4 * Copyright (C) 2008 Red Hat Inc, Steven Rostedt <srostedt@redhat.com>
6 * - Added format output of fields of the trace point.
7 * This was based off of work by Tom Zanussi <tzanussi@gmail.com>.
11 #define pr_fmt(fmt) fmt
13 #include <linux/workqueue.h>
14 #include <linux/spinlock.h>
15 #include <linux/kthread.h>
16 #include <linux/tracefs.h>
17 #include <linux/uaccess.h>
18 #include <linux/module.h>
19 #include <linux/ctype.h>
20 #include <linux/slab.h>
21 #include <linux/delay.h>
23 #include <asm/setup.h>
25 #include "trace_output.h"
28 #define TRACE_SYSTEM "TRACE_SYSTEM"
30 DEFINE_MUTEX(event_mutex
);
32 LIST_HEAD(ftrace_events
);
33 static LIST_HEAD(ftrace_common_fields
);
35 #define GFP_TRACE (GFP_KERNEL | __GFP_ZERO)
37 static struct kmem_cache
*field_cachep
;
38 static struct kmem_cache
*file_cachep
;
40 #define SYSTEM_FL_FREE_NAME (1 << 31)
42 static inline int system_refcount(struct event_subsystem
*system
)
44 return system
->ref_count
& ~SYSTEM_FL_FREE_NAME
;
47 static int system_refcount_inc(struct event_subsystem
*system
)
49 return (system
->ref_count
++) & ~SYSTEM_FL_FREE_NAME
;
52 static int system_refcount_dec(struct event_subsystem
*system
)
54 return (--system
->ref_count
) & ~SYSTEM_FL_FREE_NAME
;
57 /* Double loops, do not use break, only goto's work */
58 #define do_for_each_event_file(tr, file) \
59 list_for_each_entry(tr, &ftrace_trace_arrays, list) { \
60 list_for_each_entry(file, &tr->events, list)
62 #define do_for_each_event_file_safe(tr, file) \
63 list_for_each_entry(tr, &ftrace_trace_arrays, list) { \
64 struct trace_event_file *___n; \
65 list_for_each_entry_safe(file, ___n, &tr->events, list)
67 #define while_for_each_event_file() \
70 static struct list_head
*
71 trace_get_fields(struct trace_event_call
*event_call
)
73 if (!event_call
->class->get_fields
)
74 return &event_call
->class->fields
;
75 return event_call
->class->get_fields(event_call
);
78 static struct ftrace_event_field
*
79 __find_event_field(struct list_head
*head
, char *name
)
81 struct ftrace_event_field
*field
;
83 list_for_each_entry(field
, head
, link
) {
84 if (!strcmp(field
->name
, name
))
91 struct ftrace_event_field
*
92 trace_find_event_field(struct trace_event_call
*call
, char *name
)
94 struct ftrace_event_field
*field
;
95 struct list_head
*head
;
97 field
= __find_event_field(&ftrace_common_fields
, name
);
101 head
= trace_get_fields(call
);
102 return __find_event_field(head
, name
);
105 static int __trace_define_field(struct list_head
*head
, const char *type
,
106 const char *name
, int offset
, int size
,
107 int is_signed
, int filter_type
)
109 struct ftrace_event_field
*field
;
111 field
= kmem_cache_alloc(field_cachep
, GFP_TRACE
);
118 if (filter_type
== FILTER_OTHER
)
119 field
->filter_type
= filter_assign_type(type
);
121 field
->filter_type
= filter_type
;
123 field
->offset
= offset
;
125 field
->is_signed
= is_signed
;
127 list_add(&field
->link
, head
);
132 int trace_define_field(struct trace_event_call
*call
, const char *type
,
133 const char *name
, int offset
, int size
, int is_signed
,
136 struct list_head
*head
;
138 if (WARN_ON(!call
->class))
141 head
= trace_get_fields(call
);
142 return __trace_define_field(head
, type
, name
, offset
, size
,
143 is_signed
, filter_type
);
145 EXPORT_SYMBOL_GPL(trace_define_field
);
147 #define __common_field(type, item) \
148 ret = __trace_define_field(&ftrace_common_fields, #type, \
150 offsetof(typeof(ent), item), \
152 is_signed_type(type), FILTER_OTHER); \
156 static int trace_define_common_fields(void)
159 struct trace_entry ent
;
161 __common_field(unsigned short, type
);
162 __common_field(unsigned char, flags
);
163 __common_field(unsigned char, preempt_count
);
164 __common_field(int, pid
);
169 static void trace_destroy_fields(struct trace_event_call
*call
)
171 struct ftrace_event_field
*field
, *next
;
172 struct list_head
*head
;
174 head
= trace_get_fields(call
);
175 list_for_each_entry_safe(field
, next
, head
, link
) {
176 list_del(&field
->link
);
177 kmem_cache_free(field_cachep
, field
);
181 int trace_event_raw_init(struct trace_event_call
*call
)
185 id
= register_trace_event(&call
->event
);
191 EXPORT_SYMBOL_GPL(trace_event_raw_init
);
193 void *trace_event_buffer_reserve(struct trace_event_buffer
*fbuffer
,
194 struct trace_event_file
*trace_file
,
197 struct trace_event_call
*event_call
= trace_file
->event_call
;
199 local_save_flags(fbuffer
->flags
);
200 fbuffer
->pc
= preempt_count();
201 fbuffer
->trace_file
= trace_file
;
204 trace_event_buffer_lock_reserve(&fbuffer
->buffer
, trace_file
,
205 event_call
->event
.type
, len
,
206 fbuffer
->flags
, fbuffer
->pc
);
210 fbuffer
->entry
= ring_buffer_event_data(fbuffer
->event
);
211 return fbuffer
->entry
;
213 EXPORT_SYMBOL_GPL(trace_event_buffer_reserve
);
215 static DEFINE_SPINLOCK(tracepoint_iter_lock
);
217 static void output_printk(struct trace_event_buffer
*fbuffer
)
219 struct trace_event_call
*event_call
;
220 struct trace_event
*event
;
222 struct trace_iterator
*iter
= tracepoint_print_iter
;
227 event_call
= fbuffer
->trace_file
->event_call
;
228 if (!event_call
|| !event_call
->event
.funcs
||
229 !event_call
->event
.funcs
->trace
)
232 event
= &fbuffer
->trace_file
->event_call
->event
;
234 spin_lock_irqsave(&tracepoint_iter_lock
, flags
);
235 trace_seq_init(&iter
->seq
);
236 iter
->ent
= fbuffer
->entry
;
237 event_call
->event
.funcs
->trace(iter
, 0, event
);
238 trace_seq_putc(&iter
->seq
, 0);
239 printk("%s", iter
->seq
.buffer
);
241 spin_unlock_irqrestore(&tracepoint_iter_lock
, flags
);
244 void trace_event_buffer_commit(struct trace_event_buffer
*fbuffer
)
246 if (tracepoint_printk
)
247 output_printk(fbuffer
);
249 event_trigger_unlock_commit(fbuffer
->trace_file
, fbuffer
->buffer
,
250 fbuffer
->event
, fbuffer
->entry
,
251 fbuffer
->flags
, fbuffer
->pc
);
253 EXPORT_SYMBOL_GPL(trace_event_buffer_commit
);
255 int trace_event_reg(struct trace_event_call
*call
,
256 enum trace_reg type
, void *data
)
258 struct trace_event_file
*file
= data
;
260 WARN_ON(!(call
->flags
& TRACE_EVENT_FL_TRACEPOINT
));
262 case TRACE_REG_REGISTER
:
263 return tracepoint_probe_register(call
->tp
,
266 case TRACE_REG_UNREGISTER
:
267 tracepoint_probe_unregister(call
->tp
,
272 #ifdef CONFIG_PERF_EVENTS
273 case TRACE_REG_PERF_REGISTER
:
274 return tracepoint_probe_register(call
->tp
,
275 call
->class->perf_probe
,
277 case TRACE_REG_PERF_UNREGISTER
:
278 tracepoint_probe_unregister(call
->tp
,
279 call
->class->perf_probe
,
282 case TRACE_REG_PERF_OPEN
:
283 case TRACE_REG_PERF_CLOSE
:
284 case TRACE_REG_PERF_ADD
:
285 case TRACE_REG_PERF_DEL
:
291 EXPORT_SYMBOL_GPL(trace_event_reg
);
293 void trace_event_enable_cmd_record(bool enable
)
295 struct trace_event_file
*file
;
296 struct trace_array
*tr
;
298 mutex_lock(&event_mutex
);
299 do_for_each_event_file(tr
, file
) {
301 if (!(file
->flags
& EVENT_FILE_FL_ENABLED
))
305 tracing_start_cmdline_record();
306 set_bit(EVENT_FILE_FL_RECORDED_CMD_BIT
, &file
->flags
);
308 tracing_stop_cmdline_record();
309 clear_bit(EVENT_FILE_FL_RECORDED_CMD_BIT
, &file
->flags
);
311 } while_for_each_event_file();
312 mutex_unlock(&event_mutex
);
315 static int __ftrace_event_enable_disable(struct trace_event_file
*file
,
316 int enable
, int soft_disable
)
318 struct trace_event_call
*call
= file
->event_call
;
325 * When soft_disable is set and enable is cleared, the sm_ref
326 * reference counter is decremented. If it reaches 0, we want
327 * to clear the SOFT_DISABLED flag but leave the event in the
328 * state that it was. That is, if the event was enabled and
329 * SOFT_DISABLED isn't set, then do nothing. But if SOFT_DISABLED
330 * is set we do not want the event to be enabled before we
333 * When soft_disable is not set but the SOFT_MODE flag is,
334 * we do nothing. Do not disable the tracepoint, otherwise
335 * "soft enable"s (clearing the SOFT_DISABLED bit) wont work.
