xref: /linux/kernel/trace/trace.c (revision 96f30c8f0aa9923aa39b30bcaefeacf88b490231)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * ring buffer based function tracer
4  *
5  * Copyright (C) 2007-2012 Steven Rostedt <srostedt@redhat.com>
6  * Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
7  *
8  * Originally taken from the RT patch by:
9  *    Arnaldo Carvalho de Melo <acme@redhat.com>
10  *
11  * Based on code from the latency_tracer, that is:
12  *  Copyright (C) 2004-2006 Ingo Molnar
13  *  Copyright (C) 2004 Nadia Yvette Chambers
14  */
15 #include <linux/ring_buffer.h>
16 #include <linux/utsname.h>
17 #include <linux/stacktrace.h>
18 #include <linux/writeback.h>
19 #include <linux/kallsyms.h>
20 #include <linux/security.h>
21 #include <linux/seq_file.h>
22 #include <linux/irqflags.h>
23 #include <linux/debugfs.h>
24 #include <linux/tracefs.h>
25 #include <linux/pagemap.h>
26 #include <linux/hardirq.h>
27 #include <linux/linkage.h>
28 #include <linux/uaccess.h>
29 #include <linux/vmalloc.h>
30 #include <linux/ftrace.h>
31 #include <linux/module.h>
32 #include <linux/percpu.h>
33 #include <linux/splice.h>
34 #include <linux/kdebug.h>
35 #include <linux/string.h>
36 #include <linux/mount.h>
37 #include <linux/rwsem.h>
38 #include <linux/slab.h>
39 #include <linux/ctype.h>
40 #include <linux/init.h>
41 #include <linux/panic_notifier.h>
42 #include <linux/poll.h>
43 #include <linux/nmi.h>
44 #include <linux/fs.h>
45 #include <linux/trace.h>
46 #include <linux/sched/clock.h>
47 #include <linux/sched/rt.h>
48 #include <linux/fsnotify.h>
49 #include <linux/irq_work.h>
50 #include <linux/workqueue.h>
51 
52 #include <asm/setup.h> /* COMMAND_LINE_SIZE */
53 
54 #include "trace.h"
55 #include "trace_output.h"
56 
57 #ifdef CONFIG_FTRACE_STARTUP_TEST
58 /*
59  * We need to change this state when a selftest is running.
60  * A selftest will lurk into the ring-buffer to count the
61  * entries inserted during the selftest although some concurrent
62  * insertions into the ring-buffer such as trace_printk could occurred
63  * at the same time, giving false positive or negative results.
64  */
65 static bool __read_mostly tracing_selftest_running;
66 
67 /*
68  * If boot-time tracing including tracers/events via kernel cmdline
69  * is running, we do not want to run SELFTEST.
70  */
71 bool __read_mostly tracing_selftest_disabled;
72 
73 void __init disable_tracing_selftest(const char *reason)
74 {
75 	if (!tracing_selftest_disabled) {
76 		tracing_selftest_disabled = true;
77 		pr_info("Ftrace startup test is disabled due to %s\n", reason);
78 	}
79 }
80 #else
81 #define tracing_selftest_running	0
82 #define tracing_selftest_disabled	0
83 #endif
84 
85 /* Pipe tracepoints to printk */
86 static struct trace_iterator *tracepoint_print_iter;
87 int tracepoint_printk;
88 static bool tracepoint_printk_stop_on_boot __initdata;
89 static DEFINE_STATIC_KEY_FALSE(tracepoint_printk_key);
90 
91 /* For tracers that don't implement custom flags */
92 static struct tracer_opt dummy_tracer_opt[] = {
93 	{ }
94 };
95 
96 static int
97 dummy_set_flag(struct trace_array *tr, u32 old_flags, u32 bit, int set)
98 {
99 	return 0;
100 }
101 
102 /*
103  * To prevent the comm cache from being overwritten when no
104  * tracing is active, only save the comm when a trace event
105  * occurred.
106  */
107 DEFINE_PER_CPU(bool, trace_taskinfo_save);
108 
109 /*
110  * Kill all tracing for good (never come back).
111  * It is initialized to 1 but will turn to zero if the initialization
112  * of the tracer is successful. But that is the only place that sets
113  * this back to zero.
114  */
115 static int tracing_disabled = 1;
116 
117 cpumask_var_t __read_mostly	tracing_buffer_mask;
118 
119 /*
120  * ftrace_dump_on_oops - variable to dump ftrace buffer on oops
121  *
122  * If there is an oops (or kernel panic) and the ftrace_dump_on_oops
123  * is set, then ftrace_dump is called. This will output the contents
124  * of the ftrace buffers to the console.  This is very useful for
125  * capturing traces that lead to crashes and outputing it to a
126  * serial console.
127  *
128  * It is default off, but you can enable it with either specifying
129  * "ftrace_dump_on_oops" in the kernel command line, or setting
130  * /proc/sys/kernel/ftrace_dump_on_oops
131  * Set 1 if you want to dump buffers of all CPUs
132  * Set 2 if you want to dump the buffer of the CPU that triggered oops
133  * Set instance name if you want to dump the specific trace instance
134  * Multiple instance dump is also supported, and instances are seperated
135  * by commas.
136  */
137 /* Set to string format zero to disable by default */
138 char ftrace_dump_on_oops[MAX_TRACER_SIZE] = "0";
139 
140 /* When set, tracing will stop when a WARN*() is hit */
141 int __disable_trace_on_warning;
142 
143 #ifdef CONFIG_TRACE_EVAL_MAP_FILE
144 /* Map of enums to their values, for "eval_map" file */
145 struct trace_eval_map_head {
146 	struct module			*mod;
147 	unsigned long			length;
148 };
149 
150 union trace_eval_map_item;
151 
152 struct trace_eval_map_tail {
153 	/*
154 	 * "end" is first and points to NULL as it must be different
155 	 * than "mod" or "eval_string"
156 	 */
157 	union trace_eval_map_item	*next;
158 	const char			*end;	/* points to NULL */
159 };
160 
161 static DEFINE_MUTEX(trace_eval_mutex);
162 
163 /*
164  * The trace_eval_maps are saved in an array with two extra elements,
165  * one at the beginning, and one at the end. The beginning item contains
166  * the count of the saved maps (head.length), and the module they
167  * belong to if not built in (head.mod). The ending item contains a
168  * pointer to the next array of saved eval_map items.
169  */
170 union trace_eval_map_item {
171 	struct trace_eval_map		map;
172 	struct trace_eval_map_head	head;
173 	struct trace_eval_map_tail	tail;
174 };
175 
176 static union trace_eval_map_item *trace_eval_maps;
177 #endif /* CONFIG_TRACE_EVAL_MAP_FILE */
178 
179 int tracing_set_tracer(struct trace_array *tr, const char *buf);
180 static void ftrace_trace_userstack(struct trace_array *tr,
181 				   struct trace_buffer *buffer,
182 				   unsigned int trace_ctx);
183 
184 static char bootup_tracer_buf[MAX_TRACER_SIZE] __initdata;
185 static char *default_bootup_tracer;
186 
187 static bool allocate_snapshot;
188 static bool snapshot_at_boot;
189 
190 static char boot_instance_info[COMMAND_LINE_SIZE] __initdata;
191 static int boot_instance_index;
192 
193 static char boot_snapshot_info[COMMAND_LINE_SIZE] __initdata;
194 static int boot_snapshot_index;
195 
196 static int __init set_cmdline_ftrace(char *str)
197 {
198 	strscpy(bootup_tracer_buf, str, MAX_TRACER_SIZE);
199 	default_bootup_tracer = bootup_tracer_buf;
200 	/* We are using ftrace early, expand it */
201 	trace_set_ring_buffer_expanded(NULL);
202 	return 1;
203 }
204 __setup("ftrace=", set_cmdline_ftrace);
205 
206 int ftrace_dump_on_oops_enabled(void)
207 {
208 	if (!strcmp("0", ftrace_dump_on_oops))
209 		return 0;
210 	else
211 		return 1;
212 }
213 
214 static int __init set_ftrace_dump_on_oops(char *str)
215 {
216 	if (!*str) {
217 		strscpy(ftrace_dump_on_oops, "1", MAX_TRACER_SIZE);
218 		return 1;
219 	}
220 
221 	if (*str == ',') {
222 		strscpy(ftrace_dump_on_oops, "1", MAX_TRACER_SIZE);
223 		strscpy(ftrace_dump_on_oops + 1, str, MAX_TRACER_SIZE - 1);
224 		return 1;
225 	}
226 
227 	if (*str++ == '=') {
228 		strscpy(ftrace_dump_on_oops, str, MAX_TRACER_SIZE);
229 		return 1;
230 	}
231 
232 	return 0;
233 }
234 __setup("ftrace_dump_on_oops", set_ftrace_dump_on_oops);
235 
236 static int __init stop_trace_on_warning(char *str)
237 {
238 	if ((strcmp(str, "=0") != 0 && strcmp(str, "=off") != 0))
239 		__disable_trace_on_warning = 1;
240 	return 1;
241 }
242 __setup("traceoff_on_warning", stop_trace_on_warning);
243 
244 static int __init boot_alloc_snapshot(char *str)
245 {
246 	char *slot = boot_snapshot_info + boot_snapshot_index;
247 	int left = sizeof(boot_snapshot_info) - boot_snapshot_index;
248 	int ret;
249 
250 	if (str[0] == '=') {
251 		str++;
252 		if (strlen(str) >= left)
253 			return -1;
254 
255 		ret = snprintf(slot, left, "%s\t", str);
256 		boot_snapshot_index += ret;
257 	} else {
258 		allocate_snapshot = true;
259 		/* We also need the main ring buffer expanded */
260 		trace_set_ring_buffer_expanded(NULL);
261 	}
262 	return 1;
263 }
264 __setup("alloc_snapshot", boot_alloc_snapshot);
265 
266 
267 static int __init boot_snapshot(char *str)
268 {
269 	snapshot_at_boot = true;
270 	boot_alloc_snapshot(str);
271 	return 1;
272 }
273 __setup("ftrace_boot_snapshot", boot_snapshot);
274 
275 
276 static int __init boot_instance(char *str)
277 {
278 	char *slot = boot_instance_info + boot_instance_index;
279 	int left = sizeof(boot_instance_info) - boot_instance_index;
280 	int ret;
281 
282 	if (strlen(str) >= left)
283 		return -1;
284 
285 	ret = snprintf(slot, left, "%s\t", str);
286 	boot_instance_index += ret;
287 
288 	return 1;
289 }
290 __setup("trace_instance=", boot_instance);
291 
292 
293 static char trace_boot_options_buf[MAX_TRACER_SIZE] __initdata;
294 
295 static int __init set_trace_boot_options(char *str)
296 {
297 	strscpy(trace_boot_options_buf, str, MAX_TRACER_SIZE);
298 	return 1;
299 }
300 __setup("trace_options=", set_trace_boot_options);
301 
302 static char trace_boot_clock_buf[MAX_TRACER_SIZE] __initdata;
303 static char *trace_boot_clock __initdata;
304 
305 static int __init set_trace_boot_clock(char *str)
306 {
307 	strscpy(trace_boot_clock_buf, str, MAX_TRACER_SIZE);
308 	trace_boot_clock = trace_boot_clock_buf;
309 	return 1;
310 }
311 __setup("trace_clock=", set_trace_boot_clock);
312 
313 static int __init set_tracepoint_printk(char *str)
314 {
315 	/* Ignore the "tp_printk_stop_on_boot" param */
316 	if (*str == '_')
317 		return 0;
318 
319 	if ((strcmp(str, "=0") != 0 && strcmp(str, "=off") != 0))
320 		tracepoint_printk = 1;
321 	return 1;
322 }
323 __setup("tp_printk", set_tracepoint_printk);
324 
325 static int __init set_tracepoint_printk_stop(char *str)
326 {
327 	tracepoint_printk_stop_on_boot = true;
328 	return 1;
329 }
330 __setup("tp_printk_stop_on_boot", set_tracepoint_printk_stop);
331 
332 unsigned long long ns2usecs(u64 nsec)
333 {
334 	nsec += 500;
335 	do_div(nsec, 1000);
336 	return nsec;
337 }
338 
339 static void
340 trace_process_export(struct trace_export *export,
341 	       struct ring_buffer_event *event, int flag)
342 {
343 	struct trace_entry *entry;
344 	unsigned int size = 0;
345 
346 	if (export->flags & flag) {
347 		entry = ring_buffer_event_data(event);
348 		size = ring_buffer_event_length(event);
349 		export->write(export, entry, size);
350 	}
351 }
352 
353 static DEFINE_MUTEX(ftrace_export_lock);
354 
355 static struct trace_export __rcu *ftrace_exports_list __read_mostly;
356 
357 static DEFINE_STATIC_KEY_FALSE(trace_function_exports_enabled);
358 static DEFINE_STATIC_KEY_FALSE(trace_event_exports_enabled);
359 static DEFINE_STATIC_KEY_FALSE(trace_marker_exports_enabled);
360 
361 static inline void ftrace_exports_enable(struct trace_export *export)
362 {
363 	if (export->flags & TRACE_EXPORT_FUNCTION)
364 		static_branch_inc(&trace_function_exports_enabled);
365 
366 	if (export->flags & TRACE_EXPORT_EVENT)
367 		static_branch_inc(&trace_event_exports_enabled);
368 
369 	if (export->flags & TRACE_EXPORT_MARKER)
370 		static_branch_inc(&trace_marker_exports_enabled);
371 }
372 
373 static inline void ftrace_exports_disable(struct trace_export *export)
374 {
375 	if (export->flags & TRACE_EXPORT_FUNCTION)
376 		static_branch_dec(&trace_function_exports_enabled);
377 
378 	if (export->flags & TRACE_EXPORT_EVENT)
379 		static_branch_dec(&trace_event_exports_enabled);
380 
381 	if (export->flags & TRACE_EXPORT_MARKER)
382 		static_branch_dec(&trace_marker_exports_enabled);
383 }
384 
385 static void ftrace_exports(struct ring_buffer_event *event, int flag)
386 {
387 	struct trace_export *export;
388 
389 	preempt_disable_notrace();
390 
391 	export = rcu_dereference_raw_check(ftrace_exports_list);
392 	while (export) {
393 		trace_process_export(export, event, flag);
394 		export = rcu_dereference_raw_check(export->next);
395 	}
396 
397 	preempt_enable_notrace();
398 }
399 
400 static inline void
401 add_trace_export(struct trace_export **list, struct trace_export *export)
402 {
403 	rcu_assign_pointer(export->next, *list);
404 	/*
405 	 * We are entering export into the list but another
406 	 * CPU might be walking that list. We need to make sure
407 	 * the export->next pointer is valid before another CPU sees
408 	 * the export pointer included into the list.
409 	 */
410 	rcu_assign_pointer(*list, export);
411 }
412 
413 static inline int
414 rm_trace_export(struct trace_export **list, struct trace_export *export)
415 {
416 	struct trace_export **p;
417 
418 	for (p = list; *p != NULL; p = &(*p)->next)
419 		if (*p == export)
420 			break;
421 
422 	if (*p != export)
423 		return -1;
424 
425 	rcu_assign_pointer(*p, (*p)->next);
426 
427 	return 0;
428 }
429 
430 static inline void
431 add_ftrace_export(struct trace_export **list, struct trace_export *export)
432 {
433 	ftrace_exports_enable(export);
434 
435 	add_trace_export(list, export);
436 }
437 
438 static inline int
439 rm_ftrace_export(struct trace_export **list, struct trace_export *export)
440 {
441 	int ret;
442 
443 	ret = rm_trace_export(list, export);
444 	ftrace_exports_disable(export);
445 
446 	return ret;
447 }
448 
449 int register_ftrace_export(struct trace_export *export)
450 {
451 	if (WARN_ON_ONCE(!export->write))
452 		return -1;
453 
454 	mutex_lock(&ftrace_export_lock);
455 
456 	add_ftrace_export(&ftrace_exports_list, export);
457 
458 	mutex_unlock(&ftrace_export_lock);
459 
460 	return 0;
461 }
462 EXPORT_SYMBOL_GPL(register_ftrace_export);
463 
464 int unregister_ftrace_export(struct trace_export *export)
465 {
466 	int ret;
467 
468 	mutex_lock(&ftrace_export_lock);
469 
470 	ret = rm_ftrace_export(&ftrace_exports_list, export);
471 
472 	mutex_unlock(&ftrace_export_lock);
473 
474 	return ret;
475 }
476 EXPORT_SYMBOL_GPL(unregister_ftrace_export);
477 
478 /* trace_flags holds trace_options default values */
479 #define TRACE_DEFAULT_FLAGS						\
480 	(FUNCTION_DEFAULT_FLAGS |					\
481 	 TRACE_ITER_PRINT_PARENT | TRACE_ITER_PRINTK |			\
482 	 TRACE_ITER_ANNOTATE | TRACE_ITER_CONTEXT_INFO |		\
483 	 TRACE_ITER_RECORD_CMD | TRACE_ITER_OVERWRITE |			\
484 	 TRACE_ITER_IRQ_INFO | TRACE_ITER_MARKERS |			\
485 	 TRACE_ITER_HASH_PTR)
486 
487 /* trace_options that are only supported by global_trace */
488 #define TOP_LEVEL_TRACE_FLAGS (TRACE_ITER_PRINTK |			\
489 	       TRACE_ITER_PRINTK_MSGONLY | TRACE_ITER_RECORD_CMD)
490 
491 /* trace_flags that are default zero for instances */
492 #define ZEROED_TRACE_FLAGS \
493 	(TRACE_ITER_EVENT_FORK | TRACE_ITER_FUNC_FORK)
494 
495 /*
496  * The global_trace is the descriptor that holds the top-level tracing
497  * buffers for the live tracing.
