Commit | Line | Data |
---|---|---|
97f20251 SR |
1 | #ifndef _LINUX_FTRACE_EVENT_H |
2 | #define _LINUX_FTRACE_EVENT_H | |
3 | ||
97f20251 | 4 | #include <linux/ring_buffer.h> |
16bb8eb1 | 5 | #include <linux/trace_seq.h> |
be74b73a | 6 | #include <linux/percpu.h> |
20ab4425 | 7 | #include <linux/hardirq.h> |
430ad5a6 | 8 | #include <linux/perf_event.h> |
97f20251 SR |
9 | |
10 | struct trace_array; | |
12883efb | 11 | struct trace_buffer; |
97f20251 | 12 | struct tracer; |
6d723736 | 13 | struct dentry; |
97f20251 | 14 | |
be74b73a SR |
15 | struct trace_print_flags { |
16 | unsigned long mask; | |
17 | const char *name; | |
18 | }; | |
19 | ||
2fc1b6f0 | 20 | struct trace_print_flags_u64 { |
21 | unsigned long long mask; | |
22 | const char *name; | |
23 | }; | |
24 | ||
be74b73a SR |
25 | const char *ftrace_print_flags_seq(struct trace_seq *p, const char *delim, |
26 | unsigned long flags, | |
27 | const struct trace_print_flags *flag_array); | |
28 | ||
0f4fc29d SR |
29 | const char *ftrace_print_symbols_seq(struct trace_seq *p, unsigned long val, |
30 | const struct trace_print_flags *symbol_array); | |
31 | ||
2fc1b6f0 | 32 | #if BITS_PER_LONG == 32 |
33 | const char *ftrace_print_symbols_seq_u64(struct trace_seq *p, | |
34 | unsigned long long val, | |
35 | const struct trace_print_flags_u64 | |
36 | *symbol_array); | |
37 | #endif | |
38 | ||
5a2e3995 KT |
39 | const char *ftrace_print_hex_seq(struct trace_seq *p, |
40 | const unsigned char *buf, int len); | |
41 | ||
f71130de LZ |
42 | struct trace_iterator; |
43 | struct trace_event; | |
44 | ||
45 | int ftrace_raw_output_prep(struct trace_iterator *iter, | |
46 | struct trace_event *event); | |
47 | ||
97f20251 SR |
48 | /* |
49 | * The trace entry - the most basic unit of tracing. This is what | |
50 | * is printed in the end as a single line in the trace output, such as: | |
51 | * | |
52 | * bash-15816 [01] 235.197585: idle_cpu <- irq_enter | |
53 | */ | |
54 | struct trace_entry { | |
89ec0dee | 55 | unsigned short type; |
97f20251 SR |
56 | unsigned char flags; |
57 | unsigned char preempt_count; | |
58 | int pid; | |
97f20251 SR |
59 | }; |
60 | ||
89ec0dee SR |
61 | #define FTRACE_MAX_EVENT \ |
62 | ((1 << (sizeof(((struct trace_entry *)0)->type) * 8)) - 1) | |
63 | ||
97f20251 SR |
64 | /* |
65 | * Trace iterator - used by printout routines who present trace | |
66 | * results to users and which routines might sleep, etc: | |
67 | */ | |
68 | struct trace_iterator { | |
69 | struct trace_array *tr; | |
70 | struct tracer *trace; | |
12883efb | 71 | struct trace_buffer *trace_buffer; |
97f20251 SR |
72 | void *private; |
73 | int cpu_file; | |
74 | struct mutex mutex; | |
6d158a81 | 75 | struct ring_buffer_iter **buffer_iter; |
112f38a7 | 76 | unsigned long iter_flags; |
97f20251 | 77 | |
bc289ae9 LJ |
78 | /* trace_seq for __print_flags() and __print_symbolic() etc. */ |
79 | struct trace_seq tmp_seq; | |
80 | ||
ed5467da AV |
81 | cpumask_var_t started; |
82 | ||