338 if (atomic_dec_return(&file
->sm_ref
) > 0)
340 disable
= file
->flags
& EVENT_FILE_FL_SOFT_DISABLED
;
341 clear_bit(EVENT_FILE_FL_SOFT_MODE_BIT
, &file
->flags
);
343 disable
= !(file
->flags
& EVENT_FILE_FL_SOFT_MODE
);
345 if (disable
&& (file
->flags
& EVENT_FILE_FL_ENABLED
)) {
346 clear_bit(EVENT_FILE_FL_ENABLED_BIT
, &file
->flags
);
347 if (file
->flags
& EVENT_FILE_FL_RECORDED_CMD
) {
348 tracing_stop_cmdline_record();
349 clear_bit(EVENT_FILE_FL_RECORDED_CMD_BIT
, &file
->flags
);
351 call
->class->reg(call
, TRACE_REG_UNREGISTER
, file
);
353 /* If in SOFT_MODE, just set the SOFT_DISABLE_BIT, else clear it */
354 if (file
->flags
& EVENT_FILE_FL_SOFT_MODE
)
355 set_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &file
->flags
);
357 clear_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &file
->flags
);
361 * When soft_disable is set and enable is set, we want to
362 * register the tracepoint for the event, but leave the event
363 * as is. That means, if the event was already enabled, we do
364 * nothing (but set SOFT_MODE). If the event is disabled, we
365 * set SOFT_DISABLED before enabling the event tracepoint, so
366 * it still seems to be disabled.
369 clear_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &file
->flags
);
371 if (atomic_inc_return(&file
->sm_ref
) > 1)
373 set_bit(EVENT_FILE_FL_SOFT_MODE_BIT
, &file
->flags
);
376 if (!(file
->flags
& EVENT_FILE_FL_ENABLED
)) {
378 /* Keep the event disabled, when going to SOFT_MODE. */
380 set_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &file
->flags
);
382 if (trace_flags
& TRACE_ITER_RECORD_CMD
) {
383 tracing_start_cmdline_record();
384 set_bit(EVENT_FILE_FL_RECORDED_CMD_BIT
, &file
->flags
);
386 ret
= call
->class->reg(call
, TRACE_REG_REGISTER
, file
);
388 tracing_stop_cmdline_record();
389 pr_info("event trace: Could not enable event "
390 "%s\n", trace_event_name(call
));
393 set_bit(EVENT_FILE_FL_ENABLED_BIT
, &file
->flags
);
395 /* WAS_ENABLED gets set but never cleared. */
396 call
->flags
|= TRACE_EVENT_FL_WAS_ENABLED
;
404 int trace_event_enable_disable(struct trace_event_file
*file
,
405 int enable
, int soft_disable
)
407 return __ftrace_event_enable_disable(file
, enable
, soft_disable
);
410 static int ftrace_event_enable_disable(struct trace_event_file
*file
,
413 return __ftrace_event_enable_disable(file
, enable
, 0);
416 static void ftrace_clear_events(struct trace_array
*tr
)
418 struct trace_event_file
*file
;
420 mutex_lock(&event_mutex
);
421 list_for_each_entry(file
, &tr
->events
, list
) {
422 ftrace_event_enable_disable(file
, 0);
424 mutex_unlock(&event_mutex
);
427 static void __put_system(struct event_subsystem
*system
)
429 struct event_filter
*filter
= system
->filter
;
431 WARN_ON_ONCE(system_refcount(system
) == 0);
432 if (system_refcount_dec(system
))
435 list_del(&system
->list
);
438 kfree(filter
->filter_string
);
441 if (system
->ref_count
& SYSTEM_FL_FREE_NAME
)
446 static void __get_system(struct event_subsystem
*system
)
448 WARN_ON_ONCE(system_refcount(system
) == 0);
449 system_refcount_inc(system
);
452 static void __get_system_dir(struct trace_subsystem_dir
*dir
)
454 WARN_ON_ONCE(dir
->ref_count
== 0);
456 __get_system(dir
->subsystem
);
459 static void __put_system_dir(struct trace_subsystem_dir
*dir
)
461 WARN_ON_ONCE(dir
->ref_count
== 0);
462 /* If the subsystem is about to be freed, the dir must be too */
463 WARN_ON_ONCE(system_refcount(dir
->subsystem
) == 1 && dir
->ref_count
!= 1);
465 __put_system(dir
->subsystem
);
466 if (!--dir
->ref_count
)
470 static void put_system(struct trace_subsystem_dir
*dir
)
472 mutex_lock(&event_mutex
);
473 __put_system_dir(dir
);
474 mutex_unlock(&event_mutex
);
477 static void remove_subsystem(struct trace_subsystem_dir
*dir
)
482 if (!--dir
->nr_events
) {
483 tracefs_remove_recursive(dir
->entry
);
484 list_del(&dir
->list
);
485 __put_system_dir(dir
);
489 static void remove_event_file_dir(struct trace_event_file
*file
)
491 struct dentry
*dir
= file
->dir
;
492 struct dentry
*child
;
495 spin_lock(&dir
->d_lock
); /* probably unneeded */
496 list_for_each_entry(child
, &dir
->d_subdirs
, d_child
) {
497 if (d_really_is_positive(child
)) /* probably unneeded */
498 d_inode(child
)->i_private
= NULL
;
500 spin_unlock(&dir
->d_lock
);
502 tracefs_remove_recursive(dir
);
505 list_del(&file
->list
);
506 remove_subsystem(file
->system
);
507 free_event_filter(file
->filter
);
508 kmem_cache_free(file_cachep
, file
);
512 * __ftrace_set_clr_event(NULL, NULL, NULL, set) will set/unset all events.
515 __ftrace_set_clr_event_nolock(struct trace_array
*tr
, const char *match
,
516 const char *sub
, const char *event
, int set
)
518 struct trace_event_file
*file
;
519 struct trace_event_call
*call
;
523 list_for_each_entry(file
, &tr
->events
, list
) {
525 call
= file
->event_call
;
526 name
= trace_event_name(call
);
528 if (!name
|| !call
->class || !call
->class->reg
)
531 if (call
->flags
& TRACE_EVENT_FL_IGNORE_ENABLE
)
535 strcmp(match
, name
) != 0 &&
536 strcmp(match
, call
->class->system
) != 0)
539 if (sub
&& strcmp(sub
, call
->class->system
) != 0)
542 if (event
&& strcmp(event
, name
) != 0)
545 ftrace_event_enable_disable(file
, set
);
553 static int __ftrace_set_clr_event(struct trace_array
*tr
, const char *match
,
554 const char *sub
, const char *event
, int set
)
558 mutex_lock(&event_mutex
);
559 ret
= __ftrace_set_clr_event_nolock(tr
, match
, sub
, event
, set
);
560 mutex_unlock(&event_mutex
);
565 static int ftrace_set_clr_event(struct trace_array
*tr
, char *buf
, int set
)
567 char *event
= NULL
, *sub
= NULL
, *match
;
571 * The buf format can be <subsystem>:<event-name>
572 * *:<event-name> means any event by that name.
573 * :<event-name> is the same.
575 * <subsystem>:* means all events in that subsystem
576 * <subsystem>: means the same.
578 * <name> (no ':') means all events in a subsystem with
579 * the name <name> or any event that matches <name>
582 match
= strsep(&buf
, ":");
588 if (!strlen(sub
) || strcmp(sub
, "*") == 0)
590 if (!strlen(event
) || strcmp(event
, "*") == 0)
594 ret
= __ftrace_set_clr_event(tr
, match
, sub
, event
, set
);
596 /* Put back the colon to allow this to be called again */
604 * trace_set_clr_event - enable or disable an event
605 * @system: system name to match (NULL for any system)
606 * @event: event name to match (NULL for all events, within system)
607 * @set: 1 to enable, 0 to disable
609 * This is a way for other parts of the kernel to enable or disable
612 * Returns 0 on success, -EINVAL if the parameters do not match any
615 int trace_set_clr_event(const char *system
, const char *event
, int set
)
617 struct trace_array
*tr
= top_trace_array();
622 return __ftrace_set_clr_event(tr
, NULL
, system
, event
, set
);
624 EXPORT_SYMBOL_GPL(trace_set_clr_event
);
626 /* 128 should be much more than enough */
627 #define EVENT_BUF_SIZE 127
630 ftrace_event_write(struct file
*file
, const char __user
*ubuf
,
631 size_t cnt
, loff_t
*ppos
)
633 struct trace_parser parser
;
634 struct seq_file
*m
= file
->private_data
;
635 struct trace_array
*tr
= m
->private;
641 ret
= tracing_update_buffers();
645 if (trace_parser_get_init(&parser
, EVENT_BUF_SIZE
+ 1))
648 read
= trace_get_user(&parser
, ubuf
, cnt
, ppos
);
650 if (read
>= 0 && trace_parser_loaded((&parser
))) {
653 if (*parser
.buffer
== '!')