498  */
499 static struct trace_array global_trace = {
500 	.trace_flags = TRACE_DEFAULT_FLAGS,
501 };
502 
503 void trace_set_ring_buffer_expanded(struct trace_array *tr)
504 {
505 	if (!tr)
506 		tr = &global_trace;
507 	tr->ring_buffer_expanded = true;
508 }
509 
510 LIST_HEAD(ftrace_trace_arrays);
511 
512 int trace_array_get(struct trace_array *this_tr)
513 {
514 	struct trace_array *tr;
515 	int ret = -ENODEV;
516 
517 	mutex_lock(&trace_types_lock);
518 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
519 		if (tr == this_tr) {
520 			tr->ref++;
521 			ret = 0;
522 			break;
523 		}
524 	}
525 	mutex_unlock(&trace_types_lock);
526 
527 	return ret;
528 }
529 
530 static void __trace_array_put(struct trace_array *this_tr)
531 {
532 	WARN_ON(!this_tr->ref);
533 	this_tr->ref--;
534 }
535 
536 /**
537  * trace_array_put - Decrement the reference counter for this trace array.
538  * @this_tr : pointer to the trace array
539  *
540  * NOTE: Use this when we no longer need the trace array returned by
541  * trace_array_get_by_name(). This ensures the trace array can be later
542  * destroyed.
543  *
544  */
545 void trace_array_put(struct trace_array *this_tr)
546 {
547 	if (!this_tr)
548 		return;
549 
550 	mutex_lock(&trace_types_lock);
551 	__trace_array_put(this_tr);
552 	mutex_unlock(&trace_types_lock);
553 }
554 EXPORT_SYMBOL_GPL(trace_array_put);
555 
556 int tracing_check_open_get_tr(struct trace_array *tr)
557 {
558 	int ret;
559 
560 	ret = security_locked_down(LOCKDOWN_TRACEFS);
561 	if (ret)
562 		return ret;
563 
564 	if (tracing_disabled)
565 		return -ENODEV;
566 
567 	if (tr && trace_array_get(tr) < 0)
568 		return -ENODEV;
569 
570 	return 0;
571 }
572 
573 int call_filter_check_discard(struct trace_event_call *call, void *rec,
574 			      struct trace_buffer *buffer,
575 			      struct ring_buffer_event *event)
576 {
577 	if (unlikely(call->flags & TRACE_EVENT_FL_FILTERED) &&
578 	    !filter_match_preds(call->filter, rec)) {
579 		__trace_event_discard_commit(buffer, event);
580 		return 1;
581 	}
582 
583 	return 0;
584 }
585 
586 /**
587  * trace_find_filtered_pid - check if a pid exists in a filtered_pid list
588  * @filtered_pids: The list of pids to check
589  * @search_pid: The PID to find in @filtered_pids
590  *
591  * Returns true if @search_pid is found in @filtered_pids, and false otherwise.
592  */
593 bool
594 trace_find_filtered_pid(struct trace_pid_list *filtered_pids, pid_t search_pid)
595 {
596 	return trace_pid_list_is_set(filtered_pids, search_pid);
597 }
598 
599 /**
600  * trace_ignore_this_task - should a task be ignored for tracing
601  * @filtered_pids: The list of pids to check
602  * @filtered_no_pids: The list of pids not to be traced
603  * @task: The task that should be ignored if not filtered
604  *
605  * Checks if @task should be traced or not from @filtered_pids.
606  * Returns true if @task should *NOT* be traced.
607  * Returns false if @task should be traced.
608  */
609 bool
610 trace_ignore_this_task(struct trace_pid_list *filtered_pids,
611 		       struct trace_pid_list *filtered_no_pids,
612 		       struct task_struct *task)
613 {
614 	/*
615 	 * If filtered_no_pids is not empty, and the task's pid is listed
616 	 * in filtered_no_pids, then return true.
617 	 * Otherwise, if filtered_pids is empty, that means we can
618 	 * trace all tasks. If it has content, then only trace pids
619 	 * within filtered_pids.
620 	 */
621 
622 	return (filtered_pids &&
623 		!trace_find_filtered_pid(filtered_pids, task->pid)) ||
624 		(filtered_no_pids &&
625 		 trace_find_filtered_pid(filtered_no_pids, task->pid));
626 }
627 
628 /**
629  * trace_filter_add_remove_task - Add or remove a task from a pid_list
630  * @pid_list: The list to modify
631  * @self: The current task for fork or NULL for exit
632  * @task: The task to add or remove
633  *
634  * If adding a task, if @self is defined, the task is only added if @self
635  * is also included in @pid_list. This happens on fork and tasks should
636  * only be added when the parent is listed. If @self is NULL, then the
637  * @task pid will be removed from the list, which would happen on exit
638  * of a task.
639  */
640 void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
641 				  struct task_struct *self,
642 				  struct task_struct *task)
643 {
644 	if (!pid_list)
645 		return;
646 
647 	/* For forks, we only add if the forking task is listed */
648 	if (self) {
649 		if (!trace_find_filtered_pid(pid_list, self->pid))
650 			return;
651 	}
652 
653 	/* "self" is set for forks, and NULL for exits */
654 	if (self)
655 		trace_pid_list_set(pid_list, task->pid);
656 	else
657 		trace_pid_list_clear(pid_list, task->pid);
658 }
659 
660 /**
661  * trace_pid_next - Used for seq_file to get to the next pid of a pid_list
662  * @pid_list: The pid list to show
663  * @v: The last pid that was shown (+1 the actual pid to let zero be displayed)
664  * @pos: The position of the file
665  *
666  * This is used by the seq_file "next" operation to iterate the pids
667  * listed in a trace_pid_list structure.
668  *
669  * Returns the pid+1 as we want to display pid of zero, but NULL would
670  * stop the iteration.
671  */
672 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos)
673 {
674 	long pid = (unsigned long)v;
675 	unsigned int next;
676 
677 	(*pos)++;
678 
679 	/* pid already is +1 of the actual previous bit */
680 	if (trace_pid_list_next(pid_list, pid, &next) < 0)
681 		return NULL;
682 
683 	pid = next;
684 
685 	/* Return pid + 1 to allow zero to be represented */
686 	return (void *)(pid + 1);
687 }
688 
689 /**
690  * trace_pid_start - Used for seq_file to start reading pid lists
691  * @pid_list: The pid list to show
692  * @pos: The position of the file
693  *
694  * This is used by seq_file "start" operation to start the iteration
695  * of listing pids.
696  *
697  * Returns the pid+1 as we want to display pid of zero, but NULL would
698  * stop the iteration.
699  */
700 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos)
701 {
702 	unsigned long pid;
703 	unsigned int first;
704 	loff_t l = 0;
705 
706 	if (trace_pid_list_first(pid_list, &first) < 0)
707 		return NULL;
708 
709 	pid = first;
710 
711 	/* Return pid + 1 so that zero can be the exit value */
712 	for (pid++; pid && l < *pos;
713 	     pid = (unsigned long)trace_pid_next(pid_list, (void *)pid, &l))
714 		;
715 	return (void *)pid;
716 }
717 
718 /**
719  * trace_pid_show - show the current pid in seq_file processing
720  * @m: The seq_file structure to write into
721  * @v: A void pointer of the pid (+1) value to display
722  *
723  * Can be directly used by seq_file operations to display the current
724  * pid value.
725  */
726 int trace_pid_show(struct seq_file *m, void *v)
727 {
728 	unsigned long pid = (unsigned long)v - 1;
729 
730 	seq_printf(m, "%lu\n", pid);
731 	return 0;
732 }
733 
734 /* 128 should be much more than enough */
735 #define PID_BUF_SIZE		127
736 
737 int trace_pid_write(struct trace_pid_list *filtered_pids,
738 		    struct trace_pid_list **new_pid_list,
739 		    const char __user *ubuf, size_t cnt)
740 {
741 	struct trace_pid_list *pid_list;
742 	struct trace_parser parser;
743 	unsigned long val;
744 	int nr_pids = 0;
745 	ssize_t read = 0;
746 	ssize_t ret;
747 	loff_t pos;
748 	pid_t pid;
749 
750 	if (trace_parser_get_init(&parser, PID_BUF_SIZE + 1))
751 		return -ENOMEM;
752 
753 	/*
754 	 * Always recreate a new array. The write is an all or nothing
755 	 * operation. Always create a new array when adding new pids by
756 	 * the user. If the operation fails, then the current list is
757 	 * not modified.
758 	 */
759 	pid_list = trace_pid_list_alloc();
760 	if (!pid_list) {
761 		trace_parser_put(&parser);
762 		return -ENOMEM;
763 	}
764 
765 	if (filtered_pids) {
766 		/* copy the current bits to the new max */
767 		ret = trace_pid_list_first(filtered_pids, &pid);
768 		while (!ret) {
769 			trace_pid_list_set(pid_list, pid);
770 			ret = trace_pid_list_next(filtered_pids, pid + 1, &pid);
771 			nr_pids++;
772 		}
773 	}
774 
775 	ret = 0;
776 	while (cnt > 0) {
777 
778 		pos = 0;
779 
780 		ret = trace_get_user(&parser, ubuf, cnt, &pos);
781 		if (ret < 0)
782 			break;
783 
784 		read += ret;
785 		ubuf += ret;
786 		cnt -= ret;
787 
788 		if (!trace_parser_loaded(&parser))
789 			break;
790 
791 		ret = -EINVAL;
792 		if (kstrtoul(parser.buffer, 0, &val))
793 			break;
794 
795 		pid = (pid_t)val;
796 
797 		if (trace_pid_list_set(pid_list, pid) < 0) {
798 			ret = -1;
799 			break;
800 		}
801 		nr_pids++;
802 
803 		trace_parser_clear(&parser);
804 		ret = 0;
805 	}
806 	trace_parser_put(&parser);
807 
808 	if (ret < 0) {
809 		trace_pid_list_free(pid_list);
810 		return ret;
811 	}
812 
813 	if (!nr_pids) {
814 		/* Cleared the list of pids */
815 		trace_pid_list_free(pid_list);
816 		pid_list = NULL;
817 	}
818 
819 	*new_pid_list = pid_list;
820 
821 	return read;
822 }
823 
824 static u64 buffer_ftrace_now(struct array_buffer *buf, int cpu)
825 {
826 	u64 ts;
827 
828 	/* Early boot up does not have a buffer yet */
829 	if (!buf->buffer)
830 		return trace_clock_local();
831 
832 	ts = ring_buffer_time_stamp(buf->buffer);
833 	ring_buffer_normalize_time_stamp(buf->buffer, cpu, &ts);
834 
835 	return ts;
836 }
837 
838 u64 ftrace_now(int cpu)
839 {
840 	return buffer_ftrace_now(&global_trace.array_buffer, cpu);
841 }
842 
843 /**
844  * tracing_is_enabled - Show if global_trace has been enabled
845  *
846  * Shows if the global trace has been enabled or not. It uses the
847  * mirror flag "buffer_disabled" to be used in fast paths such as for
848  * the irqsoff tracer. But it may be inaccurate due to races. If you
849  * need to know the accurate state, use tracing_is_on() which is a little
850  * slower, but accurate.
851  */
852 int tracing_is_enabled(void)
853 {
854 	/*
855 	 * For quick access (irqsoff uses this in fast path), just
856 	 * return the mirror variable of the state of the ring buffer.
857 	 * It's a little racy, but we don't really care.
858 	 */
859 	smp_rmb();
860 	return !global_trace.buffer_disabled;
861 }
862 
863 /*
864  * trace_buf_size is the size in bytes that is allocated
865  * for a buffer. Note, the number of bytes is always rounded
866  * to page size.
867  *
868  * This number is purposely set to a low number of 16384.
869  * If the dump on oops happens, it will be much appreciated
870  * to not have to wait for all that output. Anyway this can be
871  * boot time and run time configurable.
872  */
873 #define TRACE_BUF_SIZE_DEFAULT	1441792UL /* 16384 * 88 (sizeof(entry)) */
874 
875 static unsigned long		trace_buf_size = TRACE_BUF_SIZE_DEFAULT;
876 
877 /* trace_types holds a link list of available tracers. */
878 static struct tracer		*trace_types __read_mostly;
879 
880 /*
881  * trace_types_lock is used to protect the trace_types list.
882  */
883 DEFINE_MUTEX(trace_types_lock);
884 
885 /*
886  * serialize the access of the ring buffer
887  *
888  * ring buffer serializes readers, but it is low level protection.
889  * The validity of the events (which returns by ring_buffer_peek() ..etc)
890  * are not protected by ring buffer.
891  *
892  * The content of events may become garbage if we allow other process consumes
893  * these events concurrently:
894  *   A) the page of the consumed events may become a normal page
895  *      (not reader page) in ring buffer, and this page will be rewritten
896  *      by events producer.
897  *   B) The page of the consumed events may become a page for splice_read,
898  *      and this page will be returned to system.
899  *
900  * These primitives allow multi process access to different cpu ring buffer
901  * concurrently.
902  *
903  * These primitives don't distinguish read-only and read-consume access.
904  * Multi read-only access are also serialized.
905  */
906 
907 #ifdef CONFIG_SMP
908 static DECLARE_RWSEM(all_cpu_access_lock);
909 static DEFINE_PER_CPU(struct mutex, cpu_access_lock);
910 
911 static inline void trace_access_lock(int cpu)
912 {
913 	if (cpu == RING_BUFFER_ALL_CPUS) {
914 		/* gain it for accessing the whole ring buffer. */
915 		down_write(&all_cpu_access_lock);
916 	} else {
917 		/* gain it for accessing a cpu ring buffer. */
918 
919 		/* Firstly block other trace_access_lock(RING_BUFFER_ALL_CPUS). */
920 		down_read(&all_cpu_access_lock);
921 
922 		/* Secondly block other access to this @cpu ring buffer. */
923 		mutex_lock(&per_cpu(cpu_access_lock, cpu));
924 	}
925 }
926 
927 static inline void trace_access_unlock(int cpu)
928 {
929 	if (cpu == RING_BUFFER_ALL_CPUS) {
930 		up_write(&all_cpu_access_lock);
931 	} else {
932 		mutex_unlock(&per_cpu(cpu_access_lock, cpu));
933 		up_read(&all_cpu_access_lock);
934 	}
935 }
936 
937 static inline void trace_access_lock_init(void)
938 {
939 	int cpu;
940 
941 	for_each_possible_cpu(cpu)
942 		mutex_init(&per_cpu(cpu_access_lock, cpu));
943 }
944 
945 #else
946 
947 static DEFINE_MUTEX(access_lock);
948 
949 static inline void trace_access_lock(int cpu)
950 {
951 	(void)cpu;
952 	mutex_lock(&access_lock);
953 }
954 
955 static inline void trace_access_unlock(int cpu)
956 {
957 	(void)cpu;
958 	mutex_unlock(&access_lock);
959 }
960 
961 static inline void trace_access_lock_init(void)
962 {
963 }
964 
965 #endif
966 
967 #ifdef CONFIG_STACKTRACE
968 static void __ftrace_trace_stack(struct trace_buffer *buffer,
969 				 unsigned int trace_ctx,
970 				 int skip, struct pt_regs *regs);
971 static inline void ftrace_trace_stack(struct trace_array *tr,
972 				      struct trace_buffer *buffer,
973 				      unsigned int trace_ctx,
974 				      int skip, struct pt_regs *regs);
975 
976 #else
977 static inline void __ftrace_trace_stack(struct trace_buffer *buffer,
978 					unsigned int trace_ctx,
979 					int skip, struct pt_regs *regs)
980 {
981 }
982 static inline void ftrace_trace_stack(struct trace_array *tr,
983 				      struct trace_buffer *buffer,
984 				      unsigned long trace_ctx,
985 				      int skip, struct pt_regs *regs)
986 {
987 }
988 
989 #endif
990 
991 static __always_inline void
992 trace_event_setup(struct ring_buffer_event *event,
993 		  int type, unsigned int trace_ctx)
994 {
995 	struct trace_entry *ent = ring_buffer_event_data(event);
996 
997 	tracing_generic_entry_update(ent, type, trace_ctx);
998 }
999 
1000 static __always_inline struct ring_buffer_event *
1001 __trace_buffer_lock_reserve(struct trace_buffer *buffer,
1002 			  int type,
1003 			  unsigned long len,
1004 			  unsigned int trace_ctx)
1005 {
1006 	struct ring_buffer_event *event;
1007 
1008 	event = ring_buffer_lock_reserve(buffer, len);
1009 	if (event != NULL)
1010 		trace_event_setup(event, type, trace_ctx);
1011 
1012 	return event;
1013 }
1014 
1015 void tracer_tracing_on(struct trace_array *tr)
1016 {
1017 	if (tr->array_buffer.buffer)
1018 		ring_buffer_record_on(tr->array_buffer.buffer);
1019 	/*
1020 	 * This flag is looked at when buffers haven't been allocated
1021 	 * yet, or by some tracers (like irqsoff), that just want to
1022 	 * know if the ring buffer has been disabled, but it can handle
1023 	 * races of where it gets disabled but we still do a record.
