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.
233 5.2.1 Filter limitations
234 ------------------------
236 If a filter is placed on a string pointer ``(char *)`` that does not point
237 to a string on the ring buffer, but instead points to kernel or user space
238 memory, then, for safety reasons, at most 1024 bytes of the content is
239 copied onto a temporary buffer to do the compare. If the copy of the memory
240 faults (the pointer points to memory that should not be accessed), then the
241 string compare will be treated as not matching.
246 To clear the filter for an event, write a '0' to the event's filter
249 To clear the filters for all events in a subsystem, write a '0' to the
250 subsystem's filter file.
252 5.3 Subsystem filters
253 ---------------------
255 For convenience, filters for every event in a subsystem can be set or
256 cleared as a group by writing a filter expression into the filter file
257 at the root of the subsystem. Note however, that if a filter for any
258 event within the subsystem lacks a field specified in the subsystem
259 filter, or if the filter can't be applied for any other reason, the
260 filter for that event will retain its previous setting. This can
261 result in an unintended mixture of filters which could lead to
262 confusing (to the user who might think different filters are in
263 effect) trace output. Only filters that reference just the common
264 fields can be guaranteed to propagate successfully to all events.
266 Here are a few subsystem filter examples that also illustrate the
269 Clear the filters on all events in the sched subsystem::
271 # cd /sys/kernel/debug/tracing/events/sched
273 # cat sched_switch/filter
275 # cat sched_wakeup/filter
278 Set a filter using only common fields for all events in the sched
279 subsystem (all events end up with the same filter)::
281 # cd /sys/kernel/debug/tracing/events/sched
282 # echo common_pid == 0 > filter
283 # cat sched_switch/filter
285 # cat sched_wakeup/filter
288 Attempt to set a filter using a non-common field for all events in the
289 sched subsystem (all events but those that have a prev_pid field retain
292 # cd /sys/kernel/debug/tracing/events/sched
293 # echo prev_pid == 0 > filter
294 # cat sched_switch/filter
296 # cat sched_wakeup/filter
302 The set_event_pid file in the same directory as the top events directory
303 exists, will filter all events from tracing any task that does not have the
304 PID listed in the set_event_pid file.
307 # cd /sys/kernel/debug/tracing
308 # echo $$ > set_event_pid
309 # echo 1 > events/enable
311 Will only trace events for the current task.
313 To add more PIDs without losing the PIDs already included, use '>>'.
316 # echo 123 244 1 >> set_event_pid
322 Trace events can be made to conditionally invoke trigger 'commands'
323 which can take various forms and are described in detail below;
324 examples would be enabling or disabling other trace events or invoking
325 a stack trace whenever the trace event is hit. Whenever a trace event
326 with attached triggers is invoked, the set of trigger commands
327 associated with that event is invoked. Any given trigger can
328 additionally have an event filter of the same form as described in
329 section 5 (Event filtering) associated with it - the command will only
330 be invoked if the event being invoked passes the associated filter.
331 If no filter is associated with the trigger, it always passes.
333 Triggers are added to and removed from a particular event by writing
334 trigger expressions to the 'trigger' file for the given event.
336 A given event can have any number of triggers associated with it,
337 subject to any restrictions that individual commands may have in that
340 Event triggers are implemented on top of "soft" mode, which means that
341 whenever a trace event has one or more triggers associated with it,
342 the event is activated even if it isn't actually enabled, but is
343 disabled in a "soft" mode. That is, the tracepoint will be called,
344 but just will not be traced, unless of course it's actually enabled.
345 This scheme allows triggers to be invoked even for events that aren't
346 enabled, and also allows the current event filter implementation to be
347 used for conditionally invoking triggers.
349 The syntax for event triggers is roughly based on the syntax for
350 set_ftrace_filter 'ftrace filter commands' (see the 'Filter commands'
351 section of Documentation/trace/ftrace.rst), but there are major
352 differences and the implementation isn't currently tied to it in any
353 way, so beware about making generalizations between the two.