656 parser
.buffer
[parser
.idx
] = 0;
658 ret
= ftrace_set_clr_event(tr
, parser
.buffer
+ !set
, set
);
666 trace_parser_put(&parser
);
672 t_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
674 struct trace_event_file
*file
= v
;
675 struct trace_event_call
*call
;
676 struct trace_array
*tr
= m
->private;
680 list_for_each_entry_continue(file
, &tr
->events
, list
) {
681 call
= file
->event_call
;
683 * The ftrace subsystem is for showing formats only.
684 * They can not be enabled or disabled via the event files.
686 if (call
->class && call
->class->reg
)
693 static void *t_start(struct seq_file
*m
, loff_t
*pos
)
695 struct trace_event_file
*file
;
696 struct trace_array
*tr
= m
->private;
699 mutex_lock(&event_mutex
);
701 file
= list_entry(&tr
->events
, struct trace_event_file
, list
);
702 for (l
= 0; l
<= *pos
; ) {
703 file
= t_next(m
, file
, &l
);
711 s_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
713 struct trace_event_file
*file
= v
;
714 struct trace_array
*tr
= m
->private;
718 list_for_each_entry_continue(file
, &tr
->events
, list
) {
719 if (file
->flags
& EVENT_FILE_FL_ENABLED
)
726 static void *s_start(struct seq_file
*m
, loff_t
*pos
)
728 struct trace_event_file
*file
;
729 struct trace_array
*tr
= m
->private;
732 mutex_lock(&event_mutex
);
734 file
= list_entry(&tr
->events
, struct trace_event_file
, list
);
735 for (l
= 0; l
<= *pos
; ) {
736 file
= s_next(m
, file
, &l
);
743 static int t_show(struct seq_file
*m
, void *v
)
745 struct trace_event_file
*file
= v
;
746 struct trace_event_call
*call
= file
->event_call
;
748 if (strcmp(call
->class->system
, TRACE_SYSTEM
) != 0)
749 seq_printf(m
, "%s:", call
->class->system
);
750 seq_printf(m
, "%s\n", trace_event_name(call
));
755 static void t_stop(struct seq_file
*m
, void *p
)
757 mutex_unlock(&event_mutex
);
761 event_enable_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
,
764 struct trace_event_file
*file
;
768 mutex_lock(&event_mutex
);
769 file
= event_file_data(filp
);
772 mutex_unlock(&event_mutex
);
777 if (flags
& EVENT_FILE_FL_ENABLED
&&
778 !(flags
& EVENT_FILE_FL_SOFT_DISABLED
))
781 if (flags
& EVENT_FILE_FL_SOFT_DISABLED
||
782 flags
& EVENT_FILE_FL_SOFT_MODE
)
787 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, strlen(buf
));
791 event_enable_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
794 struct trace_event_file
*file
;
798 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
802 ret
= tracing_update_buffers();
810 mutex_lock(&event_mutex
);
811 file
= event_file_data(filp
);
813 ret
= ftrace_event_enable_disable(file
, val
);
814 mutex_unlock(&event_mutex
);
823 return ret
? ret
: cnt
;
827 system_enable_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
,
830 const char set_to_char
[4] = { '?', '0', '1', 'X' };
831 struct trace_subsystem_dir
*dir
= filp
->private_data
;
832 struct event_subsystem
*system
= dir
->subsystem
;
833 struct trace_event_call
*call
;
834 struct trace_event_file
*file
;
835 struct trace_array
*tr
= dir
->tr
;
840 mutex_lock(&event_mutex
);
841 list_for_each_entry(file
, &tr
->events
, list
) {
842 call
= file
->event_call
;
843 if (!trace_event_name(call
) || !call
->class || !call
->class->reg
)
846 if (system
&& strcmp(call
->class->system
, system
->name
) != 0)
850 * We need to find out if all the events are set
851 * or if all events or cleared, or if we have
854 set
|= (1 << !!(file
->flags
& EVENT_FILE_FL_ENABLED
));
857 * If we have a mixture, no need to look further.
862 mutex_unlock(&event_mutex
);
864 buf
[0] = set_to_char
[set
];
867 ret
= simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, 2);
873 system_enable_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
876 struct trace_subsystem_dir
*dir
= filp
->private_data
;
877 struct event_subsystem
*system
= dir
->subsystem
;
878 const char *name
= NULL
;
882 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
886 ret
= tracing_update_buffers();
890 if (val
!= 0 && val
!= 1)
894 * Opening of "enable" adds a ref count to system,
895 * so the name is safe to use.
900 ret
= __ftrace_set_clr_event(dir
->tr
, NULL
, name
, NULL
, val
);
914 FORMAT_FIELD_SEPERATOR
= 2,
918 static void *f_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
920 struct trace_event_call
*call
= event_file_data(m
->private);
921 struct list_head
*common_head
= &ftrace_common_fields
;
922 struct list_head
*head
= trace_get_fields(call
);
923 struct list_head
*node
= v
;
927 switch ((unsigned long)v
) {
932 case FORMAT_FIELD_SEPERATOR
:
936 case FORMAT_PRINTFMT
:
942 if (node
== common_head
)
943 return (void *)FORMAT_FIELD_SEPERATOR
;
944 else if (node
== head
)
945 return (void *)FORMAT_PRINTFMT
;
950 static int f_show(struct seq_file
*m
, void *v
)
952 struct trace_event_call
*call
= event_file_data(m
->private);
953 struct ftrace_event_field
*field
;
954 const char *array_descriptor
;
956 switch ((unsigned long)v
) {
958 seq_printf(m
, "name: %s\n", trace_event_name(call
));
959 seq_printf(m
, "ID: %d\n", call
->event
.type
);
960 seq_puts(m
, "format:\n");
963 case FORMAT_FIELD_SEPERATOR
:
967 case FORMAT_PRINTFMT
:
968 seq_printf(m
, "\nprint fmt: %s\n",
973 field
= list_entry(v
, struct ftrace_event_field
, link
);
975 * Smartly shows the array type(except dynamic array).