1024 	 * As the check is in the fast path of the tracers, it is more
1025 	 * important to be fast than accurate.
1026 	 */
1027 	tr->buffer_disabled = 0;
1028 	/* Make the flag seen by readers */
1029 	smp_wmb();
1030 }
1031 
1032 /**
1033  * tracing_on - enable tracing buffers
1034  *
1035  * This function enables tracing buffers that may have been
1036  * disabled with tracing_off.
1037  */
1038 void tracing_on(void)
1039 {
1040 	tracer_tracing_on(&global_trace);
1041 }
1042 EXPORT_SYMBOL_GPL(tracing_on);
1043 
1044 
1045 static __always_inline void
1046 __buffer_unlock_commit(struct trace_buffer *buffer, struct ring_buffer_event *event)
1047 {
1048 	__this_cpu_write(trace_taskinfo_save, true);
1049 
1050 	/* If this is the temp buffer, we need to commit fully */
1051 	if (this_cpu_read(trace_buffered_event) == event) {
1052 		/* Length is in event->array[0] */
1053 		ring_buffer_write(buffer, event->array[0], &event->array[1]);
1054 		/* Release the temp buffer */
1055 		this_cpu_dec(trace_buffered_event_cnt);
1056 		/* ring_buffer_unlock_commit() enables preemption */
1057 		preempt_enable_notrace();
1058 	} else
1059 		ring_buffer_unlock_commit(buffer);
1060 }
1061 
1062 int __trace_array_puts(struct trace_array *tr, unsigned long ip,
1063 		       const char *str, int size)
1064 {
1065 	struct ring_buffer_event *event;
1066 	struct trace_buffer *buffer;
1067 	struct print_entry *entry;
1068 	unsigned int trace_ctx;
1069 	int alloc;
1070 
1071 	if (!(tr->trace_flags & TRACE_ITER_PRINTK))
1072 		return 0;
1073 
1074 	if (unlikely(tracing_selftest_running && tr == &global_trace))
1075 		return 0;
1076 
1077 	if (unlikely(tracing_disabled))
1078 		return 0;
1079 
1080 	alloc = sizeof(*entry) + size + 2; /* possible \n added */
1081 
1082 	trace_ctx = tracing_gen_ctx();
1083 	buffer = tr->array_buffer.buffer;
1084 	ring_buffer_nest_start(buffer);
1085 	event = __trace_buffer_lock_reserve(buffer, TRACE_PRINT, alloc,
1086 					    trace_ctx);
1087 	if (!event) {
1088 		size = 0;
1089 		goto out;
1090 	}
1091 
1092 	entry = ring_buffer_event_data(event);
1093 	entry->ip = ip;
1094 
1095 	memcpy(&entry->buf, str, size);
1096 
1097 	/* Add a newline if necessary */
1098 	if (entry->buf[size - 1] != '\n') {
1099 		entry->buf[size] = '\n';
1100 		entry->buf[size + 1] = '\0';
1101 	} else
1102 		entry->buf[size] = '\0';
1103 
1104 	__buffer_unlock_commit(buffer, event);
1105 	ftrace_trace_stack(tr, buffer, trace_ctx, 4, NULL);
1106  out:
1107 	ring_buffer_nest_end(buffer);
1108 	return size;
1109 }
1110 EXPORT_SYMBOL_GPL(__trace_array_puts);
1111 
1112 /**
1113  * __trace_puts - write a constant string into the trace buffer.
1114  * @ip:	   The address of the caller
1115  * @str:   The constant string to write
1116  * @size:  The size of the string.
1117  */
1118 int __trace_puts(unsigned long ip, const char *str, int size)
1119 {
1120 	return __trace_array_puts(&global_trace, ip, str, size);
1121 }
1122 EXPORT_SYMBOL_GPL(__trace_puts);
1123 
1124 /**
1125  * __trace_bputs - write the pointer to a constant string into trace buffer
1126  * @ip:	   The address of the caller
1127  * @str:   The constant string to write to the buffer to
1128  */
1129 int __trace_bputs(unsigned long ip, const char *str)
1130 {
1131 	struct ring_buffer_event *event;
1132 	struct trace_buffer *buffer;
1133 	struct bputs_entry *entry;
1134 	unsigned int trace_ctx;
1135 	int size = sizeof(struct bputs_entry);
1136 	int ret = 0;
1137 
1138 	if (!(global_trace.trace_flags & TRACE_ITER_PRINTK))
1139 		return 0;
1140 
1141 	if (unlikely(tracing_selftest_running || tracing_disabled))
1142 		return 0;
1143 
1144 	trace_ctx = tracing_gen_ctx();
1145 	buffer = global_trace.array_buffer.buffer;
1146 
1147 	ring_buffer_nest_start(buffer);
1148 	event = __trace_buffer_lock_reserve(buffer, TRACE_BPUTS, size,
1149 					    trace_ctx);
1150 	if (!event)
1151 		goto out;
1152 
1153 	entry = ring_buffer_event_data(event);
1154 	entry->ip			= ip;
1155 	entry->str			= str;
1156 
1157 	__buffer_unlock_commit(buffer, event);
1158 	ftrace_trace_stack(&global_trace, buffer, trace_ctx, 4, NULL);
1159 
1160 	ret = 1;
1161  out:
1162 	ring_buffer_nest_end(buffer);
1163 	return ret;
1164 }
1165 EXPORT_SYMBOL_GPL(__trace_bputs);
1166 
1167 #ifdef CONFIG_TRACER_SNAPSHOT
1168 static void tracing_snapshot_instance_cond(struct trace_array *tr,
1169 					   void *cond_data)
1170 {
1171 	struct tracer *tracer = tr->current_trace;
1172 	unsigned long flags;
1173 
1174 	if (in_nmi()) {
1175 		trace_array_puts(tr, "*** SNAPSHOT CALLED FROM NMI CONTEXT ***\n");
1176 		trace_array_puts(tr, "*** snapshot is being ignored        ***\n");
1177 		return;
1178 	}
1179 
1180 	if (!tr->allocated_snapshot) {
1181 		trace_array_puts(tr, "*** SNAPSHOT NOT ALLOCATED ***\n");
1182 		trace_array_puts(tr, "*** stopping trace here!   ***\n");
1183 		tracer_tracing_off(tr);
1184 		return;
1185 	}
1186 
1187 	/* Note, snapshot can not be used when the tracer uses it */
1188 	if (tracer->use_max_tr) {
1189 		trace_array_puts(tr, "*** LATENCY TRACER ACTIVE ***\n");
1190 		trace_array_puts(tr, "*** Can not use snapshot (sorry) ***\n");
1191 		return;
1192 	}
1193 
1194 	if (tr->mapped) {
1195 		trace_array_puts(tr, "*** BUFFER MEMORY MAPPED ***\n");
1196 		trace_array_puts(tr, "*** Can not use snapshot (sorry) ***\n");
1197 		return;
1198 	}
1199 
1200 	local_irq_save(flags);
1201 	update_max_tr(tr, current, smp_processor_id(), cond_data);
1202 	local_irq_restore(flags);
1203 }
1204 
1205 void tracing_snapshot_instance(struct trace_array *tr)
1206 {
1207 	tracing_snapshot_instance_cond(tr, NULL);
1208 }
1209 
1210 /**
1211  * tracing_snapshot - take a snapshot of the current buffer.
1212  *
1213  * This causes a swap between the snapshot buffer and the current live
1214  * tracing buffer. You can use this to take snapshots of the live
1215  * trace when some condition is triggered, but continue to trace.
1216  *
1217  * Note, make sure to allocate the snapshot with either
1218  * a tracing_snapshot_alloc(), or by doing it manually
1219  * with: echo 1 > /sys/kernel/tracing/snapshot
1220  *
1221  * If the snapshot buffer is not allocated, it will stop tracing.
1222  * Basically making a permanent snapshot.
1223  */
1224 void tracing_snapshot(void)
1225 {
1226 	struct trace_array *tr = &global_trace;
1227 
1228 	tracing_snapshot_instance(tr);
1229 }
1230 EXPORT_SYMBOL_GPL(tracing_snapshot);
1231 
1232 /**
1233  * tracing_snapshot_cond - conditionally take a snapshot of the current buffer.
1234  * @tr:		The tracing instance to snapshot
1235  * @cond_data:	The data to be tested conditionally, and possibly saved
1236  *
1237  * This is the same as tracing_snapshot() except that the snapshot is
1238  * conditional - the snapshot will only happen if the
1239  * cond_snapshot.update() implementation receiving the cond_data
1240  * returns true, which means that the trace array's cond_snapshot
1241  * update() operation used the cond_data to determine whether the
1242  * snapshot should be taken, and if it was, presumably saved it along
1243  * with the snapshot.
1244  */
1245 void tracing_snapshot_cond(struct trace_array *tr, void *cond_data)
1246 {
1247 	tracing_snapshot_instance_cond(tr, cond_data);
1248 }
1249 EXPORT_SYMBOL_GPL(tracing_snapshot_cond);
1250 
1251 /**
1252  * tracing_cond_snapshot_data - get the user data associated with a snapshot
1253  * @tr:		The tracing instance
1254  *
1255  * When the user enables a conditional snapshot using
1256  * tracing_snapshot_cond_enable(), the user-defined cond_data is saved
1257  * with the snapshot.  This accessor is used to retrieve it.
1258  *
1259  * Should not be called from cond_snapshot.update(), since it takes
1260  * the tr->max_lock lock, which the code calling
1261  * cond_snapshot.update() has already done.
1262  *
1263  * Returns the cond_data associated with the trace array's snapshot.
1264  */
1265 void *tracing_cond_snapshot_data(struct trace_array *tr)
1266 {
1267 	void *cond_data = NULL;
1268 
1269 	local_irq_disable();
1270 	arch_spin_lock(&tr->max_lock);
1271 
1272 	if (tr->cond_snapshot)
1273 		cond_data = tr->cond_snapshot->cond_data;
1274 
1275 	arch_spin_unlock(&tr->max_lock);
1276 	local_irq_enable();
1277 
1278 	return cond_data;
1279 }
1280 EXPORT_SYMBOL_GPL(tracing_cond_snapshot_data);
1281 
1282 static int resize_buffer_duplicate_size(struct array_buffer *trace_buf,
1283 					struct array_buffer *size_buf, int cpu_id);
1284 static void set_buffer_entries(struct array_buffer *buf, unsigned long val);
1285 
1286 int tracing_alloc_snapshot_instance(struct trace_array *tr)
1287 {
1288 	int order;
1289 	int ret;
1290 
1291 	if (!tr->allocated_snapshot) {
1292 
1293 		/* Make the snapshot buffer have the same order as main buffer */
1294 		order = ring_buffer_subbuf_order_get(tr->array_buffer.buffer);
1295 		ret = ring_buffer_subbuf_order_set(tr->max_buffer.buffer, order);
1296 		if (ret < 0)
1297 			return ret;
1298 
1299 		/* allocate spare buffer */
1300 		ret = resize_buffer_duplicate_size(&tr->max_buffer,
1301 				   &tr->array_buffer, RING_BUFFER_ALL_CPUS);
1302 		if (ret < 0)
1303 			return ret;
1304 
1305 		tr->allocated_snapshot = true;
1306 	}
1307 
1308 	return 0;
1309 }
1310 
1311 static void free_snapshot(struct trace_array *tr)
1312 {
1313 	/*
1314 	 * We don't free the ring buffer. instead, resize it because
1315 	 * The max_tr ring buffer has some state (e.g. ring->clock) and
1316 	 * we want preserve it.
1317 	 */
1318 	ring_buffer_subbuf_order_set(tr->max_buffer.buffer, 0);
1319 	ring_buffer_resize(tr->max_buffer.buffer, 1, RING_BUFFER_ALL_CPUS);
1320 	set_buffer_entries(&tr->max_buffer, 1);
1321 	tracing_reset_online_cpus(&tr->max_buffer);
1322 	tr->allocated_snapshot = false;
1323 }
1324 
1325 static int tracing_arm_snapshot_locked(struct trace_array *tr)
1326 {
1327 	int ret;
1328 
1329 	lockdep_assert_held(&trace_types_lock);
1330 
1331 	spin_lock(&tr->snapshot_trigger_lock);
1332 	if (tr->snapshot == UINT_MAX || tr->mapped) {
1333 		spin_unlock(&tr->snapshot_trigger_lock);
1334 		return -EBUSY;
1335 	}
1336 
1337 	tr->snapshot++;
1338 	spin_unlock(&tr->snapshot_trigger_lock);
1339 
1340 	ret = tracing_alloc_snapshot_instance(tr);
1341 	if (ret) {
1342 		spin_lock(&tr->snapshot_trigger_lock);
1343 		tr->snapshot--;
1344 		spin_unlock(&tr->snapshot_trigger_lock);
1345 	}
1346 
1347 	return ret;
1348 }
1349 
1350 int tracing_arm_snapshot(struct trace_array *tr)
1351 {
1352 	int ret;
1353 
1354 	mutex_lock(&trace_types_lock);
1355 	ret = tracing_arm_snapshot_locked(tr);
1356 	mutex_unlock(&trace_types_lock);
1357 
1358 	return ret;
1359 }
1360 
1361 void tracing_disarm_snapshot(struct trace_array *tr)
1362 {
1363 	spin_lock(&tr->snapshot_trigger_lock);
1364 	if (!WARN_ON(!tr->snapshot))
1365 		tr->snapshot--;
1366 	spin_unlock(&tr->snapshot_trigger_lock);
1367 }
1368 
1369 /**
1370  * tracing_alloc_snapshot - allocate snapshot buffer.
1371  *
1372  * This only allocates the snapshot buffer if it isn't already
1373  * allocated - it doesn't also take a snapshot.
1374  *
1375  * This is meant to be used in cases where the snapshot buffer needs
1376  * to be set up for events that can't sleep but need to be able to
1377  * trigger a snapshot.
1378  */
1379 int tracing_alloc_snapshot(void)
1380 {
1381 	struct trace_array *tr = &global_trace;
1382 	int ret;
1383 
1384 	ret = tracing_alloc_snapshot_instance(tr);
1385 	WARN_ON(ret < 0);
1386 
1387 	return ret;
1388 }
1389 EXPORT_SYMBOL_GPL(tracing_alloc_snapshot);
1390 
1391 /**
1392  * tracing_snapshot_alloc - allocate and take a snapshot of the current buffer.
1393  *
1394  * This is similar to tracing_snapshot(), but it will allocate the
1395  * snapshot buffer if it isn't already allocated. Use this only
1396  * where it is safe to sleep, as the allocation may sleep.
1397  *
1398  * This causes a swap between the snapshot buffer and the current live
1399  * tracing buffer. You can use this to take snapshots of the live
1400  * trace when some condition is triggered, but continue to trace.
1401  */
1402 void tracing_snapshot_alloc(void)
1403 {
1404 	int ret;
1405 
1406 	ret = tracing_alloc_snapshot();
1407 	if (ret < 0)
1408 		return;
1409 
1410 	tracing_snapshot();
1411 }
1412 EXPORT_SYMBOL_GPL(tracing_snapshot_alloc);
1413 
1414 /**
1415  * tracing_snapshot_cond_enable - enable conditional snapshot for an instance
1416  * @tr:		The tracing instance
1417  * @cond_data:	User data to associate with the snapshot
1418  * @update:	Implementation of the cond_snapshot update function
1419  *
1420  * Check whether the conditional snapshot for the given instance has
1421  * already been enabled, or if the current tracer is already using a
1422  * snapshot; if so, return -EBUSY, else create a cond_snapshot and
1423  * save the cond_data and update function inside.
1424  *
1425  * Returns 0 if successful, error otherwise.
1426  */
1427 int tracing_snapshot_cond_enable(struct trace_array *tr, void *cond_data,
1428 				 cond_update_fn_t update)
1429 {
1430 	struct cond_snapshot *cond_snapshot;
1431 	int ret = 0;
1432 
1433 	cond_snapshot = kzalloc(sizeof(*cond_snapshot), GFP_KERNEL);
1434 	if (!cond_snapshot)
1435 		return -ENOMEM;
1436 
1437 	cond_snapshot->cond_data = cond_data;
1438 	cond_snapshot->update = update;
1439 
1440 	mutex_lock(&trace_types_lock);
1441 
1442 	if (tr->current_trace->use_max_tr) {
1443 		ret = -EBUSY;
1444 		goto fail_unlock;
1445 	}
1446 
1447 	/*
1448 	 * The cond_snapshot can only change to NULL without the
1449 	 * trace_types_lock. We don't care if we race with it going
1450 	 * to NULL, but we want to make sure that it's not set to
1451 	 * something other than NULL when we get here, which we can
1452 	 * do safely with only holding the trace_types_lock and not
1453 	 * having to take the max_lock.