356 Writing into trace_marker (See Documentation/trace/ftrace.rst)
357 can also enable triggers that are written into
358 /sys/kernel/tracing/events/ftrace/print/trigger
360 6.1 Expression syntax
361 ---------------------
363 Triggers are added by echoing the command to the 'trigger' file::
365 # echo 'command[:count] [if filter]' > trigger
367 Triggers are removed by echoing the same command but starting with '!'
368 to the 'trigger' file::
370 # echo '!command[:count] [if filter]' > trigger
372 The [if filter] part isn't used in matching commands when removing, so
373 leaving that off in a '!' command will accomplish the same thing as
376 The filter syntax is the same as that described in the 'Event
377 filtering' section above.
379 For ease of use, writing to the trigger file using '>' currently just
380 adds or removes a single trigger and there's no explicit '>>' support
381 ('>' actually behaves like '>>') or truncation support to remove all
382 triggers (you have to use '!' for each one added.)
384 6.2 Supported trigger commands
385 ------------------------------
387 The following commands are supported:
389 - enable_event/disable_event
391 These commands can enable or disable another trace event whenever
392 the triggering event is hit. When these commands are registered,
393 the other trace event is activated, but disabled in a "soft" mode.
394 That is, the tracepoint will be called, but just will not be traced.
395 The event tracepoint stays in this mode as long as there's a trigger
396 in effect that can trigger it.
398 For example, the following trigger causes kmalloc events to be
399 traced when a read system call is entered, and the :1 at the end
400 specifies that this enablement happens only once::
402 # echo 'enable_event:kmem:kmalloc:1' > \
403 /sys/kernel/debug/tracing/events/syscalls/sys_enter_read/trigger
405 The following trigger causes kmalloc events to stop being traced
406 when a read system call exits. This disablement happens on every
407 read system call exit::
409 # echo 'disable_event:kmem:kmalloc' > \
410 /sys/kernel/debug/tracing/events/syscalls/sys_exit_read/trigger
414 enable_event:<system>:<event>[:count]
415 disable_event:<system>:<event>[:count]
417 To remove the above commands::
419 # echo '!enable_event:kmem:kmalloc:1' > \
420 /sys/kernel/debug/tracing/events/syscalls/sys_enter_read/trigger
422 # echo '!disable_event:kmem:kmalloc' > \
423 /sys/kernel/debug/tracing/events/syscalls/sys_exit_read/trigger
425 Note that there can be any number of enable/disable_event triggers
426 per triggering event, but there can only be one trigger per
427 triggered event. e.g. sys_enter_read can have triggers enabling both
428 kmem:kmalloc and sched:sched_switch, but can't have two kmem:kmalloc
429 versions such as kmem:kmalloc and kmem:kmalloc:1 or 'kmem:kmalloc if
430 bytes_req == 256' and 'kmem:kmalloc if bytes_alloc == 256' (they
431 could be combined into a single filter on kmem:kmalloc though).
435 This command dumps a stacktrace in the trace buffer whenever the
436 triggering event occurs.
438 For example, the following trigger dumps a stacktrace every time the
439 kmalloc tracepoint is hit::
441 # echo 'stacktrace' > \
442 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
444 The following trigger dumps a stacktrace the first 5 times a kmalloc
445 request happens with a size >= 64K::
447 # echo 'stacktrace:5 if bytes_req >= 65536' > \
448 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
454 To remove the above commands::
456 # echo '!stacktrace' > \
457 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
459 # echo '!stacktrace:5 if bytes_req >= 65536' > \
460 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
462 The latter can also be removed more simply by the following (without
465 # echo '!stacktrace:5' > \
466 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger
468 Note that there can be only one stacktrace trigger per triggering
473 This command causes a snapshot to be triggered whenever the
474 triggering event occurs.
476 The following command creates a snapshot every time a block request
477 queue is unplugged with a depth > 1. If you were tracing a set of
478 events or functions at the time, the snapshot trace buffer would
479 capture those events when the trigger event occurred::
481 # echo 'snapshot if nr_rq > 1' > \
482 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
484 To only snapshot once::
486 # echo 'snapshot:1 if nr_rq > 1' > \
487 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
489 To remove the above commands::
491 # echo '!snapshot if nr_rq > 1' > \
492 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
494 # echo '!snapshot:1 if nr_rq > 1' > \
495 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
497 Note that there can be only one snapshot trigger per triggering
502 These commands turn tracing on and off when the specified events are
503 hit. The parameter determines how many times the tracing system is
504 turned on and off. If unspecified, there is no limit.