978 * If TYPE := TYPE[LEN], it is shown:
979 * field:TYPE VAR[LEN]
981 array_descriptor
= strchr(field
->type
, '[');
983 if (!strncmp(field
->type
, "__data_loc", 10))
984 array_descriptor
= NULL
;
986 if (!array_descriptor
)
987 seq_printf(m
, "\tfield:%s %s;\toffset:%u;\tsize:%u;\tsigned:%d;\n",
988 field
->type
, field
->name
, field
->offset
,
989 field
->size
, !!field
->is_signed
);
991 seq_printf(m
, "\tfield:%.*s %s%s;\toffset:%u;\tsize:%u;\tsigned:%d;\n",
992 (int)(array_descriptor
- field
->type
),
993 field
->type
, field
->name
,
994 array_descriptor
, field
->offset
,
995 field
->size
, !!field
->is_signed
);
1000 static void *f_start(struct seq_file
*m
, loff_t
*pos
)
1002 void *p
= (void *)FORMAT_HEADER
;
1005 /* ->stop() is called even if ->start() fails */
1006 mutex_lock(&event_mutex
);
1007 if (!event_file_data(m
->private))
1008 return ERR_PTR(-ENODEV
);
1010 while (l
< *pos
&& p
)
1011 p
= f_next(m
, p
, &l
);
1016 static void f_stop(struct seq_file
*m
, void *p
)
1018 mutex_unlock(&event_mutex
);
1021 static const struct seq_operations trace_format_seq_ops
= {
1028 static int trace_format_open(struct inode
*inode
, struct file
*file
)
1033 ret
= seq_open(file
, &trace_format_seq_ops
);
1037 m
= file
->private_data
;
1044 event_id_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
, loff_t
*ppos
)
1046 int id
= (long)event_file_data(filp
);
1056 len
= sprintf(buf
, "%d\n", id
);
1058 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, len
);
1062 event_filter_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
,
1065 struct trace_event_file
*file
;
1066 struct trace_seq
*s
;
1072 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
1079 mutex_lock(&event_mutex
);
1080 file
= event_file_data(filp
);
1082 print_event_filter(file
, s
);
1083 mutex_unlock(&event_mutex
);
1086 r
= simple_read_from_buffer(ubuf
, cnt
, ppos
,
1087 s
->buffer
, trace_seq_used(s
));
1095 event_filter_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
1098 struct trace_event_file
*file
;
1102 if (cnt
>= PAGE_SIZE
)
1105 buf
= (char *)__get_free_page(GFP_TEMPORARY
);
1109 if (copy_from_user(buf
, ubuf
, cnt
)) {
1110 free_page((unsigned long) buf
);
1115 mutex_lock(&event_mutex
);
1116 file
= event_file_data(filp
);
1118 err
= apply_event_filter(file
, buf
);
1119 mutex_unlock(&event_mutex
);
1121 free_page((unsigned long) buf
);
1130 static LIST_HEAD(event_subsystems
);
1132 static int subsystem_open(struct inode
*inode
, struct file
*filp
)
1134 struct event_subsystem
*system
= NULL
;
1135 struct trace_subsystem_dir
*dir
= NULL
; /* Initialize for gcc */
1136 struct trace_array
*tr
;
1139 if (tracing_is_disabled())
1142 /* Make sure the system still exists */
1143 mutex_lock(&trace_types_lock
);
1144 mutex_lock(&event_mutex
);
1145 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
) {
1146 list_for_each_entry(dir
, &tr
->systems
, list
) {
1147 if (dir
== inode
->i_private
) {
1148 /* Don't open systems with no events */
1149 if (dir
->nr_events
) {
1150 __get_system_dir(dir
);
1151 system
= dir
->subsystem
;
1158 mutex_unlock(&event_mutex
);
1159 mutex_unlock(&trace_types_lock
);
1164 /* Some versions of gcc think dir can be uninitialized here */
1167 /* Still need to increment the ref count of the system */
1168 if (trace_array_get(tr
) < 0) {
1173 ret
= tracing_open_generic(inode
, filp
);
1175 trace_array_put(tr
);
1182 static int system_tr_open(struct inode
*inode
, struct file
*filp
)
1184 struct trace_subsystem_dir
*dir
;
1185 struct trace_array
*tr
= inode
->i_private
;
1188 if (tracing_is_disabled())
1191 if (trace_array_get(tr
) < 0)
1194 /* Make a temporary dir that has no system but points to tr */
1195 dir
= kzalloc(sizeof(*dir
), GFP_KERNEL
);
1197 trace_array_put(tr
);
1203 ret
= tracing_open_generic(inode
, filp
);
1205 trace_array_put(tr
);
1210 filp
->private_data
= dir
;
1215 static int subsystem_release(struct inode
*inode
, struct file
*file
)
1217 struct trace_subsystem_dir
*dir
= file
->private_data
;
1219 trace_array_put(dir
->tr
);
1222 * If dir->subsystem is NULL, then this is a temporary
1223 * descriptor that was made for a trace_array to enable
1235 subsystem_filter_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
,
1238 struct trace_subsystem_dir
*dir
= filp
->private_data
;
1239 struct event_subsystem
*system
= dir
->subsystem
;
1240 struct trace_seq
*s
;
1246 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
1252 print_subsystem_event_filter(system
, s
);
1253 r
= simple_read_from_buffer(ubuf
, cnt
, ppos
,
1254 s
->buffer
, trace_seq_used(s
));
1262 subsystem_filter_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
1265 struct trace_subsystem_dir
*dir
= filp
->private_data
;
1269 if (cnt
>= PAGE_SIZE
)
1272 buf
= (char *)__get_free_page(GFP_TEMPORARY
);
1276 if (copy_from_user(buf
, ubuf
, cnt
)) {
1277 free_page((unsigned long) buf
);
1282 err
= apply_subsystem_event_filter(dir
, buf
);
1283 free_page((unsigned long) buf
);
1293 show_header(struct file
*filp
, char __user
*ubuf
, size_t cnt
, loff_t
*ppos
)
1295 int (*func
)(struct trace_seq
*s
) = filp
->private_data
;
1296 struct trace_seq
*s
;
1302 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
1309 r
= simple_read_from_buffer(ubuf
, cnt
, ppos
,
1310 s
->buffer
, trace_seq_used(s
));
1317 static int ftrace_event_avail_open(struct inode
*inode
, struct file
*file
);
1318 static int ftrace_event_set_open(struct inode
*inode
, struct file
*file
);
1319 static int ftrace_event_release(struct inode
*inode
, struct file
*file
);
1321 static const struct seq_operations show_event_seq_ops
= {
1328 static const struct seq_operations show_set_event_seq_ops
= {
1335 static const struct file_operations ftrace_avail_fops
= {
1336 .open
= ftrace_event_avail_open
,
1338 .llseek
= seq_lseek
,
1339 .release
= seq_release
,
1342 static const struct file_operations ftrace_set_event_fops
= {
1343 .open
= ftrace_event_set_open
,
1345 .write
= ftrace_event_write
,
1346 .llseek
= seq_lseek
,
1347 .release
= ftrace_event_release
,
1350 static const struct file_operations ftrace_enable_fops
= {
1351 .open
= tracing_open_generic
,
1352 .read
= event_enable_read
,
1353 .write
= event_enable_write
,
1354 .llseek
= default_llseek
,
1357 static const struct file_operations ftrace_event_format_fops
= {
1358 .open
= trace_format_open
,
1360 .llseek
= seq_lseek
,
1361 .release
= seq_release
,
1364 static const struct file_operations ftrace_event_id_fops
= {
1365 .read
= event_id_read
,
1366 .llseek
= default_llseek
,
1369 static const struct file_operations ftrace_event_filter_fops
= {
1370 .open
= tracing_open_generic
,
1371 .read
= event_filter_read
,
1372 .write
= event_filter_write
,
1373 .llseek
= default_llseek
,
1376 static const struct file_operations ftrace_subsystem_filter_fops
= {
1377 .open
= subsystem_open
,
1378 .read
= subsystem_filter_read
,
1379 .write
= subsystem_filter_write
,
1380 .llseek
= default_llseek
,
1381 .release
= subsystem_release
,
1384 static const struct file_operations ftrace_system_enable_fops
= {
1385 .open
= subsystem_open
,
1386 .read
= system_enable_read
,
1387 .write
= system_enable_write
,
1388 .llseek
= default_llseek
,
1389 .release
= subsystem_release
,
1392 static const struct file_operations ftrace_tr_enable_fops
= {
1393 .open
= system_tr_open
,
1394 .read
= system_enable_read
,
1395 .write
= system_enable_write
,
1396 .llseek
= default_llseek
,
1397 .release
= subsystem_release
,
1400 static const struct file_operations ftrace_show_header_fops
= {
1401 .open
= tracing_open_generic
,
1402 .read
= show_header
,
1403 .llseek
= default_llseek
,
1407 ftrace_event_open(struct inode
*inode
, struct file
*file
,
1408 const struct seq_operations
*seq_ops
)
1413 ret
= seq_open(file
, seq_ops
);
1416 m
= file
->private_data
;
1417 /* copy tr over to seq ops */
1418 m
->private = inode
->i_private
;
1423 static int ftrace_event_release(struct inode
*inode
, struct file
*file
)
1425 struct trace_array
*tr
= inode
->i_private
;
1427 trace_array_put(tr
);
1429 return seq_release(inode
, file
);
1433 ftrace_event_avail_open(struct inode
*inode
, struct file
*file
)
1435 const struct seq_operations
*seq_ops
= &show_event_seq_ops
;
1437 return ftrace_event_open(inode
, file
, seq_ops
);
1441 ftrace_event_set_open(struct inode
*inode
, struct file
*file
)
1443 const struct seq_operations
*seq_ops
= &show_set_event_seq_ops
;
1444 struct trace_array
*tr
= inode
->i_private
;
1447 if (trace_array_get(tr
) < 0)
1450 if ((file
->f_mode
& FMODE_WRITE
) &&
1451 (file
->f_flags
& O_TRUNC
))
1452 ftrace_clear_events(tr
);
1454 ret
= ftrace_event_open(inode
, file
, seq_ops
);
1456 trace_array_put(tr
);
1460 static struct event_subsystem
*
1461 create_new_subsystem(const char *name
)
1463 struct event_subsystem
*system
;
1465 /* need to create new entry */
1466 system
= kmalloc(sizeof(*system
), GFP_KERNEL
);
1470 system
->ref_count
= 1;
1472 /* Only allocate if dynamic (kprobes and modules) */
1473 if (!core_kernel_data((unsigned long)name
)) {
1474 system
->ref_count
|= SYSTEM_FL_FREE_NAME
;
1475 system
->name
= kstrdup(name
, GFP_KERNEL
);
1479 system
->name
= name
;
1481 system
->filter
= NULL
;
1483 system
->filter
= kzalloc(sizeof(struct event_filter
), GFP_KERNEL
);
1484 if (!system
->filter
)
1487 list_add(&system
->list
, &event_subsystems
);
1492 if (system
->ref_count
& SYSTEM_FL_FREE_NAME
)
1493 kfree(system
->name
);
1498 static struct dentry
*
1499 event_subsystem_dir(struct trace_array
*tr
, const char *name
,
1500 struct trace_event_file
*file
, struct dentry
*parent
)
1502 struct trace_subsystem_dir
*dir
;
1503 struct event_subsystem
*system
;
1504 struct dentry
*entry
;
1506 /* First see if we did not already create this dir */
1507 list_for_each_entry(dir
, &tr
->systems
, list
) {
1508 system
= dir
->subsystem
;
1509 if (strcmp(system
->name
, name
) == 0) {
1516 /* Now see if the system itself exists. */
1517 list_for_each_entry(system
, &event_subsystems
, list
) {
1518 if (strcmp(system
->name
, name
) == 0)
1521 /* Reset system variable when not found */
1522 if (&system
->list
== &event_subsystems
)
1525 dir
= kmalloc(sizeof(*dir
), GFP_KERNEL
);
1530 system
= create_new_subsystem(name
);
1534 __get_system(system
);
1536 dir
->entry
= tracefs_create_dir(name
, parent
);
1538 pr_warn("Failed to create system directory %s\n", name
);
1539 __put_system(system
);
1546 dir
->subsystem
= system
;
1549 entry
= tracefs_create_file("filter", 0644, dir
->entry
, dir
,
1550 &ftrace_subsystem_filter_fops
);
1552 kfree(system
->filter
);
1553 system
->filter
= NULL
;
1554 pr_warn("Could not create tracefs '%s/filter' entry\n", name
);
1557 trace_create_file("enable", 0644, dir
->entry
, dir
,
1558 &ftrace_system_enable_fops
);
1560 list_add(&dir
->list
, &tr
->systems
);
1567 /* Only print this message if failed on memory allocation */
1568 if (!dir
|| !system
)
1569 pr_warn("No memory to create event subsystem %s\n", name
);
1574 event_create_dir(struct dentry
*parent
, struct trace_event_file
*file
)
1576 struct trace_event_call
*call
= file
->event_call
;
1577 struct trace_array
*tr
= file
->tr
;
1578 struct list_head
*head
;
1579 struct dentry
*d_events
;
1584 * If the trace point header did not define TRACE_SYSTEM
1585 * then the system would be called "TRACE_SYSTEM".
