ed525f6e9866a31e7dfc2b8987be4c0d880f7f6a
[lttng-ust.git] / doc / man / lttng-ust.3.txt
1 lttng-ust(3)
2 ============
3 :object-type: library
4
5
6 NAME
7 ----
8 lttng-ust - LTTng user space tracing
9
10
11 SYNOPSIS
12 --------
13 [verse]
14 *#include <lttng/tracepoint.h>*
15
16 [verse]
17 #define *TRACEPOINT_ENUM*('prov_name', 'enum_name', 'mappings')
18 #define *TRACEPOINT_EVENT*('prov_name', 't_name', 'args', 'fields')
19 #define *TRACEPOINT_EVENT_CLASS*('prov_name', 'class_name', 'args', 'fields')
20 #define *TRACEPOINT_EVENT_INSTANCE*('prov_name', 'class_name', 't_name', 'args')
21 #define *TRACEPOINT_LOGLEVEL*('prov_name', 't_name', 'level')
22 #define *ctf_array*('int_type', 'field_name', 'expr', 'count')
23 #define *ctf_array_nowrite*('int_type', 'field_name', 'expr', 'count')
24 #define *ctf_array_hex*('int_type', 'field_name', 'expr', 'count')
25 #define *ctf_array_nowrite_hex*('int_type', 'field_name', 'expr', 'count')
26 #define *ctf_array_network*('int_type', 'field_name', 'expr', 'count')
27 #define *ctf_array_network_nowrite*('int_type', 'field_name', 'expr', 'count')
28 #define *ctf_array_network_hex*('int_type', 'field_name', 'expr', 'count')
29 #define *ctf_array_network_nowrite_hex*('int_type', 'field_name', 'expr', 'count')
30 #define *ctf_array_text*(char, 'field_name', 'expr', 'count')
31 #define *ctf_array_text_nowrite*(char, 'field_name', 'expr', 'count')
32 #define *ctf_enum*('prov_name', 'enum_name', 'int_type', 'field_name', 'expr')
33 #define *ctf_enum_nowrite*('prov_name', 'enum_name', 'int_type', 'field_name',
34 'expr')
35 #define *ctf_enum_value*('label', 'value')
36 #define *ctf_enum_range*('label', 'start', 'end')
37 #define *ctf_float*('float_type', 'field_name', 'expr')
38 #define *ctf_float_nowrite*('float_type', 'field_name', 'expr')
39 #define *ctf_integer*('int_type', 'field_name', 'expr')
40 #define *ctf_integer_hex*('int_type', 'field_name', 'expr')
41 #define *ctf_integer_network*('int_type', 'field_name', 'expr')
42 #define *ctf_integer_network_hex*('int_type', 'field_name', 'expr')
43 #define *ctf_integer_nowrite*('int_type', 'field_name', 'expr')
44 #define *ctf_sequence*('int_type', 'field_name', 'expr', 'len_type', 'len_expr')
45 #define *ctf_sequence_nowrite*('int_type', 'field_name', 'expr', 'len_type',
46 'len_expr')
47 #define *ctf_sequence_hex*('int_type', 'field_name', 'expr', 'len_type',
48 'len_expr')
49 #define *ctf_sequence_nowrite_hex*('int_type', 'field_name', 'expr', 'len_type',
50 'len_expr')
51 #define *ctf_sequence_network*('int_type', 'field_name', 'expr', 'len_type',
52 'len_expr')
53 #define *ctf_sequence_network_nowrite*('int_type', 'field_name', 'expr',
54 'len_type', 'len_expr')
55 #define *ctf_sequence_network_hex*('int_type', 'field_name', 'expr', 'len_type',
56 'len_expr')
57 #define *ctf_sequence_network_nowrite_hex*('int_type', 'field_name', 'expr',
58 'len_type', 'len_expr')
59 #define *ctf_sequence_text*(char, 'field_name', 'expr', 'len_type', 'len_expr')
60 #define *ctf_sequence_text_nowrite*(char, 'field_name', 'expr', 'len_type',
61 'len_expr')
62 #define *ctf_string*('field_name', 'expr')
63 #define *ctf_string_nowrite*('field_name', 'expr')
64 #define *do_tracepoint*('prov_name', 't_name', ...)
65 #define *tracepoint*('prov_name', 't_name', ...)
66 #define *tracepoint_enabled*('prov_name', 't_name')
67
68 Link with `-llttng-ust -ldl`, following this man page.
69
70
71 DESCRIPTION
72 -----------
73 The http://lttng.org/[_Linux Trace Toolkit: next generation_] is an open
74 source software package used for correlated tracing of the Linux kernel,
75 user applications, and user libraries.
76
77 LTTng-UST is the user space tracing component of the LTTng project. It
78 is a port to user space of the low-overhead tracing capabilities of the
79 LTTng Linux kernel tracer. The `liblttng-ust` library is used to trace
80 user applications and libraries.
81
82 NOTE: This man page is about the `liblttng-ust` library. The LTTng-UST
83 project also provides Java and Python packages to trace applications
84 written in those languages. How to instrument and trace Java and Python
85 applications is documented in
86 http://lttng.org/docs/[the online LTTng documentation].
87
88 There are three ways to use `liblttng-ust`:
89
90 * Using the man:tracef(3) API, which is similar to man:printf(3).
91 * Using the man:tracelog(3) API, which is man:tracef(3) with
92 a log level parameter.
93 * Defining your own tracepoints. See the
94 <<creating-tp,Creating a tracepoint provider>> section below.
95
96
97 [[creating-tp]]
98 Creating a tracepoint provider
99 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
100 Creating a tracepoint provider is the first step of using
101 `liblttng-ust`. The next steps are:
102
103 * <<tracepoint,Instrumenting your application with `tracepoint()` calls>>
104 * Building your application with LTTng-UST support, either
105 <<build-static,statically>> or <<build-dynamic,dynamically>>.
106
107 A *tracepoint provider* is a compiled object containing the event
108 probes corresponding to your custom tracepoint definitions. A tracepoint
109 provider contains the code to get the size of an event and to serialize
110 it, amongst other things.