506 The following command turns tracing off the first time a block
507 request queue is unplugged with a depth > 1. If you were tracing a
508 set of events or functions at the time, you could then examine the
509 trace buffer to see the sequence of events that led up to the
512 # echo 'traceoff:1 if nr_rq > 1' > \
513 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
515 To always disable tracing when nr_rq > 1::
517 # echo 'traceoff if nr_rq > 1' > \
518 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
520 To remove the above commands::
522 # echo '!traceoff:1 if nr_rq > 1' > \
523 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
525 # echo '!traceoff if nr_rq > 1' > \
526 /sys/kernel/debug/tracing/events/block/block_unplug/trigger
528 Note that there can be only one traceon or traceoff trigger per
533 This command aggregates event hits into a hash table keyed on one or
534 more trace event format fields (or stacktrace) and a set of running
535 totals derived from one or more trace event format fields and/or
536 event counts (hitcount).
538 See Documentation/trace/histogram.rst for details and examples.
540 7. In-kernel trace event API
541 ============================
543 In most cases, the command-line interface to trace events is more than
544 sufficient. Sometimes, however, applications might find the need for
545 more complex relationships than can be expressed through a simple
546 series of linked command-line expressions, or putting together sets of
547 commands may be simply too cumbersome. An example might be an
548 application that needs to 'listen' to the trace stream in order to
549 maintain an in-kernel state machine detecting, for instance, when an
550 illegal kernel state occurs in the scheduler.
552 The trace event subsystem provides an in-kernel API allowing modules
553 or other kernel code to generate user-defined 'synthetic' events at
554 will, which can be used to either augment the existing trace stream
555 and/or signal that a particular important state has occurred.
557 A similar in-kernel API is also available for creating kprobe and
560 Both the synthetic event and k/ret/probe event APIs are built on top
561 of a lower-level "dynevent_cmd" event command API, which is also
562 available for more specialized applications, or as the basis of other
563 higher-level trace event APIs.
565 The API provided for these purposes is describe below and allows the
568 - dynamically creating synthetic event definitions
569 - dynamically creating kprobe and kretprobe event definitions
570 - tracing synthetic events from in-kernel code
571 - the low-level "dynevent_cmd" API
573 7.1 Dyamically creating synthetic event definitions
574 ---------------------------------------------------
576 There are a couple ways to create a new synthetic event from a kernel
577 module or other kernel code.
579 The first creates the event in one step, using synth_event_create().
580 In this method, the name of the event to create and an array defining
581 the fields is supplied to synth_event_create(). If successful, a
582 synthetic event with that name and fields will exist following that
583 call. For example, to create a new "schedtest" synthetic event::
585 ret = synth_event_create("schedtest", sched_fields,
586 ARRAY_SIZE(sched_fields), THIS_MODULE);
588 The sched_fields param in this example points to an array of struct
589 synth_field_desc, each of which describes an event field by type and
592 static struct synth_field_desc sched_fields[] = {
593 { .type = "pid_t", .name = "next_pid_field" },
594 { .type = "char[16]", .name = "next_comm_field" },
595 { .type = "u64", .name = "ts_ns" },
596 { .type = "u64", .name = "ts_ms" },
597 { .type = "unsigned int", .name = "cpu" },
598 { .type = "char[64]", .name = "my_string_field" },
599 { .type = "int", .name = "my_int_field" },
602 See synth_field_size() for available types.
604 If field_name contains [n], the field is considered to be a static array.
606 If field_names contains[] (no subscript), the field is considered to
607 be a dynamic array, which will only take as much space in the event as
608 is required to hold the array.
610 Because space for an event is reserved before assigning field values
611 to the event, using dynamic arrays implies that the piecewise
612 in-kernel API described below can't be used with dynamic arrays. The
613 other non-piecewise in-kernel APIs can, however, be used with dynamic
616 If the event is created from within a module, a pointer to the module
617 must be passed to synth_event_create(). This will ensure that the
618 trace buffer won't contain unreadable events when the module is
621 At this point, the event object is ready to be used for generating new
624 In the second method, the event is created in several steps. This
625 allows events to be created dynamically and without the need to create
626 and populate an array of fields beforehand.
