6 :Updated: Li Zefan and Tom Zanussi
11 Tracepoints (see Documentation/trace/tracepoints.rst) can be used
12 without creating custom kernel modules to register probe functions
13 using the event tracing infrastructure.
15 Not all tracepoints can be traced using the event tracing system;
16 the kernel developer must provide code snippets which define how the
17 tracing information is saved into the tracing buffer, and how the
18 tracing information should be printed.
20 2. Using Event Tracing
21 ======================
23 2.1 Via the 'set_event' interface
24 ---------------------------------
26 The events which are available for tracing can be found in the file
27 /sys/kernel/debug/tracing/available_events.
29 To enable a particular event, such as 'sched_wakeup', simply echo it
30 to /sys/kernel/debug/tracing/set_event. For example::
32 # echo sched_wakeup >> /sys/kernel/debug/tracing/set_event
34 .. Note:: '>>' is necessary, otherwise it will firstly disable all the events.
36 To disable an event, echo the event name to the set_event file prefixed
37 with an exclamation point::
39 # echo '!sched_wakeup' >> /sys/kernel/debug/tracing/set_event
41 To disable all events, echo an empty line to the set_event file::
43 # echo > /sys/kernel/debug/tracing/set_event
45 To enable all events, echo ``*:*`` or ``*:`` to the set_event file::
47 # echo *:* > /sys/kernel/debug/tracing/set_event
49 The events are organized into subsystems, such as ext4, irq, sched,
50 etc., and a full event name looks like this: <subsystem>:<event>. The
51 subsystem name is optional, but it is displayed in the available_events
52 file. All of the events in a subsystem can be specified via the syntax
53 ``<subsystem>:*``; for example, to enable all irq events, you can use the
56 # echo 'irq:*' > /sys/kernel/debug/tracing/set_event
58 2.2 Via the 'enable' toggle
59 ---------------------------
61 The events available are also listed in /sys/kernel/debug/tracing/events/ hierarchy
64 To enable event 'sched_wakeup'::
66 # echo 1 > /sys/kernel/debug/tracing/events/sched/sched_wakeup/enable
70 # echo 0 > /sys/kernel/debug/tracing/events/sched/sched_wakeup/enable
72 To enable all events in sched subsystem::
74 # echo 1 > /sys/kernel/debug/tracing/events/sched/enable
76 To enable all events::
78 # echo 1 > /sys/kernel/debug/tracing/events/enable
80 When reading one of these enable files, there are four results:
82 - 0 - all events this file affects are disabled
83 - 1 - all events this file affects are enabled
84 - X - there is a mixture of events enabled and disabled
85 - ? - this file does not affect any event
90 In order to facilitate early boot debugging, use boot option::
92 trace_event=[event-list]
94 event-list is a comma separated list of events. See section 2.1 for event
97 3. Defining an event-enabled tracepoint
98 =======================================
100 See The example provided in samples/trace_events
105 Each trace event has a 'format' file associated with it that contains
106 a description of each field in a logged event. This information can
107 be used to parse the binary trace stream, and is also the place to
108 find the field names that can be used in event filters (see section 5).
110 It also displays the format string that will be used to print the
111 event in text mode, along with the event name and ID used for
114 Every event has a set of ``common`` fields associated with it; these are
115 the fields prefixed with ``common_``. The other fields vary between
116 events and correspond to the fields defined in the TRACE_EVENT
117 definition for that event.
119 Each field in the format has the form::
121 field:field-type field-name; offset:N; size:N;
123 where offset is the offset of the field in the trace record and size
124 is the size of the data item, in bytes.
126 For example, here's the information displayed for the 'sched_wakeup'
129 # cat /sys/kernel/debug/tracing/events/sched/sched_wakeup/format
134 field:unsigned short common_type; offset:0; size:2;
135 field:unsigned char common_flags; offset:2; size:1;
136 field:unsigned char common_preempt_count; offset:3; size:1;
137 field:int common_pid; offset:4; size:4;
138 field:int common_tgid; offset:8; size:4;
140 field:char comm[TASK_COMM_LEN]; offset:12; size:16;
141 field:pid_t pid; offset:28; size:4;
142 field:int prio; offset:32; size:4;
143 field:int success; offset:36; size:4;
144 field:int cpu; offset:40; size:4;
146 print fmt: "task %s:%d [%d] success=%d [%03d]", REC->comm, REC->pid,
147 REC->prio, REC->success, REC->cpu
149 This event contains 10 fields, the first 5 common and the remaining 5
150 event-specific. All the fields for this event are numeric, except for
151 'comm' which is a string, a distinction important for event filtering.