83 | /* it's true when current open file is snapshot */ | |
84 | bool snapshot; | |
85 | ||
97f20251 SR |
86 | /* The below is zeroed out in pipe_read */ |
87 | struct trace_seq seq; | |
88 | struct trace_entry *ent; | |
bc21b478 | 89 | unsigned long lost_events; |
a63ce5b3 | 90 | int leftover; |
4a9bd3f1 | 91 | int ent_size; |
97f20251 SR |
92 | int cpu; |
93 | u64 ts; | |
94 | ||
97f20251 SR |
95 | loff_t pos; |
96 | long idx; | |
97 | ||
ed5467da | 98 | /* All new field here will be zeroed out in pipe_read */ |
97f20251 SR |
99 | }; |
100 | ||
8be0709f DS |
101 | enum trace_iter_flags { |
102 | TRACE_FILE_LAT_FMT = 1, | |
103 | TRACE_FILE_ANNOTATE = 2, | |
104 | TRACE_FILE_TIME_IN_NS = 4, | |
105 | }; | |
106 | ||
97f20251 SR |
107 | |
108 | typedef enum print_line_t (*trace_print_func)(struct trace_iterator *iter, | |
a9a57763 SR |
109 | int flags, struct trace_event *event); |
110 | ||
111 | struct trace_event_functions { | |
97f20251 SR |
112 | trace_print_func trace; |
113 | trace_print_func raw; | |
114 | trace_print_func hex; | |
115 | trace_print_func binary; | |
116 | }; | |
117 | ||
a9a57763 SR |
118 | struct trace_event { |
119 | struct hlist_node node; | |
120 | struct list_head list; | |
121 | int type; | |
122 | struct trace_event_functions *funcs; | |
123 | }; | |
124 | ||
97f20251 SR |
125 | extern int register_ftrace_event(struct trace_event *event); |
126 | extern int unregister_ftrace_event(struct trace_event *event); | |
127 | ||
128 | /* Return values for print_line callback */ | |
129 | enum print_line_t { | |
130 | TRACE_TYPE_PARTIAL_LINE = 0, /* Retry after flushing the seq */ | |
131 | TRACE_TYPE_HANDLED = 1, | |
132 | TRACE_TYPE_UNHANDLED = 2, /* Relay to other output functions */ | |
133 | TRACE_TYPE_NO_CONSUME = 3 /* Handled but ask to not consume */ | |
134 | }; | |
135 | ||
f413cdb8 FW |
136 | void tracing_generic_entry_update(struct trace_entry *entry, |
137 | unsigned long flags, | |
138 | int pc); | |
ccb469a1 SR |
139 | struct ftrace_event_file; |
140 | ||
141 | struct ring_buffer_event * | |
142 | trace_event_buffer_lock_reserve(struct ring_buffer **current_buffer, | |
143 | struct ftrace_event_file *ftrace_file, | |
144 | int type, unsigned long len, | |
145 | unsigned long flags, int pc); | |
97f20251 | 146 | struct ring_buffer_event * |
e77405ad SR |
147 | trace_current_buffer_lock_reserve(struct ring_buffer **current_buffer, |
148 | int type, unsigned long len, | |
97f20251 | 149 | unsigned long flags, int pc); |
e77405ad SR |
150 | void trace_current_buffer_unlock_commit(struct ring_buffer *buffer, |
151 | struct ring_buffer_event *event, | |
97f20251 | 152 | unsigned long flags, int pc); |
0d5c6e1c SR |
153 | void trace_buffer_unlock_commit(struct ring_buffer *buffer, |
154 | struct ring_buffer_event *event, | |
155 | unsigned long flags, int pc); | |
156 | void trace_buffer_unlock_commit_regs(struct ring_buffer *buffer, | |
157 | struct ring_buffer_event *event, | |
158 | unsigned long flags, int pc, | |
