projects/linux·blob v3.18 · refs

1313 lines · 40.2 KiB

1
2#ifndef _LINUX_KERNEL_TRACE_H
3#define _LINUX_KERNEL_TRACE_H
4
5#include <linux/fs.h>
6#include <linux/atomic.h>
7#include <linux/sched.h>
8#include <linux/clocksource.h>
9#include <linux/ring_buffer.h>
10#include <linux/mmiotrace.h>
11#include <linux/tracepoint.h>
12#include <linux/ftrace.h>
13#include <linux/hw_breakpoint.h>
14#include <linux/trace_seq.h>
15#include <linux/ftrace_event.h>
16#include <linux/compiler.h>
17
18#ifdef CONFIG_FTRACE_SYSCALLS
19#include <asm/unistd.h> /* For NR_SYSCALLS */
20#include <asm/syscall.h> /* some archs define it here */
21#endif
22
23enum trace_type {
24 __TRACE_FIRST_TYPE = 0,
25
26 TRACE_FN,
27 TRACE_CTX,
28 TRACE_WAKE,
29 TRACE_STACK,
30 TRACE_PRINT,
31 TRACE_BPRINT,
32 TRACE_MMIO_RW,
33 TRACE_MMIO_MAP,
34 TRACE_BRANCH,
35 TRACE_GRAPH_RET,
36 TRACE_GRAPH_ENT,
37 TRACE_USER_STACK,
38 TRACE_BLK,
39 TRACE_BPUTS,
40
41 __TRACE_LAST_TYPE,
42};
43
44
45#undef __field
46#define __field(type, item) type item;
47
48#undef __field_struct
49#define __field_struct(type, item) __field(type, item)
50
51#undef __field_desc
52#define __field_desc(type, container, item)
53
54#undef __array
55#define __array(type, item, size) type item[size];
56
57#undef __array_desc
58#define __array_desc(type, container, item, size)
59
60#undef __dynamic_array
61#define __dynamic_array(type, item) type item[];
62
63#undef F_STRUCT
64#define F_STRUCT(args...) args
65
66#undef FTRACE_ENTRY
67#define FTRACE_ENTRY(name, struct_name, id, tstruct, print, filter) \
68 struct struct_name { \
69 struct trace_entry ent; \
70 tstruct \
71 }
72
73#undef TP_ARGS
74#define TP_ARGS(args...) args
75
76#undef FTRACE_ENTRY_DUP
77#define FTRACE_ENTRY_DUP(name, name_struct, id, tstruct, printk, filter)
78
79#undef FTRACE_ENTRY_REG
80#define FTRACE_ENTRY_REG(name, struct_name, id, tstruct, print, \
81 filter, regfn) \
82 FTRACE_ENTRY(name, struct_name, id, PARAMS(tstruct), PARAMS(print), \
83 filter)
84
85#include "trace_entries.h"
86
87/*
88 * syscalls are special, and need special handling, this is why
89 * they are not included in trace_entries.h
90 */
91struct syscall_trace_enter {
92 struct trace_entry ent;
93 int nr;
94 unsigned long args[];
95};
96
97struct syscall_trace_exit {
98 struct trace_entry ent;
99 int nr;
100 long ret;
101};
102
103struct kprobe_trace_entry_head {
104 struct trace_entry ent;
105 unsigned long ip;
106};
107
108struct kretprobe_trace_entry_head {
109 struct trace_entry ent;
110 unsigned long func;
111 unsigned long ret_ip;
112};
113
114/*
115 * trace_flag_type is an enumeration that holds different
116 * states when a trace occurs. These are:
117 * IRQS_OFF - interrupts were disabled
118 * IRQS_NOSUPPORT - arch does not support irqs_disabled_flags
119 * NEED_RESCHED - reschedule is requested
120 * HARDIRQ - inside an interrupt handler
121 * SOFTIRQ - inside a softirq handler
122 */
123enum trace_flag_type {
124 TRACE_FLAG_IRQS_OFF = 0x01,
125 TRACE_FLAG_IRQS_NOSUPPORT = 0x02,
126 TRACE_FLAG_NEED_RESCHED = 0x04,
127 TRACE_FLAG_HARDIRQ = 0x08,
128 TRACE_FLAG_SOFTIRQ = 0x10,
129 TRACE_FLAG_PREEMPT_RESCHED = 0x20,
130};
131
132#define TRACE_BUF_SIZE 1024
133
134struct trace_array;
135
136/*
137 * The CPU trace array - it consists of thousands of trace entries
138 * plus some other descriptor data: (for example which task started
139 * the trace, etc.)
140 */
141struct trace_array_cpu {
142 atomic_t disabled;
143 void *buffer_page; /* ring buffer spare */
144
145 unsigned long entries;
146 unsigned long saved_latency;
147 unsigned long critical_start;
148 unsigned long critical_end;
149 unsigned long critical_sequence;
150 unsigned long nice;
151 unsigned long policy;
152 unsigned long rt_priority;
153 unsigned long skipped_entries;
154 cycle_t preempt_timestamp;
155 pid_t pid;
156 kuid_t uid;
157 char comm[TASK_COMM_LEN];
158};
159
160struct tracer;
161
162struct trace_buffer {
163 struct trace_array *tr;
164 struct ring_buffer *buffer;
165 struct trace_array_cpu __percpu *data;
166 cycle_t time_start;
167 int cpu;
168};
169
170/*
171 * The trace array - an array of per-CPU trace arrays. This is the
172 * highest level data structure that individual tracers deal with.
173 * They have on/off state as well:
174 */
175struct trace_array {
176 struct list_head list;
177 char *name;
178 struct trace_buffer trace_buffer;
179#ifdef CONFIG_TRACER_MAX_TRACE
180 /*
181 * The max_buffer is used to snapshot the trace when a maximum
182 * latency is reached, or when the user initiates a snapshot.
183 * Some tracers will use this to store a maximum trace while
184 * it continues examining live traces.
185 *
186 * The buffers for the max_buffer are set up the same as the trace_buffer
187 * When a snapshot is taken, the buffer of the max_buffer is swapped
188 * with the buffer of the trace_buffer and the buffers are reset for
189 * the trace_buffer so the tracing can continue.