628 To use this method, an empty or partially empty synthetic event should
629 first be created using synth_event_gen_cmd_start() or
630 synth_event_gen_cmd_array_start(). For synth_event_gen_cmd_start(),
631 the name of the event along with one or more pairs of args each pair
632 representing a 'type field_name;' field specification should be
633 supplied. For synth_event_gen_cmd_array_start(), the name of the
634 event along with an array of struct synth_field_desc should be
635 supplied. Before calling synth_event_gen_cmd_start() or
636 synth_event_gen_cmd_array_start(), the user should create and
637 initialize a dynevent_cmd object using synth_event_cmd_init().
639 For example, to create a new "schedtest" synthetic event with two
642 struct dynevent_cmd cmd;
645 /* Create a buffer to hold the generated command */
646 buf = kzalloc(MAX_DYNEVENT_CMD_LEN, GFP_KERNEL);
648 /* Before generating the command, initialize the cmd object */
649 synth_event_cmd_init(&cmd, buf, MAX_DYNEVENT_CMD_LEN);
651 ret = synth_event_gen_cmd_start(&cmd, "schedtest", THIS_MODULE,
652 "pid_t", "next_pid_field",
655 Alternatively, using an array of struct synth_field_desc fields
656 containing the same information::
658 ret = synth_event_gen_cmd_array_start(&cmd, "schedtest", THIS_MODULE,
661 Once the synthetic event object has been created, it can then be
662 populated with more fields. Fields are added one by one using
663 synth_event_add_field(), supplying the dynevent_cmd object, a field
664 type, and a field name. For example, to add a new int field named
665 "intfield", the following call should be made::
667 ret = synth_event_add_field(&cmd, "int", "intfield");
669 See synth_field_size() for available types. If field_name contains [n]
670 the field is considered to be an array.
672 A group of fields can also be added all at once using an array of
673 synth_field_desc with add_synth_fields(). For example, this would add
674 just the first four sched_fields::
676 ret = synth_event_add_fields(&cmd, sched_fields, 4);
678 If you already have a string of the form 'type field_name',
679 synth_event_add_field_str() can be used to add it as-is; it will
680 also automatically append a ';' to the string.
682 Once all the fields have been added, the event should be finalized and
683 registered by calling the synth_event_gen_cmd_end() function::
685 ret = synth_event_gen_cmd_end(&cmd);
687 At this point, the event object is ready to be used for tracing new
690 7.2 Tracing synthetic events from in-kernel code
691 ------------------------------------------------
693 To trace a synthetic event, there are several options. The first
694 option is to trace the event in one call, using synth_event_trace()
695 with a variable number of values, or synth_event_trace_array() with an
696 array of values to be set. A second option can be used to avoid the
697 need for a pre-formed array of values or list of arguments, via
698 synth_event_trace_start() and synth_event_trace_end() along with
699 synth_event_add_next_val() or synth_event_add_val() to add the values
702 7.2.1 Tracing a synthetic event all at once
703 -------------------------------------------
705 To trace a synthetic event all at once, the synth_event_trace() or
706 synth_event_trace_array() functions can be used.
708 The synth_event_trace() function is passed the trace_event_file
709 representing the synthetic event (which can be retrieved using
710 trace_get_event_file() using the synthetic event name, "synthetic" as
711 the system name, and the trace instance name (NULL if using the global
712 trace array)), along with an variable number of u64 args, one for each
713 synthetic event field, and the number of values being passed.
715 So, to trace an event corresponding to the synthetic event definition
716 above, code like the following could be used::
718 ret = synth_event_trace(create_synth_test, 7, /* number of values */
719 444, /* next_pid_field */
720 (u64)"clackers", /* next_comm_field */
723 smp_processor_id(),/* cpu */
724 (u64)"Thneed", /* my_string_field */
725 999); /* my_int_field */
727 All vals should be cast to u64, and string vals are just pointers to
728 strings, cast to u64. Strings will be copied into space reserved in
729 the event for the string, using these pointers.