159 | struct pt_regs *regs); | |
e77405ad SR |
160 | void trace_current_buffer_discard_commit(struct ring_buffer *buffer, |
161 | struct ring_buffer_event *event); | |
97f20251 SR |
162 | |
163 | void tracing_record_cmdline(struct task_struct *tsk); | |
164 | ||
1f9963cb LZ |
165 | struct event_filter; |
166 | ||
2239291a SR |
167 | enum trace_reg { |
168 | TRACE_REG_REGISTER, | |
169 | TRACE_REG_UNREGISTER, | |
37d73998 | 170 | #ifdef CONFIG_PERF_EVENTS |
2239291a SR |
171 | TRACE_REG_PERF_REGISTER, |
172 | TRACE_REG_PERF_UNREGISTER, | |
ceec0b6f JO |
173 | TRACE_REG_PERF_OPEN, |
174 | TRACE_REG_PERF_CLOSE, | |
489c75c3 JO |
175 | TRACE_REG_PERF_ADD, |
176 | TRACE_REG_PERF_DEL, | |
37d73998 | 177 | #endif |
2239291a SR |
178 | }; |
179 | ||
180 | struct ftrace_event_call; | |
181 | ||
8f082018 SR |
182 | struct ftrace_event_class { |
183 | char *system; | |
2239291a SR |
184 | void *probe; |
185 | #ifdef CONFIG_PERF_EVENTS | |
186 | void *perf_probe; | |
187 | #endif | |
188 | int (*reg)(struct ftrace_event_call *event, | |
ceec0b6f | 189 | enum trace_reg type, void *data); |
2e33af02 SR |
190 | int (*define_fields)(struct ftrace_event_call *); |
191 | struct list_head *(*get_fields)(struct ftrace_event_call *); | |
192 | struct list_head fields; | |
0405ab80 | 193 | int (*raw_init)(struct ftrace_event_call *); |
8f082018 SR |
194 | }; |
195 | ||
a1d0ce82 | 196 | extern int ftrace_event_reg(struct ftrace_event_call *event, |
ceec0b6f | 197 | enum trace_reg type, void *data); |
a1d0ce82 | 198 | |
553552ce | 199 | enum { |
553552ce | 200 | TRACE_EVENT_FL_FILTERED_BIT, |
61c32659 | 201 | TRACE_EVENT_FL_CAP_ANY_BIT, |
27b14b56 | 202 | TRACE_EVENT_FL_NO_SET_FILTER_BIT, |
9b63776f | 203 | TRACE_EVENT_FL_IGNORE_ENABLE_BIT, |
575380da | 204 | TRACE_EVENT_FL_WAS_ENABLED_BIT, |
553552ce SR |
205 | }; |
206 | ||
ae63b31e SR |
207 | /* |
208 | * Event flags: | |
209 | * FILTERED - The event has a filter attached | |
210 | * CAP_ANY - Any user can enable for perf | |
211 | * NO_SET_FILTER - Set when filter has error and is to be ignored | |
2a30c11f | 212 | * IGNORE_ENABLE - For ftrace internal events, do not enable with debugfs file |
575380da SRRH |
213 | * WAS_ENABLED - Set and stays set when an event was ever enabled |
214 | * (used for module unloading, if a module event is enabled, | |
215 | * it is best to clear the buffers that used it). | |
ae63b31e | 216 | */ |
553552ce | 217 | enum { |
e870e9a1 | 218 | TRACE_EVENT_FL_FILTERED = (1 << TRACE_EVENT_FL_FILTERED_BIT), |
61c32659 | 219 | TRACE_EVENT_FL_CAP_ANY = (1 << TRACE_EVENT_FL_CAP_ANY_BIT), |
27b14b56 | 220 | TRACE_EVENT_FL_NO_SET_FILTER = (1 << TRACE_EVENT_FL_NO_SET_FILTER_BIT), |
9b63776f | 221 | TRACE_EVENT_FL_IGNORE_ENABLE = (1 << TRACE_EVENT_FL_IGNORE_ENABLE_BIT), |
575380da | 222 | TRACE_EVENT_FL_WAS_ENABLED = (1 << TRACE_EVENT_FL_WAS_ENABLED_BIT), |
553552ce SR |
223 | }; |
224 | ||
97f20251 | 225 | struct ftrace_event_call { |
a59fd602 | 226 | struct list_head list; |