111
112 To create a tracepoint provider, start with the following
113 _tracepoint provider header_ template:
114
115 ------------------------------------------------------------------------
116 #undef TRACEPOINT_PROVIDER
117 #define TRACEPOINT_PROVIDER my_provider
118
119 #undef TRACEPOINT_INCLUDE
120 #define TRACEPOINT_INCLUDE "./tp.h"
121
122 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
123 #define _TP_H
124
125 #include <lttng/tracepoint.h>
126
127 /*
128 * TRACEPOINT_EVENT(), TRACEPOINT_EVENT_CLASS(),
129 * TRACEPOINT_EVENT_INSTANCE(), TRACEPOINT_LOGLEVEL(),
130 * and `TRACEPOINT_ENUM()` are used here.
131 */
132
133 #endif /* _TP_H */
134
135 #include <lttng/tracepoint-event.h>
136 ------------------------------------------------------------------------
137
138 In this template, the tracepoint provider is named `my_provider`
139 (`TRACEPOINT_PROVIDER` definition). The file needs to bear the
140 name of the `TRACEPOINT_INCLUDE` definition (`tp.h` in this case).
141 Between `#include <lttng/tracepoint.h>` and `#endif` go
142 the invocations of the <<tracepoint-event,`TRACEPOINT_EVENT()`>>,
143 <<tracepoint-event-class,`TRACEPOINT_EVENT_CLASS()`>>,
144 <<tracepoint-event-class,`TRACEPOINT_EVENT_INSTANCE()`>>,
145 <<tracepoint-loglevel,`TRACEPOINT_LOGLEVEL()`>>, and
146 <<tracepoint-enum,`TRACEPOINT_ENUM()`>> macros.
147
148 NOTE: You can avoid writing the prologue and epilogue boilerplate in the
149 template file above by using the man:lttng-gen-tp(1) tool shipped with
150 LTTng-UST.
151
152 The tracepoint provider header file needs to be included in a source
153 file which looks like this:
154
155 ------------------------------------------------------------------------
156 #define TRACEPOINT_CREATE_PROBES
157
158 #include "tp.h"
159 ------------------------------------------------------------------------
160
161 Together, those two files (let's call them `tp.h` and `tp.c`) form the
162 tracepoint provider sources, ready to be compiled.
163
164 You can create multiple tracepoint providers to be used in a single
165 application, but each one must have its own header file.
166
167 The <<tracepoint-event,`TRACEPOINT_EVENT()` usage>> section below
168 shows how to use the `TRACEPOINT_EVENT()` macro to define the actual
169 tracepoints in the tracepoint provider header file.
170
171 See the <<example,EXAMPLE>> section below for a complete example.
172
173
174 [[tracepoint-event]]
175 `TRACEPOINT_EVENT()` usage
176 ~~~~~~~~~~~~~~~~~~~~~~~~~~
177 The `TRACEPOINT_EVENT()` macro is used in a template provider
178 header file (see the <<creating-tp,Creating a tracepoint provider>>
179 section above) to define LTTng-UST tracepoints.
180
181 The `TRACEPOINT_EVENT()` usage template is as follows:
182
183 ------------------------------------------------------------------------
184 TRACEPOINT_EVENT(
185 /* Tracepoint provider name */
186 my_provider,
187
188 /* Tracepoint/event name */
189 my_tracepoint,
190
191 /* List of tracepoint arguments (input) */
192 TP_ARGS(
193 ...
194 ),
195
196 /* List of fields of eventual event (output) */
197 TP_FIELDS(
198 ...
199 )
200 )
201 ------------------------------------------------------------------------
202
203 The `TP_ARGS()` macro contains the input arguments of the tracepoint.
204 Those arguments can be used in the argument expressions of the output
205 fields defined in `TP_FIELDS()`.
206
207 The format of the `TP_ARGS()` parameters is: C type, then argument name;
208 repeat as needed, up to ten times. For example:
209
210 ------------------------------------------------------------------------
211 TP_ARGS(
212 int, my_int,
213 const char *, my_string,
214 FILE *, my_file,
215 double, my_float,
216 struct my_data *, my_data
217 )
218 ------------------------------------------------------------------------
219
220 The `TP_FIELDS()` macro contains the output fields of the tracepoint,
221 that is, the actual data that can be recorded in the payload of an
222 event emitted by this tracepoint.
223
224 The `TP_FIELDS()` macro contains a list of `ctf_*()` macros
225 :not: separated by commas. The available macros are documented in the
226 <<ctf-macros,Available `ctf_*()` field type macros>> section below.
227
228
229 [[ctf-macros]]
230 Available `ctf_*()` field type macros
231 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
232 This section documents the available `ctf_*()` macros that can be
233 inserted in the `TP_FIELDS()` macro of the
234 <<tracepoint-event,`TRACEPOINT_EVENT()` macro>>.
