blob: 447bd96ee658aacc085875cd014a6bcb782e6a96 [file] [log] [blame]
Andrew Scullb4b6d4a2019-01-02 15:54:55 +00001// SPDX-License-Identifier: GPL-2.0
2
3#ifndef _LINUX_KERNEL_TRACE_H
4#define _LINUX_KERNEL_TRACE_H
5
6#include <linux/fs.h>
7#include <linux/atomic.h>
8#include <linux/sched.h>
9#include <linux/clocksource.h>
10#include <linux/ring_buffer.h>
11#include <linux/mmiotrace.h>
12#include <linux/tracepoint.h>
13#include <linux/ftrace.h>
14#include <linux/hw_breakpoint.h>
15#include <linux/trace_seq.h>
16#include <linux/trace_events.h>
17#include <linux/compiler.h>
18#include <linux/trace_seq.h>
19#include <linux/glob.h>
20
21#ifdef CONFIG_FTRACE_SYSCALLS
22#include <asm/unistd.h> /* For NR_SYSCALLS */
23#include <asm/syscall.h> /* some archs define it here */
24#endif
25
26enum trace_type {
27 __TRACE_FIRST_TYPE = 0,
28
29 TRACE_FN,
30 TRACE_CTX,
31 TRACE_WAKE,
32 TRACE_STACK,
33 TRACE_PRINT,
34 TRACE_BPRINT,
35 TRACE_MMIO_RW,
36 TRACE_MMIO_MAP,
37 TRACE_BRANCH,
38 TRACE_GRAPH_RET,
39 TRACE_GRAPH_ENT,
40 TRACE_USER_STACK,
41 TRACE_BLK,
42 TRACE_BPUTS,
43 TRACE_HWLAT,
44 TRACE_RAW_DATA,
45
46 __TRACE_LAST_TYPE,
47};
48
49
50#undef __field
51#define __field(type, item) type item;
52
53#undef __field_struct
54#define __field_struct(type, item) __field(type, item)
55
56#undef __field_desc
57#define __field_desc(type, container, item)
58
59#undef __array
60#define __array(type, item, size) type item[size];
61
62#undef __array_desc
63#define __array_desc(type, container, item, size)
64
65#undef __dynamic_array
66#define __dynamic_array(type, item) type item[];
67
68#undef F_STRUCT
69#define F_STRUCT(args...) args
70
71#undef FTRACE_ENTRY
72#define FTRACE_ENTRY(name, struct_name, id, tstruct, print, filter) \
73 struct struct_name { \
74 struct trace_entry ent; \
75 tstruct \
76 }
77
78#undef FTRACE_ENTRY_DUP
79#define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk, filter)
80
81#undef FTRACE_ENTRY_REG
82#define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print, \
83 filter, regfn) \
84 FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
85 filter)
86
87#undef FTRACE_ENTRY_PACKED
88#define FTRACE_ENTRY_PACKED(name, struct_name, id, tstruct, print, \
89 filter) \
90 FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
91 filter) __packed
92
93#include "trace_entries.h"
94
95/*
96 * syscalls are special, and need special handling, this is why
97 * they are not included in trace_entries.h
98 */
99struct syscall_trace_enter {
100 struct trace_entry ent;
101 int nr;
102 unsigned long args[];
103};
104
105struct syscall_trace_exit {
106 struct trace_entry ent;
107 int nr;
108 long ret;
109};
110
111struct kprobe_trace_entry_head {
112 struct trace_entry ent;
113 unsigned long ip;
114};
115
116struct kretprobe_trace_entry_head {
117 struct trace_entry ent;
118 unsigned long func;
119 unsigned long ret_ip;
120};
121
122/*
123 * trace_flag_type is an enumeration that holds different
124 * states when a trace occurs. These are:
125 * IRQS_OFF - interrupts were disabled
126 * IRQS_NOSUPPORT - arch does not support irqs_disabled_flags
127 * NEED_RESCHED - reschedule is requested
128 * HARDIRQ - inside an interrupt handler
129 * SOFTIRQ - inside a softirq handler
130 */
131enum trace_flag_type {
132 TRACE_FLAG_IRQS_OFF = 0x01,
133 TRACE_FLAG_IRQS_NOSUPPORT = 0x02,
134 TRACE_FLAG_NEED_RESCHED = 0x04,
135 TRACE_FLAG_HARDIRQ = 0x08,
136 TRACE_FLAG_SOFTIRQ = 0x10,
137 TRACE_FLAG_PREEMPT_RESCHED = 0x20,
138 TRACE_FLAG_NMI = 0x40,
139};
140
141#define TRACE_BUF_SIZE 1024
142
143struct trace_array;
144
145/*
146 * The CPU trace array - it consists of thousands of trace entries
147 * plus some other descriptor data: (for example which task started
148 * the trace, etc.)
149 */
150struct trace_array_cpu {
151 atomic_t disabled;
152 void *buffer_page; /* ring buffer spare */
153
154 unsigned long entries;
155 unsigned long saved_latency;
156 unsigned long critical_start;
157 unsigned long critical_end;
158 unsigned long critical_sequence;
159 unsigned long nice;
160 unsigned long policy;
161 unsigned long rt_priority;
162 unsigned long skipped_entries;
163 u64 preempt_timestamp;
164 pid_t pid;
165 kuid_t uid;
166 char comm[TASK_COMM_LEN];
167
168 bool ignore_pid;
169#ifdef CONFIG_FUNCTION_TRACER
170 bool ftrace_ignore_pid;
171#endif
172};
173
174struct tracer;
175struct trace_option_dentry;
176
177struct trace_buffer {
178 struct trace_array *tr;
179 struct ring_buffer *buffer;
180 struct trace_array_cpu __percpu *data;
181 u64 time_start;
182 int cpu;
183};
184
185#define TRACE_FLAGS_MAX_SIZE 32
186
187struct trace_options {
188 struct tracer *tracer;
189 struct trace_option_dentry *topts;
190};
191
192struct trace_pid_list {
193 int pid_max;
194 unsigned long *pids;
195};
196
197/*
198 * The trace array - an array of per-CPU trace arrays. This is the
199 * highest level data structure that individual tracers deal with.
200 * They have on/off state as well:
201 */
202struct trace_array {
203 struct list_head list;
204 char *name;
205 struct trace_buffer trace_buffer;
206#ifdef CONFIG_TRACER_MAX_TRACE
207 /*
208 * The max_buffer is used to snapshot the trace when a maximum
209 * latency is reached, or when the user initiates a snapshot.
210 * Some tracers will use this to store a maximum trace while
211 * it continues examining live traces.
212 *
213 * The buffers for the max_buffer are set up the same as the trace_buffer
214 * When a snapshot is taken, the buffer of the max_buffer is swapped
215 * with the buffer of the trace_buffer and the buffers are reset for
216 * the trace_buffer so the tracing can continue.
217 */
218 struct trace_buffer max_buffer;
219 bool allocated_snapshot;
220#endif
221#if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
222 unsigned long max_latency;
223#endif
224 struct trace_pid_list __rcu *filtered_pids;
225 /*
226 * max_lock is used to protect the swapping of buffers
227 * when taking a max snapshot. The buffers themselves are
228 * protected by per_cpu spinlocks. But the action of the swap
229 * needs its own lock.
230 *
231 * This is defined as a arch_spinlock_t in order to help
232 * with performance when lockdep debugging is enabled.
233 *
234 * It is also used in other places outside the update_max_tr
235 * so it needs to be defined outside of the
236 * CONFIG_TRACER_MAX_TRACE.