156 Trace events can be filtered in the kernel by associating boolean
157 'filter expressions' with them. As soon as an event is logged into
158 the trace buffer, its fields are checked against the filter expression
159 associated with that event type. An event with field values that
160 'match' the filter will appear in the trace output, and an event whose
161 values don't match will be discarded. An event with no filter
162 associated with it matches everything, and is the default when no
163 filter has been set for an event.
165 5.1 Expression syntax
166 ---------------------
168 A filter expression consists of one or more 'predicates' that can be
169 combined using the logical operators '&&' and '||'. A predicate is
170 simply a clause that compares the value of a field contained within a
171 logged event with a constant value and returns either 0 or 1 depending
172 on whether the field value matched (1) or didn't match (0)::
174 field-name relational-operator value
176 Parentheses can be used to provide arbitrary logical groupings and
177 double-quotes can be used to prevent the shell from interpreting
178 operators as shell metacharacters.
180 The field-names available for use in filters can be found in the
181 'format' files for trace events (see section 4).
183 The relational-operators depend on the type of the field being tested:
185 The operators available for numeric fields are:
187 ==, !=, <, <=, >, >=, &
189 And for string fields they are:
193 The glob (~) accepts a wild card character (\*,?) and character classes
204 A filter for an individual event is set by writing a filter expression
205 to the 'filter' file for the given event.
209 # cd /sys/kernel/debug/tracing/events/sched/sched_wakeup
210 # echo "common_preempt_count > 4" > filter
212 A slightly more involved example::
214 # cd /sys/kernel/debug/tracing/events/signal/signal_generate
215 # echo "((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter
217 If there is an error in the expression, you'll get an 'Invalid
218 argument' error when setting it, and the erroneous string along with
219 an error message can be seen by looking at the filter e.g.::
221 # cd /sys/kernel/debug/tracing/events/signal/signal_generate
222 # echo "((sig >= 10 && sig < 15) || dsig == 17) && comm != bash" > filter
223 -bash: echo: write error: Invalid argument
225 ((sig >= 10 && sig < 15) || dsig == 17) && comm != bash
227 parse_error: Field not found
229 Currently the caret ('^') for an error always appears at the beginning of
230 the filter string; the error message should still be useful though
231 even without more accurate position info.
236 To clear the filter for an event, write a '0' to the event's filter
239 To clear the filters for all events in a subsystem, write a '0' to the
240 subsystem's filter file.
242 5.3 Subsystem filters
243 ---------------------
245 For convenience, filters for every event in a subsystem can be set or
246 cleared as a group by writing a filter expression into the filter file
247 at the root of the subsystem. Note however, that if a filter for any
248 event within the subsystem lacks a field specified in the subsystem
249 filter, or if the filter can't be applied for any other reason, the
250 filter for that event will retain its previous setting. This can
251 result in an unintended mixture of filters which could lead to
252 confusing (to the user who might think different filters are in
253 effect) trace output. Only filters that reference just the common
254 fields can be guaranteed to propagate successfully to all events.
256 Here are a few subsystem filter examples that also illustrate the
259 Clear the filters on all events in the sched subsystem::
261 # cd /sys/kernel/debug/tracing/events/sched
263 # cat sched_switch/filter
265 # cat sched_wakeup/filter
268 Set a filter using only common fields for all events in the sched
269 subsystem (all events end up with the same filter)::
271 # cd /sys/kernel/debug/tracing/events/sched
272 # echo common_pid == 0 > filter
273 # cat sched_switch/filter
275 # cat sched_wakeup/filter
278 Attempt to set a filter using a non-common field for all events in the
279 sched subsystem (all events but those that have a prev_pid field retain
282 # cd /sys/kernel/debug/tracing/events/sched
283 # echo prev_pid == 0 > filter
284 # cat sched_switch/filter
286 # cat sched_wakeup/filter
292 The set_event_pid file in the same directory as the top events directory
293 exists, will filter all events from tracing any task that does not have the
294 PID listed in the set_event_pid file.
297 # cd /sys/kernel/debug/tracing
298 # echo $$ > set_event_pid
299 # echo 1 > events/enable
301 Will only trace events for the current task.
303 To add more PIDs without losing the PIDs already included, use '>>'.
306 # echo 123 244 1 >> set_event_pid
312 Trace events can be made to conditionally invoke trigger 'commands'
313 which can take various forms and are described in detail below;
314 examples would be enabling or disabling other trace events or invoking
315 a stack trace whenever the trace event is hit. Whenever a trace event
316 with attached triggers is invoked, the set of trigger commands
317 associated with that event is invoked. Any given trigger can
318 additionally have an event filter of the same form as described in
319 section 5 (Event filtering) associated with it - the command will only
320 be invoked if the event being invoked passes the associated filter.