235
236 Standard integer, displayed in base 10:
237
238 [verse]
239 *ctf_integer*('int_type', 'field_name', 'expr')
240 *ctf_integer_nowrite*('int_type', 'field_name', 'expr')
241
242 Standard integer, displayed in base 16:
243
244 [verse]
245 *ctf_integer_hex*('int_type', 'field_name', 'expr')
246
247 Integer in network byte order (big endian), displayed in base 10:
248
249 [verse]
250 *ctf_integer_network*('int_type', 'field_name', 'expr')
251
252 Integer in network byte order, displayed in base 16:
253
254 [verse]
255 *ctf_integer_network_hex*('int_type', 'field_name', 'expr')
256
257 Floating point number:
258
259 [verse]
260 *ctf_float*('float_type', 'field_name', 'expr')
261 *ctf_float_nowrite*('float_type', 'field_name', 'expr')
262
263 Null-terminated string:
264
265 [verse]
266 *ctf_string*('field_name', 'expr')
267 *ctf_string_nowrite*('field_name', 'expr')
268
269 Statically-sized array of integers (`_hex` versions displayed in
270 hexadecimal, `_network` versions in network byte order):
271
272 [verse]
273 *ctf_array*('int_type', 'field_name', 'expr', 'count')
274 *ctf_array_nowrite*('int_type', 'field_name', 'expr', 'count')
275 *ctf_array_hex*('int_type', 'field_name', 'expr', 'count')
276 *ctf_array_nowrite_hex*('int_type', 'field_name', 'expr', 'count')
277 *ctf_array_network*('int_type', 'field_name', 'expr', 'count')
278 *ctf_array_network_nowrite*('int_type', 'field_name', 'expr', 'count')
279 *ctf_array_network_hex*('int_type', 'field_name', 'expr', 'count')
280 *ctf_array_network_nowrite_hex*('int_type', 'field_name', 'expr', 'count')
281
282 Statically-sized array, printed as text; no need to be null-terminated:
283
284 [verse]
285 *ctf_array_text*(char, 'field_name', 'expr', 'count')
286 *ctf_array_text_nowrite*(char, 'field_name', 'expr', 'count')
287
288 Dynamically-sized array of integers (`_hex` versions displayed in
289 hexadecimal, `_network` versions in network byte order):
290
291 [verse]
292 *ctf_sequence*('int_type', 'field_name', 'expr', 'len_type', 'len_expr')
293 *ctf_sequence_nowrite*('int_type', 'field_name', 'expr', 'len_type', 'len_expr')
294 *ctf_sequence_hex*('int_type', 'field_name', 'expr', 'len_type', 'len_expr')
295 *ctf_sequence_nowrite_hex*('int_type', 'field_name', 'expr', 'len_type',
296 'len_expr')
297 *ctf_sequence_network*('int_type', 'field_name', 'expr', 'len_type', 'len_expr')
298 *ctf_sequence_network_nowrite*('int_type', 'field_name', 'expr', 'len_type',
299 'len_expr')
300 *ctf_sequence_network_hex*('int_type', 'field_name', 'expr', 'len_type',
301 'len_expr')
302 *ctf_sequence_network_nowrite_hex*('int_type', 'field_name', 'expr',
303 'len_type', 'len_expr')
304
305 Dynamically-sized array, displayed as text; no need to be null-terminated:
306
307 [verse]
308 *ctf_sequence_text*(char, 'field_name', 'expr', 'len_type', 'len_expr')
309 *ctf_sequence_text_nowrite*(char, 'field_name', 'expr', 'len_type', 'len_expr')
310
311 Enumeration. The enumeration field must be defined before using this
312 macro with the `TRACEPOINT_ENUM()` macro. See the
313 <<tracepoint-enum,`TRACEPOINT_ENUM()` usage>> section for more
314 information.
315
316 [verse]
317 *ctf_enum*('prov_name', 'enum_name', 'int_type', 'field_name', 'expr')
318 *ctf_enum_nowrite*('prov_name', 'enum_name', 'int_type', 'field_name', 'expr')
319
320 The parameters are:
321
322 'count'::
323 Number of elements in array/sequence. This must be known at
324 compile time.
325
326 'enum_name'::
327 Name of an enumeration field previously defined with the
328 `TRACEPOINT_ENUM()` macro. See the
329 <<tracepoint-enum,`TRACEPOINT_ENUM()` usage>> section for more
330 information.
331
332 'expr'::
333 C expression resulting in the field's value. This expression can
334 use one or more arguments passed to the tracepoint. The arguments
335 of a given tracepoint are defined in the `TP_ARGS()` macro (see
336 the <<creating-tp,Creating a tracepoint provider>> section above).
337
338 'field_name'::
339 Event field name (C identifier syntax, :not: a literal string).
340
341 'float_type'::
342 Float C type (`float` or `double`). The size of this type determines
343 the size of the floating point number field.
344
345 'int_type'::
346 Integer C type. The size of this type determines the size of the
347 integer/enumeration field.
348
349 'len_expr'::
350 C expression resulting in the sequence's length. This expression
351 can use one or more arguments passed to the tracepoint.
352
353 'len_type'::
354 Unsigned integer C type of sequence's length.
355
356 'prov_name'::
357 Tracepoint provider name. This must be the same as the tracepoint
358 provider name used in a previous field definition.
359
360 The `_nowrite` versions omit themselves from the recorded trace, but are
361 otherwise identical. Their primary purpose is to make some of the
362 event context available to the event filters without having to commit
363 the data to sub-buffers. See man:lttng-enable-event(1) to learn more
364 about dynamic event filtering.
365
366 See the <<example,EXAMPLE>> section below for a complete example.
367
368
369 [[tracepoint-enum]]
370 `TRACEPOINT_ENUM()` usage
371 ~~~~~~~~~~~~~~~~~~~~~~~~~
372 An enumeration field is a list of mappings between an integers, or a
373 range of integers, and strings (sometimes called _labels_ or
374 _enumerators_). Enumeration fields can be used to have a more compact
375 trace when the possible values for a field are limited.
376
377 An enumeration field is defined with the `TRACEPOINT_ENUM()` macro:
378
379 ------------------------------------------------------------------------
380 TRACEPOINT_ENUM(
381 /* Tracepoint provider name */
382 my_provider,
383
384 /* Enumeration name (unique in the whole tracepoint provider) */
385 my_enum,
386
387 /* Enumeration mappings */
388 TP_ENUM_VALUES(
389 ...
390 )
391 )
392 ------------------------------------------------------------------------
393
394 `TP_ENUM_VALUES()` contains a list of enumeration mappings, :not:
395 separated by commas. Two macros can be used in the `TP_ENUM_VALUES()`:
396 `ctf_enum_value()` and `ctf_enum_range()`.
397
398 `ctf_enum_value()` is a single value mapping:
399
400 [verse]
401 *ctf_enum_value*('label', 'value')
402
403 This macro maps the given 'label' string to the value 'value'.
404
405 `ctf_enum_range()` is a range mapping:
406
407 [verse]
408 *ctf_enum_range*('label', 'start', 'end')
409
410 This macro maps the given 'label' string to the range of integers from
411 'start' to 'end', inclusively. Range mappings may overlap, but the
412 behaviour is implementation-defined: each trace reader handles
413 overlapping ranges as it wishes.
414
415 See the <<example,EXAMPLE>> section below for a complete example.
416
417
418 [[tracepoint-event-class]]
419 `TRACEPOINT_EVENT_CLASS()` usage
420 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
421 A *tracepoint class* is a class of tracepoints sharing the
422 same field types and names. A tracepoint instance is one instance of
423 such a declared tracepoint class, with its own event name.