237 */
238 arch_spinlock_t max_lock;
239 int buffer_disabled;
240#ifdef CONFIG_FTRACE_SYSCALLS
241 int sys_refcount_enter;
242 int sys_refcount_exit;
243 struct trace_event_file __rcu *enter_syscall_files[NR_syscalls];
244 struct trace_event_file __rcu *exit_syscall_files[NR_syscalls];
245#endif
246 int stop_count;
247 int clock_id;
248 int nr_topts;
249 bool clear_trace;
250 struct tracer *current_trace;
251 unsigned int trace_flags;
252 unsigned char trace_flags_index[TRACE_FLAGS_MAX_SIZE];
253 unsigned int flags;
254 raw_spinlock_t start_lock;
255 struct dentry *dir;
256 struct dentry *options;
257 struct dentry *percpu_dir;
258 struct dentry *event_dir;
259 struct trace_options *topts;
260 struct list_head systems;
261 struct list_head events;
262 struct trace_event_file *trace_marker_file;
263 cpumask_var_t tracing_cpumask; /* only trace on set CPUs */
264 int ref;
265#ifdef CONFIG_FUNCTION_TRACER
266 struct ftrace_ops *ops;
267 struct trace_pid_list __rcu *function_pids;
268#ifdef CONFIG_DYNAMIC_FTRACE
269 /* All of these are protected by the ftrace_lock */
270 struct list_head func_probes;
271 struct list_head mod_trace;
272 struct list_head mod_notrace;
273#endif
274 /* function tracing enabled */
275 int function_enabled;
276#endif
277 int time_stamp_abs_ref;
278 struct list_head hist_vars;
279};
280
281enum {
282 TRACE_ARRAY_FL_GLOBAL = (1 << 0)
283};
284
285extern struct list_head ftrace_trace_arrays;
286
287extern struct mutex trace_types_lock;
288
289extern int trace_array_get(struct trace_array *tr);
290extern void trace_array_put(struct trace_array *tr);
291
292extern int tracing_set_time_stamp_abs(struct trace_array *tr, bool abs);
293extern int tracing_set_clock(struct trace_array *tr, const char *clockstr);
294
295extern bool trace_clock_in_ns(struct trace_array *tr);
296
297/*
298 * The global tracer (top) should be the first trace array added,
299 * but we check the flag anyway.
300 */
301static inline struct trace_array *top_trace_array(void)
302{
303 struct trace_array *tr;
304
305 if (list_empty(&ftrace_trace_arrays))
306 return NULL;
307
308 tr = list_entry(ftrace_trace_arrays.prev,
309 typeof(*tr), list);
310 WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
311 return tr;
312}
313
314#define FTRACE_CMP_TYPE(var, type) \
315 __builtin_types_compatible_p(typeof(var), type *)
316
317#undef IF_ASSIGN
318#define IF_ASSIGN(var, entry, etype, id) \
319 if (FTRACE_CMP_TYPE(var, etype)) { \
320 var = (typeof(var))(entry); \
321 WARN_ON(id && (entry)->type != id); \
322 break; \
323 }
324
325/* Will cause compile errors if type is not found. */
326extern void __ftrace_bad_type(void);
327
328/*
329 * The trace_assign_type is a verifier that the entry type is
330 * the same as the type being assigned. To add new types simply
331 * add a line with the following format:
332 *
333 * IF_ASSIGN(var, ent, type, id);
334 *
335 * Where "type" is the trace type that includes the trace_entry
336 * as the "ent" item. And "id" is the trace identifier that is
337 * used in the trace_type enum.
338 *
339 * If the type can have more than one id, then use zero.
340 */
341#define trace_assign_type(var, ent) \
342 do { \
343 IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN); \
344 IF_ASSIGN(var, ent, struct ctx_switch_entry, 0); \
345 IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK); \
346 IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
347 IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT); \
348 IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT); \
349 IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS); \
350 IF_ASSIGN(var, ent, struct hwlat_entry, TRACE_HWLAT); \
351 IF_ASSIGN(var, ent, struct raw_data_entry, TRACE_RAW_DATA);\
352 IF_ASSIGN(var, ent, struct trace_mmiotrace_rw, \
353 TRACE_MMIO_RW); \
354 IF_ASSIGN(var, ent, struct trace_mmiotrace_map, \
355 TRACE_MMIO_MAP); \
356 IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
357 IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry, \
358 TRACE_GRAPH_ENT); \
359 IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry, \
360 TRACE_GRAPH_RET); \
361 __ftrace_bad_type(); \
362 } while (0)
363
364/*
365 * An option specific to a tracer. This is a boolean value.
366 * The bit is the bit index that sets its value on the
367 * flags value in struct tracer_flags.
368 */
369struct tracer_opt {
370 const char *name; /* Will appear on the trace_options file */
371 u32 bit; /* Mask assigned in val field in tracer_flags */
372};
373
374/*
375 * The set of specific options for a tracer. Your tracer
376 * have to set the initial value of the flags val.
377 */
378struct tracer_flags {
379 u32 val;
380 struct tracer_opt *opts;
381 struct tracer *trace;
382};
383
384/* Makes more easy to define a tracer opt */
385#define TRACER_OPT(s, b) .name = #s, .bit = b
386
387
388struct trace_option_dentry {
389 struct tracer_opt *opt;
390 struct tracer_flags *flags;
391 struct trace_array *tr;
392 struct dentry *entry;
393};
394
395/**
396 * struct tracer - a specific tracer and its callbacks to interact with tracefs
397 * @name: the name chosen to select it on the available_tracers file
398 * @init: called when one switches to this tracer (echo name > current_tracer)
399 * @reset: called when one switches to another tracer
400 * @start: called when tracing is unpaused (echo 1 > tracing_on)
401 * @stop: called when tracing is paused (echo 0 > tracing_on)
402 * @update_thresh: called when tracing_thresh is updated
403 * @open: called when the trace file is opened
404 * @pipe_open: called when the trace_pipe file is opened
405 * @close: called when the trace file is released
406 * @pipe_close: called when the trace_pipe file is released
407 * @read: override the default read callback on trace_pipe
408 * @splice_read: override the default splice_read callback on trace_pipe
409 * @selftest: selftest to run on boot (see trace_selftest.c)
410 * @print_headers: override the first lines that describe your columns
411 * @print_line: callback that prints a trace
412 * @set_flag: signals one of your private flags changed (trace_options file)
413 * @flags: your private flags
414 */
415struct tracer {
416 const char *name;
417 int (*init)(struct trace_array *tr);
418 void (*reset)(struct trace_array *tr);
419 void (*start)(struct trace_array *tr);
420 void (*stop)(struct trace_array *tr);
421 int (*update_thresh)(struct trace_array *tr);
422 void (*open)(struct trace_iterator *iter);
423 void (*pipe_open)(struct trace_iterator *iter);
424 void (*close)(struct trace_iterator *iter);
425 void (*pipe_close)(struct trace_iterator *iter);
426 ssize_t (*read)(struct trace_iterator *iter,
427 struct file *filp, char __user *ubuf,
428 size_t cnt, loff_t *ppos);
429 ssize_t (*splice_read)(struct trace_iterator *iter,
430 struct file *filp,
431 loff_t *ppos,
432 struct pipe_inode_info *pipe,
433 size_t len,
434 unsigned int flags);
435#ifdef CONFIG_FTRACE_STARTUP_TEST
436 int (*selftest)(struct tracer *trace,
437 struct trace_array *tr);
438#endif
439 void (*print_header)(struct seq_file *m);
440 enum print_line_t (*print_line)(struct trace_iterator *iter);
441 /* If you handled the flag setting, return 0 */
442 int (*set_flag)(struct trace_array *tr,
443 u32 old_flags, u32 bit, int set);
444 /* Return 0 if OK with change, else return non-zero */
445 int (*flag_changed)(struct trace_array *tr,
446 u32 mask, int set);
447 struct tracer *next;
448 struct tracer_flags *flags;
449 int enabled;
450 int ref;
451 bool print_max;
452 bool allow_instances;
453#ifdef CONFIG_TRACER_MAX_TRACE
454 bool use_max_tr;
455#endif
456 /* True if tracer cannot be enabled in kernel param */
457 bool noboot;
458};
459
460
461/* Only current can touch trace_recursion */
462
463/*
464 * For function tracing recursion:
465 * The order of these bits are important.
466 *
467 * When function tracing occurs, the following steps are made:
468 * If arch does not support a ftrace feature:
469 * call internal function (uses INTERNAL bits) which calls...
470 * If callback is registered to the "global" list, the list
471 * function is called and recursion checks the GLOBAL bits.
472 * then this function calls...
473 * The function callback, which can use the FTRACE bits to
474 * check for recursion.
475 *
476 * Now if the arch does not suppport a feature, and it calls
477 * the global list function which calls the ftrace callback
478 * all three of these steps will do a recursion protection.
479 * There's no reason to do one if the previous caller already
480 * did. The recursion that we are protecting against will
481 * go through the same steps again.
482 *
483 * To prevent the multiple recursion checks, if a recursion
484 * bit is set that is higher than the MAX bit of the current
485 * check, then we know that the check was made by the previous
486 * caller, and we can skip the current check.