731 Alternatively, the synth_event_trace_array() function can be used to
732 accomplish the same thing. It is passed the trace_event_file
733 representing the synthetic event (which can be retrieved using
734 trace_get_event_file() using the synthetic event name, "synthetic" as
735 the system name, and the trace instance name (NULL if using the global
736 trace array)), along with an array of u64, one for each synthetic
739 To trace an event corresponding to the synthetic event definition
740 above, code like the following could be used::
744 vals[0] = 777; /* next_pid_field */
745 vals[1] = (u64)"tiddlywinks"; /* next_comm_field */
746 vals[2] = 1000000; /* ts_ns */
747 vals[3] = 1000; /* ts_ms */
748 vals[4] = smp_processor_id(); /* cpu */
749 vals[5] = (u64)"thneed"; /* my_string_field */
750 vals[6] = 398; /* my_int_field */
752 The 'vals' array is just an array of u64, the number of which must
753 match the number of field in the synthetic event, and which must be in
754 the same order as the synthetic event fields.
756 All vals should be cast to u64, and string vals are just pointers to
757 strings, cast to u64. Strings will be copied into space reserved in
758 the event for the string, using these pointers.
760 In order to trace a synthetic event, a pointer to the trace event file
761 is needed. The trace_get_event_file() function can be used to get
762 it - it will find the file in the given trace instance (in this case
763 NULL since the top trace array is being used) while at the same time
764 preventing the instance containing it from going away::
766 schedtest_event_file = trace_get_event_file(NULL, "synthetic",
769 Before tracing the event, it should be enabled in some way, otherwise
770 the synthetic event won't actually show up in the trace buffer.
772 To enable a synthetic event from the kernel, trace_array_set_clr_event()
773 can be used (which is not specific to synthetic events, so does need
774 the "synthetic" system name to be specified explicitly).
776 To enable the event, pass 'true' to it::
778 trace_array_set_clr_event(schedtest_event_file->tr,
779 "synthetic", "schedtest", true);
781 To disable it pass false::
783 trace_array_set_clr_event(schedtest_event_file->tr,
784 "synthetic", "schedtest", false);
786 Finally, synth_event_trace_array() can be used to actually trace the
787 event, which should be visible in the trace buffer afterwards::
789 ret = synth_event_trace_array(schedtest_event_file, vals,
792 To remove the synthetic event, the event should be disabled, and the
793 trace instance should be 'put' back using trace_put_event_file()::
795 trace_array_set_clr_event(schedtest_event_file->tr,
796 "synthetic", "schedtest", false);
797 trace_put_event_file(schedtest_event_file);
799 If those have been successful, synth_event_delete() can be called to
802 ret = synth_event_delete("schedtest");
804 7.2.2 Tracing a synthetic event piecewise
805 -----------------------------------------
807 To trace a synthetic using the piecewise method described above, the
808 synth_event_trace_start() function is used to 'open' the synthetic
811 struct synth_event_trace_state trace_state;
813 ret = synth_event_trace_start(schedtest_event_file, &trace_state);
815 It's passed the trace_event_file representing the synthetic event
816 using the same methods as described above, along with a pointer to a
817 struct synth_event_trace_state object, which will be zeroed before use and
818 used to maintain state between this and following calls.