424
425 LTTng-UST creates one event serialization function per tracepoint
426 class. Using `TRACEPOINT_EVENT()` creates one tracepoint class per
427 tracepoint definition, whereas using `TRACEPOINT_EVENT_CLASS()` and
428 `TRACEPOINT_EVENT_INSTANCE()` creates one tracepoint class, and one or
429 more tracepoint instances of this class. In other words, many
430 tracepoints can reuse the same serialization code. Reusing the same
431 code, when possible, can reduce cache pollution, thus improve
432 performance.
433
434 The `TRACEPOINT_EVENT_CLASS()` macro accepts the same parameters as
435 the `TRACEPOINT_EVENT()` macro, except that instead of an event name,
436 its second parameter is the _tracepoint class name_:
437
438 ------------------------------------------------------------------------
439 TRACEPOINT_EVENT_CLASS(
440 /* Tracepoint provider name */
441 my_provider,
442
443 /* Tracepoint class name */
444 my_tracepoint_class,
445
446 /* List of tracepoint arguments (input) */
447 TP_ARGS(
448 ...
449 ),
450
451 /* List of fields of eventual event (output) */
452 TP_FIELDS(
453 ...
454 )
455 )
456 ------------------------------------------------------------------------
457
458 Once the tracepoint class is defined, you can create as many tracepoint
459 instances as needed:
460
461 -------------------------------------------------------------------------
462 TRACEPOINT_EVENT_INSTANCE(
463 /* Tracepoint provider name */
464 my_provider,
465
466 /* Tracepoint class name */
467 my_tracepoint_class,
468
469 /* Tracepoint/event name */
470 my_tracepoint,
471
472 /* List of tracepoint arguments (input) */
473 TP_ARGS(
474 ...
475 )
476 )
477 ------------------------------------------------------------------------
478
479 As you can see, the `TRACEPOINT_EVENT_INSTANCE()` does not contain
480 the `TP_FIELDS()` macro, because they are defined at the
481 `TRACEPOINT_EVENT_CLASS()` level.
482
483 See the <<example,EXAMPLE>> section below for a complete example.
484
485
486 [[tracepoint-loglevel]]
487 `TRACEPOINT_LOGLEVEL()` usage
488 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
489 Optionally, a *log level* can be assigned to a defined tracepoint.
490 Assigning different levels of severity to tracepoints can be useful:
491 when controlling tracing sessions, you can choose to only enable
492 events falling into a specific log level range using the
493 nloption:--loglevel and nloption:--loglevel-only options of the
494 man:lttng-enable-event(1) command.
495
496 Log levels are assigned to tracepoints that are already defined using
497 the `TRACEPOINT_LOGLEVEL()` macro. The latter must be used after having
498 used `TRACEPOINT_EVENT()` or `TRACEPOINT_EVENT_INSTANCE()` for a given
499 tracepoint. The `TRACEPOINT_LOGLEVEL()` macro is used as follows:
500
501 ------------------------------------------------------------------------
502 TRACEPOINT_LOGLEVEL(
503 /* Tracepoint provider name */
504 my_provider,
505
506 /* Tracepoint/event name */
507 my_tracepoint,
508
509 /* Log level */
510 TRACE_INFO
511 )
512 ------------------------------------------------------------------------
513
514 The available log level definitions are:
515
516 include::log-levels.txt[]
517
518 See the <<example,EXAMPLE>> section below for a complete example.
519
520
521 [[tracepoint]]
522 Instrumenting your application
523 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
524 Once the tracepoint provider is created (see the
525 <<creating-tp,Creating a tracepoint provider>> section above), you can
526 instrument your application with the defined tracepoints thanks to the
527 `tracepoint()` macro:
528
529 [verse]
530 #define *tracepoint*('prov_name', 't_name', ...)
531
532 With:
533
534 'prov_name'::
535 Tracepoint provider name.
536
537 't_name'::
538 Tracepoint/event name.
539
540 `...`::
541 Tracepoint arguments, if any.
542
543 Make sure to include the tracepoint provider header file anywhere you
544 use `tracepoint()` for this provider.
545
546 NOTE: Even though LTTng-UST supports `tracepoint()` call site duplicates
547 having the same provider and tracepoint names, it is recommended to use
548 a provider/tracepoint name pair only once within the application source
549 code to help map events back to their call sites when analyzing the
550 trace.
551
552 Sometimes, arguments to the tracepoint are expensive to compute (take
553 call stack, for example). To avoid the computation when the tracepoint
554 is disabled, you can use the `tracepoint_enabled()` and
555 `do_tracepoint()` macros:
556
557 [verse]
558 #define *tracepoint_enabled*('prov_name', 't_name')
559 #define *do_tracepoint*('prov_name', 't_name', ...)
560
561 `tracepoint_enabled()` returns a non-zero value if the tracepoint
562 named 't_name' from the provider named 'prov_name' is enabled at
563 run time.
564
565 `do_tracepoint()` is like `tracepoint()`, except that it doesn't check
566 if the tracepoint is enabled. Using `tracepoint()` with
567 `tracepoint_enabled()` is dangerous since `tracepoint()` also contains
568 the `tracepoint_enabled()` check, thus a race condition is possible
569 in this situation:
570
571 ------------------------------------------------------------------------
572 if (tracepoint_enabled(my_provider, my_tracepoint)) {
573 stuff = prepare_stuff();
574 }
575
576 tracepoint(my_provider, my_tracepoint, stuff);
577 ------------------------------------------------------------------------
578
579 If the tracepoint is enabled after the condition, then `stuff` is not
580 prepared: the emitted event will either contain wrong data, or the
581 whole application could crash (segmentation fault, for example).
582
583 NOTE: Neither `tracepoint_enabled()` nor `do_tracepoint()` have
584 a `STAP_PROBEV()` call, so if you need it, you should emit this call
585 yourself.
586
587
588 [[build-static]]
589 Statically linking the tracepoint provider
590 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
591 With the static linking method, compiled tracepoint providers are copied
592 into the target application.
593
594 Define `TRACEPOINT_DEFINE` definition below the
595 `TRACEPOINT_CREATE_PROBES` definition in the tracepoint provider
596 source:
597
598 ------------------------------------------------------------------------
599 #define TRACEPOINT_CREATE_PROBES
600 #define TRACEPOINT_DEFINE
601
602 #include "tp.h"
603 ------------------------------------------------------------------------
604
605 Create the tracepoint provider object file:
606
607 [role="term"]
608 --------------
609 cc -c -I. tp.c
610 --------------
611
612 NOTE: Although an application instrumented with LTTng-UST tracepoints
613 can be compiled with a C++ compiler, tracepoint probes should be
614 compiled with a C compiler.