487 */
488enum {
489 TRACE_BUFFER_BIT,
490 TRACE_BUFFER_NMI_BIT,
491 TRACE_BUFFER_IRQ_BIT,
492 TRACE_BUFFER_SIRQ_BIT,
493
494 /* Start of function recursion bits */
495 TRACE_FTRACE_BIT,
496 TRACE_FTRACE_NMI_BIT,
497 TRACE_FTRACE_IRQ_BIT,
498 TRACE_FTRACE_SIRQ_BIT,
499
500 /* INTERNAL_BITs must be greater than FTRACE_BITs */
501 TRACE_INTERNAL_BIT,
502 TRACE_INTERNAL_NMI_BIT,
503 TRACE_INTERNAL_IRQ_BIT,
504 TRACE_INTERNAL_SIRQ_BIT,
505
506 TRACE_BRANCH_BIT,
507/*
508 * Abuse of the trace_recursion.
509 * As we need a way to maintain state if we are tracing the function
510 * graph in irq because we want to trace a particular function that
511 * was called in irq context but we have irq tracing off. Since this
512 * can only be modified by current, we can reuse trace_recursion.
513 */
514 TRACE_IRQ_BIT,
515
516 /* Set if the function is in the set_graph_function file */
517 TRACE_GRAPH_BIT,
518
519 /*
520 * In the very unlikely case that an interrupt came in
521 * at a start of graph tracing, and we want to trace
522 * the function in that interrupt, the depth can be greater
523 * than zero, because of the preempted start of a previous
524 * trace. In an even more unlikely case, depth could be 2
525 * if a softirq interrupted the start of graph tracing,
526 * followed by an interrupt preempting a start of graph
527 * tracing in the softirq, and depth can even be 3
528 * if an NMI came in at the start of an interrupt function
529 * that preempted a softirq start of a function that
530 * preempted normal context!!!! Luckily, it can't be
531 * greater than 3, so the next two bits are a mask
532 * of what the depth is when we set TRACE_GRAPH_BIT
533 */
534
535 TRACE_GRAPH_DEPTH_START_BIT,
536 TRACE_GRAPH_DEPTH_END_BIT,
537};
538
539#define trace_recursion_set(bit) do { (current)->trace_recursion |= (1<<(bit)); } while (0)
540#define trace_recursion_clear(bit) do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
541#define trace_recursion_test(bit) ((current)->trace_recursion & (1<<(bit)))
542
543#define trace_recursion_depth() \
544 (((current)->trace_recursion >> TRACE_GRAPH_DEPTH_START_BIT) & 3)
545#define trace_recursion_set_depth(depth) \
546 do { \
547 current->trace_recursion &= \
548 ~(3 << TRACE_GRAPH_DEPTH_START_BIT); \
549 current->trace_recursion |= \
550 ((depth) & 3) << TRACE_GRAPH_DEPTH_START_BIT; \
551 } while (0)
552
553#define TRACE_CONTEXT_BITS 4
554
555#define TRACE_FTRACE_START TRACE_FTRACE_BIT
556#define TRACE_FTRACE_MAX ((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
557
558#define TRACE_LIST_START TRACE_INTERNAL_BIT
559#define TRACE_LIST_MAX ((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
560
561#define TRACE_CONTEXT_MASK TRACE_LIST_MAX
562
563static __always_inline int trace_get_context_bit(void)
564{
565 int bit;
566
567 if (in_interrupt()) {
568 if (in_nmi())
569 bit = 0;
570
571 else if (in_irq())
572 bit = 1;
573 else
574 bit = 2;
575 } else
576 bit = 3;
577
578 return bit;
579}
580
581static __always_inline int trace_test_and_set_recursion(int start, int max)
582{
583 unsigned int val = current->trace_recursion;
584 int bit;
585
586 /* A previous recursion check was made */
587 if ((val & TRACE_CONTEXT_MASK) > max)
588 return 0;
589
590 bit = trace_get_context_bit() + start;
591 if (unlikely(val & (1 << bit)))
592 return -1;
593
594 val |= 1 << bit;
595 current->trace_recursion = val;
596 barrier();
597
598 return bit;
599}
600
601static __always_inline void trace_clear_recursion(int bit)
602{
603 unsigned int val = current->trace_recursion;
604
605 if (!bit)
606 return;
607
608 bit = 1 << bit;
609 val &= ~bit;
610
611 barrier();
612 current->trace_recursion = val;
613}
614
615static inline struct ring_buffer_iter *
616trace_buffer_iter(struct trace_iterator *iter, int cpu)
617{
618 return iter->buffer_iter ? iter->buffer_iter[cpu] : NULL;
619}
620
621int tracer_init(struct tracer *t, struct trace_array *tr);
622int tracing_is_enabled(void);
623void tracing_reset(struct trace_buffer *buf, int cpu);
624void tracing_reset_online_cpus(struct trace_buffer *buf);
625void tracing_reset_current(int cpu);
626void tracing_reset_all_online_cpus(void);
627int tracing_open_generic(struct inode *inode, struct file *filp);
628bool tracing_is_disabled(void);
629bool tracer_tracing_is_on(struct trace_array *tr);
630void tracer_tracing_on(struct trace_array *tr);
631void tracer_tracing_off(struct trace_array *tr);
632struct dentry *trace_create_file(const char *name,
633 umode_t mode,
634 struct dentry *parent,
635 void *data,
636 const struct file_operations *fops);
637
638struct dentry *tracing_init_dentry(void);
639
640struct ring_buffer_event;
641
642struct ring_buffer_event *
643trace_buffer_lock_reserve(struct ring_buffer *buffer,
644 int type,
645 unsigned long len,
646 unsigned long flags,
647 int pc);
648
649struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
650 struct trace_array_cpu *data);
651
652struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
653 int *ent_cpu, u64 *ent_ts);
654
655void trace_buffer_unlock_commit_nostack(struct ring_buffer *buffer,
656 struct ring_buffer_event *event);
657
658int trace_empty(struct trace_iterator *iter);
659
660void *trace_find_next_entry_inc(struct trace_iterator *iter);
661
662void trace_init_global_iter(struct trace_iterator *iter);
663
664void tracing_iter_reset(struct trace_iterator *iter, int cpu);
665
666void trace_function(struct trace_array *tr,
667 unsigned long ip,
668 unsigned long parent_ip,
669 unsigned long flags, int pc);
670void trace_graph_function(struct trace_array *tr,
671 unsigned long ip,
672 unsigned long parent_ip,
673 unsigned long flags, int pc);
674void trace_latency_header(struct seq_file *m);
675void trace_default_header(struct seq_file *m);
676void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
677int trace_empty(struct trace_iterator *iter);
678
679void trace_graph_return(struct ftrace_graph_ret *trace);
680int trace_graph_entry(struct ftrace_graph_ent *trace);
681void set_graph_array(struct trace_array *tr);
682
683void tracing_start_cmdline_record(void);
684void tracing_stop_cmdline_record(void);
685void tracing_start_tgid_record(void);
686void tracing_stop_tgid_record(void);
687
688int register_tracer(struct tracer *type);
689int is_tracing_stopped(void);
690
691loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
692
693extern cpumask_var_t __read_mostly tracing_buffer_mask;
694
695#define for_each_tracing_cpu(cpu) \
696 for_each_cpu(cpu, tracing_buffer_mask)
697
698extern unsigned long nsecs_to_usecs(unsigned long nsecs);
699
700extern unsigned long tracing_thresh;
701
702/* PID filtering */
703
704extern int pid_max;
705
706bool trace_find_filtered_pid(struct trace_pid_list *filtered_pids,
707 pid_t search_pid);
708bool trace_ignore_this_task(struct trace_pid_list *filtered_pids,
709 struct task_struct *task);
710void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
711 struct task_struct *self,
712 struct task_struct *task);
713void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos);
714void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos);
715int trace_pid_show(struct seq_file *m, void *v);
716void trace_free_pid_list(struct trace_pid_list *pid_list);
717int trace_pid_write(struct trace_pid_list *filtered_pids,
718 struct trace_pid_list **new_pid_list,
719 const char __user *ubuf, size_t cnt);
720
721#ifdef CONFIG_TRACER_MAX_TRACE
722void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
723void update_max_tr_single(struct trace_array *tr,
724 struct task_struct *tsk, int cpu);
725#endif /* CONFIG_TRACER_MAX_TRACE */
726
727#ifdef CONFIG_STACKTRACE
728void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
729 int pc);
730
731void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
732 int pc);
733#else
734static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
735 unsigned long flags, int pc)
736{
737}
738
739static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
740 int skip, int pc)
741{
742}
743#endif /* CONFIG_STACKTRACE */
744
745extern u64 ftrace_now(int cpu);
746
747extern void trace_find_cmdline(int pid, char comm[]);
748extern int trace_find_tgid(int pid);
749extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
750
751#ifdef CONFIG_DYNAMIC_FTRACE
752extern unsigned long ftrace_update_tot_cnt;
753void ftrace_init_trace_array(struct trace_array *tr);
754#else
755static inline void ftrace_init_trace_array(struct trace_array *tr) { }
756#endif
757#define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
758extern int DYN_FTRACE_TEST_NAME(void);
759#define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
760extern int DYN_FTRACE_TEST_NAME2(void);
761
762extern bool ring_buffer_expanded;
763extern bool tracing_selftest_disabled;
764
765#ifdef CONFIG_FTRACE_STARTUP_TEST
766extern int trace_selftest_startup_function(struct tracer *trace,
767 struct trace_array *tr);
768extern int trace_selftest_startup_function_graph(struct tracer *trace,
769 struct trace_array *tr);
770extern int trace_selftest_startup_irqsoff(struct tracer *trace,
771 struct trace_array *tr);
772extern int trace_selftest_startup_preemptoff(struct tracer *trace,
773 struct trace_array *tr);
774extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
775 struct trace_array *tr);
776extern int trace_selftest_startup_wakeup(struct tracer *trace,
777 struct trace_array *tr);
778extern int trace_selftest_startup_nop(struct tracer *trace,
779 struct trace_array *tr);
780extern int trace_selftest_startup_branch(struct tracer *trace,
781 struct trace_array *tr);
782/*
783 * Tracer data references selftest functions that only occur
784 * on boot up. These can be __init functions. Thus, when selftests
785 * are enabled, then the tracers need to reference __init functions.