321 If no filter is associated with the trigger, it always passes.
323 Triggers are added to and removed from a particular event by writing
324 trigger expressions to the 'trigger' file for the given event.
326 A given event can have any number of triggers associated with it,
327 subject to any restrictions that individual commands may have in that
330 Event triggers are implemented on top of "soft" mode, which means that
331 whenever a trace event has one or more triggers associated with it,
332 the event is activated even if it isn't actually enabled, but is
333 disabled in a "soft" mode. That is, the tracepoint will be called,
334 but just will not be traced, unless of course it's actually enabled.
335 This scheme allows triggers to be invoked even for events that aren't
336 enabled, and also allows the current event filter implementation to be
337 used for conditionally invoking triggers.
339 The syntax for event triggers is roughly based on the syntax for
340 set_ftrace_filter 'ftrace filter commands' (see the 'Filter commands'
341 section of Documentation/trace/ftrace.rst), but there are major
342 differences and the implementation isn't currently tied to it in any
343 way, so beware about making generalizations between the two.
345 Note: Writing into trace_marker (See Documentation/trace/ftrace.rst)
346 can also enable triggers that are written into
347 /sys/kernel/tracing/events/ftrace/print/trigger
349 6.1 Expression syntax
350 ---------------------
352 Triggers are added by echoing the command to the 'trigger' file::
354 # echo 'command[:count] [if filter]' > trigger
356 Triggers are removed by echoing the same command but starting with '!'
357 to the 'trigger' file::
359 # echo '!command[:count] [if filter]' > trigger
361 The [if filter] part isn't used in matching commands when removing, so
362 leaving that off in a '!' command will accomplish the same thing as
365 The filter syntax is the same as that described in the 'Event
366 filtering' section above.
368 For ease of use, writing to the trigger file using '>' currently just
369 adds or removes a single trigger and there's no explicit '>>' support
370 ('>' actually behaves like '>>') or truncation support to remove all
371 triggers (you have to use '!' for each one added.)
373 6.2 Supported trigger commands
374 ------------------------------
376 The following commands are supported:
378 - enable_event/disable_event
380 These commands can enable or disable another trace event whenever
381 the triggering event is hit. When these commands are registered,
382 the other trace event is activated, but disabled in a "soft" mode.
383 That is, the tracepoint will be called, but just will not be traced.
384 The event tracepoint stays in this mode as long as there's a trigger
385 in effect that can trigger it.
387 For example, the following trigger causes kmalloc events to be
388 traced when a read system call is entered, and the :1 at the end
389 specifies that this enablement happens only once::
391 # echo 'enable_event:kmem:kmalloc:1' > \
392 /sys/kernel/debug/tracing/events/syscalls/sys_enter_read/trigger
394 The following trigger causes kmalloc events to stop being traced
395 when a read system call exits. This disablement happens on every
396 read system call exit::
398 # echo 'disable_event:kmem:kmalloc' > \
399 /sys/kernel/debug/tracing/events/syscalls/sys_exit_read/trigger
403 enable_event:<system>:<event>[:count]
404 disable_event:<system>:<event>[:count]
406 To remove the above commands::
408 # echo '!enable_event:kmem:kmalloc:1' > \
409 /sys/kernel/debug/tracing/events/syscalls/sys_enter_read/trigger
411 # echo '!disable_event:kmem:kmalloc' > \
412 /sys/kernel/debug/tracing/events/syscalls/sys_exit_read/trigger
414 Note that there can be any number of enable/disable_event triggers
415 per triggering event, but there can only be one trigger per
416 triggered event. e.g. sys_enter_read can have triggers enabling both
417 kmem:kmalloc and sched:sched_switch, but can't have two kmem:kmalloc
418 versions such as kmem:kmalloc and kmem:kmalloc:1 or 'kmem:kmalloc if
419 bytes_req == 256' and 'kmem:kmalloc if bytes_alloc == 256' (they
420 could be combined into a single filter on kmem:kmalloc though).