615
616 At this point, you _can_ archive this tracepoint provider object file,
617 possibly with other object files of your application or with other
618 tracepoint provider object files, as a static library:
619
620 [role="term"]
621 ---------------
622 ar rc tp.a tp.o
623 ---------------
624
625 Using a static library does have the advantage of centralising the
626 tracepoint providers objects so they can be shared between multiple
627 applications. This way, when the tracepoint provider is modified, the
628 source code changes don't have to be patched into each application's
629 source code tree. The applications need to be relinked after each
630 change, but need not to be otherwise recompiled (unless the tracepoint
631 provider's API changes).
632
633 Then, link your application with this object file (or with the static
634 library containing it) and with `liblttng-ust` and `libdl`
635 (`libc` on a BSD system):
636
637 [role="term"]
638 -------------------------------------
639 cc -o app tp.o app.o -llttng-ust -ldl
640 -------------------------------------
641
642
643 [[build-dynamic]]
644 Dynamically loading the tracepoint provider
645 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
646 The second approach to package the tracepoint provider is to use the
647 dynamic loader: the library and its member functions are explicitly
648 sought, loaded at run time.
649
650 In this scenario, the tracepoint provider is compiled as a shared
651 object.
652
653 The process to create the tracepoint provider shared object is pretty
654 much the same as the <<build-static,static linking method>>, except
655 that:
656
657 * Since the tracepoint provider is not part of the application,
658 `TRACEPOINT_DEFINE` must be defined, for each tracepoint
659 provider, in exactly one source file of the
660 _application_
661 * `TRACEPOINT_PROBE_DYNAMIC_LINKAGE` must be defined next
662 to `TRACEPOINT_DEFINE`
663
664 Regarding `TRACEPOINT_DEFINE` and `TRACEPOINT_PROBE_DYNAMIC_LINKAGE`,
665 the recommended practice is to use a separate C source file in your
666 application to define them, then include the tracepoint provider header
667 files afterwards. For example, as `tp-define.c`:
668
669 ------------------------------------------------------------------------
670 #define TRACEPOINT_DEFINE
671 #define TRACEPOINT_PROBE_DYNAMIC_LINKAGE
672
673 #include "tp.h"
674 ------------------------------------------------------------------------
675
676 The tracepoint provider object file used to create the shared library is
677 built like it is using the static linking method, but with the
678 nloption:-fpic option:
679
680 [role="term"]
681 --------------------
682 cc -c -fpic -I. tp.c
683 --------------------
684
685 It is then linked as a shared library like this:
686
687 [role="term"]
688 -------------------------------------------------------
689 cc -shared -Wl,--no-as-needed -o tp.so tp.o -llttng-ust
690 -------------------------------------------------------
691
692 This tracepoint provider shared object isn't linked with the user
693 application: it must be loaded manually. This is why the application is
694 built with no mention of this tracepoint provider, but still needs
695 libdl:
696
697 [role="term"]
698 --------------------------------
699 cc -o app app.o tp-define.o -ldl
700 --------------------------------
701
702 There are two ways to dynamically load the tracepoint provider shared
703 object:
704
705 * Load it manually from the application using man:dlopen(3)
706 * Make the dynamic loader load it with the `LD_PRELOAD`
707 environment variable (see man:ld.so(8))
708
709 If the application does not dynamically load the tracepoint provider
710 shared object using one of the methods above, tracing is disabled for
711 this application, and the events are not listed in the output of
712 man:lttng-list(1).
713
714 Note that it is not safe to use man:dlclose(3) on a tracepoint provider
715 shared object that is being actively used for tracing, due to a lack of
716 reference counting from LTTng-UST to the shared object.
717
718 For example, statically linking a tracepoint provider to a shared object
719 which is to be dynamically loaded by an application (a plugin, for
720 example) is not safe: the shared object, which contains the tracepoint
721 provider, could be dynamically closed (man:dlclose(3)) at any time by
722 the application.
723
724 To instrument a shared object, either:
725
726 * Statically link the tracepoint provider to the application, or
727 * Build the tracepoint provider as a shared object (following the
728 procedure shown in this section), and preload it when tracing is
729 needed using the `LD_PRELOAD` environment variable.
730
731
732 Using LTTng-UST with daemons
733 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
734 Some extra care is needed when using `liblttng-ust` with daemon
735 applications that call man:fork(2), man:clone(2), or BSD's man:rfork(2)
736 without a following man:exec(3) family system call. The library
737 `liblttng-ust-fork.so` needs to be preloaded before starting the
738 application with the `LD_PRELOAD` environment variable (see
739 man:ld.so(8)).
740
741 To use `liblttng-ust` with a daemon application which closes file
742 descriptors that were not opened by it, preload the `liblttng-ust-fd.so`
743 library before you start the application. Typical use cases include
744 daemons closing all file descriptors after man:fork(2), and buggy
745 applications doing ``double-closes''.
746
747
748 Context information
749 ~~~~~~~~~~~~~~~~~~~
750 Context information can be prepended by the LTTng-UST tracer before
751 each event, or before specific events.
752
753 Context fields can be added to specific channels using
754 man:lttng-add-context(1).
755
756 The following context fields are supported by LTTng-UST:
757
758 `cpu_id`::
759 CPU ID.
760 +
761 NOTE: This context field is always enabled, and it cannot be added
762 with man:lttng-add-context(1). Its main purpose is to be used for
763 dynamic event filtering. See man:lttng-enable-event(1) for more
764 information about event filtering.
765
766 `ip`::
767 Instruction pointer: enables recording the exact address from which
768 an event was emitted. This context field can be used to
769 reverse-lookup the source location that caused the event
770 to be emitted.
771
772 +perf:thread:COUNTER+::
773 perf counter named 'COUNTER'. Use `lttng add-context --list` to
774 list the available perf counters.
775 +
776 Only available on IA-32 and x86-64 architectures.
777
778 `pthread_id`::
779 POSIX thread identifier. Can be used on architectures where
780 `pthread_t` maps nicely to an `unsigned long` type.