786 */
787#define __tracer_data __refdata
788#else
789/* Tracers are seldom changed. Optimize when selftests are disabled. */
790#define __tracer_data __read_mostly
791#endif /* CONFIG_FTRACE_STARTUP_TEST */
792
793extern void *head_page(struct trace_array_cpu *data);
794extern unsigned long long ns2usecs(u64 nsec);
795extern int
796trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
797extern int
798trace_vprintk(unsigned long ip, const char *fmt, va_list args);
799extern int
800trace_array_vprintk(struct trace_array *tr,
801 unsigned long ip, const char *fmt, va_list args);
802int trace_array_printk(struct trace_array *tr,
803 unsigned long ip, const char *fmt, ...);
804int trace_array_printk_buf(struct ring_buffer *buffer,
805 unsigned long ip, const char *fmt, ...);
806void trace_printk_seq(struct trace_seq *s);
807enum print_line_t print_trace_line(struct trace_iterator *iter);
808
809extern char trace_find_mark(unsigned long long duration);
810
811struct ftrace_hash;
812
813struct ftrace_mod_load {
814 struct list_head list;
815 char *func;
816 char *module;
817 int enable;
818};
819
820enum {
821 FTRACE_HASH_FL_MOD = (1 << 0),
822};
823
824struct ftrace_hash {
825 unsigned long size_bits;
826 struct hlist_head *buckets;
827 unsigned long count;
828 unsigned long flags;
829 struct rcu_head rcu;
830};
831
832struct ftrace_func_entry *
833ftrace_lookup_ip(struct ftrace_hash *hash, unsigned long ip);
834
835static __always_inline bool ftrace_hash_empty(struct ftrace_hash *hash)
836{
837 return !hash || !(hash->count || (hash->flags & FTRACE_HASH_FL_MOD));
838}
839
840/* Standard output formatting function used for function return traces */
841#ifdef CONFIG_FUNCTION_GRAPH_TRACER
842
843/* Flag options */
844#define TRACE_GRAPH_PRINT_OVERRUN 0x1
845#define TRACE_GRAPH_PRINT_CPU 0x2
846#define TRACE_GRAPH_PRINT_OVERHEAD 0x4
847#define TRACE_GRAPH_PRINT_PROC 0x8
848#define TRACE_GRAPH_PRINT_DURATION 0x10
849#define TRACE_GRAPH_PRINT_ABS_TIME 0x20
850#define TRACE_GRAPH_PRINT_IRQS 0x40
851#define TRACE_GRAPH_PRINT_TAIL 0x80
852#define TRACE_GRAPH_SLEEP_TIME 0x100
853#define TRACE_GRAPH_GRAPH_TIME 0x200
854#define TRACE_GRAPH_PRINT_FILL_SHIFT 28
855#define TRACE_GRAPH_PRINT_FILL_MASK (0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
856
857extern void ftrace_graph_sleep_time_control(bool enable);
858extern void ftrace_graph_graph_time_control(bool enable);
859
860extern enum print_line_t
861print_graph_function_flags(struct trace_iterator *iter, u32 flags);
862extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
863extern void
864trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
865extern void graph_trace_open(struct trace_iterator *iter);
866extern void graph_trace_close(struct trace_iterator *iter);
867extern int __trace_graph_entry(struct trace_array *tr,
868 struct ftrace_graph_ent *trace,
869 unsigned long flags, int pc);
870extern void __trace_graph_return(struct trace_array *tr,
871 struct ftrace_graph_ret *trace,
872 unsigned long flags, int pc);
873
874#ifdef CONFIG_DYNAMIC_FTRACE
875extern struct ftrace_hash *ftrace_graph_hash;
876extern struct ftrace_hash *ftrace_graph_notrace_hash;
877
878static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
879{
880 unsigned long addr = trace->func;
881 int ret = 0;
882
883 preempt_disable_notrace();
884
885 if (ftrace_hash_empty(ftrace_graph_hash)) {
886 ret = 1;
887 goto out;
888 }
889
890 if (ftrace_lookup_ip(ftrace_graph_hash, addr)) {
891
892 /*
893 * This needs to be cleared on the return functions
894 * when the depth is zero.
895 */
896 trace_recursion_set(TRACE_GRAPH_BIT);
897 trace_recursion_set_depth(trace->depth);
898
899 /*
900 * If no irqs are to be traced, but a set_graph_function
901 * is set, and called by an interrupt handler, we still
902 * want to trace it.
903 */
904 if (in_irq())
905 trace_recursion_set(TRACE_IRQ_BIT);
906 else
907 trace_recursion_clear(TRACE_IRQ_BIT);
908 ret = 1;
909 }
910
911out:
912 preempt_enable_notrace();
913 return ret;
914}
915
916static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace)
917{
918 if (trace_recursion_test(TRACE_GRAPH_BIT) &&
919 trace->depth == trace_recursion_depth())
920 trace_recursion_clear(TRACE_GRAPH_BIT);
921}
922
923static inline int ftrace_graph_notrace_addr(unsigned long addr)
924{
925 int ret = 0;
926
927 preempt_disable_notrace();
928
929 if (ftrace_lookup_ip(ftrace_graph_notrace_hash, addr))
930 ret = 1;
931
932 preempt_enable_notrace();
933 return ret;
934}
935#else
936static inline int ftrace_graph_addr(struct ftrace_graph_ent *trace)
937{
938 return 1;
939}
940
941static inline int ftrace_graph_notrace_addr(unsigned long addr)
942{
943 return 0;
944}
945static inline void ftrace_graph_addr_finish(struct ftrace_graph_ret *trace)
946{ }
947#endif /* CONFIG_DYNAMIC_FTRACE */
948
949extern unsigned int fgraph_max_depth;
950
951static inline bool ftrace_graph_ignore_func(struct ftrace_graph_ent *trace)
952{
953 /* trace it when it is-nested-in or is a function enabled. */
954 return !(trace_recursion_test(TRACE_GRAPH_BIT) ||
955 ftrace_graph_addr(trace)) ||
956 (trace->depth < 0) ||
957 (fgraph_max_depth && trace->depth >= fgraph_max_depth);
958}
959
960#else /* CONFIG_FUNCTION_GRAPH_TRACER */
961static inline enum print_line_t
962print_graph_function_flags(struct trace_iterator *iter, u32 flags)
963{
964 return TRACE_TYPE_UNHANDLED;
965}
966#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
967
968extern struct list_head ftrace_pids;
969
970#ifdef CONFIG_FUNCTION_TRACER
971struct ftrace_func_command {
972 struct list_head list;
973 char *name;
974 int (*func)(struct trace_array *tr,
975 struct ftrace_hash *hash,
976 char *func, char *cmd,
977 char *params, int enable);
978};
979extern bool ftrace_filter_param __initdata;
980static inline int ftrace_trace_task(struct trace_array *tr)
981{
982 return !this_cpu_read(tr->trace_buffer.data->ftrace_ignore_pid);
983}
984extern int ftrace_is_dead(void);
985int ftrace_create_function_files(struct trace_array *tr,
986 struct dentry *parent);
987void ftrace_destroy_function_files(struct trace_array *tr);
988void ftrace_init_global_array_ops(struct trace_array *tr);
989void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
990void ftrace_reset_array_ops(struct trace_array *tr);
991void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
992void ftrace_init_tracefs_toplevel(struct trace_array *tr,
993 struct dentry *d_tracer);
994void ftrace_clear_pids(struct trace_array *tr);
995int init_function_trace(void);
996void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
997#else
998static inline int ftrace_trace_task(struct trace_array *tr)
999{
1000 return 1;
1001}
1002static inline int ftrace_is_dead(void) { return 0; }
1003static inline int
1004ftrace_create_function_files(struct trace_array *tr,
1005 struct dentry *parent)
1006{
1007 return 0;
1008}