820 Once the event has been opened, which means space for it has been
821 reserved in the trace buffer, the individual fields can be set. There
822 are two ways to do that, either one after another for each field in
823 the event, which requires no lookups, or by name, which does. The
824 tradeoff is flexibility in doing the assignments vs the cost of a
827 To assign the values one after the other without lookups,
828 synth_event_add_next_val() should be used. Each call is passed the
829 same synth_event_trace_state object used in the synth_event_trace_start(),
830 along with the value to set the next field in the event. After each
831 field is set, the 'cursor' points to the next field, which will be set
832 by the subsequent call, continuing until all the fields have been set
833 in order. The same sequence of calls as in the above examples using
834 this method would be (without error-handling code)::
837 ret = synth_event_add_next_val(777, &trace_state);
839 /* next_comm_field */
840 ret = synth_event_add_next_val((u64)"slinky", &trace_state);
843 ret = synth_event_add_next_val(1000000, &trace_state);
846 ret = synth_event_add_next_val(1000, &trace_state);
849 ret = synth_event_add_next_val(smp_processor_id(), &trace_state);
851 /* my_string_field */
852 ret = synth_event_add_next_val((u64)"thneed_2.01", &trace_state);
855 ret = synth_event_add_next_val(395, &trace_state);
857 To assign the values in any order, synth_event_add_val() should be
858 used. Each call is passed the same synth_event_trace_state object used in
859 the synth_event_trace_start(), along with the field name of the field
860 to set and the value to set it to. The same sequence of calls as in
861 the above examples using this method would be (without error-handling
864 ret = synth_event_add_val("next_pid_field", 777, &trace_state);
865 ret = synth_event_add_val("next_comm_field", (u64)"silly putty",
867 ret = synth_event_add_val("ts_ns", 1000000, &trace_state);
868 ret = synth_event_add_val("ts_ms", 1000, &trace_state);
869 ret = synth_event_add_val("cpu", smp_processor_id(), &trace_state);
870 ret = synth_event_add_val("my_string_field", (u64)"thneed_9",
872 ret = synth_event_add_val("my_int_field", 3999, &trace_state);
874 Note that synth_event_add_next_val() and synth_event_add_val() are
875 incompatible if used within the same trace of an event - either one
876 can be used but not both at the same time.
878 Finally, the event won't be actually traced until it's 'closed',
879 which is done using synth_event_trace_end(), which takes only the
880 struct synth_event_trace_state object used in the previous calls::
882 ret = synth_event_trace_end(&trace_state);
884 Note that synth_event_trace_end() must be called at the end regardless
885 of whether any of the add calls failed (say due to a bad field name
888 7.3 Dyamically creating kprobe and kretprobe event definitions
889 --------------------------------------------------------------
891 To create a kprobe or kretprobe trace event from kernel code, the
892 kprobe_event_gen_cmd_start() or kretprobe_event_gen_cmd_start()
893 functions can be used.
895 To create a kprobe event, an empty or partially empty kprobe event
896 should first be created using kprobe_event_gen_cmd_start(). The name
897 of the event and the probe location should be specfied along with one
898 or args each representing a probe field should be supplied to this
899 function. Before calling kprobe_event_gen_cmd_start(), the user
900 should create and initialize a dynevent_cmd object using
901 kprobe_event_cmd_init().
903 For example, to create a new "schedtest" kprobe event with two fields::
905 struct dynevent_cmd cmd;
908 /* Create a buffer to hold the generated command */
909 buf = kzalloc(MAX_DYNEVENT_CMD_LEN, GFP_KERNEL);
911 /* Before generating the command, initialize the cmd object */
912 kprobe_event_cmd_init(&cmd, buf, MAX_DYNEVENT_CMD_LEN);
915 * Define the gen_kprobe_test event with the first 2 kprobe
918 ret = kprobe_event_gen_cmd_start(&cmd, "gen_kprobe_test", "do_sys_open",
919 "dfd=%ax", "filename=%dx");
921 Once the kprobe event object has been created, it can then be
922 populated with more fields. Fields can be added using
923 kprobe_event_add_fields(), supplying the dynevent_cmd object along
924 with a variable arg list of probe fields. For example, to add a
925 couple additional fields, the following call could be made::
927 ret = kprobe_event_add_fields(&cmd, "flags=%cx", "mode=+4($stack)");
929 Once all the fields have been added, the event should be finalized and
930 registered by calling the kprobe_event_gen_cmd_end() or
931 kretprobe_event_gen_cmd_end() functions, depending on whether a kprobe
932 or kretprobe command was started::
934 ret = kprobe_event_gen_cmd_end(&cmd);
938 ret = kretprobe_event_gen_cmd_end(&cmd);
940 At this point, the event object is ready to be used for tracing new
943 Similarly, a kretprobe event can be created using
944 kretprobe_event_gen_cmd_start() with a probe name and location and
945 additional params such as $retval::
947 ret = kretprobe_event_gen_cmd_start(&cmd, "gen_kretprobe_test",
948 "do_sys_open", "$retval");
950 Similar to the synthetic event case, code like the following can be
951 used to enable the newly created kprobe event::
953 gen_kprobe_test = trace_get_event_file(NULL, "kprobes", "gen_kprobe_test");
955 ret = trace_array_set_clr_event(gen_kprobe_test->tr,
956 "kprobes", "gen_kprobe_test", true);
958 Finally, also similar to synthetic events, the following code can be
959 used to give the kprobe event file back and delete the event::
961 trace_put_event_file(gen_kprobe_test);
963 ret = kprobe_event_delete("gen_kprobe_test");
965 7.4 The "dynevent_cmd" low-level API
966 ------------------------------------
968 Both the in-kernel synthetic event and kprobe interfaces are built on
969 top of a lower-level "dynevent_cmd" interface. This interface is
970 meant to provide the basis for higher-level interfaces such as the
971 synthetic and kprobe interfaces, which can be used as examples.