8f082018 | 227 | struct ftrace_event_class *class; |
97f20251 | 228 | char *name; |
80decc70 | 229 | struct trace_event event; |
509e760c | 230 | const char *print_fmt; |
1f9963cb | 231 | struct event_filter *filter; |
ae63b31e | 232 | struct list_head *files; |
6d723736 | 233 | void *mod; |
69fd4f0e | 234 | void *data; |
57d01ad0 SRRH |
235 | /* |
236 | * bit 0: filter_active | |
237 | * bit 1: allow trace by non root (cap any) | |
238 | * bit 2: failed to apply filter | |
239 | * bit 3: ftrace internal event (do not enable) | |
240 | * bit 4: Event was enabled by module | |
241 | */ | |
ae63b31e SR |
242 | int flags; /* static flags of different events */ |
243 | ||
244 | #ifdef CONFIG_PERF_EVENTS | |
245 | int perf_refcount; | |
246 | struct hlist_head __percpu *perf_events; | |
247 | #endif | |
248 | }; | |
249 | ||
250 | struct trace_array; | |
251 | struct ftrace_subsystem_dir; | |
252 | ||
253 | enum { | |
254 | FTRACE_EVENT_FL_ENABLED_BIT, | |
255 | FTRACE_EVENT_FL_RECORDED_CMD_BIT, | |
417944c4 SRRH |
256 | FTRACE_EVENT_FL_SOFT_MODE_BIT, |
257 | FTRACE_EVENT_FL_SOFT_DISABLED_BIT, | |
ae63b31e SR |
258 | }; |
259 | ||
260 | /* | |
261 | * Ftrace event file flags: | |
57d01ad0 | 262 | * ENABLED - The event is enabled |
ae63b31e | 263 | * RECORDED_CMD - The comms should be recorded at sched_switch |
417944c4 SRRH |
264 | * SOFT_MODE - The event is enabled/disabled by SOFT_DISABLED |
265 | * SOFT_DISABLED - When set, do not trace the event (even though its | |
266 | * tracepoint may be enabled) | |
ae63b31e SR |
267 | */ |
268 | enum { | |
269 | FTRACE_EVENT_FL_ENABLED = (1 << FTRACE_EVENT_FL_ENABLED_BIT), | |
270 | FTRACE_EVENT_FL_RECORDED_CMD = (1 << FTRACE_EVENT_FL_RECORDED_CMD_BIT), | |
417944c4 SRRH |
271 | FTRACE_EVENT_FL_SOFT_MODE = (1 << FTRACE_EVENT_FL_SOFT_MODE_BIT), |
272 | FTRACE_EVENT_FL_SOFT_DISABLED = (1 << FTRACE_EVENT_FL_SOFT_DISABLED_BIT), | |
ae63b31e SR |
273 | }; |
274 | ||
275 | struct ftrace_event_file { | |
276 | struct list_head list; | |
277 | struct ftrace_event_call *event_call; | |
278 | struct dentry *dir; | |
279 | struct trace_array *tr; | |
280 | struct ftrace_subsystem_dir *system; | |
97f20251 | 281 | |
553552ce SR |
282 | /* |
283 | * 32 bit flags: | |
57d01ad0 SRRH |
284 | * bit 0: enabled |
285 | * bit 1: enabled cmd record | |
417944c4 SRRH |
286 | * bit 2: enable/disable with the soft disable bit |
287 | * bit 3: soft disabled | |
553552ce | 288 | * |
417944c4 SRRH |
289 | * Note: The bits must be set atomically to prevent races |
290 | * from other writers. Reads of flags do not need to be in | |
291 | * sync as they occur in critical sections. But the way flags | |
ae63b31e | 292 | * is currently used, these changes do not affect the code |
1eaa4787 SR |
293 | * except that when a change is made, it may have a slight |
294 | * delay in propagating the changes to other CPUs due to | |
417944c4 | 295 | * caching and such. Which is mostly OK ;-) |
553552ce | 296 | */ |
417944c4 | 297 | unsigned long flags; |