781
782 `procname`::
783 Thread name, as set by man:exec(3) or man:prctl(2). It is
784 recommended that programs set their thread name with man:prctl(2)
785 before hitting the first tracepoint for that thread.
786
787 `vpid`::
788 Virtual process ID: process ID as seen from the point of view of
789 the process namespace.
790
791 `vtid`::
792 Virtual thread ID: thread ID as seen from the point of view of
793 the process namespace.
794
795
796 [[state-dump]]
797 LTTng-UST state dump
798 ~~~~~~~~~~~~~~~~~~~~
799 If an application that uses `liblttng-ust` becomes part of a tracing
800 session, information about its currently loaded shared objects, their
801 build IDs, and their debug link information are emitted as events
802 by the tracer.
803
804 The following LTTng-UST state dump events exist and must be enabled
805 to record application state dumps. Note that, during the state dump
806 phase, LTTng-UST can also emit _shared library load/unload_ events
807 (see <<ust-lib,Shared library load/unload tracking>> below).
808
809 `lttng_ust_statedump:start`::
810 Emitted when the state dump begins.
811 +
812 This event has no fields.
813
814 `lttng_ust_statedump:end`::
815 Emitted when the state dump ends. Once this event is emitted, it
816 is guaranteed that, for a given process, the state dump is
817 complete.
818 +
819 This event has no fields.
820
821 `lttng_ust_statedump:bin_info`::
822 Emitted when information about a currently loaded executable or
823 shared object is found.
824 +
825 Fields:
826 +
827 [options="header"]
828 |===
829 |Field name |Description
830
831 |`baddr`
832 |Base address of loaded executable.
833
834 |`memsz`
835 |Size of loaded executable in memory.
836
837 |`path`
838 |Path to loaded executable file.
839
840 |`is_pic`
841 |Whether or not the executable is position-independent code.
842
843 |`has_build_id`
844 |Whether or not the executable has a build ID. If this field is 1, you
845 can expect that an `lttng_ust_statedump:build_id` event record follows
846 this one (not necessarily immediately after).
847
848 |`has_debug_link`
849 |Whether or not the executable has debug link information. If this field
850 is 1, you can expect that an `lttng_ust_statedump:debug_link` event
851 record follows this one (not necessarily immediately after).
852 |===
853
854 `lttng_ust_statedump:build_id`::
855 Emitted when a build ID is found in a currently loaded shared
856 library. See
857 https://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.html[Debugging Information in Separate Files]
858 for more information about build IDs.
859 +
860 Fields:
861 +
862 [options="header"]
863 |===
864 |Field name |Description
865
866 |`baddr`
867 |Base address of loaded library.
868
869 |`build_id`
870 |Build ID.
871 |===
872
873 `lttng_ust_statedump:debug_link`::
874 Emitted when debug link information is found in a currently loaded
875 shared library. See
876 https://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.html[Debugging Information in Separate Files]
877 for more information about debug links.
878 +
879 Fields:
880 +
881 [options="header"]
882 |===
883 |Field name |Description
884
885 |`baddr`
886 |Base address of loaded library.
887
888 |`crc`
889 |Debug link file's CRC.
890
891 |`filename`
892 |Debug link file name.
893 |===
894
895
896 [[ust-lib]]
897 Shared library load/unload tracking
898 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
899 The <<state-dump,LTTng-UST state dump>> and the LTTng-UST helper library
900 to instrument the dynamic linker (see man:liblttng-ust-dl(3)) can emit
901 **shared library load/unload tracking** events.
902
903 The following shared library load/unload tracking events exist and must
904 be enabled to track the loading and unloading of shared libraries:
905
906 `lttng_ust_lib:load`::
907 Emitted when a shared library (shared object) is loaded.
908 +
909 Fields:
910 +
911 [options="header"]
912 |===
913 |Field name |Description
914
915 |`baddr`
916 |Base address of loaded library.
917
918 |`memsz`
919 |Size of loaded library in memory.
920
921 |`path`
922 |Path to loaded library file.
923
924 |`has_build_id`
925 |Whether or not the library has a build ID. If this field is 1, you
926 can expect that an `lttng_ust_lib:build_id` event record follows
927 this one (not necessarily immediately after).
928
929 |`has_debug_link`
930 |Whether or not the library has debug link information. If this field
931 is 1, you can expect that an `lttng_ust_lib:debug_link` event
932 record follows this one (not necessarily immediately after).
933 |===
934
935 `lttng_ust_lib:unload`::
936 Emitted when a shared library (shared object) is unloaded.
937 +
938 Fields:
939 +
940 [options="header"]
941 |===
942 |Field name |Description
943
944 |`baddr`
945 |Base address of unloaded library.
946 |===
947
948 `lttng_ust_lib:build_id`::
949 Emitted when a build ID is found in a loaded shared library (shared
950 object). See
951 https://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.html[Debugging Information in Separate Files]
952 for more information about build IDs.
953 +
954 Fields:
955 +
956 [options="header"]
957 |===
958 |Field name |Description
959
960 |`baddr`
961 |Base address of loaded library.
962
963 |`build_id`
964 |Build ID.
965 |===
966
967 `lttng_ust_lib:debug_link`::
968 Emitted when debug link information is found in a loaded
969 shared library (shared object). See
970 https://sourceware.org/gdb/onlinedocs/gdb/Separate-Debug-Files.html[Debugging Information in Separate Files]
971 for more information about debug links.
972 +
973 Fields:
974 +
975 [options="header"]
976 |===
977 |Field name |Description
978
979 |`baddr`
980 |Base address of loaded library.
981
982 |`crc`
983 |Debug link file's CRC.
984
985 |`filename`
986 |Debug link file name.
987 |===
988
989
990 [[example]]
991 EXAMPLE
992 -------
993 NOTE: A few examples are available in the
994 https://github.com/lttng/lttng-ust/tree/master/doc/examples[`doc/examples`]
995 directory of LTTng-UST's source tree.
996
997 This example shows all the features documented in the previous
998 sections. The <<build-static,static linking>> method is chosen here
999 to link the application with the tracepoint provider.
