1 \input texinfo @c -*-texinfo-*-
4 @settitle LTTng Userspace Tracer (UST) Manual
8 This manual is for program, version version.
10 Copyright @copyright{} copyright-owner.
13 Permission is granted to ...
18 @title LTTng Userspace Tracer (UST) Manual
19 @c @subtitle subtitle-if-any
20 @c @subtitle second-subtitle
23 @c The following two commands
24 @c start the copyright page.
26 @c @vskip 0pt plus 1filll
32 @c So the toc is printed at the start.
37 @top LTTng Userspace Tracer
39 This manual is for UST 0.11.
46 * Instrumenting an application::
51 * List of environment variables detected by libust::
53 @c * Copying:: Your rights and freedoms.
62 * Supported platforms::
68 The LTTng Userspace Tracer (UST) is a library accompanied by a set of tools to
71 Code may be instrumented with either markers or tracepoints. A highly efficient
72 lockless tracer records these events to a trace buffers. These buffers are reaped
73 by a deamon which writes trace data to disk.
75 High performance is achieved by the use of lockless buffering algorithms, RCU and
76 per-cpu buffers. In addition, special care is taken to minize cache impact.
80 The LTTng Userspace Tracer is intended to be linkable to open source software
81 as well as to proprietary applications. This was accomplished by licensing
82 the code that needs to be linked to the traced program as @acronym{LGPL}.
84 Components licensed as LGPL v2.1:
91 Components licensed as GPL v2:
98 @node Supported platforms
99 @section Supported platforms
101 UST can currently trace applications running on Linux, on the x86-32, x86-64
102 and PowerPC 32 architectures.
105 @chapter Installation
107 The LTTng userspace tracer is a library and a set of userspace tools.
109 The following packages are required:
115 This contains the tracing library, the ust-consumerd daemon, trace control tools
116 and other helper tools.
118 Repository: @url{http://git.dorsal.polymtl.ca}
123 This is the userspace read-copy update library by Mathieu Desnoyers.
125 Available in Debian as package liburcu-dev.
127 Home page: @url{http://lttng.org/urcu}
132 LTTV is a graphical (and text) viewer for LTTng traces.
134 Home page: @url{http://lttng.org}
138 Liburcu should be installed first. UST may then be compiled and installed. LTTV
139 has no dependency on the other packages; it may therefore be installed on a
140 system which does not have UST installed.
142 Refer to the README in each of these packages for installation instructions.
150 First, instrument a program with a marker.
155 #include <ust/marker.h>
157 int main(int argc, char **argv)
162 /* ... set values of v and st ... */
165 trace_mark(ust, myevent, "firstarg %d secondarg %s", v, st);
167 /* a marker without arguments: */
168 trace_mark(ust, myotherevent, MARK_NOARGS);
176 Then compile it in the regular way, linking it with libust. For example:
179 gcc -o foo -lust foo.c
182 Run the program with @command{usttrace}. The @command{usttrace} output says where the trace
189 Finally, open the trace in LTTV.
192 lttv-gui -t /path/to/trace
195 The trace can also be dumped as text in the console:
198 lttv -m textDump -t /path/to/trace
201 @node Instrumenting an application
202 @chapter Instrumenting an application
204 In order to record a trace of events occurring in a application, the
205 application must be instrumented. Instrumentation points resemble function
206 calls. When the program reaches an instrumentation point, an event is
209 There are no limitations on the type of code that may be instrumented.
210 Multi-threaded programs may be instrumented without problem. Signal handlers
211 may be instrumented as well.
213 There are two APIs to instrument programs: markers and tracepoints. Markers are
214 quick to add and are usually used for temporary instrumentation. Tracepoints
215 provide a way to instrument code more cleanly and are suited for permanent
218 In addition to executable programs, shared libraries may also be instrumented
219 with the methods described in this chapter.
229 Adding a marker is simply a matter of inserting one line in the program.
233 #include <ust/marker.h>
235 int main(int argc, char **argv)
240 /* ... set values of v and st ... */
243 trace_mark(main, myevent, "firstarg %d secondarg %s", v, st);
245 /* another marker without arguments: */
246 trace_mark(main, myotherevent, MARK_NOARGS);
253 The invocation of the trace_mark() macro requires at least 3 arguments. The
254 first, here "main", is the name of the event category. It is also the name of
255 the channel the event will go in. The second, here "myevent" is the name of the
256 event. The third is a format string that announces the names and the types of
257 the event arguments. Its format resembles that of a printf() format string; it
258 is described thoroughly in Appendix x.
260 A given Marker may appear more than once in the same program. Other Markers may
261 have the same name and a different format string, although this might induce
262 some confusion at analysis time.
267 The Tracepoints API uses the Markers, but provides a higher-level abstraction.
268 Whereas the markers API provides limited type checking, the Tracepoints API
269 provides more thorough type checking and discharges from the need to insert
270 format strings directly in the code and to have format strings appear more than
271 once if a given marker is reused.