190 */
191 struct trace_buffer max_buffer;
192 bool allocated_snapshot;
193 unsigned long max_latency;
194#endif
195 /*
196 * max_lock is used to protect the swapping of buffers
197 * when taking a max snapshot. The buffers themselves are
198 * protected by per_cpu spinlocks. But the action of the swap
199 * needs its own lock.
200 *
201 * This is defined as a arch_spinlock_t in order to help
202 * with performance when lockdep debugging is enabled.
203 *
204 * It is also used in other places outside the update_max_tr
205 * so it needs to be defined outside of the
206 * CONFIG_TRACER_MAX_TRACE.
207 */
208 arch_spinlock_t max_lock;
209 int buffer_disabled;
210#ifdef CONFIG_FTRACE_SYSCALLS
211 int sys_refcount_enter;
212 int sys_refcount_exit;
213 struct ftrace_event_file __rcu *enter_syscall_files[NR_syscalls];
214 struct ftrace_event_file __rcu *exit_syscall_files[NR_syscalls];
215#endif
216 int stop_count;
217 int clock_id;
218 struct tracer *current_trace;
219 unsigned int flags;
220 raw_spinlock_t start_lock;
221 struct dentry *dir;
222 struct dentry *options;
223 struct dentry *percpu_dir;
224 struct dentry *event_dir;
225 struct list_head systems;
226 struct list_head events;
227 cpumask_var_t tracing_cpumask; /* only trace on set CPUs */
228 int ref;
229#ifdef CONFIG_FUNCTION_TRACER
230 struct ftrace_ops *ops;
231 /* function tracing enabled */
232 int function_enabled;
233#endif
234};
235
236enum {
237 TRACE_ARRAY_FL_GLOBAL = (1 << 0)
238};
239
240extern struct list_head ftrace_trace_arrays;
241
242extern struct mutex trace_types_lock;
243
244extern int trace_array_get(struct trace_array *tr);
245extern void trace_array_put(struct trace_array *tr);
246
247/*
248 * The global tracer (top) should be the first trace array added,
249 * but we check the flag anyway.
250 */
251static inline struct trace_array *top_trace_array(void)
252{
253 struct trace_array *tr;
254
255 if (list_empty(&ftrace_trace_arrays))
256 return NULL;
257
258 tr = list_entry(ftrace_trace_arrays.prev,
259 typeof(*tr), list);
260 WARN_ON(!(tr->flags & TRACE_ARRAY_FL_GLOBAL));
261 return tr;
262}
263
264#define FTRACE_CMP_TYPE(var, type) \
265 __builtin_types_compatible_p(typeof(var), type *)
266
267#undef IF_ASSIGN
268#define IF_ASSIGN(var, entry, etype, id) \
269 if (FTRACE_CMP_TYPE(var, etype)) { \
270 var = (typeof(var))(entry); \
271 WARN_ON(id && (entry)->type != id); \
272 break; \
273 }
274
275/* Will cause compile errors if type is not found. */
276extern void __ftrace_bad_type(void);
277
278/*
279 * The trace_assign_type is a verifier that the entry type is
280 * the same as the type being assigned. To add new types simply
281 * add a line with the following format:
282 *
283 * IF_ASSIGN(var, ent, type, id);
284 *
285 * Where "type" is the trace type that includes the trace_entry
286 * as the "ent" item. And "id" is the trace identifier that is
287 * used in the trace_type enum.
288 *
289 * If the type can have more than one id, then use zero.
290 */
291#define trace_assign_type(var, ent) \
292 do { \
293 IF_ASSIGN(var, ent, struct ftrace_entry, TRACE_FN); \
294 IF_ASSIGN(var, ent, struct ctx_switch_entry, 0); \
295 IF_ASSIGN(var, ent, struct stack_entry, TRACE_STACK); \
296 IF_ASSIGN(var, ent, struct userstack_entry, TRACE_USER_STACK);\
297 IF_ASSIGN(var, ent, struct print_entry, TRACE_PRINT); \
298 IF_ASSIGN(var, ent, struct bprint_entry, TRACE_BPRINT); \
299 IF_ASSIGN(var, ent, struct bputs_entry, TRACE_BPUTS); \
300 IF_ASSIGN(var, ent, struct trace_mmiotrace_rw, \
301 TRACE_MMIO_RW); \
302 IF_ASSIGN(var, ent, struct trace_mmiotrace_map, \
303 TRACE_MMIO_MAP); \
304 IF_ASSIGN(var, ent, struct trace_branch, TRACE_BRANCH); \
305 IF_ASSIGN(var, ent, struct ftrace_graph_ent_entry, \
306 TRACE_GRAPH_ENT); \
307 IF_ASSIGN(var, ent, struct ftrace_graph_ret_entry, \
308 TRACE_GRAPH_RET); \
309 __ftrace_bad_type(); \
310 } while (0)
311
312/*
313 * An option specific to a tracer. This is a boolean value.
314 * The bit is the bit index that sets its value on the
315 * flags value in struct tracer_flags.
316 */
317struct tracer_opt {
318 const char *name; /* Will appear on the trace_options file */
319 u32 bit; /* Mask assigned in val field in tracer_flags */
320};
321
322/*
323 * The set of specific options for a tracer. Your tracer
324 * have to set the initial value of the flags val.