1587 if (strcmp(call
->class->system
, TRACE_SYSTEM
) != 0) {
1588 d_events
= event_subsystem_dir(tr
, call
->class->system
, file
, parent
);
1594 name
= trace_event_name(call
);
1595 file
->dir
= tracefs_create_dir(name
, d_events
);
1597 pr_warn("Could not create tracefs '%s' directory\n", name
);
1601 if (call
->class->reg
&& !(call
->flags
& TRACE_EVENT_FL_IGNORE_ENABLE
))
1602 trace_create_file("enable", 0644, file
->dir
, file
,
1603 &ftrace_enable_fops
);
1605 #ifdef CONFIG_PERF_EVENTS
1606 if (call
->event
.type
&& call
->class->reg
)
1607 trace_create_file("id", 0444, file
->dir
,
1608 (void *)(long)call
->event
.type
,
1609 &ftrace_event_id_fops
);
1613 * Other events may have the same class. Only update
1614 * the fields if they are not already defined.
1616 head
= trace_get_fields(call
);
1617 if (list_empty(head
)) {
1618 ret
= call
->class->define_fields(call
);
1620 pr_warn("Could not initialize trace point events/%s\n",
1625 trace_create_file("filter", 0644, file
->dir
, file
,
1626 &ftrace_event_filter_fops
);
1628 trace_create_file("trigger", 0644, file
->dir
, file
,
1629 &event_trigger_fops
);
1631 trace_create_file("format", 0444, file
->dir
, call
,
1632 &ftrace_event_format_fops
);
1637 static void remove_event_from_tracers(struct trace_event_call
*call
)
1639 struct trace_event_file
*file
;
1640 struct trace_array
*tr
;
1642 do_for_each_event_file_safe(tr
, file
) {
1643 if (file
->event_call
!= call
)
1646 remove_event_file_dir(file
);
1648 * The do_for_each_event_file_safe() is
1649 * a double loop. After finding the call for this
1650 * trace_array, we use break to jump to the next
1654 } while_for_each_event_file();
1657 static void event_remove(struct trace_event_call
*call
)
1659 struct trace_array
*tr
;
1660 struct trace_event_file
*file
;
1662 do_for_each_event_file(tr
, file
) {
1663 if (file
->event_call
!= call
)
1665 ftrace_event_enable_disable(file
, 0);
1667 * The do_for_each_event_file() is
1668 * a double loop. After finding the call for this
1669 * trace_array, we use break to jump to the next
1673 } while_for_each_event_file();
1675 if (call
->event
.funcs
)
1676 __unregister_trace_event(&call
->event
);
1677 remove_event_from_tracers(call
);
1678 list_del(&call
->list
);
1681 static int event_init(struct trace_event_call
*call
)
1686 name
= trace_event_name(call
);
1690 if (call
->class->raw_init
) {
1691 ret
= call
->class->raw_init(call
);
1692 if (ret
< 0 && ret
!= -ENOSYS
)
1693 pr_warn("Could not initialize trace events/%s\n", name
);
1700 __register_event(struct trace_event_call
*call
, struct module
*mod
)
1704 ret
= event_init(call
);
1708 list_add(&call
->list
, &ftrace_events
);
1714 static char *enum_replace(char *ptr
, struct trace_enum_map
*map
, int len
)
1719 /* Find the length of the enum value as a string */
1720 elen
= snprintf(ptr
, 0, "%ld", map
->enum_value
);
1721 /* Make sure there's enough room to replace the string with the value */
1725 snprintf(ptr
, elen
+ 1, "%ld", map
->enum_value
);
1727 /* Get the rest of the string of ptr */
1728 rlen
= strlen(ptr
+ len
);
1729 memmove(ptr
+ elen
, ptr
+ len
, rlen
);
1730 /* Make sure we end the new string */
1731 ptr
[elen
+ rlen
] = 0;
1736 static void update_event_printk(struct trace_event_call
*call
,
1737 struct trace_enum_map
*map
)
1741 int len
= strlen(map
->enum_string
);
1743 for (ptr
= call
->print_fmt
; *ptr
; ptr
++) {
1757 if (isdigit(*ptr
)) {
1761 /* Check for alpha chars like ULL */
1762 } while (isalnum(*ptr
));
1766 * A number must have some kind of delimiter after
1767 * it, and we can ignore that too.
1771 if (isalpha(*ptr
) || *ptr
== '_') {
1772 if (strncmp(map
->enum_string
, ptr
, len
) == 0 &&
1773 !isalnum(ptr
[len
]) && ptr
[len
] != '_') {
1774 ptr
= enum_replace(ptr
, map
, len
);
1775 /* Hmm, enum string smaller than value */
1776 if (WARN_ON_ONCE(!ptr
))
1779 * No need to decrement here, as enum_replace()
1780 * returns the pointer to the character passed
1781 * the enum, and two enums can not be placed
1782 * back to back without something in between.
1783 * We can skip that something in between.
1790 } while (isalnum(*ptr
) || *ptr
== '_');
1794 * If what comes after this variable is a '.' or
1795 * '->' then we can continue to ignore that string.
1797 if (*ptr
== '.' || (ptr
[0] == '-' && ptr
[1] == '>')) {
1798 ptr
+= *ptr
== '.' ? 1 : 2;
1804 * Once again, we can skip the delimiter that came
1812 void trace_event_enum_update(struct trace_enum_map
**map
, int len
)
1814 struct trace_event_call
*call
, *p
;
1815 const char *last_system
= NULL
;
1819 down_write(&trace_event_sem
);
1820 list_for_each_entry_safe(call
, p
, &ftrace_events
, list
) {
1821 /* events are usually grouped together with systems */
1822 if (!last_system
|| call
->class->system
!= last_system
) {
1824 last_system
= call
->class->system
;
1827 for (i
= last_i
; i
< len
; i
++) {
1828 if (call
->class->system
== map
[i
]->system
) {
1829 /* Save the first system if need be */
1832 update_event_printk(call
, map
[i
]);
1836 up_write(&trace_event_sem
);
1839 static struct trace_event_file
*
1840 trace_create_new_event(struct trace_event_call
*call
,
1841 struct trace_array
*tr
)
1843 struct trace_event_file
*file
;
1845 file
= kmem_cache_alloc(file_cachep
, GFP_TRACE
);
1849 file
->event_call
= call
;
1851 atomic_set(&file
->sm_ref
, 0);
1852 atomic_set(&file
->tm_ref
, 0);
1853 INIT_LIST_HEAD(&file
->triggers
);
1854 list_add(&file
->list
, &tr
->events
);
1859 /* Add an event to a trace directory */
1861 __trace_add_new_event(struct trace_event_call
*call
, struct trace_array
*tr
)
1863 struct trace_event_file
*file
;
1865 file
= trace_create_new_event(call
, tr
);
1869 return event_create_dir(tr
->event_dir
, file
);
1873 * Just create a decriptor for early init. A descriptor is required
1874 * for enabling events at boot. We want to enable events before
1875 * the filesystem is initialized.