274 Although this example uses @emph{mychannel} as the channel, the
275 only channel name currently supported with early tracing is @strong{ust}. The
276 @command{usttrace} tool always uses the early tracing mode. When using manual
277 mode without early tracing, any channel name may be used.
280 A function instrumented with a tracepoint looks like this:
291 /* ... set values of v and st ... */
294 trace_mychannel_myevent(v, st);
299 Another file, here tp.h, contains declarations for the tracepoint.
303 #include <ust/tracepoint.h>
305 DECLARE_TRACE(mychannel_myevent, TP_PROTO(int v, char *st),
310 A third file, here tp.c, contains definitions for the tracepoint.
314 #include <ust/marker.h>
317 DEFINE_TRACE(mychannel_myevent);
319 void mychannel_myevent_probe(int v, char *st)
321 trace_mark(mychannel, myevent, "v %d st %s", v, st);
324 static void __attribute__((constructor)) init()
326 register_trace_mychannel_myevent(mychannel_myevent_probe);
331 Here, tp.h and tp.c could contain declarations and definitions for other
332 tracepoints. The constructor would contain other register_* calls.
334 @node Recording a trace
335 @chapter Recording a trace
338 * Using @command{usttrace}::
339 * Setting up the recording manually::
340 * Using early tracing::
342 * Tracing across @code{fork()} and @code{clone()}::
343 * Tracing programs and libraries that were not linked to libust::
346 @node Using @command{usttrace}
347 @section Using @command{usttrace}
349 The simplest way to record a trace is to use the @command{usttrace} script. An
350 example is given in the quickstart above.
352 The @command{usttrace} script automatically:
354 @item creates a daemon
355 @item enables all markers
356 @item runs the command specified on the command line
357 @item after the command ends, prints the location where the trace was saved
360 Each subdirectory of the save location contains the trace of one process that
361 was generated by the command. The name of a subdirectory consists in the the PID
362 of the process, followed by the timestamp of its creation.
364 The save location also contains logs of the tracing.
366 When using @command{usttrace}, the early tracing is always active, which means
367 that the tracing is guaranteed to be started by the time the process enters its
368 @code{main()} function.
370 Several @command{usttrace}'s may be run simultaneously without risk of
371 conflict. This facilitates the use of the tracer by idependent users on a
372 system. Each instance of @command{usttrace} starts its own daemon which
373 collects the events of the processes it creates.
375 @node Setting up the recording manually
376 @section Setting up the recording manually
378 Instead of using @command{usttrace}, a trace may be recorded on an already
381 First the daemon must be started.
385 # Make sure the directory for the communication sockets exists.
386 $ mkdir /tmp/ustsocks
388 # Make sure the directory where ust-consumerd will write the trace exists.
394 # We assume the program we want to trace is already running and that
397 # List the available markers
398 $ ustctl list-markers 1234
399 # A column indicates 0 for an inactive marker and 1 for an active marker.
402 $ ustctl enable-marker 1234 auto ust/mymark
405 $ ustctl create-trace 1234 auto
408 $ ustctl start-trace 1234 auto
413 $ ustctl stop-trace 1234 auto
416 $ ustctl destroy-trace 1234 auto
420 For more information about the manual mode, see the ustctl(1) man page.
422 @node Using early tracing
423 @section Using early tracing
425 Early tracing consists in starting the tracing as early as possible in the
426 program, so no events are lost between program start and the point where the
427 command to start the tracing is given. When using early tracing, it is
428 guaranteed that by the time the traced program enters its @code{main()}
429 function, the tracing will be started.
431 When using @command{usttrace}, the early tracing is always active.
433 When using the manual mode (@command{ustctl}), early tracing is enabled using
434 environment variables. Setting @env{UST_TRACE} to @code{1}, enables early
435 tracing, while setting @env{UST_AUTOPROBE} to @code{1} enables all markers
440 @section Crash recovery
442 When a process being traced crashes, the daemon is able to recover all the
443 events in its buffers that were successfully commited. This is possible because
444 the buffers are in a shared memory segment which remains available to the
445 daemon even after the termination of the traced process.
447 @node Tracing across @code{fork()} and @code{clone()}
448 @section Tracing across @code{fork()} and @code{clone()}
450 Tracing across @code{clone()} when the @code{CLONE_VM} flag is specified is
451 supported without any particular action.
453 When @code{clone()} is called without @code{CLONE_VM} or @code{fork()} is
454 called, a new address space is created and the tracer must be notified to
455 create new buffers for it.
457 This can be done automatically, by @env{LD_PRELOAD}'ing @file{libinterfork.so}.
458 This library intercepts calls to @code{fork()} and informs the tracer it is
459 being called. When using @command{usttrace}, this is accomplied by specifying
460 the @option{-f} command line argument.