325 */
326struct tracer_flags {
327 u32 val;
328 struct tracer_opt *opts;
329};
330
331/* Makes more easy to define a tracer opt */
332#define TRACER_OPT(s, b) .name = #s, .bit = b
333
334
335/**
336 * struct tracer - a specific tracer and its callbacks to interact with debugfs
337 * @name: the name chosen to select it on the available_tracers file
338 * @init: called when one switches to this tracer (echo name > current_tracer)
339 * @reset: called when one switches to another tracer
340 * @start: called when tracing is unpaused (echo 1 > tracing_enabled)
341 * @stop: called when tracing is paused (echo 0 > tracing_enabled)
342 * @update_thresh: called when tracing_thresh is updated
343 * @open: called when the trace file is opened
344 * @pipe_open: called when the trace_pipe file is opened
345 * @close: called when the trace file is released
346 * @pipe_close: called when the trace_pipe file is released
347 * @read: override the default read callback on trace_pipe
348 * @splice_read: override the default splice_read callback on trace_pipe
349 * @selftest: selftest to run on boot (see trace_selftest.c)
350 * @print_headers: override the first lines that describe your columns
351 * @print_line: callback that prints a trace
352 * @set_flag: signals one of your private flags changed (trace_options file)
353 * @flags: your private flags
354 */
355struct tracer {
356 const char *name;
357 int (*init)(struct trace_array *tr);
358 void (*reset)(struct trace_array *tr);
359 void (*start)(struct trace_array *tr);
360 void (*stop)(struct trace_array *tr);
361 int (*update_thresh)(struct trace_array *tr);
362 void (*open)(struct trace_iterator *iter);
363 void (*pipe_open)(struct trace_iterator *iter);
364 void (*close)(struct trace_iterator *iter);
365 void (*pipe_close)(struct trace_iterator *iter);
366 ssize_t (*read)(struct trace_iterator *iter,
367 struct file *filp, char __user *ubuf,
368 size_t cnt, loff_t *ppos);
369 ssize_t (*splice_read)(struct trace_iterator *iter,
370 struct file *filp,
371 loff_t *ppos,
372 struct pipe_inode_info *pipe,
373 size_t len,
374 unsigned int flags);
375#ifdef CONFIG_FTRACE_STARTUP_TEST
376 int (*selftest)(struct tracer *trace,
377 struct trace_array *tr);
378#endif
379 void (*print_header)(struct seq_file *m);
380 enum print_line_t (*print_line)(struct trace_iterator *iter);
381 /* If you handled the flag setting, return 0 */
382 int (*set_flag)(struct trace_array *tr,
383 u32 old_flags, u32 bit, int set);
384 /* Return 0 if OK with change, else return non-zero */
385 int (*flag_changed)(struct trace_array *tr,
386 u32 mask, int set);
387 struct tracer *next;
388 struct tracer_flags *flags;
389 int enabled;
390 bool print_max;
391 bool allow_instances;
392#ifdef CONFIG_TRACER_MAX_TRACE
393 bool use_max_tr;
394#endif
395};
396
397
398/* Only current can touch trace_recursion */
399
400/*
401 * For function tracing recursion:
402 * The order of these bits are important.
403 *
404 * When function tracing occurs, the following steps are made:
405 * If arch does not support a ftrace feature:
406 * call internal function (uses INTERNAL bits) which calls...
407 * If callback is registered to the "global" list, the list
408 * function is called and recursion checks the GLOBAL bits.
409 * then this function calls...
410 * The function callback, which can use the FTRACE bits to
411 * check for recursion.
412 *
413 * Now if the arch does not suppport a feature, and it calls
414 * the global list function which calls the ftrace callback
415 * all three of these steps will do a recursion protection.
416 * There's no reason to do one if the previous caller already
417 * did. The recursion that we are protecting against will
418 * go through the same steps again.
419 *
420 * To prevent the multiple recursion checks, if a recursion
421 * bit is set that is higher than the MAX bit of the current
422 * check, then we know that the check was made by the previous
423 * caller, and we can skip the current check.
424 */
425enum {
426 TRACE_BUFFER_BIT,
427 TRACE_BUFFER_NMI_BIT,
428 TRACE_BUFFER_IRQ_BIT,
429 TRACE_BUFFER_SIRQ_BIT,
430
431 /* Start of function recursion bits */
432 TRACE_FTRACE_BIT,
433 TRACE_FTRACE_NMI_BIT,
434 TRACE_FTRACE_IRQ_BIT,
435 TRACE_FTRACE_SIRQ_BIT,
436
437 /* INTERNAL_BITs must be greater than FTRACE_BITs */
438 TRACE_INTERNAL_BIT,
439 TRACE_INTERNAL_NMI_BIT,
440 TRACE_INTERNAL_IRQ_BIT,
441 TRACE_INTERNAL_SIRQ_BIT,
442
443 TRACE_CONTROL_BIT,
444
445/*
446 * Abuse of the trace_recursion.
447 * As we need a way to maintain state if we are tracing the function
448 * graph in irq because we want to trace a particular function that
449 * was called in irq context but we have irq tracing off. Since this
450 * can only be modified by current, we can reuse trace_recursion.
451 */
452 TRACE_IRQ_BIT,
453};
454
455#define trace_recursion_set(bit) do { (current)->trace_recursion |= (1<<(bit)); } while (0)
456#define trace_recursion_clear(bit) do { (current)->trace_recursion &= ~(1<<(bit)); } while (0)
457#define trace_recursion_test(bit) ((current)->trace_recursion & (1<<(bit)))
458
459#define TRACE_CONTEXT_BITS 4
460
461#define TRACE_FTRACE_START TRACE_FTRACE_BIT
462#define TRACE_FTRACE_MAX ((1 << (TRACE_FTRACE_START + TRACE_CONTEXT_BITS)) - 1)
463
464#define TRACE_LIST_START TRACE_INTERNAL_BIT
465#define TRACE_LIST_MAX ((1 << (TRACE_LIST_START + TRACE_CONTEXT_BITS)) - 1)
466
467#define TRACE_CONTEXT_MASK TRACE_LIST_MAX
468
469static __always_inline int trace_get_context_bit(void)
470{
471 int bit;
472
473 if (in_interrupt()) {
474 if (in_nmi())
475 bit = 0;
476
477 else if (in_irq())
478 bit = 1;
479 else
480 bit = 2;
481 } else
482 bit = 3;
483
484 return bit;
485}
486
487static __always_inline int trace_test_and_set_recursion(int start, int max)
488{
489 unsigned int val = current->trace_recursion;
490 int bit;
491
492 /* A previous recursion check was made */
493 if ((val & TRACE_CONTEXT_MASK) > max)
494 return 0;
495
496 bit = trace_get_context_bit() + start;
497 if (unlikely(val & (1 << bit)))
498 return -1;
499
500 val |= 1 << bit;
501 current->trace_recursion = val;
502 barrier();
503
504 return bit;
505}
506
507static __always_inline void trace_clear_recursion(int bit)
508{
509 unsigned int val = current->trace_recursion;
510
511 if (!bit)
512 return;
513
514 bit = 1 << bit;