424 This command dumps a stacktrace in the trace buffer whenever the
425 triggering event occurs.
427 For example, the following trigger dumps a stacktrace every time the
428 kmalloc tracepoint is hit::
430 # echo 'stacktrace' > \
431 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
433 The following trigger dumps a stacktrace the first 5 times a kmalloc
434 request happens with a size >= 64K::
436 # echo 'stacktrace:5 if bytes_req >= 65536' > \
437 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
443 To remove the above commands::
445 # echo '!stacktrace' > \
446 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
448 # echo '!stacktrace:5 if bytes_req >= 65536' > \
449 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
451 The latter can also be removed more simply by the following (without
454 # echo '!stacktrace:5' > \
455 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
457 Note that there can be only one stacktrace trigger per triggering
462 This command causes a snapshot to be triggered whenever the
463 triggering event occurs.
465 The following command creates a snapshot every time a block request
466 queue is unplugged with a depth > 1. If you were tracing a set of
467 events or functions at the time, the snapshot trace buffer would
468 capture those events when the trigger event occurred::
470 # echo 'snapshot if nr_rq > 1' > \
471 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
473 To only snapshot once::
475 # echo 'snapshot:1 if nr_rq > 1' > \
476 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
478 To remove the above commands::
480 # echo '!snapshot if nr_rq > 1' > \
481 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
483 # echo '!snapshot:1 if nr_rq > 1' > \
484 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
486 Note that there can be only one snapshot trigger per triggering
491 These commands turn tracing on and off when the specified events are
492 hit. The parameter determines how many times the tracing system is
493 turned on and off. If unspecified, there is no limit.
495 The following command turns tracing off the first time a block
496 request queue is unplugged with a depth > 1. If you were tracing a
497 set of events or functions at the time, you could then examine the
498 trace buffer to see the sequence of events that led up to the
501 # echo 'traceoff:1 if nr_rq > 1' > \
502 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
504 To always disable tracing when nr_rq > 1::
506 # echo 'traceoff if nr_rq > 1' > \
507 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
509 To remove the above commands::
511 # echo '!traceoff:1 if nr_rq > 1' > \
512 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
514 # echo '!traceoff if nr_rq > 1' > \
515 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
517 Note that there can be only one traceon or traceoff trigger per
522 This command aggregates event hits into a hash table keyed on one or
523 more trace event format fields (or stacktrace) and a set of running
524 totals derived from one or more trace event format fields and/or
525 event counts (hitcount).
527 See Documentation/trace/histogram.rst for details and examples.
529 6.3 In-kernel trace event API
530 -----------------------------
532 In most cases, the command-line interface to trace events is more than
533 sufficient. Sometimes, however, applications might find the need for
534 more complex relationships than can be expressed through a simple
535 series of linked command-line expressions, or putting together sets of
536 commands may be simply too cumbersome. An example might be an
537 application that needs to 'listen' to the trace stream in order to
538 maintain an in-kernel state machine detecting, for instance, when an
539 illegal kernel state occurs in the scheduler.
541 The trace event subsystem provides an in-kernel API allowing modules
542 or other kernel code to generate user-defined 'synthetic' events at
543 will, which can be used to either augment the existing trace stream
544 and/or signal that a particular important state has occurred.
546 A similar in-kernel API is also available for creating kprobe and
549 Both the synthetic event and k/ret/probe event APIs are built on top
550 of a lower-level "dynevent_cmd" event command API, which is also
551 available for more specialized applications, or as the basis of other
552 higher-level trace event APIs.
554 The API provided for these purposes is describe below and allows the
557 - dynamically creating synthetic event definitions
558 - dynamically creating kprobe and kretprobe event definitions
559 - tracing synthetic events from in-kernel code
560 - the low-level "dynevent_cmd" API
562 6.3.1 Dyamically creating synthetic event definitions
563 -----------------------------------------------------
565 There are a couple ways to create a new synthetic event from a kernel
566 module or other kernel code.