1454 	 */
1455 	if (tr->cond_snapshot) {
1456 		ret = -EBUSY;
1457 		goto fail_unlock;
1458 	}
1459 
1460 	ret = tracing_arm_snapshot_locked(tr);
1461 	if (ret)
1462 		goto fail_unlock;
1463 
1464 	local_irq_disable();
1465 	arch_spin_lock(&tr->max_lock);
1466 	tr->cond_snapshot = cond_snapshot;
1467 	arch_spin_unlock(&tr->max_lock);
1468 	local_irq_enable();
1469 
1470 	mutex_unlock(&trace_types_lock);
1471 
1472 	return ret;
1473 
1474  fail_unlock:
1475 	mutex_unlock(&trace_types_lock);
1476 	kfree(cond_snapshot);
1477 	return ret;
1478 }
1479 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_enable);
1480 
1481 /**
1482  * tracing_snapshot_cond_disable - disable conditional snapshot for an instance
1483  * @tr:		The tracing instance
1484  *
1485  * Check whether the conditional snapshot for the given instance is
1486  * enabled; if so, free the cond_snapshot associated with it,
1487  * otherwise return -EINVAL.
1488  *
1489  * Returns 0 if successful, error otherwise.
1490  */
1491 int tracing_snapshot_cond_disable(struct trace_array *tr)
1492 {
1493 	int ret = 0;
1494 
1495 	local_irq_disable();
1496 	arch_spin_lock(&tr->max_lock);
1497 
1498 	if (!tr->cond_snapshot)
1499 		ret = -EINVAL;
1500 	else {
1501 		kfree(tr->cond_snapshot);
1502 		tr->cond_snapshot = NULL;
1503 	}
1504 
1505 	arch_spin_unlock(&tr->max_lock);
1506 	local_irq_enable();
1507 
1508 	tracing_disarm_snapshot(tr);
1509 
1510 	return ret;
1511 }
1512 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_disable);
1513 #else
1514 void tracing_snapshot(void)
1515 {
1516 	WARN_ONCE(1, "Snapshot feature not enabled, but internal snapshot used");
1517 }
1518 EXPORT_SYMBOL_GPL(tracing_snapshot);
1519 void tracing_snapshot_cond(struct trace_array *tr, void *cond_data)
1520 {
1521 	WARN_ONCE(1, "Snapshot feature not enabled, but internal conditional snapshot used");
1522 }
1523 EXPORT_SYMBOL_GPL(tracing_snapshot_cond);
1524 int tracing_alloc_snapshot(void)
1525 {
1526 	WARN_ONCE(1, "Snapshot feature not enabled, but snapshot allocation used");
1527 	return -ENODEV;
1528 }
1529 EXPORT_SYMBOL_GPL(tracing_alloc_snapshot);
1530 void tracing_snapshot_alloc(void)
1531 {
1532 	/* Give warning */
1533 	tracing_snapshot();
1534 }
1535 EXPORT_SYMBOL_GPL(tracing_snapshot_alloc);
1536 void *tracing_cond_snapshot_data(struct trace_array *tr)
1537 {
1538 	return NULL;
1539 }
1540 EXPORT_SYMBOL_GPL(tracing_cond_snapshot_data);
1541 int tracing_snapshot_cond_enable(struct trace_array *tr, void *cond_data, cond_update_fn_t update)
1542 {
1543 	return -ENODEV;
1544 }
1545 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_enable);
1546 int tracing_snapshot_cond_disable(struct trace_array *tr)
1547 {
1548 	return false;
1549 }
1550 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_disable);
1551 #define free_snapshot(tr)	do { } while (0)
1552 #define tracing_arm_snapshot_locked(tr) ({ -EBUSY; })
1553 #endif /* CONFIG_TRACER_SNAPSHOT */
1554 
1555 void tracer_tracing_off(struct trace_array *tr)
1556 {
1557 	if (tr->array_buffer.buffer)
1558 		ring_buffer_record_off(tr->array_buffer.buffer);
1559 	/*
1560 	 * This flag is looked at when buffers haven't been allocated
1561 	 * yet, or by some tracers (like irqsoff), that just want to
1562 	 * know if the ring buffer has been disabled, but it can handle
1563 	 * races of where it gets disabled but we still do a record.
1564 	 * As the check is in the fast path of the tracers, it is more
1565 	 * important to be fast than accurate.
1566 	 */
1567 	tr->buffer_disabled = 1;
1568 	/* Make the flag seen by readers */
1569 	smp_wmb();
1570 }
1571 
1572 /**
1573  * tracing_off - turn off tracing buffers
1574  *
1575  * This function stops the tracing buffers from recording data.
1576  * It does not disable any overhead the tracers themselves may
1577  * be causing. This function simply causes all recording to
1578  * the ring buffers to fail.
1579  */
1580 void tracing_off(void)
1581 {
1582 	tracer_tracing_off(&global_trace);
1583 }
1584 EXPORT_SYMBOL_GPL(tracing_off);
1585 
1586 void disable_trace_on_warning(void)
1587 {
1588 	if (__disable_trace_on_warning) {
1589 		trace_array_printk_buf(global_trace.array_buffer.buffer, _THIS_IP_,
1590 			"Disabling tracing due to warning\n");
1591 		tracing_off();
1592 	}
1593 }
1594 
1595 /**
1596  * tracer_tracing_is_on - show real state of ring buffer enabled
1597  * @tr : the trace array to know if ring buffer is enabled
1598  *
1599  * Shows real state of the ring buffer if it is enabled or not.
1600  */
1601 bool tracer_tracing_is_on(struct trace_array *tr)
1602 {
1603 	if (tr->array_buffer.buffer)
1604 		return ring_buffer_record_is_set_on(tr->array_buffer.buffer);
1605 	return !tr->buffer_disabled;
1606 }
1607 
1608 /**
1609  * tracing_is_on - show state of ring buffers enabled
1610  */
1611 int tracing_is_on(void)
1612 {
1613 	return tracer_tracing_is_on(&global_trace);
1614 }
1615 EXPORT_SYMBOL_GPL(tracing_is_on);
1616 
1617 static int __init set_buf_size(char *str)
1618 {
1619 	unsigned long buf_size;
1620 
1621 	if (!str)
1622 		return 0;
1623 	buf_size = memparse(str, &str);
1624 	/*
1625 	 * nr_entries can not be zero and the startup
1626 	 * tests require some buffer space. Therefore
1627 	 * ensure we have at least 4096 bytes of buffer.
1628 	 */
1629 	trace_buf_size = max(4096UL, buf_size);
1630 	return 1;
1631 }
1632 __setup("trace_buf_size=", set_buf_size);
1633 
1634 static int __init set_tracing_thresh(char *str)
1635 {
1636 	unsigned long threshold;
1637 	int ret;
1638 
1639 	if (!str)
1640 		return 0;
1641 	ret = kstrtoul(str, 0, &threshold);
1642 	if (ret < 0)
1643 		return 0;
1644 	tracing_thresh = threshold * 1000;
1645 	return 1;
1646 }
1647 __setup("tracing_thresh=", set_tracing_thresh);
1648 
1649 unsigned long nsecs_to_usecs(unsigned long nsecs)
1650 {
1651 	return nsecs / 1000;
1652 }
1653 
1654 /*
1655  * TRACE_FLAGS is defined as a tuple matching bit masks with strings.
1656  * It uses C(a, b) where 'a' is the eval (enum) name and 'b' is the string that
1657  * matches it. By defining "C(a, b) b", TRACE_FLAGS becomes a list
1658  * of strings in the order that the evals (enum) were defined.
1659  */
1660 #undef C
1661 #define C(a, b) b
1662 
1663 /* These must match the bit positions in trace_iterator_flags */
1664 static const char *trace_options[] = {
1665 	TRACE_FLAGS
1666 	NULL
1667 };
1668 
1669 static struct {
1670 	u64 (*func)(void);
1671 	const char *name;
1672 	int in_ns;		/* is this clock in nanoseconds? */
1673 } trace_clocks[] = {
1674 	{ trace_clock_local,		"local",	1 },
1675 	{ trace_clock_global,		"global",	1 },
1676 	{ trace_clock_counter,		"counter",	0 },
1677 	{ trace_clock_jiffies,		"uptime",	0 },
1678 	{ trace_clock,			"perf",		1 },
1679 	{ ktime_get_mono_fast_ns,	"mono",		1 },
1680 	{ ktime_get_raw_fast_ns,	"mono_raw",	1 },
1681 	{ ktime_get_boot_fast_ns,	"boot",		1 },
1682 	{ ktime_get_tai_fast_ns,	"tai",		1 },
1683 	ARCH_TRACE_CLOCKS
1684 };
1685 
1686 bool trace_clock_in_ns(struct trace_array *tr)
1687 {
1688 	if (trace_clocks[tr->clock_id].in_ns)
1689 		return true;
1690 
1691 	return false;
1692 }
1693 
1694 /*
1695  * trace_parser_get_init - gets the buffer for trace parser
1696  */
1697 int trace_parser_get_init(struct trace_parser *parser, int size)
1698 {
1699 	memset(parser, 0, sizeof(*parser));
1700 
1701 	parser->buffer = kmalloc(size, GFP_KERNEL);
1702 	if (!parser->buffer)
1703 		return 1;
1704 
1705 	parser->size = size;
1706 	return 0;
1707 }
1708 
1709 /*
1710  * trace_parser_put - frees the buffer for trace parser
1711  */
1712 void trace_parser_put(struct trace_parser *parser)
1713 {
1714 	kfree(parser->buffer);
1715 	parser->buffer = NULL;
1716 }
1717 
1718 /*
1719  * trace_get_user - reads the user input string separated by  space
1720  * (matched by isspace(ch))
1721  *
1722  * For each string found the 'struct trace_parser' is updated,
1723  * and the function returns.
1724  *
1725  * Returns number of bytes read.
1726  *
1727  * See kernel/trace/trace.h for 'struct trace_parser' details.
1728  */
1729 int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
1730 	size_t cnt, loff_t *ppos)
1731 {
1732 	char ch;
1733 	size_t read = 0;
1734 	ssize_t ret;
1735 
1736 	if (!*ppos)
1737 		trace_parser_clear(parser);
1738 
1739 	ret = get_user(ch, ubuf++);
1740 	if (ret)
1741 		goto out;
1742 
1743 	read++;
1744 	cnt--;
1745 
1746 	/*
1747 	 * The parser is not finished with the last write,
1748 	 * continue reading the user input without skipping spaces.
1749 	 */
1750 	if (!parser->cont) {
1751 		/* skip white space */
1752 		while (cnt && isspace(ch)) {
1753 			ret = get_user(ch, ubuf++);
1754 			if (ret)
1755 				goto out;
1756 			read++;
1757 			cnt--;
1758 		}
1759 
1760 		parser->idx = 0;
1761 
1762 		/* only spaces were written */
1763 		if (isspace(ch) || !ch) {
1764 			*ppos += read;
1765 			ret = read;
1766 			goto out;
1767 		}
1768 	}
1769 
1770 	/* read the non-space input */
1771 	while (cnt && !isspace(ch) && ch) {
1772 		if (parser->idx < parser->size - 1)
1773 			parser->buffer[parser->idx++] = ch;
1774 		else {
1775 			ret = -EINVAL;
1776 			goto out;
1777 		}
1778 		ret = get_user(ch, ubuf++);
1779 		if (ret)
1780 			goto out;
1781 		read++;
1782 		cnt--;
1783 	}
1784 
1785 	/* We either got finished input or we have to wait for another call. */
1786 	if (isspace(ch) || !ch) {
1787 		parser->buffer[parser->idx] = 0;
1788 		parser->cont = false;
1789 	} else if (parser->idx < parser->size - 1) {
1790 		parser->cont = true;
1791 		parser->buffer[parser->idx++] = ch;
1792 		/* Make sure the parsed string always terminates with '\0'. */
1793 		parser->buffer[parser->idx] = 0;
1794 	} else {
1795 		ret = -EINVAL;
1796 		goto out;
1797 	}
1798 
1799 	*ppos += read;
1800 	ret = read;
1801 
1802 out:
1803 	return ret;
1804 }
1805 
1806 /* TODO add a seq_buf_to_buffer() */
1807 static ssize_t trace_seq_to_buffer(struct trace_seq *s, void *buf, size_t cnt)
1808 {
1809 	int len;
1810 
1811 	if (trace_seq_used(s) <= s->readpos)
1812 		return -EBUSY;
1813 
1814 	len = trace_seq_used(s) - s->readpos;
1815 	if (cnt > len)
1816 		cnt = len;
1817 	memcpy(buf, s->buffer + s->readpos, cnt);
1818 
1819 	s->readpos += cnt;
1820 	return cnt;
1821 }
1822 
1823 unsigned long __read_mostly	tracing_thresh;
1824 
1825 #ifdef CONFIG_TRACER_MAX_TRACE
1826 static const struct file_operations tracing_max_lat_fops;
1827 
1828 #ifdef LATENCY_FS_NOTIFY
1829 
1830 static struct workqueue_struct *fsnotify_wq;
1831 
1832 static void latency_fsnotify_workfn(struct work_struct *work)
1833 {
1834 	struct trace_array *tr = container_of(work, struct trace_array,
1835 					      fsnotify_work);
1836 	fsnotify_inode(tr->d_max_latency->d_inode, FS_MODIFY);
1837 }
1838 
1839 static void latency_fsnotify_workfn_irq(struct irq_work *iwork)
1840 {
1841 	struct trace_array *tr = container_of(iwork, struct trace_array,
1842 					      fsnotify_irqwork);
1843 	queue_work(fsnotify_wq, &tr->fsnotify_work);
1844 }
1845 
1846 static void trace_create_maxlat_file(struct trace_array *tr,
1847 				     struct dentry *d_tracer)
1848 {
1849 	INIT_WORK(&tr->fsnotify_work, latency_fsnotify_workfn);
1850 	init_irq_work(&tr->fsnotify_irqwork, latency_fsnotify_workfn_irq);
1851 	tr->d_max_latency = trace_create_file("tracing_max_latency",
1852 					      TRACE_MODE_WRITE,
1853 					      d_tracer, tr,
1854 					      &tracing_max_lat_fops);
1855 }
1856 
1857 __init static int latency_fsnotify_init(void)
1858 {
1859 	fsnotify_wq = alloc_workqueue("tr_max_lat_wq",
1860 				      WQ_UNBOUND | WQ_HIGHPRI, 0);
1861 	if (!fsnotify_wq) {
1862 		pr_err("Unable to allocate tr_max_lat_wq\n");
1863 		return -ENOMEM;
1864 	}
1865 	return 0;
1866 }
1867 
1868 late_initcall_sync(latency_fsnotify_init);
1869 
1870 void latency_fsnotify(struct trace_array *tr)
1871 {
1872 	if (!fsnotify_wq)
1873 		return;
1874 	/*
1875 	 * We cannot call queue_work(&tr->fsnotify_work) from here because it's
1876 	 * possible that we are called from __schedule() or do_idle(), which
1877 	 * could cause a deadlock.
1878 	 */
1879 	irq_work_queue(&tr->fsnotify_irqwork);
1880 }
1881 
1882 #else /* !LATENCY_FS_NOTIFY */
1883 
1884 #define trace_create_maxlat_file(tr, d_tracer)				\
1885 	trace_create_file("tracing_max_latency", TRACE_MODE_WRITE,	\
1886 			  d_tracer, tr, &tracing_max_lat_fops)
1887 
1888 #endif
1889 
1890 /*
1891  * Copy the new maximum trace into the separate maximum-trace
1892  * structure. (this way the maximum trace is permanently saved,
1893  * for later retrieval via /sys/kernel/tracing/tracing_max_latency)
1894  */
1895 static void
1896 __update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
1897 {
1898 	struct array_buffer *trace_buf = &tr->array_buffer;
1899 	struct array_buffer *max_buf = &tr->max_buffer;
1900 	struct trace_array_cpu *data = per_cpu_ptr(trace_buf->data, cpu);
1901 	struct trace_array_cpu *max_data = per_cpu_ptr(max_buf->data, cpu);
1902 
1903 	max_buf->cpu = cpu;
1904 	max_buf->time_start = data->preempt_timestamp;
1905 
1906 	max_data->saved_latency = tr->max_latency;
1907 	max_data->critical_start = data->critical_start;
1908 	max_data->critical_end = data->critical_end;
1909 
1910 	strncpy(max_data->comm, tsk->comm, TASK_COMM_LEN);
1911 	max_data->pid = tsk->pid;
1912 	/*
1913 	 * If tsk == current, then use current_uid(), as that does not use
1914 	 * RCU. The irq tracer can be called out of RCU scope.
1915 	 */
1916 	if (tsk == current)
1917 		max_data->uid = current_uid();
1918 	else
1919 		max_data->uid = task_uid(tsk);
1920 
1921 	max_data->nice = tsk->static_prio - 20 - MAX_RT_PRIO;
1922 	max_data->policy = tsk->policy;
1923 	max_data->rt_priority = tsk->rt_priority;
1924 
1925 	/* record this tasks comm */
1926 	tracing_record_cmdline(tsk);
1927 	latency_fsnotify(tr);
1928 }
1929 
1930 /**
1931  * update_max_tr - snapshot all trace buffers from global_trace to max_tr
1932  * @tr: tracer
1933  * @tsk: the task with the latency
1934  * @cpu: The cpu that initiated the trace.
1935  * @cond_data: User data associated with a conditional snapshot
1936  *
1937  * Flip the buffers between the @tr and the max_tr and record information
1938  * about which task was the cause of this latency.