515 val &= ~bit;
516
517 barrier();
518 current->trace_recursion = val;
519}
520
521static inline struct ring_buffer_iter *
522trace_buffer_iter(struct trace_iterator *iter, int cpu)
523{
524 if (iter->buffer_iter && iter->buffer_iter[cpu])
525 return iter->buffer_iter[cpu];
526 return NULL;
527}
528
529int tracer_init(struct tracer *t, struct trace_array *tr);
530int tracing_is_enabled(void);
531void tracing_reset(struct trace_buffer *buf, int cpu);
532void tracing_reset_online_cpus(struct trace_buffer *buf);
533void tracing_reset_current(int cpu);
534void tracing_reset_all_online_cpus(void);
535int tracing_open_generic(struct inode *inode, struct file *filp);
536bool tracing_is_disabled(void);
537struct dentry *trace_create_file(const char *name,
538 umode_t mode,
539 struct dentry *parent,
540 void *data,
541 const struct file_operations *fops);
542
543struct dentry *tracing_init_dentry_tr(struct trace_array *tr);
544struct dentry *tracing_init_dentry(void);
545
546struct ring_buffer_event;
547
548struct ring_buffer_event *
549trace_buffer_lock_reserve(struct ring_buffer *buffer,
550 int type,
551 unsigned long len,
552 unsigned long flags,
553 int pc);
554
555struct trace_entry *tracing_get_trace_entry(struct trace_array *tr,
556 struct trace_array_cpu *data);
557
558struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
559 int *ent_cpu, u64 *ent_ts);
560
561void __buffer_unlock_commit(struct ring_buffer *buffer,
562 struct ring_buffer_event *event);
563
564int trace_empty(struct trace_iterator *iter);
565
566void *trace_find_next_entry_inc(struct trace_iterator *iter);
567
568void trace_init_global_iter(struct trace_iterator *iter);
569
570void tracing_iter_reset(struct trace_iterator *iter, int cpu);
571
572void tracing_sched_switch_trace(struct trace_array *tr,
573 struct task_struct *prev,
574 struct task_struct *next,
575 unsigned long flags, int pc);
576
577void tracing_sched_wakeup_trace(struct trace_array *tr,
578 struct task_struct *wakee,
579 struct task_struct *cur,
580 unsigned long flags, int pc);
581void trace_function(struct trace_array *tr,
582 unsigned long ip,
583 unsigned long parent_ip,
584 unsigned long flags, int pc);
585void trace_graph_function(struct trace_array *tr,
586 unsigned long ip,
587 unsigned long parent_ip,
588 unsigned long flags, int pc);
589void trace_latency_header(struct seq_file *m);
590void trace_default_header(struct seq_file *m);
591void print_trace_header(struct seq_file *m, struct trace_iterator *iter);
592int trace_empty(struct trace_iterator *iter);
593
594void trace_graph_return(struct ftrace_graph_ret *trace);
595int trace_graph_entry(struct ftrace_graph_ent *trace);
596void set_graph_array(struct trace_array *tr);
597
598void tracing_start_cmdline_record(void);
599void tracing_stop_cmdline_record(void);
600void tracing_sched_switch_assign_trace(struct trace_array *tr);
601void tracing_stop_sched_switch_record(void);
602void tracing_start_sched_switch_record(void);
603int register_tracer(struct tracer *type);
604int is_tracing_stopped(void);
605
606loff_t tracing_lseek(struct file *file, loff_t offset, int whence);
607
608extern cpumask_var_t __read_mostly tracing_buffer_mask;
609
610#define for_each_tracing_cpu(cpu) \
611 for_each_cpu(cpu, tracing_buffer_mask)
612
613extern unsigned long nsecs_to_usecs(unsigned long nsecs);
614
615extern unsigned long tracing_thresh;
616
617#ifdef CONFIG_TRACER_MAX_TRACE
618void update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu);
619void update_max_tr_single(struct trace_array *tr,
620 struct task_struct *tsk, int cpu);
621#endif /* CONFIG_TRACER_MAX_TRACE */
622
623#ifdef CONFIG_STACKTRACE
624void ftrace_trace_stack(struct ring_buffer *buffer, unsigned long flags,
625 int skip, int pc);
626
627void ftrace_trace_stack_regs(struct ring_buffer *buffer, unsigned long flags,
628 int skip, int pc, struct pt_regs *regs);
629
630void ftrace_trace_userstack(struct ring_buffer *buffer, unsigned long flags,
631 int pc);
632
633void __trace_stack(struct trace_array *tr, unsigned long flags, int skip,
634 int pc);
635#else
636static inline void ftrace_trace_stack(struct ring_buffer *buffer,
637 unsigned long flags, int skip, int pc)
638{
639}
640
641static inline void ftrace_trace_stack_regs(struct ring_buffer *buffer,
642 unsigned long flags, int skip,
643 int pc, struct pt_regs *regs)
644{
645}
646
647static inline void ftrace_trace_userstack(struct ring_buffer *buffer,
648 unsigned long flags, int pc)
649{
650}
651
652static inline void __trace_stack(struct trace_array *tr, unsigned long flags,
653 int skip, int pc)
654{
655}
656#endif /* CONFIG_STACKTRACE */
657
658extern cycle_t ftrace_now(int cpu);
659
660extern void trace_find_cmdline(int pid, char comm[]);
661
662#ifdef CONFIG_DYNAMIC_FTRACE
663extern unsigned long ftrace_update_tot_cnt;
664#endif
665#define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
666extern int DYN_FTRACE_TEST_NAME(void);
667#define DYN_FTRACE_TEST_NAME2 trace_selftest_dynamic_test_func2
668extern int DYN_FTRACE_TEST_NAME2(void);
669
670extern bool ring_buffer_expanded;
671extern bool tracing_selftest_disabled;
672DECLARE_PER_CPU(int, ftrace_cpu_disabled);
673
674#ifdef CONFIG_FTRACE_STARTUP_TEST
675extern int trace_selftest_startup_function(struct tracer *trace,
676 struct trace_array *tr);
677extern int trace_selftest_startup_function_graph(struct tracer *trace,
678 struct trace_array *tr);
679extern int trace_selftest_startup_irqsoff(struct tracer *trace,
680 struct trace_array *tr);
681extern int trace_selftest_startup_preemptoff(struct tracer *trace,
682 struct trace_array *tr);
683extern int trace_selftest_startup_preemptirqsoff(struct tracer *trace,
684 struct trace_array *tr);
685extern int trace_selftest_startup_wakeup(struct tracer *trace,
686 struct trace_array *tr);
687extern int trace_selftest_startup_nop(struct tracer *trace,
688 struct trace_array *tr);
689extern int trace_selftest_startup_sched_switch(struct tracer *trace,
690 struct trace_array *tr);
691extern int trace_selftest_startup_branch(struct tracer *trace,
692 struct trace_array *tr);
693/*
694 * Tracer data references selftest functions that only occur
695 * on boot up. These can be __init functions. Thus, when selftests
696 * are enabled, then the tracers need to reference __init functions.
