-live,LTTng live>>::
7787   A communication protocol between the <<lttng-relayd,relay daemon>> and
7788   live viewers which makes it possible to see <<def-event-record,event
7789   records>> "live", as they are received by the relay daemon.
7790
7791 <<lttng-modules,LTTng-modules>>::
7792   The https://github.com/lttng/lttng-modules[LTTng-modules] project,
7793   which contains the Linux kernel modules to make the Linux kernel
7794   <<def-instrumentation-point,instrumentation points>> available for
7795   LTTng tracing.
7796
7797 cmd:lttng-relayd::
7798   The name of the <<lttng-relayd,relay daemon>> program.
7799
7800 cmd:lttng-sessiond::
7801   The name of the <<lttng-sessiond,session daemon>> program.
7802
7803 LTTng-tools::
7804   The https://github.com/lttng/lttng-tools[LTTng-tools] project, which
7805   contains the various programs and libraries used to
7806   <<controlling-tracing,control tracing>>.
7807
7808 <<lttng-ust,LTTng-UST>>::
7809   The https://github.com/lttng/lttng-ust[LTTng-UST] project, which
7810   contains libraries to instrument
7811   <<def-user-application,user applications>>.
7812
7813 <<lttng-ust-agents,LTTng-UST Java agent>>::
7814   A Java package provided by the LTTng-UST project to allow the
7815   LTTng instrumentation of `java.util.logging` and Apache log4j{nbsp}1.2
7816   logging statements.
7817
7818 <<lttng-ust-agents,LTTng-UST Python agent>>::
7819   A Python package provided by the LTTng-UST project to allow the
7820   LTTng instrumentation of Python logging statements.
7821
7822 <<channel-overwrite-mode-vs-discard-mode,overwrite mode>>::
7823   The <<def-event-record-loss-mode,event record loss mode>> in which new
7824   <<def-event-record,event records>> _overwrite_ older event records
7825   when there's no <<def-sub-buffer,sub-buffer>> space left to store
7826   them.
7827
7828 <<channel-buffering-schemes,per-process buffering>>::
7829   A <<def-buffering-scheme,buffering scheme>> in which each instrumented
7830   process has its own <<def-sub-buffer,sub-buffers>> for a given user
7831   space <<channel,channel>>.
7832
7833 <<channel-buffering-schemes,per-user buffering>>::
7834   A <<def-buffering-scheme,buffering scheme>> in which all the processes
7835   of a Unix user share the same <<def-sub-buffer,sub-buffers>> for a
7836   given user space <<channel,channel>>.
7837
7838 <<lttng-relayd,relay daemon>>::
7839   A process which is responsible for receiving the <<def-trace,trace>>
7840   data sent by a distant <<lttng-consumerd,consumer daemon>>.
7841
7842 [[def-ring-buffer]]ring buffer::
7843   A set of <<def-sub-buffer,sub-buffers>>.
7844
7845 rotation::
7846   See _<<def-tracing-session-rotation,tracing session rotation>>_.
7847
7848 <<lttng-sessiond,session daemon>>::
7849   A process which receives control commands from you and orchestrates
7850   the <<def-tracer,tracers>> and various LTTng daemons.
7851
7852 <<taking-a-snapshot,snapshot>>::
7853   A copy of the current data of all the <<def-sub-buffer,sub-buffers>>
7854   of a given <<tracing-session,tracing session>>, saved as
7855   <<def-trace,trace>> files.
7856
7857 [[def-sub-buffer]]sub-buffer::
7858   One part of an LTTng <<def-ring-buffer,ring buffer>> which contains
7859   <<def-event-record,event records>>.
7860
7861 timestamp::
7862   The time information attached to an
7863   <<def-event,event>> when it is emitted.
7864
7865 [[def-trace]]trace (_noun_)::
7866   A set of:
7867 +
7868 * One http://diamon.org/ctf/[CTF] metadata stream file.
7869 * One or more CTF data stream files which are the concatenations of one
7870   or more flushed <<def-sub-buffer,sub-buffers>>.
7871
7872 trace (_verb_)::
7873   The action of recording the <<def-event,events>> emitted by an
7874   application or by a system, or to initiate such recording by
7875   controlling a tracer.
7876
7877 [[def-trace-chunk]]trace chunk::
7878   A self-contained trace which is part of a <<tracing-session,tracing
7879   session>>. Each <<session-rotation, tracing session rotation>>
7880   produces a trace chunk archive.
7881
7882 trace chunk archive::
7883   The result of a <<session-rotation, tracing session rotation>>. A
7884   trace chunk archive is not managed by LTTng, even if its containing
7885   <<tracing-session,tracing session>> is still active: you are free to
7886   read it, modify it, move it, or remove it.
7887
7888 Trace Compass::
7889   The http://tracecompass.org[Trace Compass] project and application.
7890
7891 [[def-tracepoint]]tracepoint::
7892   An instrumentation point using the tracepoint mechanism of the Linux
7893   kernel or of LTTng-UST.
7894
7895 tracepoint definition::
7896   The definition of a single tracepoint.
7897
7898 tracepoint name::
7899   The name of a tracepoint.
7900
7901 tracepoint provider::
7902   A set of functions providing tracepoints to an instrumented user
7903   application.
7904 +
7905 Not to be confused with a _tracepoint provider package_: many tracepoint
7906 providers can exist within a tracepoint provider package.
7907
7908 tracepoint provider package::
7909   One or more tracepoint providers compiled as an
7910   https://en.wikipedia.org/wiki/Object_file[object file] or as
7911   a link:https://en.wikipedia.org/wiki/Library_(computing)#Shared_libraries[shared library].
7912
7913 [[def-tracer]]tracer::
7914   A software which records emitted <<def-event,events>>.
7915
7916 <<domain,tracing domain>>::
7917   A namespace for <<def-event,event>> sources.
7918
7919 <<tracing-group,tracing group>>::
7920   The Unix group in which a Unix user can be to be allowed to trace the
7921   Linux kernel.
7922
7923 [[def-tracing-session-rotation]]<<tracing-session,tracing session>>::
7924   A stateful dialogue between you and a <<lttng-sessiond,session
7925   daemon>>.
7926
7927 <<session-rotation,tracing session rotation>>::
7928   The action of archiving the
7929   <<def-current-trace-chunk,current trace chunk>> of a
7930   <<tracing-session,tracing session>>.
7931
7932 [[def-user-application]]user application::
7933   An application running in user space, as opposed to a Linux kernel
7934   module, for example.