1cf4c073 | 298 | atomic_t sm_ref; /* soft-mode reference counter */ |
97f20251 SR |
299 | }; |
300 | ||
53cf810b FW |
301 | #define __TRACE_EVENT_FLAGS(name, value) \ |
302 | static int __init trace_init_flags_##name(void) \ | |
303 | { \ | |
304 | event_##name.flags = value; \ | |
305 | return 0; \ | |
306 | } \ | |
307 | early_initcall(trace_init_flags_##name); | |
308 | ||
97d5a220 | 309 | #define PERF_MAX_TRACE_SIZE 2048 |
20ab4425 | 310 | |
16bb8eb1 | 311 | #define MAX_FILTER_STR_VAL 256 /* Should handle KSYM_SYMBOL_LEN */ |
97f20251 | 312 | |
2df75e41 | 313 | extern void destroy_preds(struct ftrace_event_call *call); |
6fb2915d | 314 | extern int filter_match_preds(struct event_filter *filter, void *rec); |
e77405ad SR |
315 | extern int filter_current_check_discard(struct ring_buffer *buffer, |
316 | struct ftrace_event_call *call, | |
97f20251 SR |
317 | void *rec, |
318 | struct ring_buffer_event *event); | |
319 | ||
43b51ead LZ |
320 | enum { |
321 | FILTER_OTHER = 0, | |
322 | FILTER_STATIC_STRING, | |
323 | FILTER_DYN_STRING, | |
87a342f5 | 324 | FILTER_PTR_STRING, |
02aa3162 | 325 | FILTER_TRACE_FN, |
43b51ead LZ |
326 | }; |
327 | ||
04295780 SR |
328 | #define EVENT_STORAGE_SIZE 128 |
329 | extern struct mutex event_storage_mutex; | |
330 | extern char event_storage[EVENT_STORAGE_SIZE]; | |
331 | ||
87d9b4e1 | 332 | extern int trace_event_raw_init(struct ftrace_event_call *call); |
aeaeae11 FW |
333 | extern int trace_define_field(struct ftrace_event_call *call, const char *type, |
334 | const char *name, int offset, int size, | |
335 | int is_signed, int filter_type); | |
bd1a5c84 | 336 | extern int trace_add_event_call(struct ftrace_event_call *call); |
2816c551 | 337 | extern int trace_remove_event_call(struct ftrace_event_call *call); |
97f20251 | 338 | |
d2802d07 | 339 | #define is_signed_type(type) (((type)(-1)) < (type)1) |
97f20251 | 340 | |
4671c794 SR |
341 | int trace_set_clr_event(const char *system, const char *event, int set); |
342 | ||
97f20251 SR |
343 | /* |
344 | * The double __builtin_constant_p is because gcc will give us an error | |
345 | * if we try to allocate the static variable to fmt if it is not a | |
346 | * constant. Even with the outer if statement optimizing out. | |
347 | */ | |
348 | #define event_trace_printk(ip, fmt, args...) \ | |
349 | do { \ | |
350 | __trace_printk_check_format(fmt, ##args); \ | |
351 | tracing_record_cmdline(current); \ | |
352 | if (__builtin_constant_p(fmt)) { \ | |
353 | static const char *trace_printk_fmt \ | |
354 | __attribute__((section("__trace_printk_fmt"))) = \ | |
355 | __builtin_constant_p(fmt) ? fmt : NULL; \ | |
356 | \ | |
357 | __trace_bprintk(ip, trace_printk_fmt, ##args); \ | |
358 | } else \ | |
359 | __trace_printk(ip, fmt, ##args); \ | |
360 | } while (0) | |
361 | ||
102c9323 SRRH |
362 | /** |
363 | * tracepoint_string - register constant persistent string to trace system | |
364 | * @str - a constant persistent string that will be referenced in tracepoints | |