697 */
698#define __tracer_data __refdata
699#else
700/* Tracers are seldom changed. Optimize when selftests are disabled. */
701#define __tracer_data __read_mostly
702#endif /* CONFIG_FTRACE_STARTUP_TEST */
703
704extern void *head_page(struct trace_array_cpu *data);
705extern unsigned long long ns2usecs(cycle_t nsec);
706extern int
707trace_vbprintk(unsigned long ip, const char *fmt, va_list args);
708extern int
709trace_vprintk(unsigned long ip, const char *fmt, va_list args);
710extern int
711trace_array_vprintk(struct trace_array *tr,
712 unsigned long ip, const char *fmt, va_list args);
713int trace_array_printk(struct trace_array *tr,
714 unsigned long ip, const char *fmt, ...);
715int trace_array_printk_buf(struct ring_buffer *buffer,
716 unsigned long ip, const char *fmt, ...);
717void trace_printk_seq(struct trace_seq *s);
718enum print_line_t print_trace_line(struct trace_iterator *iter);
719
720extern unsigned long trace_flags;
721
722/* Standard output formatting function used for function return traces */
723#ifdef CONFIG_FUNCTION_GRAPH_TRACER
724
725/* Flag options */
726#define TRACE_GRAPH_PRINT_OVERRUN 0x1
727#define TRACE_GRAPH_PRINT_CPU 0x2
728#define TRACE_GRAPH_PRINT_OVERHEAD 0x4
729#define TRACE_GRAPH_PRINT_PROC 0x8
730#define TRACE_GRAPH_PRINT_DURATION 0x10
731#define TRACE_GRAPH_PRINT_ABS_TIME 0x20
732#define TRACE_GRAPH_PRINT_IRQS 0x40
733#define TRACE_GRAPH_PRINT_TAIL 0x80
734#define TRACE_GRAPH_PRINT_FILL_SHIFT 28
735#define TRACE_GRAPH_PRINT_FILL_MASK (0x3 << TRACE_GRAPH_PRINT_FILL_SHIFT)
736
737extern enum print_line_t
738print_graph_function_flags(struct trace_iterator *iter, u32 flags);
739extern void print_graph_headers_flags(struct seq_file *s, u32 flags);
740extern enum print_line_t
741trace_print_graph_duration(unsigned long long duration, struct trace_seq *s);
742extern void graph_trace_open(struct trace_iterator *iter);
743extern void graph_trace_close(struct trace_iterator *iter);
744extern int __trace_graph_entry(struct trace_array *tr,
745 struct ftrace_graph_ent *trace,
746 unsigned long flags, int pc);
747extern void __trace_graph_return(struct trace_array *tr,
748 struct ftrace_graph_ret *trace,
749 unsigned long flags, int pc);
750
751
752#ifdef CONFIG_DYNAMIC_FTRACE
753/* TODO: make this variable */
754#define FTRACE_GRAPH_MAX_FUNCS 32
755extern int ftrace_graph_count;
756extern unsigned long ftrace_graph_funcs[FTRACE_GRAPH_MAX_FUNCS];
757extern int ftrace_graph_notrace_count;
758extern unsigned long ftrace_graph_notrace_funcs[FTRACE_GRAPH_MAX_FUNCS];
759
760static inline int ftrace_graph_addr(unsigned long addr)
761{
762 int i;
763
764 if (!ftrace_graph_count)
765 return 1;
766
767 for (i = 0; i < ftrace_graph_count; i++) {
768 if (addr == ftrace_graph_funcs[i]) {
769 /*
770 * If no irqs are to be traced, but a set_graph_function
771 * is set, and called by an interrupt handler, we still
772 * want to trace it.
773 */
774 if (in_irq())
775 trace_recursion_set(TRACE_IRQ_BIT);
776 else
777 trace_recursion_clear(TRACE_IRQ_BIT);
778 return 1;
779 }
780 }
781
782 return 0;
783}
784
785static inline int ftrace_graph_notrace_addr(unsigned long addr)
786{
787 int i;
788
789 if (!ftrace_graph_notrace_count)
790 return 0;
791
792 for (i = 0; i < ftrace_graph_notrace_count; i++) {
793 if (addr == ftrace_graph_notrace_funcs[i])
794 return 1;
795 }
796
797 return 0;
798}
799#else
800static inline int ftrace_graph_addr(unsigned long addr)
801{
802 return 1;
803}
804
805static inline int ftrace_graph_notrace_addr(unsigned long addr)
806{
807 return 0;
808}
809#endif /* CONFIG_DYNAMIC_FTRACE */
810#else /* CONFIG_FUNCTION_GRAPH_TRACER */
811static inline enum print_line_t
812print_graph_function_flags(struct trace_iterator *iter, u32 flags)
813{
814 return TRACE_TYPE_UNHANDLED;
815}
816#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
817
818extern struct list_head ftrace_pids;
819
820#ifdef CONFIG_FUNCTION_TRACER
821extern bool ftrace_filter_param __initdata;
822static inline int ftrace_trace_task(struct task_struct *task)
823{
824 if (list_empty(&ftrace_pids))
825 return 1;
826
827 return test_tsk_trace_trace(task);
828}
829extern int ftrace_is_dead(void);
830int ftrace_create_function_files(struct trace_array *tr,
831 struct dentry *parent);
832void ftrace_destroy_function_files(struct trace_array *tr);
833void ftrace_init_global_array_ops(struct trace_array *tr);
834void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func);
835void ftrace_reset_array_ops(struct trace_array *tr);
836int using_ftrace_ops_list_func(void);
837#else
838static inline int ftrace_trace_task(struct task_struct *task)
839{
840 return 1;
841}
842static inline int ftrace_is_dead(void) { return 0; }
843static inline int
844ftrace_create_function_files(struct trace_array *tr,
845 struct dentry *parent)
846{
847 return 0;
848}
849static inline void ftrace_destroy_function_files(struct trace_array *tr) { }
850static inline __init void
851ftrace_init_global_array_ops(struct trace_array *tr) { }
852static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
853/* ftace_func_t type is not defined, use macro instead of static inline */
854#define ftrace_init_array_ops(tr, func) do { } while (0)
855#endif /* CONFIG_FUNCTION_TRACER */
856
857#if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
858void ftrace_create_filter_files(struct ftrace_ops *ops,
859 struct dentry *parent);
860void ftrace_destroy_filter_files(struct ftrace_ops *ops);
861#else
862/*
863 * The ops parameter passed in is usually undefined.