1878 __trace_early_add_new_event(struct trace_event_call
*call
,
1879 struct trace_array
*tr
)
1881 struct trace_event_file
*file
;
1883 file
= trace_create_new_event(call
, tr
);
1890 struct ftrace_module_file_ops
;
1891 static void __add_event_to_tracers(struct trace_event_call
*call
);
1893 /* Add an additional event_call dynamically */
1894 int trace_add_event_call(struct trace_event_call
*call
)
1897 mutex_lock(&trace_types_lock
);
1898 mutex_lock(&event_mutex
);
1900 ret
= __register_event(call
, NULL
);
1902 __add_event_to_tracers(call
);
1904 mutex_unlock(&event_mutex
);
1905 mutex_unlock(&trace_types_lock
);
1910 * Must be called under locking of trace_types_lock, event_mutex and
1913 static void __trace_remove_event_call(struct trace_event_call
*call
)
1916 trace_destroy_fields(call
);
1917 free_event_filter(call
->filter
);
1918 call
->filter
= NULL
;
1921 static int probe_remove_event_call(struct trace_event_call
*call
)
1923 struct trace_array
*tr
;
1924 struct trace_event_file
*file
;
1926 #ifdef CONFIG_PERF_EVENTS
1927 if (call
->perf_refcount
)
1930 do_for_each_event_file(tr
, file
) {
1931 if (file
->event_call
!= call
)
1934 * We can't rely on ftrace_event_enable_disable(enable => 0)
1935 * we are going to do, EVENT_FILE_FL_SOFT_MODE can suppress
1936 * TRACE_REG_UNREGISTER.
1938 if (file
->flags
& EVENT_FILE_FL_ENABLED
)
1941 * The do_for_each_event_file_safe() is
1942 * a double loop. After finding the call for this
1943 * trace_array, we use break to jump to the next
1947 } while_for_each_event_file();
1949 __trace_remove_event_call(call
);
1954 /* Remove an event_call */
1955 int trace_remove_event_call(struct trace_event_call
*call
)
1959 mutex_lock(&trace_types_lock
);
1960 mutex_lock(&event_mutex
);
1961 down_write(&trace_event_sem
);
1962 ret
= probe_remove_event_call(call
);
1963 up_write(&trace_event_sem
);
1964 mutex_unlock(&event_mutex
);
1965 mutex_unlock(&trace_types_lock
);
1970 #define for_each_event(event, start, end) \
1971 for (event = start; \
1972 (unsigned long)event < (unsigned long)end; \
1975 #ifdef CONFIG_MODULES
1977 static void trace_module_add_events(struct module
*mod
)
1979 struct trace_event_call
**call
, **start
, **end
;
1981 if (!mod
->num_trace_events
)
1984 /* Don't add infrastructure for mods without tracepoints */
1985 if (trace_module_has_bad_taint(mod
)) {
1986 pr_err("%s: module has bad taint, not creating trace events\n",
1991 start
= mod
->trace_events
;
1992 end
= mod
->trace_events
+ mod
->num_trace_events
;
1994 for_each_event(call
, start
, end
) {
1995 __register_event(*call
, mod
);
1996 __add_event_to_tracers(*call
);
2000 static void trace_module_remove_events(struct module
*mod
)
2002 struct trace_event_call
*call
, *p
;
2003 bool clear_trace
= false;
2005 down_write(&trace_event_sem
);
2006 list_for_each_entry_safe(call
, p
, &ftrace_events
, list
) {
2007 if (call
->mod
== mod
) {
2008 if (call
->flags
& TRACE_EVENT_FL_WAS_ENABLED
)
2010 __trace_remove_event_call(call
);
2013 up_write(&trace_event_sem
);
2016 * It is safest to reset the ring buffer if the module being unloaded
2017 * registered any events that were used. The only worry is if
2018 * a new module gets loaded, and takes on the same id as the events
2019 * of this module. When printing out the buffer, traced events left
2020 * over from this module may be passed to the new module events and
2021 * unexpected results may occur.
2024 tracing_reset_all_online_cpus();
2027 static int trace_module_notify(struct notifier_block
*self
,
2028 unsigned long val
, void *data
)
2030 struct module
*mod
= data
;
2032 mutex_lock(&trace_types_lock
);
2033 mutex_lock(&event_mutex
);
2035 case MODULE_STATE_COMING
:
2036 trace_module_add_events(mod
);
2038 case MODULE_STATE_GOING
:
2039 trace_module_remove_events(mod
);
2042 mutex_unlock(&event_mutex
);
2043 mutex_unlock(&trace_types_lock
);
2048 static struct notifier_block trace_module_nb
= {
2049 .notifier_call
= trace_module_notify
,
2050 .priority
= 1, /* higher than trace.c module notify */
2052 #endif /* CONFIG_MODULES */
2054 /* Create a new event directory structure for a trace directory. */
2056 __trace_add_event_dirs(struct trace_array
*tr
)
2058 struct trace_event_call
*call
;
2061 list_for_each_entry(call
, &ftrace_events
, list
) {
2062 ret
= __trace_add_new_event(call
, tr
);
2064 pr_warn("Could not create directory for event %s\n",
2065 trace_event_name(call
));
2069 struct trace_event_file
*
2070 find_event_file(struct trace_array
*tr
, const char *system
, const char *event
)
2072 struct trace_event_file
*file
;
2073 struct trace_event_call
*call
;
2076 list_for_each_entry(file
, &tr
->events
, list
) {
2078 call
= file
->event_call
;
2079 name
= trace_event_name(call
);
2081 if (!name
|| !call
->class || !call
->class->reg
)
2084 if (call
->flags
& TRACE_EVENT_FL_IGNORE_ENABLE
)
2087 if (strcmp(event
, name
) == 0 &&
2088 strcmp(system
, call
->class->system
) == 0)
2094 #ifdef CONFIG_DYNAMIC_FTRACE
2097 #define ENABLE_EVENT_STR "enable_event"
2098 #define DISABLE_EVENT_STR "disable_event"
2100 struct event_probe_data
{
2101 struct trace_event_file
*file
;
2102 unsigned long count
;
2108 event_enable_probe(unsigned long ip
, unsigned long parent_ip
, void **_data
)
2110 struct event_probe_data
**pdata
= (struct event_probe_data
**)_data
;
2111 struct event_probe_data
*data
= *pdata
;
2117 clear_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &data
->file
->flags
);
2119 set_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &data
->file
->flags
);
2123 event_enable_count_probe(unsigned long ip
, unsigned long parent_ip
, void **_data
)
2125 struct event_probe_data
**pdata
= (struct event_probe_data
**)_data
;
2126 struct event_probe_data
*data
= *pdata
;
2134 /* Skip if the event is in a state we want to switch to */
2135 if (data
->enable
== !(data
->file
->flags
& EVENT_FILE_FL_SOFT_DISABLED
))
2138 if (data
->count
!= -1)
2141 event_enable_probe(ip
, parent_ip
, _data
);
2145 event_enable_print(struct seq_file
*m
, unsigned long ip
,
2146 struct ftrace_probe_ops
*ops
, void *_data
)
2148 struct event_probe_data
*data
= _data
;
2150 seq_printf(m
, "%ps:", (void *)ip
);
2152 seq_printf(m
, "%s:%s:%s",
2153 data
->enable
? ENABLE_EVENT_STR
: DISABLE_EVENT_STR
,
2154 data
->file
->event_call
->class->system
,
2155 trace_event_name(data
->file
->event_call
));
2157 if (data
->count
== -1)
2158 seq_puts(m
, ":unlimited\n");
2160 seq_printf(m
, ":count=%ld\n", data
->count
);
2166 event_enable_init(struct ftrace_probe_ops
*ops
, unsigned long ip
,
2169 struct event_probe_data
**pdata
= (struct event_probe_data
**)_data
;
2170 struct event_probe_data
*data
= *pdata
;
2177 event_enable_free(struct ftrace_probe_ops
*ops
, unsigned long ip
,
2180 struct event_probe_data
**pdata
= (struct event_probe_data
**)_data
;
2181 struct event_probe_data
*data
= *pdata
;
2183 if (WARN_ON_ONCE(data
->ref
<= 0))
2188 /* Remove the SOFT_MODE flag */
2189 __ftrace_event_enable_disable(data
->file
, 0, 1);
2190 module_put(data
->file
->event_call
->mod
);
2196 static struct ftrace_probe_ops event_enable_probe_ops
= {
2197 .func
= event_enable_probe
,
2198 .print
= event_enable_print
,
2199 .init
= event_enable_init
,
2200 .free
= event_enable_free
,
2203 static struct ftrace_probe_ops event_enable_count_probe_ops
= {
2204 .func
= event_enable_count_probe
,
2205 .print
= event_enable_print
,
2206 .init
= event_enable_init
,
2207 .free
= event_enable_free
,
2210 static struct ftrace_probe_ops event_disable_probe_ops
= {
2211 .func
= event_enable_probe
,
2212 .print
= event_enable_print
,
2213 .init
= event_enable_init
,
2214 .free
= event_enable_free
,
2217 static struct ftrace_probe_ops event_disable_count_probe_ops
= {
2218 .func
= event_enable_count_probe
,
2219 .print
= event_enable_print
,
2220 .init
= event_enable_init
,
2221 .free
= event_enable_free
,
2225 event_enable_func(struct ftrace_hash
*hash
,
2226 char *glob
, char *cmd
, char *param
, int enabled
)
2228 struct trace_array
*tr
= top_trace_array();
2229 struct trace_event_file
*file
;
2230 struct ftrace_probe_ops
*ops
;
2231 struct event_probe_data
*data
;
2241 /* hash funcs only work with set_ftrace_filter */
2242 if (!enabled
|| !param
)
2245 system
= strsep(¶m
, ":");
2249 event
= strsep(¶m
, ":");
2251 mutex_lock(&event_mutex
);
2254 file
= find_event_file(tr
, system
, event
);
2258 enable
= strcmp(cmd
, ENABLE_EVENT_STR
) == 0;
2261 ops
= param
? &event_enable_count_probe_ops
: &event_enable_probe_ops
;
2263 ops
= param
? &event_disable_count_probe_ops
: &event_disable_probe_ops
;
2265 if (glob
[0] == '!') {
2266 unregister_ftrace_function_probe_func(glob
+1, ops
);
2272 data
= kzalloc(sizeof(*data
), GFP_KERNEL
);
2276 data
->enable
= enable
;
2283 number
= strsep(¶m
, ":");
2286 if (!strlen(number
))
2290 * We use the callback data field (which is a pointer)
2293 ret
= kstrtoul(number
, 0, &data
->count
);
2298 /* Don't let event modules unload while probe registered */
2299 ret
= try_module_get(file
->event_call
->mod
);
2305 ret
= __ftrace_event_enable_disable(file
, 1, 1);
2308 ret
= register_ftrace_function_probe(glob
, ops
, data
);
2310 * The above returns on success the # of functions enabled,
2311 * but if it didn't find any functions it returns zero.