462 Alternatively, the program can call @code{ust_before_fork()} before calling
463 @code{fork()} or @code{clone()} with @code{CLONE_VM}. After the call,
464 @code{ust_after_fork_parent()} must be called in the parent process and
465 @code{ust_after_fork_child()} must be called in the child process.
468 @node Tracing programs and libraries that were not linked to libust
469 @section Tracing programs and libraries that were not linked to libust
471 Some programs need to be traced even though they were not linked to libust
472 either because they were not instrumented or because it was not practical.
474 An executable that is not instrumented can still yield interesting traces when
475 at least one of its dynamic libraries is instrumented. It is also possible to
476 trace certain function calls by intercepting them with a specially crafted
477 library that is linked with @env{LD_PRELOAD} at program start.
479 In any case, a program that was not linked to libust at compile time must be
480 linked to it at run time with @env{LD_PRELOAD}. This can be accomplished with
481 @command{usttrace}'s @option{-l} option. It can also be done by setting the
482 @env{LD_PRELOAD} environment variable on the command line. For example:
486 # Run ls with usttrace, LD_PRELOAD'ing libust
487 # (assuming one of the libraries used by ls is instrumented).
490 # Run ls, manually adding the LD_PRELOAD.
491 $ LD_PRELOAD=/usr/local/lib/libust.so.0 ls
502 @chapter Viewing traces
504 Traces may be viewed with LTTV. An example of command for launching LTTV is
505 given in the quickstart.
508 * Viewing multiple traces::
509 * Combined kernel-userspace tracing::
512 @node Viewing multiple traces
513 @section Viewing multiple traces
515 When tracing multi-process applications or several applications simultaneously,
516 more than one trace will be obtained. LTTV can open and display all these
517 traces simultaneously.
519 @node Combined kernel-userspace tracing
520 @section Combined kernel-userspace tracing
522 In addition to multiple userspace traces, LTTV can open a kernel trace recorded
523 with the LTTng kernel tracer. This provides events that enable the rendering of
524 the Control Flow View and the Resource View.
526 When doing so, it is necessary to use the same time source for the kernel
527 tracer as well as the userspace tracer. Currently, the recommended method is to
528 use the timestamp counter for both. The TSC can however only be used on architectures
529 where it is synchronized across cores.
532 @chapter Resource Usage
534 The purpose of this section is to give an overview of the resource usage of libust. For
535 a developer, knowing this can be important: because libust is linked with applications, it
536 needs to share some resources with it. Some applications may make some assumptions that are in
537 conflict with libust's usage of resources.
539 In practice however, libust is designed to be transparent and is compatible
540 with the vast majority of applications. This means no changes are required in
541 the application (or library) being linked to libust.
543 Libust is initialized by a constructor, which by definition runs before the
544 @code{main()} function of the application starts. This constructor creates a
545 thread called the @emph{listener thread}. The listener thread initializes a
546 named socket and waits for connections for ust-consumerd or ustctl.
548 Libust-specific code may:
550 @item use @code{malloc()} and @code{free()}
551 @item map shared memory segment in the process adress space
552 @item intercept some library calls, specifically @code{fork()} and @code{clone()}
553 @item do interprocess communication with the daemon or ustctl
554 @item create and open named sockets
560 @item handle any signal (all signals are blocked in the listener thread)
561 @item change any process-wide setting that could confuse the application
564 @node List of environment variables detected by libust
565 @appendix List of environment variables detected by libust
567 The behavior of tracing can be influenced by setting special environment
568 variables in the environment of the traced application. This section
569 describes these variables.
576 If set to 1, start tracing as soon as the program starts. Tracing is
577 guaranteed to be started by the time the @code{main()} function starts.
582 If set to @code{1}, enable all markers by the time the @code{main()} function starts.
585 @env{UST_AUTOCOLLECT}
587 If set to @code{0}, disable notification of daemon on trace start. Useful for
593 If set to @code{1}, enable overwriting of buffers on overrun.
598 If set, defines the default number of subbuffers per buffer.
601 @env{UST_SUBBUF_SIZE}
603 If set, defines the default size of subbuffers, in bytes.
607 @node GDB integration
608 @appendix GDB integration
610 GDB, the GNU Debugger, can use UST markers as GDB tracepoints (note GDB has its
611 own concept of tracepoint). This feature is called GDB Static Tracepoints. When
612 a GDB tracepoint is hit, GDB collects the marker arguments, as well as the
613 state of the registers.
615 In UST, support for GDB integration is not compiled in by default because of
616 the cost of saving registers when a marker is hit. To enable it, run the
617 @command{./configure} script with the @code{-DCONFIG_UST_GDB_INTEGRATION} flag
618 in the @env{CFLAGS} environment variable. For example:
623 CFLAGS=-DCONFIG_UST_GDB_INTEGRATION ./configure
628 As of this writing, GDB Static Tracepoints have been submitted
629 (@url{http://sourceware.org/ml/gdb-patches/2010-06/msg00592.html}) to the GDB
632 GDB integration is currently only supported on x86-32 and x86-64.