864 * This must be a macro.
865 */
866#define ftrace_create_filter_files(ops, parent) do { } while (0)
867#define ftrace_destroy_filter_files(ops) do { } while (0)
868#endif /* CONFIG_FUNCTION_TRACER && CONFIG_DYNAMIC_FTRACE */
869
870int ftrace_event_is_function(struct ftrace_event_call *call);
871
872/*
873 * struct trace_parser - servers for reading the user input separated by spaces
874 * @cont: set if the input is not complete - no final space char was found
875 * @buffer: holds the parsed user input
876 * @idx: user input length
877 * @size: buffer size
878 */
879struct trace_parser {
880 bool cont;
881 char *buffer;
882 unsigned idx;
883 unsigned size;
884};
885
886static inline bool trace_parser_loaded(struct trace_parser *parser)
887{
888 return (parser->idx != 0);
889}
890
891static inline bool trace_parser_cont(struct trace_parser *parser)
892{
893 return parser->cont;
894}
895
896static inline void trace_parser_clear(struct trace_parser *parser)
897{
898 parser->cont = false;
899 parser->idx = 0;
900}
901
902extern int trace_parser_get_init(struct trace_parser *parser, int size);
903extern void trace_parser_put(struct trace_parser *parser);
904extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
905 size_t cnt, loff_t *ppos);
906
907/*
908 * trace_iterator_flags is an enumeration that defines bit
909 * positions into trace_flags that controls the output.
910 *
911 * NOTE: These bits must match the trace_options array in
912 * trace.c.
913 */
914enum trace_iterator_flags {
915 TRACE_ITER_PRINT_PARENT = 0x01,
916 TRACE_ITER_SYM_OFFSET = 0x02,
917 TRACE_ITER_SYM_ADDR = 0x04,
918 TRACE_ITER_VERBOSE = 0x08,
919 TRACE_ITER_RAW = 0x10,
920 TRACE_ITER_HEX = 0x20,
921 TRACE_ITER_BIN = 0x40,
922 TRACE_ITER_BLOCK = 0x80,
923 TRACE_ITER_STACKTRACE = 0x100,
924 TRACE_ITER_PRINTK = 0x200,
925 TRACE_ITER_PREEMPTONLY = 0x400,
926 TRACE_ITER_BRANCH = 0x800,
927 TRACE_ITER_ANNOTATE = 0x1000,
928 TRACE_ITER_USERSTACKTRACE = 0x2000,
929 TRACE_ITER_SYM_USEROBJ = 0x4000,
930 TRACE_ITER_PRINTK_MSGONLY = 0x8000,
931 TRACE_ITER_CONTEXT_INFO = 0x10000, /* Print pid/cpu/time */
932 TRACE_ITER_LATENCY_FMT = 0x20000,
933 TRACE_ITER_SLEEP_TIME = 0x40000,
934 TRACE_ITER_GRAPH_TIME = 0x80000,
935 TRACE_ITER_RECORD_CMD = 0x100000,
936 TRACE_ITER_OVERWRITE = 0x200000,
937 TRACE_ITER_STOP_ON_FREE = 0x400000,
938 TRACE_ITER_IRQ_INFO = 0x800000,
939 TRACE_ITER_MARKERS = 0x1000000,
940 TRACE_ITER_FUNCTION = 0x2000000,
941};
942
943/*
944 * TRACE_ITER_SYM_MASK masks the options in trace_flags that
945 * control the output of kernel symbols.
946 */
947#define TRACE_ITER_SYM_MASK \
948 (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR)
949
950extern struct tracer nop_trace;
951
952#ifdef CONFIG_BRANCH_TRACER
953extern int enable_branch_tracing(struct trace_array *tr);
954extern void disable_branch_tracing(void);
955static inline int trace_branch_enable(struct trace_array *tr)
956{
957 if (trace_flags & TRACE_ITER_BRANCH)
958 return enable_branch_tracing(tr);
959 return 0;
960}
961static inline void trace_branch_disable(void)
962{
963 /* due to races, always disable */
964 disable_branch_tracing();
965}
966#else
967static inline int trace_branch_enable(struct trace_array *tr)
968{
969 return 0;
970}
971static inline void trace_branch_disable(void)
972{
973}
974#endif /* CONFIG_BRANCH_TRACER */
975
976/* set ring buffers to default size if not already done so */
977int tracing_update_buffers(void);
978
979struct ftrace_event_field {
980 struct list_head link;
981 const char *name;
982 const char *type;
983 int filter_type;
984 int offset;
985 int size;
986 int is_signed;
987};
988
989struct event_filter {
990 int n_preds; /* Number assigned */
991 int a_preds; /* allocated */
992 struct filter_pred *preds;
993 struct filter_pred *root;
994 char *filter_string;
995};
996
997struct event_subsystem {
998 struct list_head list;
999 const char *name;
1000 struct event_filter *filter;
1001 int ref_count;
1002};
1003
1004struct ftrace_subsystem_dir {
1005 struct list_head list;
1006 struct event_subsystem *subsystem;
1007 struct trace_array *tr;
1008 struct dentry *entry;
1009 int ref_count;
1010 int nr_events;
1011};
1012
1013#define FILTER_PRED_INVALID ((unsigned short)-1)
1014#define FILTER_PRED_IS_RIGHT (1 << 15)
1015#define FILTER_PRED_FOLD (1 << 15)
1016
1017/*
1018 * The max preds is the size of unsigned short with
1019 * two flags at the MSBs. One bit is used for both the IS_RIGHT
1020 * and FOLD flags. The other is reserved.