1009static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
1010static inline __init void
1011ftrace_init_global_array_ops(struct trace_array *tr) { }
1012static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
1013static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
1014static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
1015static inline void ftrace_clear_pids(struct trace_array *tr) { }
1016static inline int init_function_trace(void) { return 0; }
1017static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
1018/* ftace_func_t type is not defined, use macro instead of static inline */
1019#define ftrace_init_array_ops(tr, func) do { } while (0)
1020#endif /* CONFIG_FUNCTION_TRACER */
1021
1022#if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
1023
1024struct ftrace_probe_ops {
1025 void (*func)(unsigned long ip,
1026 unsigned long parent_ip,
1027 struct trace_array *tr,
1028 struct ftrace_probe_ops *ops,
1029 void *data);
1030 int (*init)(struct ftrace_probe_ops *ops,
1031 struct trace_array *tr,
1032 unsigned long ip, void *init_data,
1033 void **data);
1034 void (*free)(struct ftrace_probe_ops *ops,
1035 struct trace_array *tr,
1036 unsigned long ip, void *data);
1037 int (*print)(struct seq_file *m,
1038 unsigned long ip,
1039 struct ftrace_probe_ops *ops,
1040 void *data);
1041};
1042
1043struct ftrace_func_mapper;
1044typedef int (*ftrace_mapper_func)(void *data);
1045
1046struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
1047void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
1048 unsigned long ip);
1049int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
1050 unsigned long ip, void *data);
1051void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
1052 unsigned long ip);
1053void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
1054 ftrace_mapper_func free_func);
1055
1056extern int
1057register_ftrace_function_probe(char *glob, struct trace_array *tr,
1058 struct ftrace_probe_ops *ops, void *data);
1059extern int
1060unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
1061 struct ftrace_probe_ops *ops);
1062extern void clear_ftrace_function_probes(struct trace_array *tr);
1063
1064int register_ftrace_command(struct ftrace_func_command *cmd);
1065int unregister_ftrace_command(struct ftrace_func_command *cmd);
1066
1067void ftrace_create_filter_files(struct ftrace_ops *ops,
1068 struct dentry *parent);
1069void ftrace_destroy_filter_files(struct ftrace_ops *ops);
1070#else
1071struct ftrace_func_command;
1072
1073static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
1074{
1075 return -EINVAL;
1076}
1077static inline __init int unregister_ftrace_command(char *cmd_name)
1078{
1079 return -EINVAL;
1080}
1081static inline void clear_ftrace_function_probes(struct trace_array *tr)
1082{
1083}
1084
1085/*
1086 * The ops parameter passed in is usually undefined.
1087 * This must be a macro.
1088 */
1089#define ftrace_create_filter_files(ops, parent) do { } while (0)
1090#define ftrace_destroy_filter_files(ops) do { } while (0)
1091#endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
1092
1093bool ftrace_event_is_function(struct trace_event_call *call);
1094
1095/*
1096 * struct trace_parser - servers for reading the user input separated by spaces
1097 * @cont: set if the input is not complete - no final space char was found
1098 * @buffer: holds the parsed user input
1099 * @idx: user input length
1100 * @size: buffer size
1101 */
1102struct trace_parser {
1103 bool cont;
1104 char *buffer;
1105 unsigned idx;
1106 unsigned size;
1107};
1108
1109static inline bool trace_parser_loaded(struct trace_parser *parser)
1110{
1111 return (parser->idx != 0);
1112}
1113
1114static inline bool trace_parser_cont(struct trace_parser *parser)
1115{
1116 return parser->cont;
1117}
1118
1119static inline void trace_parser_clear(struct trace_parser *parser)
1120{
1121 parser->cont = false;
1122 parser->idx = 0;
1123}
1124
1125extern int trace_parser_get_init(struct trace_parser *parser, int size);
1126extern void trace_parser_put(struct trace_parser *parser);
1127extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1128 size_t cnt, loff_t *ppos);
1129
1130/*
1131 * Only create function graph options if function graph is configured.
1132 */
1133#ifdef CONFIG_FUNCTION_GRAPH_TRACER
1134# define FGRAPH_FLAGS \
1135 C(DISPLAY_GRAPH, "display-graph"),
1136#else
1137# define FGRAPH_FLAGS
1138#endif
1139
1140#ifdef CONFIG_BRANCH_TRACER
1141# define BRANCH_FLAGS \
1142 C(BRANCH, "branch"),
1143#else
1144# define BRANCH_FLAGS
1145#endif
1146
1147#ifdef CONFIG_FUNCTION_TRACER
1148# define FUNCTION_FLAGS \
1149 C(FUNCTION, "function-trace"), \
1150 C(FUNC_FORK, "function-fork"),
1151# define FUNCTION_DEFAULT_FLAGS TRACE_ITER_FUNCTION
1152#else
1153# define FUNCTION_FLAGS
1154# define FUNCTION_DEFAULT_FLAGS 0UL
1155# define TRACE_ITER_FUNC_FORK 0UL
1156#endif
1157
1158#ifdef CONFIG_STACKTRACE
1159# define STACK_FLAGS \
1160 C(STACKTRACE, "stacktrace"),
1161#else
1162# define STACK_FLAGS
1163#endif
1164
1165/*
1166 * trace_iterator_flags is an enumeration that defines bit
1167 * positions into trace_flags that controls the output.
1168 *
1169 * NOTE: These bits must match the trace_options array in
1170 * trace.c (this macro guarantees it).
1171 */
1172#define TRACE_FLAGS \
1173 C(PRINT_PARENT, "print-parent"), \
1174 C(SYM_OFFSET, "sym-offset"), \
1175 C(SYM_ADDR, "sym-addr"), \
1176 C(VERBOSE, "verbose"), \
1177 C(RAW, "raw"), \
1178 C(HEX, "hex"), \
1179 C(BIN, "bin"), \
1180 C(BLOCK, "block"), \
1181 C(PRINTK, "trace_printk"), \
1182 C(ANNOTATE, "annotate"), \
1183 C(USERSTACKTRACE, "userstacktrace"), \
1184 C(SYM_USEROBJ, "sym-userobj"), \
1185 C(PRINTK_MSGONLY, "printk-msg-only"), \
1186 C(CONTEXT_INFO, "context-info"), /* Print pid/cpu/time */ \
1187 C(LATENCY_FMT, "latency-format"), \
1188 C(RECORD_CMD, "record-cmd"), \
1189 C(RECORD_TGID, "record-tgid"), \
1190 C(OVERWRITE, "overwrite"), \
1191 C(STOP_ON_FREE, "disable_on_free"), \
1192 C(IRQ_INFO, "irq-info"), \
1193 C(MARKERS, "markers"), \
1194 C(EVENT_FORK, "event-fork"), \
1195 FUNCTION_FLAGS \
1196 FGRAPH_FLAGS \
1197 STACK_FLAGS \
1198 BRANCH_FLAGS
1199
1200/*
1201 * By defining C, we can make TRACE_FLAGS a list of bit names
1202 * that will define the bits for the flag masks.
1203 */
1204#undef C
1205#define C(a, b) TRACE_ITER_##a##_BIT
1206
1207enum trace_iterator_bits {
1208 TRACE_FLAGS
1209 /* Make sure we don't go more than we have bits for */
1210 TRACE_ITER_LAST_BIT
1211};
1212
1213/*
1214 * By redefining C, we can make TRACE_FLAGS a list of masks that
1215 * use the bits as defined above.
1216 */
1217#undef C
1218#define C(a, b) TRACE_ITER_##a = (1 << TRACE_ITER_##a##_BIT)
1219
1220enum trace_iterator_flags { TRACE_FLAGS };
1221
1222/*
1223 * TRACE_ITER_SYM_MASK masks the options in trace_flags that
1224 * control the output of kernel symbols.