568 The first creates the event in one step, using synth_event_create().
569 In this method, the name of the event to create and an array defining
570 the fields is supplied to synth_event_create(). If successful, a
571 synthetic event with that name and fields will exist following that
572 call. For example, to create a new "schedtest" synthetic event:
574 ret = synth_event_create("schedtest", sched_fields,
575 ARRAY_SIZE(sched_fields), THIS_MODULE);
577 The sched_fields param in this example points to an array of struct
578 synth_field_desc, each of which describes an event field by type and
581 static struct synth_field_desc sched_fields[] = {
582 { .type = "pid_t", .name = "next_pid_field" },
583 { .type = "char[16]", .name = "next_comm_field" },
584 { .type = "u64", .name = "ts_ns" },
585 { .type = "u64", .name = "ts_ms" },
586 { .type = "unsigned int", .name = "cpu" },
587 { .type = "char[64]", .name = "my_string_field" },
588 { .type = "int", .name = "my_int_field" },
591 See synth_field_size() for available types. If field_name contains [n]
592 the field is considered to be an array.
594 If the event is created from within a module, a pointer to the module
595 must be passed to synth_event_create(). This will ensure that the
596 trace buffer won't contain unreadable events when the module is
599 At this point, the event object is ready to be used for generating new
602 In the second method, the event is created in several steps. This
603 allows events to be created dynamically and without the need to create
604 and populate an array of fields beforehand.
606 To use this method, an empty or partially empty synthetic event should
607 first be created using synth_event_gen_cmd_start() or
608 synth_event_gen_cmd_array_start(). For synth_event_gen_cmd_start(),
609 the name of the event along with one or more pairs of args each pair
610 representing a 'type field_name;' field specification should be
611 supplied. For synth_event_gen_cmd_array_start(), the name of the
612 event along with an array of struct synth_field_desc should be
613 supplied. Before calling synth_event_gen_cmd_start() or
614 synth_event_gen_cmd_array_start(), the user should create and
615 initialize a dynevent_cmd object using synth_event_cmd_init().
617 For example, to create a new "schedtest" synthetic event with two
620 struct dynevent_cmd cmd;
623 /* Create a buffer to hold the generated command */
624 buf = kzalloc(MAX_DYNEVENT_CMD_LEN, GFP_KERNEL);
626 /* Before generating the command, initialize the cmd object */
627 synth_event_cmd_init(&cmd, buf, MAX_DYNEVENT_CMD_LEN);
629 ret = synth_event_gen_cmd_start(&cmd, "schedtest", THIS_MODULE,
630 "pid_t", "next_pid_field",
633 Alternatively, using an array of struct synth_field_desc fields
634 containing the same information:
636 ret = synth_event_gen_cmd_array_start(&cmd, "schedtest", THIS_MODULE,
639 Once the synthetic event object has been created, it can then be
640 populated with more fields. Fields are added one by one using
641 synth_event_add_field(), supplying the dynevent_cmd object, a field
642 type, and a field name. For example, to add a new int field named
643 "intfield", the following call should be made:
645 ret = synth_event_add_field(&cmd, "int", "intfield");
647 See synth_field_size() for available types. If field_name contains [n]
648 the field is considered to be an array.
650 A group of fields can also be added all at once using an array of
651 synth_field_desc with add_synth_fields(). For example, this would add
652 just the first four sched_fields:
654 ret = synth_event_add_fields(&cmd, sched_fields, 4);
656 If you already have a string of the form 'type field_name',
657 synth_event_add_field_str() can be used to add it as-is; it will
658 also automatically append a ';' to the string.
660 Once all the fields have been added, the event should be finalized and
661 registered by calling the synth_event_gen_cmd_end() function:
663 ret = synth_event_gen_cmd_end(&cmd);
665 At this point, the event object is ready to be used for tracing new
668 6.3.3 Tracing synthetic events from in-kernel code
669 --------------------------------------------------
671 To trace a synthetic event, there are several options. The first
672 option is to trace the event in one call, using synth_event_trace()
673 with a variable number of values, or synth_event_trace_array() with an
674 array of values to be set. A second option can be used to avoid the
675 need for a pre-formed array of values or list of arguments, via
676 synth_event_trace_start() and synth_event_trace_end() along with
677 synth_event_add_next_val() or synth_event_add_val() to add the values
680 6.3.3.1 Tracing a synthetic event all at once
681 ---------------------------------------------
683 To trace a synthetic event all at once, the synth_event_trace() or
684 synth_event_trace_array() functions can be used.
686 The synth_event_trace() function is passed the trace_event_file
687 representing the synthetic event (which can be retrieved using
688 trace_get_event_file() using the synthetic event name, "synthetic" as
689 the system name, and the trace instance name (NULL if using the global
690 trace array)), along with an variable number of u64 args, one for each
691 synthetic event field, and the number of values being passed.