1021 *
1022 * 2^14 preds is way more than enough.
1023 */
1024#define MAX_FILTER_PRED 16384
1025
1026struct filter_pred;
1027struct regex;
1028
1029typedef int (*filter_pred_fn_t) (struct filter_pred *pred, void *event);
1030
1031typedef int (*regex_match_func)(char *str, struct regex *r, int len);
1032
1033enum regex_type {
1034 MATCH_FULL = 0,
1035 MATCH_FRONT_ONLY,
1036 MATCH_MIDDLE_ONLY,
1037 MATCH_END_ONLY,
1038};
1039
1040struct regex {
1041 char pattern[MAX_FILTER_STR_VAL];
1042 int len;
1043 int field_len;
1044 regex_match_func match;
1045};
1046
1047struct filter_pred {
1048 filter_pred_fn_t fn;
1049 u64 val;
1050 struct regex regex;
1051 unsigned short *ops;
1052 struct ftrace_event_field *field;
1053 int offset;
1054 int not;
1055 int op;
1056 unsigned short index;
1057 unsigned short parent;
1058 unsigned short left;
1059 unsigned short right;
1060};
1061
1062extern enum regex_type
1063filter_parse_regex(char *buff, int len, char **search, int *not);
1064extern void print_event_filter(struct ftrace_event_file *file,
1065 struct trace_seq *s);
1066extern int apply_event_filter(struct ftrace_event_file *file,
1067 char *filter_string);
1068extern int apply_subsystem_event_filter(struct ftrace_subsystem_dir *dir,
1069 char *filter_string);
1070extern void print_subsystem_event_filter(struct event_subsystem *system,
1071 struct trace_seq *s);
1072extern int filter_assign_type(const char *type);
1073extern int create_event_filter(struct ftrace_event_call *call,
1074 char *filter_str, bool set_str,
1075 struct event_filter **filterp);
1076extern void free_event_filter(struct event_filter *filter);
1077
1078struct ftrace_event_field *
1079trace_find_event_field(struct ftrace_event_call *call, char *name);
1080
1081extern void trace_event_enable_cmd_record(bool enable);
1082extern int event_trace_add_tracer(struct dentry *parent, struct trace_array *tr);
1083extern int event_trace_del_tracer(struct trace_array *tr);
1084
1085extern struct ftrace_event_file *find_event_file(struct trace_array *tr,
1086 const char *system,
1087 const char *event);
1088
1089static inline void *event_file_data(struct file *filp)
1090{
1091 return ACCESS_ONCE(file_inode(filp)->i_private);
1092}
1093
1094extern struct mutex event_mutex;
1095extern struct list_head ftrace_events;
1096
1097extern const struct file_operations event_trigger_fops;
1098
1099extern int register_trigger_cmds(void);
1100extern void clear_event_triggers(struct trace_array *tr);
1101
1102struct event_trigger_data {
1103 unsigned long count;
1104 int ref;
1105 struct event_trigger_ops *ops;
1106 struct event_command *cmd_ops;
1107 struct event_filter __rcu *filter;
1108 char *filter_str;
1109 void *private_data;
1110 struct list_head list;
1111};
1112
1113/**
1114 * struct event_trigger_ops - callbacks for trace event triggers
1115 *
1116 * The methods in this structure provide per-event trigger hooks for
1117 * various trigger operations.
1118 *
1119 * All the methods below, except for @init() and @free(), must be
1120 * implemented.
1121 *
1122 * @func: The trigger 'probe' function called when the triggering
1123 * event occurs. The data passed into this callback is the data
1124 * that was supplied to the event_command @reg() function that
1125 * registered the trigger (see struct event_command).
1126 *
1127 * @init: An optional initialization function called for the trigger
1128 * when the trigger is registered (via the event_command reg()
1129 * function). This can be used to perform per-trigger
1130 * initialization such as incrementing a per-trigger reference
1131 * count, for instance. This is usually implemented by the
1132 * generic utility function @event_trigger_init() (see
1133 * trace_event_triggers.c).
1134 *
1135 * @free: An optional de-initialization function called for the
1136 * trigger when the trigger is unregistered (via the
1137 * event_command @reg() function). This can be used to perform
1138 * per-trigger de-initialization such as decrementing a
1139 * per-trigger reference count and freeing corresponding trigger
1140 * data, for instance. This is usually implemented by the
1141 * generic utility function @event_trigger_free() (see
1142 * trace_event_triggers.c).
1143 *
1144 * @print: The callback function invoked to have the trigger print
1145 * itself. This is usually implemented by a wrapper function
1146 * that calls the generic utility function @event_trigger_print()
1147 * (see trace_event_triggers.c).
1148 */
1149struct event_trigger_ops {
1150 void (*func)(struct event_trigger_data *data);
1151 int (*init)(struct event_trigger_ops *ops,
1152 struct event_trigger_data *data);
1153 void (*free)(struct event_trigger_ops *ops,
1154 struct event_trigger_data *data);
1155 int (*print)(struct seq_file *m,
1156 struct event_trigger_ops *ops,
1157 struct event_trigger_data *data);
1158};
1159
1160/**
1161 * struct event_command - callbacks and data members for event commands
1162 *
1163 * Event commands are invoked by users by writing the command name
1164 * into the 'trigger' file associated with a trace event. The
1165 * parameters associated with a specific invocation of an event
1166 * command are used to create an event trigger instance, which is
1167 * added to the list of trigger instances associated with that trace
1168 * event. When the event is hit, the set of triggers associated with
1169 * that event is invoked.
1170 *
1171 * The data members in this structure provide per-event command data
1172 * for various event commands.
1173 *
1174 * All the data members below, except for @post_trigger, must be set
1175 * for each event command.
1176 *
1177 * @name: The unique name that identifies the event command. This is
1178 * the name used when setting triggers via trigger files.