1939  */
1940 void
1941 update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu,
1942 	      void *cond_data)
1943 {
1944 	if (tr->stop_count)
1945 		return;
1946 
1947 	WARN_ON_ONCE(!irqs_disabled());
1948 
1949 	if (!tr->allocated_snapshot) {
1950 		/* Only the nop tracer should hit this when disabling */
1951 		WARN_ON_ONCE(tr->current_trace != &nop_trace);
1952 		return;
1953 	}
1954 
1955 	arch_spin_lock(&tr->max_lock);
1956 
1957 	/* Inherit the recordable setting from array_buffer */
1958 	if (ring_buffer_record_is_set_on(tr->array_buffer.buffer))
1959 		ring_buffer_record_on(tr->max_buffer.buffer);
1960 	else
1961 		ring_buffer_record_off(tr->max_buffer.buffer);
1962 
1963 #ifdef CONFIG_TRACER_SNAPSHOT
1964 	if (tr->cond_snapshot && !tr->cond_snapshot->update(tr, cond_data)) {
1965 		arch_spin_unlock(&tr->max_lock);
1966 		return;
1967 	}
1968 #endif
1969 	swap(tr->array_buffer.buffer, tr->max_buffer.buffer);
1970 
1971 	__update_max_tr(tr, tsk, cpu);
1972 
1973 	arch_spin_unlock(&tr->max_lock);
1974 
1975 	/* Any waiters on the old snapshot buffer need to wake up */
1976 	ring_buffer_wake_waiters(tr->array_buffer.buffer, RING_BUFFER_ALL_CPUS);
1977 }
1978 
1979 /**
1980  * update_max_tr_single - only copy one trace over, and reset the rest
1981  * @tr: tracer
1982  * @tsk: task with the latency
1983  * @cpu: the cpu of the buffer to copy.
1984  *
1985  * Flip the trace of a single CPU buffer between the @tr and the max_tr.
1986  */
1987 void
1988 update_max_tr_single(struct trace_array *tr, struct task_struct *tsk, int cpu)
1989 {
1990 	int ret;
1991 
1992 	if (tr->stop_count)
1993 		return;
1994 
1995 	WARN_ON_ONCE(!irqs_disabled());
1996 	if (!tr->allocated_snapshot) {
1997 		/* Only the nop tracer should hit this when disabling */
1998 		WARN_ON_ONCE(tr->current_trace != &nop_trace);
1999 		return;
2000 	}
2001 
2002 	arch_spin_lock(&tr->max_lock);
2003 
2004 	ret = ring_buffer_swap_cpu(tr->max_buffer.buffer, tr->array_buffer.buffer, cpu);
2005 
2006 	if (ret == -EBUSY) {
2007 		/*
2008 		 * We failed to swap the buffer due to a commit taking
2009 		 * place on this CPU. We fail to record, but we reset
2010 		 * the max trace buffer (no one writes directly to it)
2011 		 * and flag that it failed.
2012 		 * Another reason is resize is in progress.
2013 		 */
2014 		trace_array_printk_buf(tr->max_buffer.buffer, _THIS_IP_,
2015 			"Failed to swap buffers due to commit or resize in progress\n");
2016 	}
2017 
2018 	WARN_ON_ONCE(ret && ret != -EAGAIN && ret != -EBUSY);
2019 
2020 	__update_max_tr(tr, tsk, cpu);
2021 	arch_spin_unlock(&tr->max_lock);
2022 }
2023 
2024 #endif /* CONFIG_TRACER_MAX_TRACE */
2025 
2026 struct pipe_wait {
2027 	struct trace_iterator		*iter;
2028 	int				wait_index;
2029 };
2030 
2031 static bool wait_pipe_cond(void *data)
2032 {
2033 	struct pipe_wait *pwait = data;
2034 	struct trace_iterator *iter = pwait->iter;
2035 
2036 	if (atomic_read_acquire(&iter->wait_index) != pwait->wait_index)
2037 		return true;
2038 
2039 	return iter->closed;
2040 }
2041 
2042 static int wait_on_pipe(struct trace_iterator *iter, int full)
2043 {
2044 	struct pipe_wait pwait;
2045 	int ret;
2046 
2047 	/* Iterators are static, they should be filled or empty */
2048 	if (trace_buffer_iter(iter, iter->cpu_file))
2049 		return 0;
2050 
2051 	pwait.wait_index = atomic_read_acquire(&iter->wait_index);
2052 	pwait.iter = iter;
2053 
2054 	ret = ring_buffer_wait(iter->array_buffer->buffer, iter->cpu_file, full,
2055 			       wait_pipe_cond, &pwait);
2056 
2057 #ifdef CONFIG_TRACER_MAX_TRACE
2058 	/*
2059 	 * Make sure this is still the snapshot buffer, as if a snapshot were
2060 	 * to happen, this would now be the main buffer.
2061 	 */
2062 	if (iter->snapshot)
2063 		iter->array_buffer = &iter->tr->max_buffer;
2064 #endif
2065 	return ret;
2066 }
2067 
2068 #ifdef CONFIG_FTRACE_STARTUP_TEST
2069 static bool selftests_can_run;
2070 
2071 struct trace_selftests {
2072 	struct list_head		list;
2073 	struct tracer			*type;
2074 };
2075 
2076 static LIST_HEAD(postponed_selftests);
2077 
2078 static int save_selftest(struct tracer *type)
2079 {
2080 	struct trace_selftests *selftest;
2081 
2082 	selftest = kmalloc(sizeof(*selftest), GFP_KERNEL);
2083 	if (!selftest)
2084 		return -ENOMEM;
2085 
2086 	selftest->type = type;
2087 	list_add(&selftest->list, &postponed_selftests);
2088 	return 0;
2089 }
2090 
2091 static int run_tracer_selftest(struct tracer *type)
2092 {
2093 	struct trace_array *tr = &global_trace;
2094 	struct tracer *saved_tracer = tr->current_trace;
2095 	int ret;
2096 
2097 	if (!type->selftest || tracing_selftest_disabled)
2098 		return 0;
2099 
2100 	/*
2101 	 * If a tracer registers early in boot up (before scheduling is
2102 	 * initialized and such), then do not run its selftests yet.
2103 	 * Instead, run it a little later in the boot process.
2104 	 */
2105 	if (!selftests_can_run)
2106 		return save_selftest(type);
2107 
2108 	if (!tracing_is_on()) {
2109 		pr_warn("Selftest for tracer %s skipped due to tracing disabled\n",
2110 			type->name);
2111 		return 0;
2112 	}
2113 
2114 	/*
2115 	 * Run a selftest on this tracer.
2116 	 * Here we reset the trace buffer, and set the current
2117 	 * tracer to be this tracer. The tracer can then run some
2118 	 * internal tracing to verify that everything is in order.
2119 	 * If we fail, we do not register this tracer.
2120 	 */
2121 	tracing_reset_online_cpus(&tr->array_buffer);
2122 
2123 	tr->current_trace = type;
2124 
2125 #ifdef CONFIG_TRACER_MAX_TRACE
2126 	if (type->use_max_tr) {
2127 		/* If we expanded the buffers, make sure the max is expanded too */
2128 		if (tr->ring_buffer_expanded)
2129 			ring_buffer_resize(tr->max_buffer.buffer, trace_buf_size,
2130 					   RING_BUFFER_ALL_CPUS);
2131 		tr->allocated_snapshot = true;
2132 	}
2133 #endif
2134 
2135 	/* the test is responsible for initializing and enabling */
2136 	pr_info("Testing tracer %s: ", type->name);
2137 	ret = type->selftest(type, tr);
2138 	/* the test is responsible for resetting too */
2139 	tr->current_trace = saved_tracer;
2140 	if (ret) {
2141 		printk(KERN_CONT "FAILED!\n");
2142 		/* Add the warning after printing 'FAILED' */
2143 		WARN_ON(1);
2144 		return -1;
2145 	}
2146 	/* Only reset on passing, to avoid touching corrupted buffers */
2147 	tracing_reset_online_cpus(&tr->array_buffer);
2148 
2149 #ifdef CONFIG_TRACER_MAX_TRACE
2150 	if (type->use_max_tr) {
2151 		tr->allocated_snapshot = false;
2152 
2153 		/* Shrink the max buffer again */
2154 		if (tr->ring_buffer_expanded)
2155 			ring_buffer_resize(tr->max_buffer.buffer, 1,
2156 					   RING_BUFFER_ALL_CPUS);
2157 	}
2158 #endif
2159 
2160 	printk(KERN_CONT "PASSED\n");
2161 	return 0;
2162 }
2163 
2164 static int do_run_tracer_selftest(struct tracer *type)
2165 {
2166 	int ret;
2167 
2168 	/*
2169 	 * Tests can take a long time, especially if they are run one after the
2170 	 * other, as does happen during bootup when all the tracers are
2171 	 * registered. This could cause the soft lockup watchdog to trigger.
2172 	 */
2173 	cond_resched();
2174 
2175 	tracing_selftest_running = true;
2176 	ret = run_tracer_selftest(type);
2177 	tracing_selftest_running = false;
2178 
2179 	return ret;
2180 }
2181 
2182 static __init int init_trace_selftests(void)
2183 {
2184 	struct trace_selftests *p, *n;
2185 	struct tracer *t, **last;
2186 	int ret;
2187 
2188 	selftests_can_run = true;
2189 
2190 	mutex_lock(&trace_types_lock);
2191 
2192 	if (list_empty(&postponed_selftests))
2193 		goto out;
2194 
2195 	pr_info("Running postponed tracer tests:\n");
2196 
2197 	tracing_selftest_running = true;
2198 	list_for_each_entry_safe(p, n, &postponed_selftests, list) {
2199 		/* This loop can take minutes when sanitizers are enabled, so
2200 		 * lets make sure we allow RCU processing.
2201 		 */
2202 		cond_resched();
2203 		ret = run_tracer_selftest(p->type);
2204 		/* If the test fails, then warn and remove from available_tracers */
2205 		if (ret < 0) {
2206 			WARN(1, "tracer: %s failed selftest, disabling\n",
2207 			     p->type->name);
2208 			last = &trace_types;
2209 			for (t = trace_types; t; t = t->next) {
2210 				if (t == p->type) {
2211 					*last = t->next;
2212 					break;
2213 				}
2214 				last = &t->next;
2215 			}
2216 		}
2217 		list_del(&p->list);
2218 		kfree(p);
2219 	}
2220 	tracing_selftest_running = false;
2221 
2222  out:
2223 	mutex_unlock(&trace_types_lock);
2224 
2225 	return 0;
2226 }
2227 core_initcall(init_trace_selftests);
2228 #else
2229 static inline int run_tracer_selftest(struct tracer *type)
2230 {
2231 	return 0;
2232 }
2233 static inline int do_run_tracer_selftest(struct tracer *type)
2234 {
2235 	return 0;
2236 }
2237 #endif /* CONFIG_FTRACE_STARTUP_TEST */
2238 
2239 static void add_tracer_options(struct trace_array *tr, struct tracer *t);
2240 
2241 static void __init apply_trace_boot_options(void);
2242 
2243 /**
2244  * register_tracer - register a tracer with the ftrace system.
2245  * @type: the plugin for the tracer
2246  *
2247  * Register a new plugin tracer.
2248  */
2249 int __init register_tracer(struct tracer *type)
2250 {
2251 	struct tracer *t;
2252 	int ret = 0;
2253 
2254 	if (!type->name) {
2255 		pr_info("Tracer must have a name\n");
2256 		return -1;
2257 	}
2258 
2259 	if (strlen(type->name) >= MAX_TRACER_SIZE) {
2260 		pr_info("Tracer has a name longer than %d\n", MAX_TRACER_SIZE);
2261 		return -1;
2262 	}
2263 
2264 	if (security_locked_down(LOCKDOWN_TRACEFS)) {
2265 		pr_warn("Can not register tracer %s due to lockdown\n",
2266 			   type->name);
2267 		return -EPERM;
2268 	}
2269 
2270 	mutex_lock(&trace_types_lock);
2271 
2272 	for (t = trace_types; t; t = t->next) {
2273 		if (strcmp(type->name, t->name) == 0) {
2274 			/* already found */
2275 			pr_info("Tracer %s already registered\n",
2276 				type->name);
2277 			ret = -1;
2278 			goto out;
2279 		}
2280 	}
2281 
2282 	if (!type->set_flag)
2283 		type->set_flag = &dummy_set_flag;
2284 	if (!type->flags) {
2285 		/*allocate a dummy tracer_flags*/
2286 		type->flags = kmalloc(sizeof(*type->flags), GFP_KERNEL);
2287 		if (!type->flags) {
2288 			ret = -ENOMEM;
2289 			goto out;
2290 		}
2291 		type->flags->val = 0;
2292 		type->flags->opts = dummy_tracer_opt;
2293 	} else
2294 		if (!type->flags->opts)
2295 			type->flags->opts = dummy_tracer_opt;
2296 
2297 	/* store the tracer for __set_tracer_option */
2298 	type->flags->trace = type;
2299 
2300 	ret = do_run_tracer_selftest(type);
2301 	if (ret < 0)
2302 		goto out;
2303 
2304 	type->next = trace_types;
2305 	trace_types = type;
2306 	add_tracer_options(&global_trace, type);
2307 
2308  out:
2309 	mutex_unlock(&trace_types_lock);
2310 
2311 	if (ret || !default_bootup_tracer)
2312 		goto out_unlock;
2313 
2314 	if (strncmp(default_bootup_tracer, type->name, MAX_TRACER_SIZE))
2315 		goto out_unlock;
2316 
2317 	printk(KERN_INFO "Starting tracer '%s'\n", type->name);
2318 	/* Do we want this tracer to start on bootup? */
2319 	tracing_set_tracer(&global_trace, type->name);
2320 	default_bootup_tracer = NULL;
2321 
2322 	apply_trace_boot_options();
2323 
2324 	/* disable other selftests, since this will break it. */
2325 	disable_tracing_selftest("running a tracer");
2326 
2327  out_unlock:
2328 	return ret;
2329 }
2330 
2331 static void tracing_reset_cpu(struct array_buffer *buf, int cpu)
2332 {
2333 	struct trace_buffer *buffer = buf->buffer;
2334 
2335 	if (!buffer)
2336 		return;
2337 
2338 	ring_buffer_record_disable(buffer);
2339 
2340 	/* Make sure all commits have finished */
2341 	synchronize_rcu();
2342 	ring_buffer_reset_cpu(buffer, cpu);
2343 
2344 	ring_buffer_record_enable(buffer);
2345 }
2346 
2347 void tracing_reset_online_cpus(struct array_buffer *buf)
2348 {
2349 	struct trace_buffer *buffer = buf->buffer;
2350 
2351 	if (!buffer)
2352 		return;
2353 
2354 	ring_buffer_record_disable(buffer);
2355 
2356 	/* Make sure all commits have finished */
2357 	synchronize_rcu();
2358 
2359 	buf->time_start = buffer_ftrace_now(buf, buf->cpu);
2360 
2361 	ring_buffer_reset_online_cpus(buffer);
2362 
2363 	ring_buffer_record_enable(buffer);
2364 }
2365 
2366 /* Must have trace_types_lock held */
2367 void tracing_reset_all_online_cpus_unlocked(void)
2368 {
2369 	struct trace_array *tr;
2370 
2371 	lockdep_assert_held(&trace_types_lock);
2372 
2373 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
2374 		if (!tr->clear_trace)
2375 			continue;
2376 		tr->clear_trace = false;
2377 		tracing_reset_online_cpus(&tr->array_buffer);
2378 #ifdef CONFIG_TRACER_MAX_TRACE
2379 		tracing_reset_online_cpus(&tr->max_buffer);
2380 #endif
2381 	}
2382 }
2383 
2384 void tracing_reset_all_online_cpus(void)
2385 {
2386 	mutex_lock(&trace_types_lock);
2387 	tracing_reset_all_online_cpus_unlocked();
2388 	mutex_unlock(&trace_types_lock);
2389 }
2390 
2391 int is_tracing_stopped(void)
2392 {
2393 	return global_trace.stop_count;
2394 }
2395 
2396 static void tracing_start_tr(struct trace_array *tr)
2397 {
2398 	struct trace_buffer *buffer;
2399 	unsigned long flags;
2400 
2401 	if (tracing_disabled)
2402 		return;
2403 
2404 	raw_spin_lock_irqsave(&tr->start_lock, flags);
2405 	if (--tr->stop_count) {
2406 		if (WARN_ON_ONCE(tr->stop_count < 0)) {
2407 			/* Someone screwed up their debugging */
2408 			tr->stop_count = 0;
2409 		}
2410 		goto out;
2411 	}
2412 
2413 	/* Prevent the buffers from switching */
2414 	arch_spin_lock(&tr->max_lock);
2415 
2416 	buffer = tr->array_buffer.buffer;
2417 	if (buffer)
2418 		ring_buffer_record_enable(buffer);
2419 
2420 #ifdef CONFIG_TRACER_MAX_TRACE
2421 	buffer = tr->max_buffer.buffer;
2422 	if (buffer)
2423 		ring_buffer_record_enable(buffer);
2424 #endif
2425 
2426 	arch_spin_unlock(&tr->max_lock);
2427 
2428  out:
2429 	raw_spin_unlock_irqrestore(&tr->start_lock, flags);
2430 }
2431 
2432 /**
2433  * tracing_start - quick start of the tracer
2434  *
2435  * If tracing is enabled but was stopped by tracing_stop,
2436  * this will start the tracer back up.