1225 */
1226#define TRACE_ITER_SYM_MASK \
1227 (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
1228
1229extern struct tracer nop_trace;
1230
1231#ifdef CONFIG_BRANCH_TRACER
1232extern int enable_branch_tracing(struct trace_array *tr);
1233extern void disable_branch_tracing(void);
1234static inline int trace_branch_enable(struct trace_array *tr)
1235{
1236 if (tr->trace_flags & TRACE_ITER_BRANCH)
1237 return enable_branch_tracing(tr);
1238 return 0;
1239}
1240static inline void trace_branch_disable(void)
1241{
1242 /* due to races, always disable */
1243 disable_branch_tracing();
1244}
1245#else
1246static inline int trace_branch_enable(struct trace_array *tr)
1247{
1248 return 0;
1249}
1250static inline void trace_branch_disable(void)
1251{
1252}
1253#endif /* CONFIG_BRANCH_TRACER */
1254
1255/* set ring buffers to default size if not already done so */
1256int tracing_update_buffers(void);
1257
1258struct ftrace_event_field {
1259 struct list_head link;
1260 const char *name;
1261 const char *type;
1262 int filter_type;
1263 int offset;
1264 int size;
1265 int is_signed;
1266};
1267
1268struct prog_entry;
1269
1270struct event_filter {
1271 struct prog_entry __rcu *prog;
1272 char *filter_string;
1273};
1274
1275struct event_subsystem {
1276 struct list_head list;
1277 const char *name;
1278 struct event_filter *filter;
1279 int ref_count;
1280};
1281
1282struct trace_subsystem_dir {
1283 struct list_head list;
1284 struct event_subsystem *subsystem;
1285 struct trace_array *tr;
1286 struct dentry *entry;
1287 int ref_count;
1288 int nr_events;
1289};
1290
1291extern int call_filter_check_discard(struct trace_event_call *call, void *rec,
1292 struct ring_buffer *buffer,
1293 struct ring_buffer_event *event);
1294
1295void trace_buffer_unlock_commit_regs(struct trace_array *tr,
1296 struct ring_buffer *buffer,
1297 struct ring_buffer_event *event,
1298 unsigned long flags, int pc,
1299 struct pt_regs *regs);
1300
1301static inline void trace_buffer_unlock_commit(struct trace_array *tr,
1302 struct ring_buffer *buffer,
1303 struct ring_buffer_event *event,
1304 unsigned long flags, int pc)
1305{
1306 trace_buffer_unlock_commit_regs(tr, buffer, event, flags, pc, NULL);
1307}
1308
1309DECLARE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
1310DECLARE_PER_CPU(int, trace_buffered_event_cnt);
1311void trace_buffered_event_disable(void);
1312void trace_buffered_event_enable(void);
1313
1314static inline void
1315__trace_event_discard_commit(struct ring_buffer *buffer,
1316 struct ring_buffer_event *event)
1317{
1318 if (this_cpu_read(trace_buffered_event) == event) {
1319 /* Simply release the temp buffer */
1320 this_cpu_dec(trace_buffered_event_cnt);
1321 return;
1322 }
1323 ring_buffer_discard_commit(buffer, event);
1324}
1325
1326/*
1327 * Helper function for event_trigger_unlock_commit{_regs}().
1328 * If there are event triggers attached to this event that requires
1329 * filtering against its fields, then they wil be called as the
1330 * entry already holds the field information of the current event.
1331 *
1332 * It also checks if the event should be discarded or not.
1333 * It is to be discarded if the event is soft disabled and the
1334 * event was only recorded to process triggers, or if the event
1335 * filter is active and this event did not match the filters.
1336 *
1337 * Returns true if the event is discarded, false otherwise.
1338 */
1339static inline bool
1340__event_trigger_test_discard(struct trace_event_file *file,
1341 struct ring_buffer *buffer,
1342 struct ring_buffer_event *event,
1343 void *entry,
1344 enum event_trigger_type *tt)
1345{
1346 unsigned long eflags = file->flags;
1347
1348 if (eflags & EVENT_FILE_FL_TRIGGER_COND)
1349 *tt = event_triggers_call(file, entry, event);
1350
1351 if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
1352 (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
1353 !filter_match_preds(file->filter, entry))) {
1354 __trace_event_discard_commit(buffer, event);
1355 return true;
1356 }
1357
1358 return false;
1359}
1360
1361/**
1362 * event_trigger_unlock_commit - handle triggers and finish event commit
1363 * @file: The file pointer assoctiated to the event
1364 * @buffer: The ring buffer that the event is being written to
1365 * @event: The event meta data in the ring buffer
1366 * @entry: The event itself
1367 * @irq_flags: The state of the interrupts at the start of the event
1368 * @pc: The state of the preempt count at the start of the event.
1369 *
1370 * This is a helper function to handle triggers that require data
1371 * from the event itself. It also tests the event against filters and
1372 * if the event is soft disabled and should be discarded.
1373 */
1374static inline void
1375event_trigger_unlock_commit(struct trace_event_file *file,
1376 struct ring_buffer *buffer,
1377 struct ring_buffer_event *event,
1378 void *entry, unsigned long irq_flags, int pc)
1379{
1380 enum event_trigger_type tt = ETT_NONE;
1381
1382 if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1383 trace_buffer_unlock_commit(file->tr, buffer, event, irq_flags, pc);
1384
1385 if (tt)
1386 event_triggers_post_call(file, tt);
1387}
1388
1389/**
1390 * event_trigger_unlock_commit_regs - handle triggers and finish event commit
1391 * @file: The file pointer assoctiated to the event
1392 * @buffer: The ring buffer that the event is being written to
1393 * @event: The event meta data in the ring buffer
1394 * @entry: The event itself
1395 * @irq_flags: The state of the interrupts at the start of the event
1396 * @pc: The state of the preempt count at the start of the event.
1397 *
1398 * This is a helper function to handle triggers that require data
1399 * from the event itself. It also tests the event against filters and
1400 * if the event is soft disabled and should be discarded.
1401 *
1402 * Same as event_trigger_unlock_commit() but calls
1403 * trace_buffer_unlock_commit_regs() instead of trace_buffer_unlock_commit().
1404 */
1405static inline void
1406event_trigger_unlock_commit_regs(struct trace_event_file *file,
1407 struct ring_buffer *buffer,
1408 struct ring_buffer_event *event,
1409 void *entry, unsigned long irq_flags, int pc,
1410 struct pt_regs *regs)
1411{
1412 enum event_trigger_type tt = ETT_NONE;
1413
1414 if (!__event_trigger_test_discard(file, buffer, event, entry, &tt))
1415 trace_buffer_unlock_commit_regs(file->tr, buffer, event,
1416 irq_flags, pc, regs);
1417
1418 if (tt)
1419 event_triggers_post_call(file, tt);
1420}
1421
1422#define FILTER_PRED_INVALID ((unsigned short)-1)
1423#define FILTER_PRED_IS_RIGHT (1 << 15)
1424#define FILTER_PRED_FOLD (1 << 15)
1425
1426/*
1427 * The max preds is the size of unsigned short with
1428 * two flags at the MSBs. One bit is used for both the IS_RIGHT
1429 * and FOLD flags. The other is reserved.
1430 *
1431 * 2^14 preds is way more than enough.