365 | * | |
366 | * If constant strings are being used in tracepoints, it is faster and | |
367 | * more efficient to just save the pointer to the string and reference | |
368 | * that with a printf "%s" instead of saving the string in the ring buffer | |
369 | * and wasting space and time. | |
370 | * | |
371 | * The problem with the above approach is that userspace tools that read | |
372 | * the binary output of the trace buffers do not have access to the string. | |
373 | * Instead they just show the address of the string which is not very | |
374 | * useful to users. | |
375 | * | |
376 | * With tracepoint_string(), the string will be registered to the tracing | |
377 | * system and exported to userspace via the debugfs/tracing/printk_formats | |
378 | * file that maps the string address to the string text. This way userspace | |
379 | * tools that read the binary buffers have a way to map the pointers to | |
380 | * the ASCII strings they represent. | |
381 | * | |
382 | * The @str used must be a constant string and persistent as it would not | |
383 | * make sense to show a string that no longer exists. But it is still fine | |
384 | * to be used with modules, because when modules are unloaded, if they | |
385 | * had tracepoints, the ring buffers are cleared too. As long as the string | |
386 | * does not change during the life of the module, it is fine to use | |
387 | * tracepoint_string() within a module. | |
388 | */ | |
389 | #define tracepoint_string(str) \ | |
390 | ({ \ | |
391 | static const char *___tp_str __tracepoint_string = str; \ | |
392 | ___tp_str; \ | |
393 | }) | |
394 | #define __tracepoint_string __attribute__((section("__tracepoint_str"))) | |
395 | ||
07b139c8 | 396 | #ifdef CONFIG_PERF_EVENTS |
6fb2915d | 397 | struct perf_event; |
c530665c FW |
398 | |
399 | DECLARE_PER_CPU(struct pt_regs, perf_trace_regs); | |
400 | ||
1c024eca PZ |
401 | extern int perf_trace_init(struct perf_event *event); |
402 | extern void perf_trace_destroy(struct perf_event *event); | |
a4eaf7f1 PZ |
403 | extern int perf_trace_add(struct perf_event *event, int flags); |
404 | extern void perf_trace_del(struct perf_event *event, int flags); | |
1c024eca | 405 | extern int ftrace_profile_set_filter(struct perf_event *event, int event_id, |
6fb2915d LZ |
406 | char *filter_str); |
407 | extern void ftrace_profile_free_filter(struct perf_event *event); | |
b7e2ecef PZ |
408 | extern void *perf_trace_buf_prepare(int size, unsigned short type, |
409 | struct pt_regs *regs, int *rctxp); | |
430ad5a6 XG |
410 | |
411 | static inline void | |
97d5a220 | 412 | perf_trace_buf_submit(void *raw_data, int size, int rctx, u64 addr, |
e6dab5ff AV |
413 | u64 count, struct pt_regs *regs, void *head, |
414 | struct task_struct *task) | |
430ad5a6 | 415 | { |
e6dab5ff | 416 | perf_tp_event(addr, count, raw_data, size, regs, head, rctx, task); |
430ad5a6 | 417 | } |
6fb2915d LZ |
418 | #endif |
419 | ||
97f20251 | 420 | #endif /* _LINUX_FTRACE_EVENT_H */ |