2437  */
2438 void tracing_start(void)
2439 
2440 {
2441 	return tracing_start_tr(&global_trace);
2442 }
2443 
2444 static void tracing_stop_tr(struct trace_array *tr)
2445 {
2446 	struct trace_buffer *buffer;
2447 	unsigned long flags;
2448 
2449 	raw_spin_lock_irqsave(&tr->start_lock, flags);
2450 	if (tr->stop_count++)
2451 		goto out;
2452 
2453 	/* Prevent the buffers from switching */
2454 	arch_spin_lock(&tr->max_lock);
2455 
2456 	buffer = tr->array_buffer.buffer;
2457 	if (buffer)
2458 		ring_buffer_record_disable(buffer);
2459 
2460 #ifdef CONFIG_TRACER_MAX_TRACE
2461 	buffer = tr->max_buffer.buffer;
2462 	if (buffer)
2463 		ring_buffer_record_disable(buffer);
2464 #endif
2465 
2466 	arch_spin_unlock(&tr->max_lock);
2467 
2468  out:
2469 	raw_spin_unlock_irqrestore(&tr->start_lock, flags);
2470 }
2471 
2472 /**
2473  * tracing_stop - quick stop of the tracer
2474  *
2475  * Light weight way to stop tracing. Use in conjunction with
2476  * tracing_start.
2477  */
2478 void tracing_stop(void)
2479 {
2480 	return tracing_stop_tr(&global_trace);
2481 }
2482 
2483 /*
2484  * Several functions return TRACE_TYPE_PARTIAL_LINE if the trace_seq
2485  * overflowed, and TRACE_TYPE_HANDLED otherwise. This helper function
2486  * simplifies those functions and keeps them in sync.
2487  */
2488 enum print_line_t trace_handle_return(struct trace_seq *s)
2489 {
2490 	return trace_seq_has_overflowed(s) ?
2491 		TRACE_TYPE_PARTIAL_LINE : TRACE_TYPE_HANDLED;
2492 }
2493 EXPORT_SYMBOL_GPL(trace_handle_return);
2494 
2495 static unsigned short migration_disable_value(void)
2496 {
2497 #if defined(CONFIG_SMP)
2498 	return current->migration_disabled;
2499 #else
2500 	return 0;
2501 #endif
2502 }
2503 
2504 unsigned int tracing_gen_ctx_irq_test(unsigned int irqs_status)
2505 {
2506 	unsigned int trace_flags = irqs_status;
2507 	unsigned int pc;
2508 
2509 	pc = preempt_count();
2510 
2511 	if (pc & NMI_MASK)
2512 		trace_flags |= TRACE_FLAG_NMI;
2513 	if (pc & HARDIRQ_MASK)
2514 		trace_flags |= TRACE_FLAG_HARDIRQ;
2515 	if (in_serving_softirq())
2516 		trace_flags |= TRACE_FLAG_SOFTIRQ;
2517 	if (softirq_count() >> (SOFTIRQ_SHIFT + 1))
2518 		trace_flags |= TRACE_FLAG_BH_OFF;
2519 
2520 	if (tif_need_resched())
2521 		trace_flags |= TRACE_FLAG_NEED_RESCHED;
2522 	if (test_preempt_need_resched())
2523 		trace_flags |= TRACE_FLAG_PREEMPT_RESCHED;
2524 	return (trace_flags << 16) | (min_t(unsigned int, pc & 0xff, 0xf)) |
2525 		(min_t(unsigned int, migration_disable_value(), 0xf)) << 4;
2526 }
2527 
2528 struct ring_buffer_event *
2529 trace_buffer_lock_reserve(struct trace_buffer *buffer,
2530 			  int type,
2531 			  unsigned long len,
2532 			  unsigned int trace_ctx)
2533 {
2534 	return __trace_buffer_lock_reserve(buffer, type, len, trace_ctx);
2535 }
2536 
2537 DEFINE_PER_CPU(struct ring_buffer_event *, trace_buffered_event);
2538 DEFINE_PER_CPU(int, trace_buffered_event_cnt);
2539 static int trace_buffered_event_ref;
2540 
2541 /**
2542  * trace_buffered_event_enable - enable buffering events
2543  *
2544  * When events are being filtered, it is quicker to use a temporary
2545  * buffer to write the event data into if there's a likely chance
2546  * that it will not be committed. The discard of the ring buffer
2547  * is not as fast as committing, and is much slower than copying
2548  * a commit.
2549  *
2550  * When an event is to be filtered, allocate per cpu buffers to
2551  * write the event data into, and if the event is filtered and discarded
2552  * it is simply dropped, otherwise, the entire data is to be committed
2553  * in one shot.
2554  */
2555 void trace_buffered_event_enable(void)
2556 {
2557 	struct ring_buffer_event *event;
2558 	struct page *page;
2559 	int cpu;
2560 
2561 	WARN_ON_ONCE(!mutex_is_locked(&event_mutex));
2562 
2563 	if (trace_buffered_event_ref++)
2564 		return;
2565 
2566 	for_each_tracing_cpu(cpu) {
2567 		page = alloc_pages_node(cpu_to_node(cpu),
2568 					GFP_KERNEL | __GFP_NORETRY, 0);
2569 		/* This is just an optimization and can handle failures */
2570 		if (!page) {
2571 			pr_err("Failed to allocate event buffer\n");
2572 			break;
2573 		}
2574 
2575 		event = page_address(page);
2576 		memset(event, 0, sizeof(*event));
2577 
2578 		per_cpu(trace_buffered_event, cpu) = event;
2579 
2580 		preempt_disable();
2581 		if (cpu == smp_processor_id() &&
2582 		    __this_cpu_read(trace_buffered_event) !=
2583 		    per_cpu(trace_buffered_event, cpu))
2584 			WARN_ON_ONCE(1);
2585 		preempt_enable();
2586 	}
2587 }
2588 
2589 static void enable_trace_buffered_event(void *data)
2590 {
2591 	/* Probably not needed, but do it anyway */
2592 	smp_rmb();
2593 	this_cpu_dec(trace_buffered_event_cnt);
2594 }
2595 
2596 static void disable_trace_buffered_event(void *data)
2597 {
2598 	this_cpu_inc(trace_buffered_event_cnt);
2599 }
2600 
2601 /**
2602  * trace_buffered_event_disable - disable buffering events
2603  *
2604  * When a filter is removed, it is faster to not use the buffered
2605  * events, and to commit directly into the ring buffer. Free up
2606  * the temp buffers when there are no more users. This requires
2607  * special synchronization with current events.
2608  */
2609 void trace_buffered_event_disable(void)
2610 {
2611 	int cpu;
2612 
2613 	WARN_ON_ONCE(!mutex_is_locked(&event_mutex));
2614 
2615 	if (WARN_ON_ONCE(!trace_buffered_event_ref))
2616 		return;
2617 
2618 	if (--trace_buffered_event_ref)
2619 		return;
2620 
2621 	/* For each CPU, set the buffer as used. */
2622 	on_each_cpu_mask(tracing_buffer_mask, disable_trace_buffered_event,
2623 			 NULL, true);
2624 
2625 	/* Wait for all current users to finish */
2626 	synchronize_rcu();
2627 
2628 	for_each_tracing_cpu(cpu) {
2629 		free_page((unsigned long)per_cpu(trace_buffered_event, cpu));
2630 		per_cpu(trace_buffered_event, cpu) = NULL;
2631 	}
2632 
2633 	/*
2634 	 * Wait for all CPUs that potentially started checking if they can use
2635 	 * their event buffer only after the previous synchronize_rcu() call and
2636 	 * they still read a valid pointer from trace_buffered_event. It must be
2637 	 * ensured they don't see cleared trace_buffered_event_cnt else they
2638 	 * could wrongly decide to use the pointed-to buffer which is now freed.
2639 	 */
2640 	synchronize_rcu();
2641 
2642 	/* For each CPU, relinquish the buffer */
2643 	on_each_cpu_mask(tracing_buffer_mask, enable_trace_buffered_event, NULL,
2644 			 true);
2645 }
2646 
2647 static struct trace_buffer *temp_buffer;
2648 
2649 struct ring_buffer_event *
2650 trace_event_buffer_lock_reserve(struct trace_buffer **current_rb,
2651 			  struct trace_event_file *trace_file,
2652 			  int type, unsigned long len,
2653 			  unsigned int trace_ctx)
2654 {
2655 	struct ring_buffer_event *entry;
2656 	struct trace_array *tr = trace_file->tr;
2657 	int val;
2658 
2659 	*current_rb = tr->array_buffer.buffer;
2660 
2661 	if (!tr->no_filter_buffering_ref &&
2662 	    (trace_file->flags & (EVENT_FILE_FL_SOFT_DISABLED | EVENT_FILE_FL_FILTERED))) {
2663 		preempt_disable_notrace();
2664 		/*
2665 		 * Filtering is on, so try to use the per cpu buffer first.
2666 		 * This buffer will simulate a ring_buffer_event,
2667 		 * where the type_len is zero and the array[0] will
2668 		 * hold the full length.
2669 		 * (see include/linux/ring-buffer.h for details on
2670 		 *  how the ring_buffer_event is structured).
2671 		 *
2672 		 * Using a temp buffer during filtering and copying it
2673 		 * on a matched filter is quicker than writing directly
2674 		 * into the ring buffer and then discarding it when
2675 		 * it doesn't match. That is because the discard
2676 		 * requires several atomic operations to get right.
2677 		 * Copying on match and doing nothing on a failed match
2678 		 * is still quicker than no copy on match, but having
2679 		 * to discard out of the ring buffer on a failed match.
2680 		 */
2681 		if ((entry = __this_cpu_read(trace_buffered_event))) {
2682 			int max_len = PAGE_SIZE - struct_size(entry, array, 1);
2683 
2684 			val = this_cpu_inc_return(trace_buffered_event_cnt);
2685 
2686 			/*
2687 			 * Preemption is disabled, but interrupts and NMIs
2688 			 * can still come in now. If that happens after
2689 			 * the above increment, then it will have to go
2690 			 * back to the old method of allocating the event
2691 			 * on the ring buffer, and if the filter fails, it
2692 			 * will have to call ring_buffer_discard_commit()
2693 			 * to remove it.
2694 			 *
2695 			 * Need to also check the unlikely case that the
2696 			 * length is bigger than the temp buffer size.
2697 			 * If that happens, then the reserve is pretty much
2698 			 * guaranteed to fail, as the ring buffer currently
2699 			 * only allows events less than a page. But that may
2700 			 * change in the future, so let the ring buffer reserve
2701 			 * handle the failure in that case.
2702 			 */
2703 			if (val == 1 && likely(len <= max_len)) {
2704 				trace_event_setup(entry, type, trace_ctx);
2705 				entry->array[0] = len;
2706 				/* Return with preemption disabled */
2707 				return entry;
2708 			}
2709 			this_cpu_dec(trace_buffered_event_cnt);
2710 		}
2711 		/* __trace_buffer_lock_reserve() disables preemption */
2712 		preempt_enable_notrace();
2713 	}
2714 
2715 	entry = __trace_buffer_lock_reserve(*current_rb, type, len,
2716 					    trace_ctx);
2717 	/*
2718 	 * If tracing is off, but we have triggers enabled
2719 	 * we still need to look at the event data. Use the temp_buffer
2720 	 * to store the trace event for the trigger to use. It's recursive
2721 	 * safe and will not be recorded anywhere.
2722 	 */
2723 	if (!entry && trace_file->flags & EVENT_FILE_FL_TRIGGER_COND) {
2724 		*current_rb = temp_buffer;
2725 		entry = __trace_buffer_lock_reserve(*current_rb, type, len,
2726 						    trace_ctx);
2727 	}
2728 	return entry;
2729 }
2730 EXPORT_SYMBOL_GPL(trace_event_buffer_lock_reserve);
2731 
2732 static DEFINE_RAW_SPINLOCK(tracepoint_iter_lock);
2733 static DEFINE_MUTEX(tracepoint_printk_mutex);
2734 
2735 static void output_printk(struct trace_event_buffer *fbuffer)
2736 {
2737 	struct trace_event_call *event_call;
2738 	struct trace_event_file *file;
2739 	struct trace_event *event;
2740 	unsigned long flags;
2741 	struct trace_iterator *iter = tracepoint_print_iter;
2742 
2743 	/* We should never get here if iter is NULL */
2744 	if (WARN_ON_ONCE(!iter))
2745 		return;
2746 
2747 	event_call = fbuffer->trace_file->event_call;
2748 	if (!event_call || !event_call->event.funcs ||
2749 	    !event_call->event.funcs->trace)
2750 		return;
2751 
2752 	file = fbuffer->trace_file;
2753 	if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &file->flags) ||
2754 	    (unlikely(file->flags & EVENT_FILE_FL_FILTERED) &&
2755 	     !filter_match_preds(file->filter, fbuffer->entry)))
2756 		return;
2757 
2758 	event = &fbuffer->trace_file->event_call->event;
2759 
2760 	raw_spin_lock_irqsave(&tracepoint_iter_lock, flags);
2761 	trace_seq_init(&iter->seq);
2762 	iter->ent = fbuffer->entry;
2763 	event_call->event.funcs->trace(iter, 0, event);
2764 	trace_seq_putc(&iter->seq, 0);
2765 	printk("%s", iter->seq.buffer);
2766 
2767 	raw_spin_unlock_irqrestore(&tracepoint_iter_lock, flags);
2768 }
2769 
2770 int tracepoint_printk_sysctl(const struct ctl_table *table, int write,
2771 			     void *buffer, size_t *lenp,
2772 			     loff_t *ppos)
2773 {
2774 	int save_tracepoint_printk;
2775 	int ret;
2776 
2777 	mutex_lock(&tracepoint_printk_mutex);
2778 	save_tracepoint_printk = tracepoint_printk;
2779 
2780 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
2781 
2782 	/*
2783 	 * This will force exiting early, as tracepoint_printk
2784 	 * is always zero when tracepoint_printk_iter is not allocated
2785 	 */
2786 	if (!tracepoint_print_iter)
2787 		tracepoint_printk = 0;
2788 
2789 	if (save_tracepoint_printk == tracepoint_printk)
2790 		goto out;
2791 
2792 	if (tracepoint_printk)
2793 		static_key_enable(&tracepoint_printk_key.key);
2794 	else
2795 		static_key_disable(&tracepoint_printk_key.key);
2796 
2797  out:
2798 	mutex_unlock(&tracepoint_printk_mutex);
2799 
2800 	return ret;
2801 }
2802 
2803 void trace_event_buffer_commit(struct trace_event_buffer *fbuffer)
2804 {
2805 	enum event_trigger_type tt = ETT_NONE;
2806 	struct trace_event_file *file = fbuffer->trace_file;
2807 
2808 	if (__event_trigger_test_discard(file, fbuffer->buffer, fbuffer->event,
2809 			fbuffer->entry, &tt))
2810 		goto discard;
2811 
2812 	if (static_key_false(&tracepoint_printk_key.key))
2813 		output_printk(fbuffer);
2814 
2815 	if (static_branch_unlikely(&trace_event_exports_enabled))
2816 		ftrace_exports(fbuffer->event, TRACE_EXPORT_EVENT);
2817 
2818 	trace_buffer_unlock_commit_regs(file->tr, fbuffer->buffer,
2819 			fbuffer->event, fbuffer->trace_ctx, fbuffer->regs);
2820 
2821 discard:
2822 	if (tt)
2823 		event_triggers_post_call(file, tt);
2824 
2825 }
2826 EXPORT_SYMBOL_GPL(trace_event_buffer_commit);
2827 
2828 /*
2829  * Skip 3:
2830  *
2831  *   trace_buffer_unlock_commit_regs()
2832  *   trace_event_buffer_commit()
2833  *   trace_event_raw_event_xxx()
2834  */
2835 # define STACK_SKIP 3
2836 
2837 void trace_buffer_unlock_commit_regs(struct trace_array *tr,
2838 				     struct trace_buffer *buffer,
2839 				     struct ring_buffer_event *event,
2840 				     unsigned int trace_ctx,
2841 				     struct pt_regs *regs)
2842 {
2843 	__buffer_unlock_commit(buffer, event);
2844 
2845 	/*
2846 	 * If regs is not set, then skip the necessary functions.
2847 	 * Note, we can still get here via blktrace, wakeup tracer
2848 	 * and mmiotrace, but that's ok if they lose a function or
2849 	 * two. They are not that meaningful.
2850 	 */
2851 	ftrace_trace_stack(tr, buffer, trace_ctx, regs ? 0 : STACK_SKIP, regs);
2852 	ftrace_trace_userstack(tr, buffer, trace_ctx);
2853 }
2854 
2855 /*
2856  * Similar to trace_buffer_unlock_commit_regs() but do not dump stack.