693 So, to trace an event corresponding to the synthetic event definition
694 above, code like the following could be used:
696 ret = synth_event_trace(create_synth_test, 7, /* number of values */
697 444, /* next_pid_field */
698 (u64)"clackers", /* next_comm_field */
701 smp_processor_id(),/* cpu */
702 (u64)"Thneed", /* my_string_field */
703 999); /* my_int_field */
705 All vals should be cast to u64, and string vals are just pointers to
706 strings, cast to u64. Strings will be copied into space reserved in
707 the event for the string, using these pointers.
709 Alternatively, the synth_event_trace_array() function can be used to
710 accomplish the same thing. It is passed the trace_event_file
711 representing the synthetic event (which can be retrieved using
712 trace_get_event_file() using the synthetic event name, "synthetic" as
713 the system name, and the trace instance name (NULL if using the global
714 trace array)), along with an array of u64, one for each synthetic
717 To trace an event corresponding to the synthetic event definition
718 above, code like the following could be used:
722 vals[0] = 777; /* next_pid_field */
723 vals[1] = (u64)"tiddlywinks"; /* next_comm_field */
724 vals[2] = 1000000; /* ts_ns */
725 vals[3] = 1000; /* ts_ms */
726 vals[4] = smp_processor_id(); /* cpu */
727 vals[5] = (u64)"thneed"; /* my_string_field */
728 vals[6] = 398; /* my_int_field */
730 The 'vals' array is just an array of u64, the number of which must
731 match the number of field in the synthetic event, and which must be in
732 the same order as the synthetic event fields.
734 All vals should be cast to u64, and string vals are just pointers to
735 strings, cast to u64. Strings will be copied into space reserved in
736 the event for the string, using these pointers.
738 In order to trace a synthetic event, a pointer to the trace event file
739 is needed. The trace_get_event_file() function can be used to get
740 it - it will find the file in the given trace instance (in this case
741 NULL since the top trace array is being used) while at the same time
742 preventing the instance containing it from going away:
744 schedtest_event_file = trace_get_event_file(NULL, "synthetic",
747 Before tracing the event, it should be enabled in some way, otherwise
748 the synthetic event won't actually show up in the trace buffer.
750 To enable a synthetic event from the kernel, trace_array_set_clr_event()
751 can be used (which is not specific to synthetic events, so does need
752 the "synthetic" system name to be specified explicitly).
754 To enable the event, pass 'true' to it:
756 trace_array_set_clr_event(schedtest_event_file->tr,
757 "synthetic", "schedtest", true);
759 To disable it pass false:
761 trace_array_set_clr_event(schedtest_event_file->tr,
762 "synthetic", "schedtest", false);
764 Finally, synth_event_trace_array() can be used to actually trace the
765 event, which should be visible in the trace buffer afterwards:
767 ret = synth_event_trace_array(schedtest_event_file, vals,
770 To remove the synthetic event, the event should be disabled, and the
771 trace instance should be 'put' back using trace_put_event_file():
773 trace_array_set_clr_event(schedtest_event_file->tr,
774 "synthetic", "schedtest", false);
775 trace_put_event_file(schedtest_event_file);
777 If those have been successful, synth_event_delete() can be called to
780 ret = synth_event_delete("schedtest");
782 6.3.3.1 Tracing a synthetic event piecewise
783 -------------------------------------------
785 To trace a synthetic using the piecewise method described above, the
786 synth_event_trace_start() function is used to 'open' the synthetic
789 struct synth_trace_state trace_state;
791 ret = synth_event_trace_start(schedtest_event_file, &trace_state);
793 It's passed the trace_event_file representing the synthetic event
794 using the same methods as described above, along with a pointer to a
795 struct synth_trace_state object, which will be zeroed before use and
796 used to maintain state between this and following calls.