1432 */
1433#define MAX_FILTER_PRED 16384
1434
1435struct filter_pred;
1436struct regex;
1437
1438typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1439
1440typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1441
1442enum regex_type {
1443 MATCH_FULL = 0,
1444 MATCH_FRONT_ONLY,
1445 MATCH_MIDDLE_ONLY,
1446 MATCH_END_ONLY,
1447 MATCH_GLOB,
1448};
1449
1450struct regex {
1451 char pattern[MAX_FILTER_STR_VAL];
1452 int len;
1453 int field_len;
1454 regex_match_func match;
1455};
1456
1457struct filter_pred {
1458 filter_pred_fn_t fn;
1459 u64 val;
1460 struct regex regex;
1461 unsigned short *ops;
1462 struct ftrace_event_field *field;
1463 int offset;
1464 int not;
1465 int op;
1466};
1467
1468static inline bool is_string_field(struct ftrace_event_field *field)
1469{
1470 return field->filter_type == FILTER_DYN_STRING ||
1471 field->filter_type == FILTER_STATIC_STRING ||
1472 field->filter_type == FILTER_PTR_STRING ||
1473 field->filter_type == FILTER_COMM;
1474}
1475
1476static inline bool is_function_field(struct ftrace_event_field *field)
1477{
1478 return field->filter_type == FILTER_TRACE_FN;
1479}
1480
1481extern enum regex_type
1482filter_parse_regex(char *buff, int len, char **search, int *not);
1483extern void print_event_filter(struct trace_event_file *file,
1484 struct trace_seq *s);
1485extern int apply_event_filter(struct trace_event_file *file,
1486 char *filter_string);
1487extern int apply_subsystem_event_filter(struct trace_subsystem_dir *dir,
1488 char *filter_string);
1489extern void print_subsystem_event_filter(struct event_subsystem *system,
1490 struct trace_seq *s);
1491extern int filter_assign_type(const char *type);
1492extern int create_event_filter(struct trace_event_call *call,
1493 char *filter_str, bool set_str,
1494 struct event_filter **filterp);
1495extern void free_event_filter(struct event_filter *filter);
1496
1497struct ftrace_event_field *
1498trace_find_event_field(struct trace_event_call *call, char *name);
1499
1500extern void trace_event_enable_cmd_record(bool enable);
1501extern void trace_event_enable_tgid_record(bool enable);
1502
1503extern int event_trace_init(void);
1504extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1505extern int event_trace_del_tracer(struct trace_array *tr);
1506
1507extern struct trace_event_file *__find_event_file(struct trace_array *tr,
1508 const char *system,
1509 const char *event);
1510extern struct trace_event_file *find_event_file(struct trace_array *tr,
1511 const char *system,
1512 const char *event);
1513
1514static inline void *event_file_data(struct file *filp)
1515{
1516 return READ_ONCE(file_inode(filp)->i_private);
1517}
1518
1519extern struct mutex event_mutex;
1520extern struct list_head ftrace_events;
1521
1522extern const struct file_operations event_trigger_fops;
1523extern const struct file_operations event_hist_fops;
1524
1525#ifdef CONFIG_HIST_TRIGGERS
1526extern int register_trigger_hist_cmd(void);
1527extern int register_trigger_hist_enable_disable_cmds(void);
1528#else
1529static inline int register_trigger_hist_cmd(void) { return 0; }
1530static inline int register_trigger_hist_enable_disable_cmds(void) { return 0; }
1531#endif
1532
1533extern int register_trigger_cmds(void);
1534extern void clear_event_triggers(struct trace_array *tr);
1535
1536struct event_trigger_data {
1537 unsigned long count;
1538 int ref;
1539 struct event_trigger_ops *ops;
1540 struct event_command *cmd_ops;
1541 struct event_filter __rcu *filter;
1542 char *filter_str;
1543 void *private_data;
1544 bool paused;
1545 bool paused_tmp;
1546 struct list_head list;
1547 char *name;
1548 struct list_head named_list;
1549 struct event_trigger_data *named_data;
1550};
1551
1552/* Avoid typos */
1553#define ENABLE_EVENT_STR "enable_event"
1554#define DISABLE_EVENT_STR "disable_event"
1555#define ENABLE_HIST_STR "enable_hist"
1556#define DISABLE_HIST_STR "disable_hist"
1557
1558struct enable_trigger_data {
1559 struct trace_event_file *file;
1560 bool enable;
1561 bool hist;
1562};
1563
1564extern int event_enable_trigger_print(struct seq_file *m,
1565 struct event_trigger_ops *ops,
1566 struct event_trigger_data *data);
1567extern void event_enable_trigger_free(struct event_trigger_ops *ops,
1568 struct event_trigger_data *data);
1569extern int event_enable_trigger_func(struct event_command *cmd_ops,
1570 struct trace_event_file *file,
1571 char *glob, char *cmd, char *param);
1572extern int event_enable_register_trigger(char *glob,
1573 struct event_trigger_ops *ops,
1574 struct event_trigger_data *data,
1575 struct trace_event_file *file);
1576extern void event_enable_unregister_trigger(char *glob,
1577 struct event_trigger_ops *ops,
1578 struct event_trigger_data *test,
1579 struct trace_event_file *file);
1580extern void trigger_data_free(struct event_trigger_data *data);
1581extern int event_trigger_init(struct event_trigger_ops *ops,
1582 struct event_trigger_data *data);
1583extern int trace_event_trigger_enable_disable(struct trace_event_file *file,
1584 int trigger_enable);
1585extern void update_cond_flag(struct trace_event_file *file);
1586extern int set_trigger_filter(char *filter_str,
1587 struct event_trigger_data *trigger_data,
1588 struct trace_event_file *file);
1589extern struct event_trigger_data *find_named_trigger(const char *name);
1590extern bool is_named_trigger(struct event_trigger_data *test);
1591extern int save_named_trigger(const char *name,
1592 struct event_trigger_data *data);
1593extern void del_named_trigger(struct event_trigger_data *data);
1594extern void pause_named_trigger(struct event_trigger_data *data);
1595extern void unpause_named_trigger(struct event_trigger_data *data);
1596extern void set_named_trigger_data(struct event_trigger_data *data,
1597 struct event_trigger_data *named_data);
1598extern struct event_trigger_data *
1599get_named_trigger_data(struct event_trigger_data *data);
1600extern int register_event_command(struct event_command *cmd);
1601extern int unregister_event_command(struct event_command *cmd);
1602extern int register_trigger_hist_enable_disable_cmds(void);
1603
1604/**
1605 * struct event_trigger_ops - callbacks for trace event triggers
1606 *
1607 * The methods in this structure provide per-event trigger hooks for
1608 * various trigger operations.
1609 *
1610 * All the methods below, except for @init() and @free(), must be
1611 * implemented.
1612 *
1613 * @func: The trigger 'probe' function called when the triggering
1614 * event occurs. The data passed into this callback is the data
1615 * that was supplied to the event_command @reg() function that
1616 * registered the trigger (see struct event_command) along with
1617 * the trace record, rec.
1618 *
1619 * @init: An optional initialization function called for the trigger
1620 * when the trigger is registered (via the event_command reg()
1621 * function). This can be used to perform per-trigger
1622 * initialization such as incrementing a per-trigger reference
1623 * count, for instance. This is usually implemented by the
1624 * generic utility function @event_trigger_init() (see
1625 * trace_event_triggers.c).
1626 *
1627 * @free: An optional de-initialization function called for the
1628 * trigger when the trigger is unregistered (via the
1629 * event_command @reg() function). This can be used to perform
1630 * per-trigger de-initialization such as decrementing a
1631 * per-trigger reference count and freeing corresponding trigger
1632 * data, for instance. This is usually implemented by the
1633 * generic utility function @event_trigger_free() (see
1634 * trace_event_triggers.c).
1635 *
1636 * @print: The callback function invoked to have the trigger print
1637 * itself. This is usually implemented by a wrapper function
1638 * that calls the generic utility function @event_trigger_print()
1639 * (see trace_event_triggers.c).
1640 */
1641struct event_trigger_ops {
1642 void (*func)(struct event_trigger_data *data,
1643 void *rec,
1644 struct ring_buffer_event *rbe);
1645 int (*init)(struct event_trigger_ops *ops,
1646 struct event_trigger_data *data);
1647 void (*free)(struct event_trigger_ops *ops,
1648 struct event_trigger_data *data);
1649 int (*print)(struct seq_file *m,
1650 struct event_trigger_ops *ops,
1651 struct event_trigger_data *data);
1652};
1653
1654/**
1655 * struct event_command - callbacks and data members for event commands
1656 *
1657 * Event commands are invoked by users by writing the command name
1658 * into the 'trigger' file associated with a trace event. The
1659 * parameters associated with a specific invocation of an event
1660 * command are used to create an event trigger instance, which is
1661 * added to the list of trigger instances associated with that trace
1662 * event. When the event is hit, the set of triggers associated with
1663 * that event is invoked.
1664 *
1665 * The data members in this structure provide per-event command data
1666 * for various event commands.
1667 *
1668 * All the data members below, except for @post_trigger, must be set
1669 * for each event command.
1670 *
1671 * @name: The unique name that identifies the event command. This is
1672 * the name used when setting triggers via trigger files.
1673 *
1674 * @trigger_type: A unique id that identifies the event command
1675 * 'type'. This value has two purposes, the first to ensure that
1676 * only one trigger of the same type can be set at a given time
1677 * for a particular event e.g. it doesn't make sense to have both
1678 * a traceon and traceoff trigger attached to a single event at
1679 * the same time, so traceon and traceoff have the same type
1680 * though they have different names. The @trigger_type value is
1681 * also used as a bit value for deferring the actual trigger
1682 * action until after the current event is finished. Some
1683 * commands need to do this if they themselves log to the trace
1684 * buffer (see the @post_trigger() member below). @trigger_type
1685 * values are defined by adding new values to the trigger_type
1686 * enum in include/linux/trace_events.h.
1687 *
1688 * @flags: See the enum event_command_flags below.
1689 *
1690 * All the methods below, except for @set_filter() and @unreg_all(),
1691 * must be implemented.