2312 * Consider no functions a failure too.
2319 /* Just return zero, not the number of enabled functions */
2322 mutex_unlock(&event_mutex
);
2326 __ftrace_event_enable_disable(file
, 0, 1);
2328 module_put(file
->event_call
->mod
);
2334 static struct ftrace_func_command event_enable_cmd
= {
2335 .name
= ENABLE_EVENT_STR
,
2336 .func
= event_enable_func
,
2339 static struct ftrace_func_command event_disable_cmd
= {
2340 .name
= DISABLE_EVENT_STR
,
2341 .func
= event_enable_func
,
2344 static __init
int register_event_cmds(void)
2348 ret
= register_ftrace_command(&event_enable_cmd
);
2349 if (WARN_ON(ret
< 0))
2351 ret
= register_ftrace_command(&event_disable_cmd
);
2352 if (WARN_ON(ret
< 0))
2353 unregister_ftrace_command(&event_enable_cmd
);
2357 static inline int register_event_cmds(void) { return 0; }
2358 #endif /* CONFIG_DYNAMIC_FTRACE */
2361 * The top level array has already had its trace_event_file
2362 * descriptors created in order to allow for early events to
2363 * be recorded. This function is called after the tracefs has been
2364 * initialized, and we now have to create the files associated
2368 __trace_early_add_event_dirs(struct trace_array
*tr
)
2370 struct trace_event_file
*file
;
2374 list_for_each_entry(file
, &tr
->events
, list
) {
2375 ret
= event_create_dir(tr
->event_dir
, file
);
2377 pr_warn("Could not create directory for event %s\n",
2378 trace_event_name(file
->event_call
));
2383 * For early boot up, the top trace array requires to have
2384 * a list of events that can be enabled. This must be done before
2385 * the filesystem is set up in order to allow events to be traced
2389 __trace_early_add_events(struct trace_array
*tr
)
2391 struct trace_event_call
*call
;
2394 list_for_each_entry(call
, &ftrace_events
, list
) {
2395 /* Early boot up should not have any modules loaded */
2396 if (WARN_ON_ONCE(call
->mod
))
2399 ret
= __trace_early_add_new_event(call
, tr
);
2401 pr_warn("Could not create early event %s\n",
2402 trace_event_name(call
));
2406 /* Remove the event directory structure for a trace directory. */
2408 __trace_remove_event_dirs(struct trace_array
*tr
)
2410 struct trace_event_file
*file
, *next
;
2412 list_for_each_entry_safe(file
, next
, &tr
->events
, list
)
2413 remove_event_file_dir(file
);
2416 static void __add_event_to_tracers(struct trace_event_call
*call
)
2418 struct trace_array
*tr
;
2420 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
)
2421 __trace_add_new_event(call
, tr
);
2424 extern struct trace_event_call
*__start_ftrace_events
[];
2425 extern struct trace_event_call
*__stop_ftrace_events
[];
2427 static char bootup_event_buf
[COMMAND_LINE_SIZE
] __initdata
;
2429 static __init
int setup_trace_event(char *str
)
2431 strlcpy(bootup_event_buf
, str
, COMMAND_LINE_SIZE
);
2432 ring_buffer_expanded
= true;
2433 tracing_selftest_disabled
= true;
2437 __setup("trace_event=", setup_trace_event
);
2439 /* Expects to have event_mutex held when called */
2441 create_event_toplevel_files(struct dentry
*parent
, struct trace_array
*tr
)
2443 struct dentry
*d_events
;
2444 struct dentry
*entry
;
2446 entry
= tracefs_create_file("set_event", 0644, parent
,
2447 tr
, &ftrace_set_event_fops
);
2449 pr_warn("Could not create tracefs 'set_event' entry\n");
2453 d_events
= tracefs_create_dir("events", parent
);
2455 pr_warn("Could not create tracefs 'events' directory\n");
2459 /* ring buffer internal formats */
2460 trace_create_file("header_page", 0444, d_events
,
2461 ring_buffer_print_page_header
,
2462 &ftrace_show_header_fops
);
2464 trace_create_file("header_event", 0444, d_events
,
2465 ring_buffer_print_entry_header
,
2466 &ftrace_show_header_fops
);
2468 trace_create_file("enable", 0644, d_events
,
2469 tr
, &ftrace_tr_enable_fops
);
2471 tr
->event_dir
= d_events
;
2477 * event_trace_add_tracer - add a instance of a trace_array to events
2478 * @parent: The parent dentry to place the files/directories for events in
2479 * @tr: The trace array associated with these events
2481 * When a new instance is created, it needs to set up its events
2482 * directory, as well as other files associated with events. It also
2483 * creates the event hierachry in the @parent/events directory.
2485 * Returns 0 on success.
2487 int event_trace_add_tracer(struct dentry
*parent
, struct trace_array
*tr
)
2491 mutex_lock(&event_mutex
);
2493 ret
= create_event_toplevel_files(parent
, tr
);
2497 down_write(&trace_event_sem
);
2498 __trace_add_event_dirs(tr
);
2499 up_write(&trace_event_sem
);
2502 mutex_unlock(&event_mutex
);
2508 * The top trace array already had its file descriptors created.
2509 * Now the files themselves need to be created.
2512 early_event_add_tracer(struct dentry
*parent
, struct trace_array
*tr
)
2516 mutex_lock(&event_mutex
);
2518 ret
= create_event_toplevel_files(parent
, tr
);
2522 down_write(&trace_event_sem
);
2523 __trace_early_add_event_dirs(tr
);
2524 up_write(&trace_event_sem
);
2527 mutex_unlock(&event_mutex
);
2532 int event_trace_del_tracer(struct trace_array
*tr
)
2534 mutex_lock(&event_mutex
);
2536 /* Disable any event triggers and associated soft-disabled events */
2537 clear_event_triggers(tr
);
2539 /* Disable any running events */
2540 __ftrace_set_clr_event_nolock(tr
, NULL
, NULL
, NULL
, 0);
2542 /* Access to events are within rcu_read_lock_sched() */
2543 synchronize_sched();
2545 down_write(&trace_event_sem
);
2546 __trace_remove_event_dirs(tr
);
2547 tracefs_remove_recursive(tr
->event_dir
);
2548 up_write(&trace_event_sem
);
2550 tr
->event_dir
= NULL
;
2552 mutex_unlock(&event_mutex
);
2557 static __init
int event_trace_memsetup(void)
2559 field_cachep
= KMEM_CACHE(ftrace_event_field
, SLAB_PANIC
);
2560 file_cachep
= KMEM_CACHE(trace_event_file
, SLAB_PANIC
);
2565 early_enable_events(struct trace_array
*tr
, bool disable_first
)
2567 char *buf
= bootup_event_buf
;
2572 token
= strsep(&buf
, ",");
2579 /* Restarting syscalls requires that we stop them first */
2581 ftrace_set_clr_event(tr
, token
, 0);
2583 ret
= ftrace_set_clr_event(tr
, token
, 1);
2585 pr_warn("Failed to enable trace event: %s\n", token
);
2587 /* Put back the comma to allow this to be called again */
2593 static __init
int event_trace_enable(void)
2595 struct trace_array
*tr
= top_trace_array();
2596 struct trace_event_call
**iter
, *call
;
2602 for_each_event(iter
, __start_ftrace_events
, __stop_ftrace_events
) {
2605 ret
= event_init(call
);
2607 list_add(&call
->list
, &ftrace_events
);
2611 * We need the top trace array to have a working set of trace
2612 * points at early init, before the debug files and directories
2613 * are created. Create the file entries now, and attach them
2614 * to the actual file dentries later.