1000
1001 You can compile the source files and link them together statically
1002 like this:
1003
1004 [role="term"]
1005 -------------------------------------
1006 cc -c -I. tp.c
1007 cc -c app.c
1008 cc -o app tp.o app.o -llttng-ust -ldl
1009 -------------------------------------
1010
1011 Using the man:lttng(1) tool, create an LTTng tracing session, enable
1012 all the events of this tracepoint provider, and start tracing:
1013
1014 [role="term"]
1015 ----------------------------------------------
1016 lttng create my-session
1017 lttng enable-event --userspace 'my_provider:*'
1018 lttng start
1019 ----------------------------------------------
1020
1021 You may also enable specific events:
1022
1023 [role="term"]
1024 ----------------------------------------------------------
1025 lttng enable-event --userspace my_provider:big_event
1026 lttng enable-event --userspace my_provider:event_instance2
1027 ----------------------------------------------------------
1028
1029 Run the application:
1030
1031 [role="term"]
1032 --------------------
1033 ./app some arguments
1034 --------------------
1035
1036 Stop the current tracing session and inspect the recorded events:
1037
1038 [role="term"]
1039 ----------
1040 lttng stop
1041 lttng view
1042 ----------
1043
1044
1045 Tracepoint provider header file
1046 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1047 `tp.h`:
1048
1049 ------------------------------------------------------------------------
1050 #undef TRACEPOINT_PROVIDER
1051 #define TRACEPOINT_PROVIDER my_provider
1052
1053 #undef TRACEPOINT_INCLUDE
1054 #define TRACEPOINT_INCLUDE "./tp.h"
1055
1056 #if !defined(_TP_H) || defined(TRACEPOINT_HEADER_MULTI_READ)
1057 #define _TP_H
1058
1059 #include <lttng/tracepoint.h>
1060 #include <stdio.h>
1061
1062 #include "app.h"
1063
1064 TRACEPOINT_EVENT(
1065 my_provider,
1066 simple_event,
1067 TP_ARGS(
1068 int, my_integer_arg,
1069 const char *, my_string_arg
1070 ),
1071 TP_FIELDS(
1072 ctf_string(argc, my_string_arg)
1073 ctf_integer(int, argv, my_integer_arg)
1074 )
1075 )
1076
1077 TRACEPOINT_ENUM(
1078 my_provider,
1079 my_enum,
1080 TP_ENUM_VALUES(
1081 ctf_enum_value("ZERO", 0)
1082 ctf_enum_value("ONE", 1)
1083 ctf_enum_value("TWO", 2)
1084 ctf_enum_range("A RANGE", 52, 125)
1085 ctf_enum_value("ONE THOUSAND", 1000)
1086 )
1087 )
1088
1089 TRACEPOINT_EVENT(
1090 my_provider,
1091 big_event,
1092 TP_ARGS(
1093 int, my_integer_arg,
1094 const char *, my_string_arg,
1095 FILE *, stream,
1096 double, flt_arg,
1097 int *, array_arg
1098 ),
1099 TP_FIELDS(
1100 ctf_integer(int, int_field1, my_integer_arg * 2)
1101 ctf_integer_hex(long int, stream_pos, ftell(stream))
1102 ctf_float(double, float_field, flt_arg)
1103 ctf_string(string_field, my_string_arg)
1104 ctf_array(int, array_field, array_arg, 7)
1105 ctf_array_text(char, array_text_field, array_arg, 5)
1106 ctf_sequence(int, seq_field, array_arg, int,
1107 my_integer_arg / 10)
1108 ctf_sequence_text(char, seq_text_field, array_arg,
1109 int, my_integer_arg / 5)
1110 ctf_enum(my_provider, my_enum, int,
1111 enum_field, array_arg[1])
1112 )
1113 )
1114
1115 TRACEPOINT_LOGLEVEL(my_provider, big_event, TRACE_WARNING)
1116
1117 TRACEPOINT_EVENT_CLASS(
1118 my_provider,
1119 my_tracepoint_class,
1120 TP_ARGS(
1121 int, my_integer_arg,
1122 struct app_struct *, app_struct_arg
1123 ),
1124 TP_FIELDS(
1125 ctf_integer(int, a, my_integer_arg)
1126 ctf_integer(unsigned long, b, app_struct_arg->b)
1127 ctf_string(c, app_struct_arg->c)
1128 )
1129 )
1130
1131 TRACEPOINT_EVENT_INSTANCE(
1132 my_provider,
1133 my_tracepoint_class,
1134 event_instance1,
1135 TP_ARGS(
1136 int, my_integer_arg,
1137 struct app_struct *, app_struct_arg
1138 )
1139 )
1140
1141 TRACEPOINT_EVENT_INSTANCE(
1142 my_provider,
1143 my_tracepoint_class,
1144 event_instance2,
1145 TP_ARGS(
1146 int, my_integer_arg,
1147 struct app_struct *, app_struct_arg
1148 )
1149 )
1150
1151 TRACEPOINT_LOGLEVEL(my_provider, event_instance2, TRACE_INFO)
1152
1153 TRACEPOINT_EVENT_INSTANCE(
1154 my_provider,
1155 my_tracepoint_class,
1156 event_instance3,
1157 TP_ARGS(
1158 int, my_integer_arg,
1159 struct app_struct *, app_struct_arg
1160 )
1161 )
1162
1163 #endif /* _TP_H */
1164
1165 #include <lttng/tracepoint-event.h>
1166 ------------------------------------------------------------------------
1167
1168
1169 Tracepoint provider source file
1170 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1171 `tp.c`:
1172
1173 ------------------------------------------------------------------------
1174 #define TRACEPOINT_CREATE_PROBES
1175 #define TRACEPOINT_DEFINE
1176
1177 #include "tp.h"
1178 ------------------------------------------------------------------------
1179
1180
1181 Application header file
1182 ~~~~~~~~~~~~~~~~~~~~~~~
1183 `app.h`:
1184
1185 ------------------------------------------------------------------------
1186 #ifndef _APP_H
1187 #define _APP_H
1188
1189 struct app_struct {
1190 unsigned long b;
1191 const char *c;
1192 double d;
1193 };
1194
1195 #endif /* _APP_H */
1196 ------------------------------------------------------------------------
1197
1198
1199 Application source file
1200 ~~~~~~~~~~~~~~~~~~~~~~~
1201 `app.c`:
1202
1203 ------------------------------------------------------------------------
1204 #include <stdlib.h>
1205 #include <stdio.h>
1206
1207 #include "tp.h"
1208 #include "app.h"
1209
1210 static int array_of_ints[] = {
1211 100, -35, 1, 23, 14, -6, 28, 1001, -3000,