798 Once the event has been opened, which means space for it has been
799 reserved in the trace buffer, the individual fields can be set. There
800 are two ways to do that, either one after another for each field in
801 the event, which requires no lookups, or by name, which does. The
802 tradeoff is flexibility in doing the assignments vs the cost of a
805 To assign the values one after the other without lookups,
806 synth_event_add_next_val() should be used. Each call is passed the
807 same synth_trace_state object used in the synth_event_trace_start(),
808 along with the value to set the next field in the event. After each
809 field is set, the 'cursor' points to the next field, which will be set
810 by the subsequent call, continuing until all the fields have been set
811 in order. The same sequence of calls as in the above examples using
812 this method would be (without error-handling code):
815 ret = synth_event_add_next_val(777, &trace_state);
817 /* next_comm_field */
818 ret = synth_event_add_next_val((u64)"slinky", &trace_state);
821 ret = synth_event_add_next_val(1000000, &trace_state);
824 ret = synth_event_add_next_val(1000, &trace_state);
827 ret = synth_event_add_next_val(smp_processor_id(), &trace_state);
829 /* my_string_field */
830 ret = synth_event_add_next_val((u64)"thneed_2.01", &trace_state);
833 ret = synth_event_add_next_val(395, &trace_state);
835 To assign the values in any order, synth_event_add_val() should be
836 used. Each call is passed the same synth_trace_state object used in
837 the synth_event_trace_start(), along with the field name of the field
838 to set and the value to set it to. The same sequence of calls as in
839 the above examples using this method would be (without error-handling
842 ret = synth_event_add_val("next_pid_field", 777, &trace_state);
843 ret = synth_event_add_val("next_comm_field", (u64)"silly putty",
845 ret = synth_event_add_val("ts_ns", 1000000, &trace_state);
846 ret = synth_event_add_val("ts_ms", 1000, &trace_state);
847 ret = synth_event_add_val("cpu", smp_processor_id(), &trace_state);
848 ret = synth_event_add_val("my_string_field", (u64)"thneed_9",
850 ret = synth_event_add_val("my_int_field", 3999, &trace_state);
852 Note that synth_event_add_next_val() and synth_event_add_val() are
853 incompatible if used within the same trace of an event - either one
854 can be used but not both at the same time.
856 Finally, the event won't be actually traced until it's 'closed',
857 which is done using synth_event_trace_end(), which takes only the
858 struct synth_trace_state object used in the previous calls:
860 ret = synth_event_trace_end(&trace_state);
862 Note that synth_event_trace_end() must be called at the end regardless
863 of whether any of the add calls failed (say due to a bad field name
866 6.3.4 Dyamically creating kprobe and kretprobe event definitions
867 ----------------------------------------------------------------
869 To create a kprobe or kretprobe trace event from kernel code, the
870 kprobe_event_gen_cmd_start() or kretprobe_event_gen_cmd_start()
871 functions can be used.
873 To create a kprobe event, an empty or partially empty kprobe event
874 should first be created using kprobe_event_gen_cmd_start(). The name
875 of the event and the probe location should be specfied along with one
876 or args each representing a probe field should be supplied to this
877 function. Before calling kprobe_event_gen_cmd_start(), the user
878 should create and initialize a dynevent_cmd object using
879 kprobe_event_cmd_init().
881 For example, to create a new "schedtest" kprobe event with two fields:
883 struct dynevent_cmd cmd;
886 /* Create a buffer to hold the generated command */
887 buf = kzalloc(MAX_DYNEVENT_CMD_LEN, GFP_KERNEL);
889 /* Before generating the command, initialize the cmd object */
890 kprobe_event_cmd_init(&cmd, buf, MAX_DYNEVENT_CMD_LEN);
893 * Define the gen_kprobe_test event with the first 2 kprobe
896 ret = kprobe_event_gen_cmd_start(&cmd, "gen_kprobe_test", "do_sys_open",
897 "dfd=%ax", "filename=%dx");
899 Once the kprobe event object has been created, it can then be
900 populated with more fields. Fields can be added using
901 kprobe_event_add_fields(), supplying the dynevent_cmd object along
902 with a variable arg list of probe fields. For example, to add a
903 couple additional fields, the following call could be made:
905 ret = kprobe_event_add_fields(&cmd, "flags=%cx", "mode=+4($stack)");
907 Once all the fields have been added, the event should be finalized and
908 registered by calling the kprobe_event_gen_cmd_end() or
909 kretprobe_event_gen_cmd_end() functions, depending on whether a kprobe
910 or kretprobe command was started:
912 ret = kprobe_event_gen_cmd_end(&cmd);
916 ret = kretprobe_event_gen_cmd_end(&cmd);
918 At this point, the event object is ready to be used for tracing new
921 Similarly, a kretprobe event can be created using
922 kretprobe_event_gen_cmd_start() with a probe name and location and
923 additional params such as $retval:
925 ret = kretprobe_event_gen_cmd_start(&cmd, "gen_kretprobe_test",
926 "do_sys_open", "$retval");
928 Similar to the synthetic event case, code like the following can be
929 used to enable the newly created kprobe event:
931 gen_kprobe_test = trace_get_event_file(NULL, "kprobes", "gen_kprobe_test");
933 ret = trace_array_set_clr_event(gen_kprobe_test->tr,
934 "kprobes", "gen_kprobe_test", true);
936 Finally, also similar to synthetic events, the following code can be
937 used to give the kprobe event file back and delete the event:
939 trace_put_event_file(gen_kprobe_test);
941 ret = kprobe_event_delete("gen_kprobe_test");
943 6.3.4 The "dynevent_cmd" low-level API
944 --------------------------------------
946 Both the in-kernel synthetic event and kprobe interfaces are built on
947 top of a lower-level "dynevent_cmd" interface. This interface is
948 meant to provide the basis for higher-level interfaces such as the
949 synthetic and kprobe interfaces, which can be used as examples.