2616 __trace_early_add_events(tr
);
2618 early_enable_events(tr
, false);
2620 trace_printk_start_comm();
2622 register_event_cmds();
2624 register_trigger_cmds();
2630 * event_trace_enable() is called from trace_event_init() first to
2631 * initialize events and perhaps start any events that are on the
2632 * command line. Unfortunately, there are some events that will not
2633 * start this early, like the system call tracepoints that need
2634 * to set the TIF_SYSCALL_TRACEPOINT flag of pid 1. But event_trace_enable()
2635 * is called before pid 1 starts, and this flag is never set, making
2636 * the syscall tracepoint never get reached, but the event is enabled
2637 * regardless (and not doing anything).
2639 static __init
int event_trace_enable_again(void)
2641 struct trace_array
*tr
;
2643 tr
= top_trace_array();
2647 early_enable_events(tr
, true);
2652 early_initcall(event_trace_enable_again
);
2654 static __init
int event_trace_init(void)
2656 struct trace_array
*tr
;
2657 struct dentry
*d_tracer
;
2658 struct dentry
*entry
;
2661 tr
= top_trace_array();
2665 d_tracer
= tracing_init_dentry();
2666 if (IS_ERR(d_tracer
))
2669 entry
= tracefs_create_file("available_events", 0444, d_tracer
,
2670 tr
, &ftrace_avail_fops
);
2672 pr_warn("Could not create tracefs 'available_events' entry\n");
2674 if (trace_define_common_fields())
2675 pr_warn("tracing: Failed to allocate common fields");
2677 ret
= early_event_add_tracer(d_tracer
, tr
);
2681 #ifdef CONFIG_MODULES
2682 ret
= register_module_notifier(&trace_module_nb
);
2684 pr_warn("Failed to register trace events module notifier\n");
2689 void __init
trace_event_init(void)
2691 event_trace_memsetup();
2692 init_ftrace_syscalls();
2693 event_trace_enable();
2696 fs_initcall(event_trace_init
);
2698 #ifdef CONFIG_FTRACE_STARTUP_TEST
2700 static DEFINE_SPINLOCK(test_spinlock
);
2701 static DEFINE_SPINLOCK(test_spinlock_irq
);
2702 static DEFINE_MUTEX(test_mutex
);
2704 static __init
void test_work(struct work_struct
*dummy
)
2706 spin_lock(&test_spinlock
);
2707 spin_lock_irq(&test_spinlock_irq
);
2709 spin_unlock_irq(&test_spinlock_irq
);
2710 spin_unlock(&test_spinlock
);
2712 mutex_lock(&test_mutex
);
2714 mutex_unlock(&test_mutex
);
2717 static __init
int event_test_thread(void *unused
)
2721 test_malloc
= kmalloc(1234, GFP_KERNEL
);
2723 pr_info("failed to kmalloc\n");
2725 schedule_on_each_cpu(test_work
);
2729 set_current_state(TASK_INTERRUPTIBLE
);
2730 while (!kthread_should_stop()) {
2732 set_current_state(TASK_INTERRUPTIBLE
);
2734 __set_current_state(TASK_RUNNING
);
2740 * Do various things that may trigger events.
2742 static __init
void event_test_stuff(void)
2744 struct task_struct
*test_thread
;
2746 test_thread
= kthread_run(event_test_thread
, NULL
, "test-events");
2748 kthread_stop(test_thread
);
2752 * For every trace event defined, we will test each trace point separately,
2753 * and then by groups, and finally all trace points.
2755 static __init
void event_trace_self_tests(void)
2757 struct trace_subsystem_dir
*dir
;
2758 struct trace_event_file
*file
;
2759 struct trace_event_call
*call
;
2760 struct event_subsystem
*system
;
2761 struct trace_array
*tr
;
2764 tr
= top_trace_array();
2768 pr_info("Running tests on trace events:\n");
2770 list_for_each_entry(file
, &tr
->events
, list
) {
2772 call
= file
->event_call
;
2774 /* Only test those that have a probe */
2775 if (!call
->class || !call
->class->probe
)
2779 * Testing syscall events here is pretty useless, but
2780 * we still do it if configured. But this is time consuming.
2781 * What we really need is a user thread to perform the
2782 * syscalls as we test.
2784 #ifndef CONFIG_EVENT_TRACE_TEST_SYSCALLS
2785 if (call
->class->system
&&
2786 strcmp(call
->class->system
, "syscalls") == 0)
2790 pr_info("Testing event %s: ", trace_event_name(call
));
2793 * If an event is already enabled, someone is using
2794 * it and the self test should not be on.
2796 if (file
->flags
& EVENT_FILE_FL_ENABLED
) {
2797 pr_warn("Enabled event during self test!\n");
2802 ftrace_event_enable_disable(file
, 1);
2804 ftrace_event_enable_disable(file
, 0);
2809 /* Now test at the sub system level */
2811 pr_info("Running tests on trace event systems:\n");
2813 list_for_each_entry(dir
, &tr
->systems
, list
) {
2815 system
= dir
->subsystem
;
2817 /* the ftrace system is special, skip it */
2818 if (strcmp(system
->name
, "ftrace") == 0)
2821 pr_info("Testing event system %s: ", system
->name
);
2823 ret
= __ftrace_set_clr_event(tr
, NULL
, system
->name
, NULL
, 1);
2824 if (WARN_ON_ONCE(ret
)) {
2825 pr_warn("error enabling system %s\n",
2832 ret
= __ftrace_set_clr_event(tr
, NULL
, system
->name
, NULL
, 0);
2833 if (WARN_ON_ONCE(ret
)) {
2834 pr_warn("error disabling system %s\n",
2842 /* Test with all events enabled */
2844 pr_info("Running tests on all trace events:\n");
2845 pr_info("Testing all events: ");
2847 ret
= __ftrace_set_clr_event(tr
, NULL
, NULL
, NULL
, 1);
2848 if (WARN_ON_ONCE(ret
)) {
2849 pr_warn("error enabling all events\n");
2856 ret
= __ftrace_set_clr_event(tr
, NULL
, NULL
, NULL
, 0);
2857 if (WARN_ON_ONCE(ret
)) {
2858 pr_warn("error disabling all events\n");
2865 #ifdef CONFIG_FUNCTION_TRACER
2867 static DEFINE_PER_CPU(atomic_t
, ftrace_test_event_disable
);
2870 function_test_events_call(unsigned long ip
, unsigned long parent_ip
,
2871 struct ftrace_ops
*op
, struct pt_regs
*pt_regs
)
2873 struct ring_buffer_event
*event
;
2874 struct ring_buffer
*buffer
;
2875 struct ftrace_entry
*entry
;
2876 unsigned long flags
;
2881 pc
= preempt_count();
2882 preempt_disable_notrace();
2883 cpu
= raw_smp_processor_id();
2884 disabled
= atomic_inc_return(&per_cpu(ftrace_test_event_disable
, cpu
));
2889 local_save_flags(flags
);
2891 event
= trace_current_buffer_lock_reserve(&buffer
,
2892 TRACE_FN
, sizeof(*entry
),
2896 entry
= ring_buffer_event_data(event
);
2898 entry
->parent_ip
= parent_ip
;
2900 trace_buffer_unlock_commit(buffer
, event
, flags
, pc
);
2903 atomic_dec(&per_cpu(ftrace_test_event_disable
, cpu
));
2904 preempt_enable_notrace();
2907 static struct ftrace_ops trace_ops __initdata
=
2909 .func
= function_test_events_call
,
2910 .flags
= FTRACE_OPS_FL_RECURSION_SAFE
,
2913 static __init
void event_trace_self_test_with_function(void)
2916 ret
= register_ftrace_function(&trace_ops
);
2917 if (WARN_ON(ret
< 0)) {
2918 pr_info("Failed to enable function tracer for event tests\n");
2921 pr_info("Running tests again, along with the function tracer\n");
2922 event_trace_self_tests();
2923 unregister_ftrace_function(&trace_ops
);
2926 static __init
void event_trace_self_test_with_function(void)
2931 static __init
int event_trace_self_tests_init(void)
2933 if (!tracing_selftest_disabled
) {
2934 event_trace_self_tests();
2935 event_trace_self_test_with_function();
2941 late_initcall(event_trace_self_tests_init
);