2857  */
2858 void
2859 trace_buffer_unlock_commit_nostack(struct trace_buffer *buffer,
2860 				   struct ring_buffer_event *event)
2861 {
2862 	__buffer_unlock_commit(buffer, event);
2863 }
2864 
2865 void
2866 trace_function(struct trace_array *tr, unsigned long ip, unsigned long
2867 	       parent_ip, unsigned int trace_ctx)
2868 {
2869 	struct trace_event_call *call = &event_function;
2870 	struct trace_buffer *buffer = tr->array_buffer.buffer;
2871 	struct ring_buffer_event *event;
2872 	struct ftrace_entry *entry;
2873 
2874 	event = __trace_buffer_lock_reserve(buffer, TRACE_FN, sizeof(*entry),
2875 					    trace_ctx);
2876 	if (!event)
2877 		return;
2878 	entry	= ring_buffer_event_data(event);
2879 	entry->ip			= ip;
2880 	entry->parent_ip		= parent_ip;
2881 
2882 	if (!call_filter_check_discard(call, entry, buffer, event)) {
2883 		if (static_branch_unlikely(&trace_function_exports_enabled))
2884 			ftrace_exports(event, TRACE_EXPORT_FUNCTION);
2885 		__buffer_unlock_commit(buffer, event);
2886 	}
2887 }
2888 
2889 #ifdef CONFIG_STACKTRACE
2890 
2891 /* Allow 4 levels of nesting: normal, softirq, irq, NMI */
2892 #define FTRACE_KSTACK_NESTING	4
2893 
2894 #define FTRACE_KSTACK_ENTRIES	(PAGE_SIZE / FTRACE_KSTACK_NESTING)
2895 
2896 struct ftrace_stack {
2897 	unsigned long		calls[FTRACE_KSTACK_ENTRIES];
2898 };
2899 
2900 
2901 struct ftrace_stacks {
2902 	struct ftrace_stack	stacks[FTRACE_KSTACK_NESTING];
2903 };
2904 
2905 static DEFINE_PER_CPU(struct ftrace_stacks, ftrace_stacks);
2906 static DEFINE_PER_CPU(int, ftrace_stack_reserve);
2907 
2908 static void __ftrace_trace_stack(struct trace_buffer *buffer,
2909 				 unsigned int trace_ctx,
2910 				 int skip, struct pt_regs *regs)
2911 {
2912 	struct trace_event_call *call = &event_kernel_stack;
2913 	struct ring_buffer_event *event;
2914 	unsigned int size, nr_entries;
2915 	struct ftrace_stack *fstack;
2916 	struct stack_entry *entry;
2917 	int stackidx;
2918 
2919 	/*
2920 	 * Add one, for this function and the call to save_stack_trace()
2921 	 * If regs is set, then these functions will not be in the way.
2922 	 */
2923 #ifndef CONFIG_UNWINDER_ORC
2924 	if (!regs)
2925 		skip++;
2926 #endif
2927 
2928 	preempt_disable_notrace();
2929 
2930 	stackidx = __this_cpu_inc_return(ftrace_stack_reserve) - 1;
2931 
2932 	/* This should never happen. If it does, yell once and skip */
2933 	if (WARN_ON_ONCE(stackidx >= FTRACE_KSTACK_NESTING))
2934 		goto out;
2935 
2936 	/*
2937 	 * The above __this_cpu_inc_return() is 'atomic' cpu local. An
2938 	 * interrupt will either see the value pre increment or post
2939 	 * increment. If the interrupt happens pre increment it will have
2940 	 * restored the counter when it returns.  We just need a barrier to
2941 	 * keep gcc from moving things around.
2942 	 */
2943 	barrier();
2944 
2945 	fstack = this_cpu_ptr(ftrace_stacks.stacks) + stackidx;
2946 	size = ARRAY_SIZE(fstack->calls);
2947 
2948 	if (regs) {
2949 		nr_entries = stack_trace_save_regs(regs, fstack->calls,
2950 						   size, skip);
2951 	} else {
2952 		nr_entries = stack_trace_save(fstack->calls, size, skip);
2953 	}
2954 
2955 	event = __trace_buffer_lock_reserve(buffer, TRACE_STACK,
2956 				    struct_size(entry, caller, nr_entries),
2957 				    trace_ctx);
2958 	if (!event)
2959 		goto out;
2960 	entry = ring_buffer_event_data(event);
2961 
2962 	entry->size = nr_entries;
2963 	memcpy(&entry->caller, fstack->calls,
2964 	       flex_array_size(entry, caller, nr_entries));
2965 
2966 	if (!call_filter_check_discard(call, entry, buffer, event))
2967 		__buffer_unlock_commit(buffer, event);
2968 
2969  out:
2970 	/* Again, don't let gcc optimize things here */
2971 	barrier();
2972 	__this_cpu_dec(ftrace_stack_reserve);
2973 	preempt_enable_notrace();
2974 
2975 }
2976 
2977 static inline void ftrace_trace_stack(struct trace_array *tr,
2978 				      struct trace_buffer *buffer,
2979 				      unsigned int trace_ctx,
2980 				      int skip, struct pt_regs *regs)
2981 {
2982 	if (!(tr->trace_flags & TRACE_ITER_STACKTRACE))
2983 		return;
2984 
2985 	__ftrace_trace_stack(buffer, trace_ctx, skip, regs);
2986 }
2987 
2988 void __trace_stack(struct trace_array *tr, unsigned int trace_ctx,
2989 		   int skip)
2990 {
2991 	struct trace_buffer *buffer = tr->array_buffer.buffer;
2992 
2993 	if (rcu_is_watching()) {
2994 		__ftrace_trace_stack(buffer, trace_ctx, skip, NULL);
2995 		return;
2996 	}
2997 
2998 	if (WARN_ON_ONCE(IS_ENABLED(CONFIG_GENERIC_ENTRY)))
2999 		return;
3000 
3001 	/*
3002 	 * When an NMI triggers, RCU is enabled via ct_nmi_enter(),
3003 	 * but if the above rcu_is_watching() failed, then the NMI
3004 	 * triggered someplace critical, and ct_irq_enter() should
3005 	 * not be called from NMI.
3006 	 */
3007 	if (unlikely(in_nmi()))
3008 		return;
3009 
3010 	ct_irq_enter_irqson();
3011 	__ftrace_trace_stack(buffer, trace_ctx, skip, NULL);
3012 	ct_irq_exit_irqson();
3013 }
3014 
3015 /**
3016  * trace_dump_stack - record a stack back trace in the trace buffer
3017  * @skip: Number of functions to skip (helper handlers)
3018  */
3019 void trace_dump_stack(int skip)
3020 {
3021 	if (tracing_disabled || tracing_selftest_running)
3022 		return;
3023 
3024 #ifndef CONFIG_UNWINDER_ORC
3025 	/* Skip 1 to skip this function. */
3026 	skip++;
3027 #endif
3028 	__ftrace_trace_stack(global_trace.array_buffer.buffer,
3029 			     tracing_gen_ctx(), skip, NULL);
3030 }
3031 EXPORT_SYMBOL_GPL(trace_dump_stack);
3032 
3033 #ifdef CONFIG_USER_STACKTRACE_SUPPORT
3034 static DEFINE_PER_CPU(int, user_stack_count);
3035 
3036 static void
3037 ftrace_trace_userstack(struct trace_array *tr,
3038 		       struct trace_buffer *buffer, unsigned int trace_ctx)
3039 {
3040 	struct trace_event_call *call = &event_user_stack;
3041 	struct ring_buffer_event *event;
3042 	struct userstack_entry *entry;
3043 
3044 	if (!(tr->trace_flags & TRACE_ITER_USERSTACKTRACE))
3045 		return;
3046 
3047 	/*
3048 	 * NMIs can not handle page faults, even with fix ups.
3049 	 * The save user stack can (and often does) fault.
3050 	 */
3051 	if (unlikely(in_nmi()))
3052 		return;
3053 
3054 	/*
3055 	 * prevent recursion, since the user stack tracing may
3056 	 * trigger other kernel events.
3057 	 */
3058 	preempt_disable();
3059 	if (__this_cpu_read(user_stack_count))
3060 		goto out;
3061 
3062 	__this_cpu_inc(user_stack_count);
3063 
3064 	event = __trace_buffer_lock_reserve(buffer, TRACE_USER_STACK,
3065 					    sizeof(*entry), trace_ctx);
3066 	if (!event)
3067 		goto out_drop_count;
3068 	entry	= ring_buffer_event_data(event);
3069 
3070 	entry->tgid		= current->tgid;
3071 	memset(&entry->caller, 0, sizeof(entry->caller));
3072 
3073 	stack_trace_save_user(entry->caller, FTRACE_STACK_ENTRIES);
3074 	if (!call_filter_check_discard(call, entry, buffer, event))
3075 		__buffer_unlock_commit(buffer, event);
3076 
3077  out_drop_count:
3078 	__this_cpu_dec(user_stack_count);
3079  out:
3080 	preempt_enable();
3081 }
3082 #else /* CONFIG_USER_STACKTRACE_SUPPORT */
3083 static void ftrace_trace_userstack(struct trace_array *tr,
3084 				   struct trace_buffer *buffer,
3085 				   unsigned int trace_ctx)
3086 {
3087 }
3088 #endif /* !CONFIG_USER_STACKTRACE_SUPPORT */
3089 
3090 #endif /* CONFIG_STACKTRACE */
3091 
3092 static inline void
3093 func_repeats_set_delta_ts(struct func_repeats_entry *entry,
3094 			  unsigned long long delta)
3095 {
3096 	entry->bottom_delta_ts = delta & U32_MAX;
3097 	entry->top_delta_ts = (delta >> 32);
3098 }
3099 
3100 void trace_last_func_repeats(struct trace_array *tr,
3101 			     struct trace_func_repeats *last_info,
3102 			     unsigned int trace_ctx)
3103 {
3104 	struct trace_buffer *buffer = tr->array_buffer.buffer;
3105 	struct func_repeats_entry *entry;
3106 	struct ring_buffer_event *event;
3107 	u64 delta;
3108 
3109 	event = __trace_buffer_lock_reserve(buffer, TRACE_FUNC_REPEATS,
3110 					    sizeof(*entry), trace_ctx);
3111 	if (!event)
3112 		return;
3113 
3114 	delta = ring_buffer_event_time_stamp(buffer, event) -
3115 		last_info->ts_last_call;
3116 
3117 	entry = ring_buffer_event_data(event);
3118 	entry->ip = last_info->ip;
3119 	entry->parent_ip = last_info->parent_ip;
3120 	entry->count = last_info->count;
3121 	func_repeats_set_delta_ts(entry, delta);
3122 
3123 	__buffer_unlock_commit(buffer, event);
3124 }
3125 
3126 /* created for use with alloc_percpu */
3127 struct trace_buffer_struct {
3128 	int nesting;
3129 	char buffer[4][TRACE_BUF_SIZE];
3130 };
3131 
3132 static struct trace_buffer_struct __percpu *trace_percpu_buffer;
3133 
3134 /*
3135  * This allows for lockless recording.  If we're nested too deeply, then
3136  * this returns NULL.
3137  */
3138 static char *get_trace_buf(void)
3139 {
3140 	struct trace_buffer_struct *buffer = this_cpu_ptr(trace_percpu_buffer);
3141 
3142 	if (!trace_percpu_buffer || buffer->nesting >= 4)
3143 		return NULL;
3144 
3145 	buffer->nesting++;
3146 
3147 	/* Interrupts must see nesting incremented before we use the buffer */
3148 	barrier();
3149 	return &buffer->buffer[buffer->nesting - 1][0];
3150 }
3151 
3152 static void put_trace_buf(void)
3153 {
3154 	/* Don't let the decrement of nesting leak before this */
3155 	barrier();
3156 	this_cpu_dec(trace_percpu_buffer->nesting);
3157 }
3158 
3159 static int alloc_percpu_trace_buffer(void)
3160 {
3161 	struct trace_buffer_struct __percpu *buffers;
3162 
3163 	if (trace_percpu_buffer)
3164 		return 0;
3165 
3166 	buffers = alloc_percpu(struct trace_buffer_struct);
3167 	if (MEM_FAIL(!buffers, "Could not allocate percpu trace_printk buffer"))
3168 		return -ENOMEM;
3169 
3170 	trace_percpu_buffer = buffers;
3171 	return 0;
3172 }
3173 
3174 static int buffers_allocated;
3175 
3176 void trace_printk_init_buffers(void)
3177 {
3178 	if (buffers_allocated)
3179 		return;
3180 
3181 	if (alloc_percpu_trace_buffer())
3182 		return;
3183 
3184 	/* trace_printk() is for debug use only. Don't use it in production. */
3185 
3186 	pr_warn("\n");
3187 	pr_warn("**********************************************************\n");
3188 	pr_warn("**   NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE   **\n");
3189 	pr_warn("**                                                      **\n");
3190 	pr_warn("** trace_printk() being used. Allocating extra memory.  **\n");
3191 	pr_warn("**                                                      **\n");
3192 	pr_warn("** This means that this is a DEBUG kernel and it is     **\n");
3193 	pr_warn("** unsafe for production use.                           **\n");
3194 	pr_warn("**                                                      **\n");
3195 	pr_warn("** If you see this message and you are not debugging    **\n");
3196 	pr_warn("** the kernel, report this immediately to your vendor!  **\n");
3197 	pr_warn("**                                                      **\n");
3198 	pr_warn("**   NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE   **\n");
3199 	pr_warn("**********************************************************\n");
3200 
3201 	/* Expand the buffers to set size */
3202 	tracing_update_buffers(&global_trace);
3203 
3204 	buffers_allocated = 1;
3205 
3206 	/*
3207 	 * trace_printk_init_buffers() can be called by modules.
3208 	 * If that happens, then we need to start cmdline recording
3209 	 * directly here. If the global_trace.buffer is already
3210 	 * allocated here, then this was called by module code.
3211 	 */
3212 	if (global_trace.array_buffer.buffer)
3213 		tracing_start_cmdline_record();
3214 }
3215 EXPORT_SYMBOL_GPL(trace_printk_init_buffers);
3216 
3217 void trace_printk_start_comm(void)
3218 {
3219 	/* Start tracing comms if trace printk is set */
3220 	if (!buffers_allocated)
3221 		return;
3222 	tracing_start_cmdline_record();
3223 }
3224 
3225 static void trace_printk_start_stop_comm(int enabled)
3226 {
3227 	if (!buffers_allocated)
3228 		return;
3229 
3230 	if (enabled)
3231 		tracing_start_cmdline_record();
3232 	else
3233 		tracing_stop_cmdline_record();
3234 }
3235 
3236 /**
3237  * trace_vbprintk - write binary msg to tracing buffer
3238  * @ip:    The address of the caller
3239  * @fmt:   The string format to write to the buffer
3240  * @args:  Arguments for @fmt
3241  */
3242 int trace_vbprintk(unsigned long ip, const char *fmt, va_list args)
3243 {
3244 	struct trace_event_call *call = &event_bprint;
3245 	struct ring_buffer_event *event;
3246 	struct trace_buffer *buffer;
3247 	struct trace_array *tr = &global_trace;
3248 	struct bprint_entry *entry;
3249 	unsigned int trace_ctx;
3250 	char *tbuffer;
3251 	int len = 0, size;
3252 
3253 	if (unlikely(tracing_selftest_running || tracing_disabled))
3254 		return 0;
3255 
3256 	/* Don't pollute graph traces with trace_vprintk internals */
3257 	pause_graph_tracing();
3258 
3259 	trace_ctx = tracing_gen_ctx();
3260 	preempt_disable_notrace();
3261 
3262 	tbuffer = get_trace_buf();
3263 	if (!tbuffer) {
3264 		len = 0;
3265 		goto out_nobuffer;
3266 	}
3267 
3268 	len = vbin_printf((u32 *)tbuffer, TRACE_BUF_SIZE/sizeof(int), fmt, args);
3269 
3270 	if (len > TRACE_BUF_SIZE/sizeof(int) || len < 0)
3271 		goto out_put;
3272 
3273 	size = sizeof(*entry) + sizeof(u32) * len;
3274 	buffer = tr->array_buffer.buffer;
3275 	ring_buffer_nest_start(buffer);
3276 	event = __trace_buffer_lock_reserve(buffer, TRACE_BPRINT, size,
3277 					    trace_ctx);
3278 	if (!event)
3279 		goto out;
3280 	entry = ring_buffer_event_data(event);
3281 	entry->ip			= ip;
3282 	entry->fmt			= fmt;
3283 
3284 	memcpy(entry->buf, tbuffer, sizeof(u32) * len);
3285 	if (!call_filter_check_discard(call, entry, buffer, event)) {
3286 		__buffer_unlock_commit(buffer, event);
3287 		ftrace_trace_stack(tr, buffer, trace_ctx, 6, NULL);
3288 	}
3289 
3290 out:
3291 	ring_buffer_nest_end(buffer);
3292 out_put:
3293 	put_trace_buf();
3294 
3295 out_nobuffer:
3296 	preempt_enable_notrace();
3297 	unpause_graph_tracing();
3298 
3299 	return len;
3300 }
3301 EXPORT_SYMBOL_GPL(trace_vbprintk);
3302 
3303 __printf(3, 0)
3304 static int
3305 __trace_array_vprintk(struct trace_buffer *buffer,
3306 		      unsigned long ip, const char *fmt, va_list args)
3307 {
3308 	struct trace_event_call *call = &event_print;
3309 	struct ring_buffer_event *event;
3310 	int len = 0, size;
3311 	struct print_entry *entry;
3312 	unsigned int trace_ctx;
3313 	char *tbuffer;
3314 
3315 	if (tracing_disabled)
3316 		return 0;
3317 
3318 	/* Don't pollute graph traces with trace_vprintk internals */
3319 	pause_graph_tracing();
3320 
3321 	trace_ctx = tracing_gen_ctx();
3322 	preempt_disable_notrace();
3323 
3324 
3325 	tbuffer = get_trace_buf();
3326 	if (!tbuffer) {
3327 		len = 0;
3328 		goto out_nobuffer;
3329 	}
3330 
3331 	len = vscnprintf(tbuffer, TRACE_BUF_SIZE, fmt, args);
3332 
3333 	size = sizeof(*entry) + len + 1;
3334 	ring_buffer_nest_start(buffer);
3335 	event = __trace_buffer_lock_reserve(buffer, TRACE_PRINT, size,
3336 					    trace_ctx);
3337 	if (!event)
3338 		goto out;
3339 	entry = ring_buffer_event_data(event);
3340 	entry->ip = ip;
3341 
3342 	memcpy(&entry->buf, tbuffer, len + 1);
3343 	if (!call_filter_check_discard(call, entry, buffer, event)) {
3344 		__buffer_unlock_commit(buffer, event);
3345 		ftrace_trace_stack(&global_trace, buffer, trace_ctx, 6, NULL);
3346 	}
3347 
3348 out:
3349 	ring_buffer_nest_end(buffer);
3350 	put_trace_buf();
3351 
3352 out_nobuffer:
3353 	preempt_enable_notrace();
3354 	unpause_graph_tracing();
3355 
3356 	return len;
3357 }
3358 
3359 __printf(3, 0)
3360 int trace_array_vprintk(struct trace_array *tr,
3361 			unsigned long ip, const char *fmt, va_list args)
3362 {
3363 	if (tracing_selftest_running && tr == &global_trace)
3364 		return 0;
3365 
3366 	return __trace_array_vprintk(tr->array_buffer.buffer, ip, fmt, args);
3367 }
3368 
3369 /**
3370  * trace_array_printk - Print a message to a specific instance
3371  * @tr: The instance trace_array descriptor
3372  * @ip: The instruction pointer that this is called from.