1212 };
1213
1214 int main(int argc, char* argv[])
1215 {
1216 FILE *stream;
1217 struct app_struct app_struct;
1218
1219 tracepoint(my_provider, simple_event, argc, argv[0]);
1220 stream = fopen("/tmp/app.txt", "w");
1221
1222 if (!stream) {
1223 fprintf(stderr,
1224 "Error: Cannot open /tmp/app.txt for writing\n");
1225 return EXIT_FAILURE;
1226 }
1227
1228 if (fprintf(stream, "0123456789") != 10) {
1229 fclose(stream);
1230 fprintf(stderr, "Error: Cannot write to /tmp/app.txt\n");
1231 return EXIT_FAILURE;
1232 }
1233
1234 tracepoint(my_provider, big_event, 35, "hello tracepoint",
1235 stream, -3.14, array_of_ints);
1236 fclose(stream);
1237 app_struct.b = argc;
1238 app_struct.c = "[the string]";
1239 tracepoint(my_provider, event_instance1, 23, &app_struct);
1240 app_struct.b = argc * 5;
1241 app_struct.c = "[other string]";
1242 tracepoint(my_provider, event_instance2, 17, &app_struct);
1243 app_struct.b = 23;
1244 app_struct.c = "nothing";
1245 tracepoint(my_provider, event_instance3, -52, &app_struct);
1246
1247 return EXIT_SUCCESS;
1248 }
1249 ------------------------------------------------------------------------
1250
1251
1252 ENVIRONMENT VARIABLES
1253 ---------------------
1254 `LTTNG_HOME`::
1255 Alternative user's home directory. This variable is useful when the
1256 user running the instrumented application has a non-writable home
1257 directory.
1258 +
1259 Unix sockets used for the communication between `liblttng-ust` and the
1260 LTTng session and consumer daemons (part of the LTTng-tools project)
1261 are located in a specific directory under `$LTTNG_HOME` (or `$HOME` if
1262 `$LTTNG_HOME` is not set).
1263
1264 `LTTNG_UST_BLOCKING_RETRY_TIMEOUT`::
1265 Maximum duration (milliseconds) to retry event tracing when
1266 there's no space left for the event record in the sub-buffer.
1267 +
1268 --
1269 `0` (default)::
1270 Never block the application.
1271
1272 Positive value::
1273 Block the application for the specified number of milliseconds. If
1274 there's no space left after this duration, discard the event
1275 record.
1276
1277 Negative value::
1278 Block the application until there's space left for the event record.
1279 --
1280 +
1281 This option can be useful in workloads generating very large trace data
1282 throughput, where blocking the application is an acceptable trade-off to
1283 prevent discarding event records.
1284 +
1285 WARNING: Setting this environment variable to a non-zero value may
1286 significantly affect application timings.
1287
1288 `LTTNG_UST_CLOCK_PLUGIN`::
1289 Path to the shared object which acts as the clock override plugin.
1290 An example of such a plugin can be found in the LTTng-UST
1291 documentation under
1292 https://github.com/lttng/lttng-ust/tree/master/doc/examples/clock-override[`examples/clock-override`].
1293
1294 `LTTNG_UST_DEBUG`::
1295 Activates `liblttng-ust`'s debug and error output if set to `1`.
1296
1297 `LTTNG_UST_GETCPU_PLUGIN`::
1298 Path to the shared object which acts as the `getcpu()` override
1299 plugin. An example of such a plugin can be found in the LTTng-UST
1300 documentation under
1301 https://github.com/lttng/lttng-ust/tree/master/doc/examples/getcpu-override[`examples/getcpu-override`].
1302
1303 `LTTNG_UST_REGISTER_TIMEOUT`::
1304 Waiting time for the _registration done_ session daemon command
1305 before proceeding to execute the main program (milliseconds).
1306 +
1307 The value `0` means _do not wait_. The value `-1` means _wait forever_.
1308 Setting this environment variable to `0` is recommended for applications
1309 with time constraints on the process startup time.
1310 +
1311 Default: {lttng_ust_register_timeout}.
1312
1313 `LTTNG_UST_BLOCKING_RETRY_TIMEOUT`::
1314 Maximum time during which event tracing retry is attempted on buffer
1315 full condition (millliseconds). Setting this environment to non-zero
1316 value effectively blocks the application on buffer full condition.
1317 Setting this environment variable to non-zero values may
1318 significantly affect application timings. Setting this to a negative
1319 value may block the application indefinitely if there is no consumer
1320 emptying the ring buffer. The delay between retry attempts is the
1321 minimum between the specified timeout value and 100ms. This option
1322 can be useful in workloads generating very large trace data
1323 throughput, where blocking the application is an acceptable
1324 trade-off to not discard events. _Use with caution_.
1325 +
1326 The value `0` means _do not retry_. The value `-1` means _retry forever_.
1327 Value > `0` means a maximum timeout of the given value.
1328 +
1329 Default: {lttng_ust_blocking_retry_timeout}.
1330
1331 `LTTNG_UST_WITHOUT_BADDR_STATEDUMP`::
1332 Prevents `liblttng-ust` from performing a base address state dump
1333 (see the <<state-dump,LTTng-UST state dump>> section above) if
1334 set to `1`.
1335
1336
1337 include::common-footer.txt[]
1338
1339 include::common-copyrights.txt[]
1340
1341 include::common-authors.txt[]
1342
1343
1344 SEE ALSO
1345 --------
1346 man:tracef(3),
1347 man:tracelog(3),
1348 man:lttng-gen-tp(1),
1349 man:lttng-ust-dl(3),
1350 man:lttng-ust-cyg-profile(3),
1351 man:lttng(1),
1352 man:lttng-enable-event(1),
1353 man:lttng-list(1),
1354 man:lttng-add-context(1),
1355 man:babeltrace(1),
1356 man:dlopen(3),
1357 man:ld.so(8)
This page took 0.053599 seconds and 3 git commands to generate.