951 The basic idea is simple and amounts to providing a general-purpose
952 layer that can be used to generate trace event commands. The
953 generated command strings can then be passed to the command-parsing
954 and event creation code that already exists in the trace event
955 subystem for creating the corresponding trace events.
957 In a nutshell, the way it works is that the higher-level interface
958 code creates a struct dynevent_cmd object, then uses a couple
959 functions, dynevent_arg_add() and dynevent_arg_pair_add() to build up
960 a command string, which finally causes the command to be executed
961 using the dynevent_create() function. The details of the interface
964 The first step in building a new command string is to create and
965 initialize an instance of a dynevent_cmd. Here, for instance, we
966 create a dynevent_cmd on the stack and initialize it:
968 struct dynevent_cmd cmd;
972 buf = kzalloc(MAX_DYNEVENT_CMD_LEN, GFP_KERNEL);
974 dynevent_cmd_init(cmd, buf, maxlen, DYNEVENT_TYPE_FOO,
975 foo_event_run_command);
977 The dynevent_cmd initialization needs to be given a user-specified
978 buffer and the length of the buffer (MAX_DYNEVENT_CMD_LEN can be used
979 for this purpose - at 2k it's generally too big to be comfortably put
980 on the stack, so is dynamically allocated), a dynevent type id, which
981 is meant to be used to check that further API calls are for the
982 correct command type, and a pointer to an event-specific run_command()
983 callback that will be called to actually execute the event-specific
986 Once that's done, the command string can by built up by successive
987 calls to argument-adding functions.
989 To add a single argument, define and initialize a struct dynevent_arg
990 or struct dynevent_arg_pair object. Here's an example of the simplest
991 possible arg addition, which is simply to append the given string as
992 a whitespace-separated argument to the command:
994 struct dynevent_arg arg;
996 dynevent_arg_init(&arg, NULL, 0);
1000 ret = dynevent_arg_add(cmd, &arg);
1002 The arg object is first initialized using dynevent_arg_init() and in
1003 this case the parameters are NULL or 0, which means there's no
1004 optional sanity-checking function or separator appended to the end of
1007 Here's another more complicated example using an 'arg pair', which is
1008 used to create an argument that consists of a couple components added
1009 together as a unit, for example, a 'type field_name;' arg or a simple
1010 expression arg e.g. 'flags=%cx':
1012 struct dynevent_arg_pair arg_pair;
1014 dynevent_arg_pair_init(&arg_pair, dynevent_foo_check_arg_fn, 0, ';');
1016 arg_pair.lhs = type;
1017 arg_pair.rhs = name;
1019 ret = dynevent_arg_pair_add(cmd, &arg_pair);
1021 Again, the arg_pair is first initialized, in this case with a callback
1022 function used to check the sanity of the args (for example, that
1023 neither part of the pair is NULL), along with a character to be used
1024 to add an operator between the pair (here none) and a separator to be
1025 appended onto the end of the arg pair (here ';').
1027 There's also a dynevent_str_add() function that can be used to simply
1028 add a string as-is, with no spaces, delimeters, or arg check.
1030 Any number of dynevent_*_add() calls can be made to build up the string
1031 (until its length surpasses cmd->maxlen). When all the arguments have
1032 been added and the command string is complete, the only thing left to
1033 do is run the command, which happens by simply calling
1036 ret = dynevent_create(&cmd);
1038 At that point, if the return value is 0, the dynamic event has been
1039 created and is ready to use.
1041 See the dynevent_cmd function definitions themselves for the details