1179 *
1180 * @trigger_type: A unique id that identifies the event command
1181 * 'type'. This value has two purposes, the first to ensure that
1182 * only one trigger of the same type can be set at a given time
1183 * for a particular event e.g. it doesn't make sense to have both
1184 * a traceon and traceoff trigger attached to a single event at
1185 * the same time, so traceon and traceoff have the same type
1186 * though they have different names. The @trigger_type value is
1187 * also used as a bit value for deferring the actual trigger
1188 * action until after the current event is finished. Some
1189 * commands need to do this if they themselves log to the trace
1190 * buffer (see the @post_trigger() member below). @trigger_type
1191 * values are defined by adding new values to the trigger_type
1192 * enum in include/linux/ftrace_event.h.
1193 *
1194 * @post_trigger: A flag that says whether or not this command needs
1195 * to have its action delayed until after the current event has
1196 * been closed. Some triggers need to avoid being invoked while
1197 * an event is currently in the process of being logged, since
1198 * the trigger may itself log data into the trace buffer. Thus
1199 * we make sure the current event is committed before invoking
1200 * those triggers. To do that, the trigger invocation is split
1201 * in two - the first part checks the filter using the current
1202 * trace record; if a command has the @post_trigger flag set, it
1203 * sets a bit for itself in the return value, otherwise it
1204 * directly invokes the trigger. Once all commands have been
1205 * either invoked or set their return flag, the current record is
1206 * either committed or discarded. At that point, if any commands
1207 * have deferred their triggers, those commands are finally
1208 * invoked following the close of the current event. In other
1209 * words, if the event_trigger_ops @func() probe implementation
1210 * itself logs to the trace buffer, this flag should be set,
1211 * otherwise it can be left unspecified.
1212 *
1213 * All the methods below, except for @set_filter(), must be
1214 * implemented.
1215 *
1216 * @func: The callback function responsible for parsing and
1217 * registering the trigger written to the 'trigger' file by the
1218 * user. It allocates the trigger instance and registers it with
1219 * the appropriate trace event. It makes use of the other
1220 * event_command callback functions to orchestrate this, and is
1221 * usually implemented by the generic utility function
1222 * @event_trigger_callback() (see trace_event_triggers.c).
1223 *
1224 * @reg: Adds the trigger to the list of triggers associated with the
1225 * event, and enables the event trigger itself, after
1226 * initializing it (via the event_trigger_ops @init() function).
1227 * This is also where commands can use the @trigger_type value to
1228 * make the decision as to whether or not multiple instances of
1229 * the trigger should be allowed. This is usually implemented by
1230 * the generic utility function @register_trigger() (see
1231 * trace_event_triggers.c).
1232 *
1233 * @unreg: Removes the trigger from the list of triggers associated
1234 * with the event, and disables the event trigger itself, after
1235 * initializing it (via the event_trigger_ops @free() function).
1236 * This is usually implemented by the generic utility function
1237 * @unregister_trigger() (see trace_event_triggers.c).
1238 *
1239 * @set_filter: An optional function called to parse and set a filter
1240 * for the trigger. If no @set_filter() method is set for the
1241 * event command, filters set by the user for the command will be
1242 * ignored. This is usually implemented by the generic utility
1243 * function @set_trigger_filter() (see trace_event_triggers.c).
1244 *
1245 * @get_trigger_ops: The callback function invoked to retrieve the
1246 * event_trigger_ops implementation associated with the command.
1247 */
1248struct event_command {
1249 struct list_head list;
1250 char *name;
1251 enum event_trigger_type trigger_type;
1252 bool post_trigger;
1253 int (*func)(struct event_command *cmd_ops,
1254 struct ftrace_event_file *file,
1255 char *glob, char *cmd, char *params);
1256 int (*reg)(char *glob,
1257 struct event_trigger_ops *ops,
1258 struct event_trigger_data *data,
1259 struct ftrace_event_file *file);
1260 void (*unreg)(char *glob,
1261 struct event_trigger_ops *ops,
1262 struct event_trigger_data *data,
1263 struct ftrace_event_file *file);
1264 int (*set_filter)(char *filter_str,
1265 struct event_trigger_data *data,
1266 struct ftrace_event_file *file);
1267 struct event_trigger_ops *(*get_trigger_ops)(char *cmd, char *param);
1268};
1269
1270extern int trace_event_enable_disable(struct ftrace_event_file *file,
1271 int enable, int soft_disable);
1272extern int tracing_alloc_snapshot(void);
1273
1274extern const char *__start___trace_bprintk_fmt[];
1275extern const char *__stop___trace_bprintk_fmt[];
1276
1277extern const char *__start___tracepoint_str[];
1278extern const char *__stop___tracepoint_str[];
1279
1280void trace_printk_init_buffers(void);
1281void trace_printk_start_comm(void);
1282int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set);
1283int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled);
1284
1285/*
1286 * Normal trace_printk() and friends allocates special buffers
1287 * to do the manipulation, as well as saves the print formats
1288 * into sections to display. But the trace infrastructure wants
1289 * to use these without the added overhead at the price of being
1290 * a bit slower (used mainly for warnings, where we don't care
1291 * about performance). The internal_trace_puts() is for such
1292 * a purpose.
1293 */
1294#define internal_trace_puts(str) __trace_puts(_THIS_IP_, str, strlen(str))
1295
1296#undef FTRACE_ENTRY
1297#define FTRACE_ENTRY(call, struct_name, id, tstruct, print, filter) \
1298 extern struct ftrace_event_call \
1299 __aligned(4) event_##call;
1300#undef FTRACE_ENTRY_DUP
1301#define FTRACE_ENTRY_DUP(call, struct_name, id, tstruct, print, filter) \
1302 FTRACE_ENTRY(call, struct_name, id, PARAMS(tstruct), PARAMS(print), \
1303 filter)
1304#include "trace_entries.h"
1305
1306#if defined(CONFIG_PERF_EVENTS) && defined(CONFIG_FUNCTION_TRACER)
1307int perf_ftrace_event_register(struct ftrace_event_call *call,
1308 enum trace_reg type, void *data);
1309#else
1310#define perf_ftrace_event_register NULL
1311#endif
1312
1313#endif /* _LINUX_KERNEL_TRACE_H */