973 The basic idea is simple and amounts to providing a general-purpose
974 layer that can be used to generate trace event commands. The
975 generated command strings can then be passed to the command-parsing
976 and event creation code that already exists in the trace event
977 subystem for creating the corresponding trace events.
979 In a nutshell, the way it works is that the higher-level interface
980 code creates a struct dynevent_cmd object, then uses a couple
981 functions, dynevent_arg_add() and dynevent_arg_pair_add() to build up
982 a command string, which finally causes the command to be executed
983 using the dynevent_create() function. The details of the interface
986 The first step in building a new command string is to create and
987 initialize an instance of a dynevent_cmd. Here, for instance, we
988 create a dynevent_cmd on the stack and initialize it::
990 struct dynevent_cmd cmd;
994 buf = kzalloc(MAX_DYNEVENT_CMD_LEN, GFP_KERNEL);
996 dynevent_cmd_init(cmd, buf, maxlen, DYNEVENT_TYPE_FOO,
997 foo_event_run_command);
999 The dynevent_cmd initialization needs to be given a user-specified
1000 buffer and the length of the buffer (MAX_DYNEVENT_CMD_LEN can be used
1001 for this purpose - at 2k it's generally too big to be comfortably put
1002 on the stack, so is dynamically allocated), a dynevent type id, which
1003 is meant to be used to check that further API calls are for the
1004 correct command type, and a pointer to an event-specific run_command()
1005 callback that will be called to actually execute the event-specific
1008 Once that's done, the command string can by built up by successive
1009 calls to argument-adding functions.
1011 To add a single argument, define and initialize a struct dynevent_arg
1012 or struct dynevent_arg_pair object. Here's an example of the simplest
1013 possible arg addition, which is simply to append the given string as
1014 a whitespace-separated argument to the command::
1016 struct dynevent_arg arg;
1018 dynevent_arg_init(&arg, NULL, 0);
1022 ret = dynevent_arg_add(cmd, &arg);
1024 The arg object is first initialized using dynevent_arg_init() and in
1025 this case the parameters are NULL or 0, which means there's no
1026 optional sanity-checking function or separator appended to the end of
1029 Here's another more complicated example using an 'arg pair', which is
1030 used to create an argument that consists of a couple components added
1031 together as a unit, for example, a 'type field_name;' arg or a simple
1032 expression arg e.g. 'flags=%cx'::
1034 struct dynevent_arg_pair arg_pair;
1036 dynevent_arg_pair_init(&arg_pair, dynevent_foo_check_arg_fn, 0, ';');
1038 arg_pair.lhs = type;
1039 arg_pair.rhs = name;
1041 ret = dynevent_arg_pair_add(cmd, &arg_pair);
1043 Again, the arg_pair is first initialized, in this case with a callback
1044 function used to check the sanity of the args (for example, that
1045 neither part of the pair is NULL), along with a character to be used
1046 to add an operator between the pair (here none) and a separator to be
1047 appended onto the end of the arg pair (here ';').
1049 There's also a dynevent_str_add() function that can be used to simply
1050 add a string as-is, with no spaces, delimeters, or arg check.
1052 Any number of dynevent_*_add() calls can be made to build up the string
1053 (until its length surpasses cmd->maxlen). When all the arguments have
1054 been added and the command string is complete, the only thing left to
1055 do is run the command, which happens by simply calling
1058 ret = dynevent_create(&cmd);
1060 At that point, if the return value is 0, the dynamic event has been
1061 created and is ready to use.
1063 See the dynevent_cmd function definitions themselves for the details