1692 *
1693 * @func: The callback function responsible for parsing and
1694 * registering the trigger written to the 'trigger' file by the
1695 * user. It allocates the trigger instance and registers it with
1696 * the appropriate trace event. It makes use of the other
1697 * event_command callback functions to orchestrate this, and is
1698 * usually implemented by the generic utility function
1699 * @event_trigger_callback() (see trace_event_triggers.c).
1700 *
1701 * @reg: Adds the trigger to the list of triggers associated with the
1702 * event, and enables the event trigger itself, after
1703 * initializing it (via the event_trigger_ops @init() function).
1704 * This is also where commands can use the @trigger_type value to
1705 * make the decision as to whether or not multiple instances of
1706 * the trigger should be allowed. This is usually implemented by
1707 * the generic utility function @register_trigger() (see
1708 * trace_event_triggers.c).
1709 *
1710 * @unreg: Removes the trigger from the list of triggers associated
1711 * with the event, and disables the event trigger itself, after
1712 * initializing it (via the event_trigger_ops @free() function).
1713 * This is usually implemented by the generic utility function
1714 * @unregister_trigger() (see trace_event_triggers.c).
1715 *
1716 * @unreg_all: An optional function called to remove all the triggers
1717 * from the list of triggers associated with the event. Called
1718 * when a trigger file is opened in truncate mode.
1719 *
1720 * @set_filter: An optional function called to parse and set a filter
1721 * for the trigger. If no @set_filter() method is set for the
1722 * event command, filters set by the user for the command will be
1723 * ignored. This is usually implemented by the generic utility
1724 * function @set_trigger_filter() (see trace_event_triggers.c).
1725 *
1726 * @get_trigger_ops: The callback function invoked to retrieve the
1727 * event_trigger_ops implementation associated with the command.
1728 */
1729struct event_command {
1730 struct list_head list;
1731 char *name;
1732 enum event_trigger_type trigger_type;
1733 int flags;
1734 int (*func)(struct event_command *cmd_ops,
1735 struct trace_event_file *file,
1736 char *glob, char *cmd, char *params);
1737 int (*reg)(char *glob,
1738 struct event_trigger_ops *ops,
1739 struct event_trigger_data *data,
1740 struct trace_event_file *file);
1741 void (*unreg)(char *glob,
1742 struct event_trigger_ops *ops,
1743 struct event_trigger_data *data,
1744 struct trace_event_file *file);
1745 void (*unreg_all)(struct trace_event_file *file);
1746 int (*set_filter)(char *filter_str,
1747 struct event_trigger_data *data,
1748 struct trace_event_file *file);
1749 struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1750};
1751
1752/**
1753 * enum event_command_flags - flags for struct event_command
1754 *
1755 * @POST_TRIGGER: A flag that says whether or not this command needs
1756 * to have its action delayed until after the current event has
1757 * been closed. Some triggers need to avoid being invoked while
1758 * an event is currently in the process of being logged, since
1759 * the trigger may itself log data into the trace buffer. Thus
1760 * we make sure the current event is committed before invoking
1761 * those triggers. To do that, the trigger invocation is split
1762 * in two - the first part checks the filter using the current
1763 * trace record; if a command has the @post_trigger flag set, it
1764 * sets a bit for itself in the return value, otherwise it
1765 * directly invokes the trigger. Once all commands have been
1766 * either invoked or set their return flag, the current record is
1767 * either committed or discarded. At that point, if any commands
1768 * have deferred their triggers, those commands are finally
1769 * invoked following the close of the current event. In other
1770 * words, if the event_trigger_ops @func() probe implementation
1771 * itself logs to the trace buffer, this flag should be set,
1772 * otherwise it can be left unspecified.
1773 *
1774 * @NEEDS_REC: A flag that says whether or not this command needs
1775 * access to the trace record in order to perform its function,
1776 * regardless of whether or not it has a filter associated with
1777 * it (filters make a trigger require access to the trace record
1778 * but are not always present).
1779 */
1780enum event_command_flags {
1781 EVENT_CMD_FL_POST_TRIGGER = 1,
1782 EVENT_CMD_FL_NEEDS_REC = 2,
1783};
1784
1785static inline bool event_command_post_trigger(struct event_command *cmd_ops)
1786{
1787 return cmd_ops->flags & EVENT_CMD_FL_POST_TRIGGER;
1788}
1789
1790static inline bool event_command_needs_rec(struct event_command *cmd_ops)
1791{
1792 return cmd_ops->flags & EVENT_CMD_FL_NEEDS_REC;
1793}
1794
1795extern int trace_event_enable_disable(struct trace_event_file *file,
1796 int enable, int soft_disable);
1797extern int tracing_alloc_snapshot(void);
1798
1799extern const char *__start___trace_bprintk_fmt[];
1800extern const char *__stop___trace_bprintk_fmt[];
1801
1802extern const char *__start___tracepoint_str[];
1803extern const char *__stop___tracepoint_str[];
1804
1805void trace_printk_control(bool enabled);
1806void trace_printk_init_buffers(void);
1807void trace_printk_start_comm(void);
1808int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1809int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1810
1811#define MAX_EVENT_NAME_LEN 64
1812
1813extern int trace_run_command(const char *buf, int (*createfn)(int, char**));
1814extern ssize_t trace_parse_run_command(struct file *file,
1815 const char __user *buffer, size_t count, loff_t *ppos,
1816 int (*createfn)(int, char**));
1817
1818/*
1819 * Normal trace_printk() and friends allocates special buffers
1820 * to do the manipulation, as well as saves the print formats
1821 * into sections to display. But the trace infrastructure wants
1822 * to use these without the added overhead at the price of being
1823 * a bit slower (used mainly for warnings, where we don't care
1824 * about performance). The internal_trace_puts() is for such
1825 * a purpose.
1826 */
1827#define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1828
1829#undef FTRACE_ENTRY
1830#define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter) \
1831 extern struct trace_event_call \
1832 __aligned(4) event_##call;
1833#undef FTRACE_ENTRY_DUP
1834#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter) \
1835 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1836 filter)
1837#undef FTRACE_ENTRY_PACKED
1838#define FTRACE_ENTRY_PACKED(call, struct_name, id, tstruct, print, filter) \
1839 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1840 filter)
1841
1842#include "trace_entries.h"
1843
1844#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1845int perf_ftrace_event_register(struct trace_event_call *call,
1846 enum trace_reg type, void *data);
1847#else
1848#define perf_ftrace_event_register NULL
1849#endif
1850
1851#ifdef CONFIG_FTRACE_SYSCALLS
1852void init_ftrace_syscalls(void);
1853const char *get_syscall_name(int syscall);
1854#else
1855static inline void init_ftrace_syscalls(void) { }
1856static inline const char *get_syscall_name(int syscall)
1857{
1858 return NULL;
1859}
1860#endif
1861
1862#ifdef CONFIG_EVENT_TRACING
1863void trace_event_init(void);
1864void trace_event_eval_update(struct trace_eval_map **map, int len);
1865#else
1866static inline void __init trace_event_init(void) { }
1867static inline void trace_event_eval_update(struct trace_eval_map **map, int len) { }
1868#endif
1869
1870#ifdef CONFIG_TRACER_SNAPSHOT
1871void tracing_snapshot_instance(struct trace_array *tr);
1872int tracing_alloc_snapshot_instance(struct trace_array *tr);
1873#else
1874static inline void tracing_snapshot_instance(struct trace_array *tr) { }
1875static inline int tracing_alloc_snapshot_instance(struct trace_array *tr)
1876{
1877 return 0;
1878}
1879#endif
1880
1881#ifdef CONFIG_PREEMPT_TRACER
1882void tracer_preempt_on(unsigned long a0, unsigned long a1);
1883void tracer_preempt_off(unsigned long a0, unsigned long a1);
1884#else
1885static inline void tracer_preempt_on(unsigned long a0, unsigned long a1) { }
1886static inline void tracer_preempt_off(unsigned long a0, unsigned long a1) { }
1887#endif
1888#ifdef CONFIG_IRQSOFF_TRACER
1889void tracer_hardirqs_on(unsigned long a0, unsigned long a1);
1890void tracer_hardirqs_off(unsigned long a0, unsigned long a1);
1891#else
1892static inline void tracer_hardirqs_on(unsigned long a0, unsigned long a1) { }
1893static inline void tracer_hardirqs_off(unsigned long a0, unsigned long a1) { }
1894#endif
1895
1896extern struct trace_iterator *tracepoint_print_iter;
1897
1898#endif /* _LINUX_KERNEL_TRACE_H */