3373  * @fmt: The format to print (printf format)
3374  *
3375  * If a subsystem sets up its own instance, they have the right to
3376  * printk strings into their tracing instance buffer using this
3377  * function. Note, this function will not write into the top level
3378  * buffer (use trace_printk() for that), as writing into the top level
3379  * buffer should only have events that can be individually disabled.
3380  * trace_printk() is only used for debugging a kernel, and should not
3381  * be ever incorporated in normal use.
3382  *
3383  * trace_array_printk() can be used, as it will not add noise to the
3384  * top level tracing buffer.
3385  *
3386  * Note, trace_array_init_printk() must be called on @tr before this
3387  * can be used.
3388  */
3389 __printf(3, 0)
3390 int trace_array_printk(struct trace_array *tr,
3391 		       unsigned long ip, const char *fmt, ...)
3392 {
3393 	int ret;
3394 	va_list ap;
3395 
3396 	if (!tr)
3397 		return -ENOENT;
3398 
3399 	/* This is only allowed for created instances */
3400 	if (tr == &global_trace)
3401 		return 0;
3402 
3403 	if (!(tr->trace_flags & TRACE_ITER_PRINTK))
3404 		return 0;
3405 
3406 	va_start(ap, fmt);
3407 	ret = trace_array_vprintk(tr, ip, fmt, ap);
3408 	va_end(ap);
3409 	return ret;
3410 }
3411 EXPORT_SYMBOL_GPL(trace_array_printk);
3412 
3413 /**
3414  * trace_array_init_printk - Initialize buffers for trace_array_printk()
3415  * @tr: The trace array to initialize the buffers for
3416  *
3417  * As trace_array_printk() only writes into instances, they are OK to
3418  * have in the kernel (unlike trace_printk()). This needs to be called
3419  * before trace_array_printk() can be used on a trace_array.
3420  */
3421 int trace_array_init_printk(struct trace_array *tr)
3422 {
3423 	if (!tr)
3424 		return -ENOENT;
3425 
3426 	/* This is only allowed for created instances */
3427 	if (tr == &global_trace)
3428 		return -EINVAL;
3429 
3430 	return alloc_percpu_trace_buffer();
3431 }
3432 EXPORT_SYMBOL_GPL(trace_array_init_printk);
3433 
3434 __printf(3, 4)
3435 int trace_array_printk_buf(struct trace_buffer *buffer,
3436 			   unsigned long ip, const char *fmt, ...)
3437 {
3438 	int ret;
3439 	va_list ap;
3440 
3441 	if (!(global_trace.trace_flags & TRACE_ITER_PRINTK))
3442 		return 0;
3443 
3444 	va_start(ap, fmt);
3445 	ret = __trace_array_vprintk(buffer, ip, fmt, ap);
3446 	va_end(ap);
3447 	return ret;
3448 }
3449 
3450 __printf(2, 0)
3451 int trace_vprintk(unsigned long ip, const char *fmt, va_list args)
3452 {
3453 	return trace_array_vprintk(&global_trace, ip, fmt, args);
3454 }
3455 EXPORT_SYMBOL_GPL(trace_vprintk);
3456 
3457 static void trace_iterator_increment(struct trace_iterator *iter)
3458 {
3459 	struct ring_buffer_iter *buf_iter = trace_buffer_iter(iter, iter->cpu);
3460 
3461 	iter->idx++;
3462 	if (buf_iter)
3463 		ring_buffer_iter_advance(buf_iter);
3464 }
3465 
3466 static struct trace_entry *
3467 peek_next_entry(struct trace_iterator *iter, int cpu, u64 *ts,
3468 		unsigned long *lost_events)
3469 {
3470 	struct ring_buffer_event *event;
3471 	struct ring_buffer_iter *buf_iter = trace_buffer_iter(iter, cpu);
3472 
3473 	if (buf_iter) {
3474 		event = ring_buffer_iter_peek(buf_iter, ts);
3475 		if (lost_events)
3476 			*lost_events = ring_buffer_iter_dropped(buf_iter) ?
3477 				(unsigned long)-1 : 0;
3478 	} else {
3479 		event = ring_buffer_peek(iter->array_buffer->buffer, cpu, ts,
3480 					 lost_events);
3481 	}
3482 
3483 	if (event) {
3484 		iter->ent_size = ring_buffer_event_length(event);
3485 		return ring_buffer_event_data(event);
3486 	}
3487 	iter->ent_size = 0;
3488 	return NULL;
3489 }
3490 
3491 static struct trace_entry *
3492 __find_next_entry(struct trace_iterator *iter, int *ent_cpu,
3493 		  unsigned long *missing_events, u64 *ent_ts)
3494 {
3495 	struct trace_buffer *buffer = iter->array_buffer->buffer;
3496 	struct trace_entry *ent, *next = NULL;
3497 	unsigned long lost_events = 0, next_lost = 0;
3498 	int cpu_file = iter->cpu_file;
3499 	u64 next_ts = 0, ts;
3500 	int next_cpu = -1;
3501 	int next_size = 0;
3502 	int cpu;
3503 
3504 	/*
3505 	 * If we are in a per_cpu trace file, don't bother by iterating over
3506 	 * all cpu and peek directly.
3507 	 */
3508 	if (cpu_file > RING_BUFFER_ALL_CPUS) {
3509 		if (ring_buffer_empty_cpu(buffer, cpu_file))
3510 			return NULL;
3511 		ent = peek_next_entry(iter, cpu_file, ent_ts, missing_events);
3512 		if (ent_cpu)
3513 			*ent_cpu = cpu_file;
3514 
3515 		return ent;
3516 	}
3517 
3518 	for_each_tracing_cpu(cpu) {
3519 
3520 		if (ring_buffer_empty_cpu(buffer, cpu))
3521 			continue;
3522 
3523 		ent = peek_next_entry(iter, cpu, &ts, &lost_events);
3524 
3525 		/*
3526 		 * Pick the entry with the smallest timestamp:
3527 		 */
3528 		if (ent && (!next || ts < next_ts)) {
3529 			next = ent;
3530 			next_cpu = cpu;
3531 			next_ts = ts;
3532 			next_lost = lost_events;
3533 			next_size = iter->ent_size;
3534 		}
3535 	}
3536 
3537 	iter->ent_size = next_size;
3538 
3539 	if (ent_cpu)
3540 		*ent_cpu = next_cpu;
3541 
3542 	if (ent_ts)
3543 		*ent_ts = next_ts;
3544 
3545 	if (missing_events)
3546 		*missing_events = next_lost;
3547 
3548 	return next;
3549 }
3550 
3551 #define STATIC_FMT_BUF_SIZE	128
3552 static char static_fmt_buf[STATIC_FMT_BUF_SIZE];
3553 
3554 char *trace_iter_expand_format(struct trace_iterator *iter)
3555 {
3556 	char *tmp;
3557 
3558 	/*
3559 	 * iter->tr is NULL when used with tp_printk, which makes
3560 	 * this get called where it is not safe to call krealloc().
3561 	 */
3562 	if (!iter->tr || iter->fmt == static_fmt_buf)
3563 		return NULL;
3564 
3565 	tmp = krealloc(iter->fmt, iter->fmt_size + STATIC_FMT_BUF_SIZE,
3566 		       GFP_KERNEL);
3567 	if (tmp) {
3568 		iter->fmt_size += STATIC_FMT_BUF_SIZE;
3569 		iter->fmt = tmp;
3570 	}
3571 
3572 	return tmp;
3573 }
3574 
3575 /* Returns true if the string is safe to dereference from an event */
3576 static bool trace_safe_str(struct trace_iterator *iter, const char *str,
3577 			   bool star, int len)
3578 {
3579 	unsigned long addr = (unsigned long)str;
3580 	struct trace_event *trace_event;
3581 	struct trace_event_call *event;
3582 
3583 	/* Ignore strings with no length */
3584 	if (star && !len)
3585 		return true;
3586 
3587 	/* OK if part of the event data */
3588 	if ((addr >= (unsigned long)iter->ent) &&
3589 	    (addr < (unsigned long)iter->ent + iter->ent_size))
3590 		return true;
3591 
3592 	/* OK if part of the temp seq buffer */
3593 	if ((addr >= (unsigned long)iter->tmp_seq.buffer) &&
3594 	    (addr < (unsigned long)iter->tmp_seq.buffer + TRACE_SEQ_BUFFER_SIZE))
3595 		return true;
3596 
3597 	/* Core rodata can not be freed */
3598 	if (is_kernel_rodata(addr))
3599 		return true;
3600 
3601 	if (trace_is_tracepoint_string(str))
3602 		return true;
3603 
3604 	/*
3605 	 * Now this could be a module event, referencing core module
3606 	 * data, which is OK.
3607 	 */
3608 	if (!iter->ent)
3609 		return false;
3610 
3611 	trace_event = ftrace_find_event(iter->ent->type);
3612 	if (!trace_event)
3613 		return false;
3614 
3615 	event = container_of(trace_event, struct trace_event_call, event);
3616 	if ((event->flags & TRACE_EVENT_FL_DYNAMIC) || !event->module)
3617 		return false;
3618 
3619 	/* Would rather have rodata, but this will suffice */
3620 	if (within_module_core(addr, event->module))
3621 		return true;
3622 
3623 	return false;
3624 }
3625 
3626 static DEFINE_STATIC_KEY_FALSE(trace_no_verify);
3627 
3628 static int test_can_verify_check(const char *fmt, ...)
3629 {
3630 	char buf[16];
3631 	va_list ap;
3632 	int ret;
3633 
3634 	/*
3635 	 * The verifier is dependent on vsnprintf() modifies the va_list
3636 	 * passed to it, where it is sent as a reference. Some architectures
3637 	 * (like x86_32) passes it by value, which means that vsnprintf()
3638 	 * does not modify the va_list passed to it, and the verifier
3639 	 * would then need to be able to understand all the values that
3640 	 * vsnprintf can use. If it is passed by value, then the verifier
3641 	 * is disabled.
3642 	 */
3643 	va_start(ap, fmt);
3644 	vsnprintf(buf, 16, "%d", ap);
3645 	ret = va_arg(ap, int);
3646 	va_end(ap);
3647 
3648 	return ret;
3649 }
3650 
3651 static void test_can_verify(void)
3652 {
3653 	if (!test_can_verify_check("%d %d", 0, 1)) {
3654 		pr_info("trace event string verifier disabled\n");
3655 		static_branch_inc(&trace_no_verify);
3656 	}
3657 }
3658 
3659 /**
3660  * trace_check_vprintf - Check dereferenced strings while writing to the seq buffer
3661  * @iter: The iterator that holds the seq buffer and the event being printed
3662  * @fmt: The format used to print the event
3663  * @ap: The va_list holding the data to print from @fmt.
3664  *
3665  * This writes the data into the @iter->seq buffer using the data from
3666  * @fmt and @ap. If the format has a %s, then the source of the string
3667  * is examined to make sure it is safe to print, otherwise it will
3668  * warn and print "[UNSAFE MEMORY]" in place of the dereferenced string
3669  * pointer.
3670  */
3671 void trace_check_vprintf(struct trace_iterator *iter, const char *fmt,
3672 			 va_list ap)
3673 {
3674 	const char *p = fmt;
3675 	const char *str;
3676 	int i, j;
3677 
3678 	if (WARN_ON_ONCE(!fmt))
3679 		return;
3680 
3681 	if (static_branch_unlikely(&trace_no_verify))
3682 		goto print;
3683 
3684 	/* Don't bother checking when doing a ftrace_dump() */
3685 	if (iter->fmt == static_fmt_buf)
3686 		goto print;
3687 
3688 	while (*p) {
3689 		bool star = false;
3690 		int len = 0;
3691 
3692 		j = 0;
3693 
3694 		/* We only care about %s and variants */
3695 		for (i = 0; p[i]; i++) {
3696 			if (i + 1 >= iter->fmt_size) {
3697 				/*
3698 				 * If we can't expand the copy buffer,
3699 				 * just print it.
3700 				 */
3701 				if (!trace_iter_expand_format(iter))
3702 					goto print;
3703 			}
3704 
3705 			if (p[i] == '\\' && p[i+1]) {
3706 				i++;
3707 				continue;
3708 			}
3709 			if (p[i] == '%') {
3710 				/* Need to test cases like %08.*s */
3711 				for (j = 1; p[i+j]; j++) {
3712 					if (isdigit(p[i+j]) ||
3713 					    p[i+j] == '.')
3714 						continue;
3715 					if (p[i+j] == '*') {
3716 						star = true;
3717 						continue;
3718 					}
3719 					break;
3720 				}
3721 				if (p[i+j] == 's')
3722 					break;
3723 				star = false;
3724 			}
3725 			j = 0;
3726 		}
3727 		/* If no %s found then just print normally */
3728 		if (!p[i])
3729 			break;
3730 
3731 		/* Copy up to the %s, and print that */
3732 		strncpy(iter->fmt, p, i);
3733 		iter->fmt[i] = '\0';
3734 		trace_seq_vprintf(&iter->seq, iter->fmt, ap);
3735 
3736 		/*
3737 		 * If iter->seq is full, the above call no longer guarantees
3738 		 * that ap is in sync with fmt processing, and further calls
3739 		 * to va_arg() can return wrong positional arguments.
3740 		 *
3741 		 * Ensure that ap is no longer used in this case.
3742 		 */
3743 		if (iter->seq.full) {
3744 			p = "";
3745 			break;
3746 		}
3747 
3748 		if (star)
3749 			len = va_arg(ap, int);
3750 
3751 		/* The ap now points to the string data of the %s */
3752 		str = va_arg(ap, const char *);
3753 
3754 		/*
3755 		 * If you hit this warning, it is likely that the
3756 		 * trace event in question used %s on a string that
3757 		 * was saved at the time of the event, but may not be
3758 		 * around when the trace is read. Use __string(),
3759 		 * __assign_str() and __get_str() helpers in the TRACE_EVENT()
3760 		 * instead. See samples/trace_events/trace-events-sample.h
3761 		 * for reference.
3762 		 */
3763 		if (WARN_ONCE(!trace_safe_str(iter, str, star, len),
3764 			      "fmt: '%s' current_buffer: '%s'",
3765 			      fmt, seq_buf_str(&iter->seq.seq))) {
3766 			int ret;
3767 
3768 			/* Try to safely read the string */
3769 			if (star) {
3770 				if (len + 1 > iter->fmt_size)
3771 					len = iter->fmt_size - 1;
3772 				if (len < 0)
3773 					len = 0;
3774 				ret = copy_from_kernel_nofault(iter->fmt, str, len);
3775 				iter->fmt[len] = 0;
3776 				star = false;
3777 			} else {
3778 				ret = strncpy_from_kernel_nofault(iter->fmt, str,
3779 								  iter->fmt_size);
3780 			}
3781 			if (ret < 0)
3782 				trace_seq_printf(&iter->seq, "(0x%px)", str);
3783 			else
3784 				trace_seq_printf(&iter->seq, "(0x%px:%s)",
3785 						 str, iter->fmt);
3786 			str = "[UNSAFE-MEMORY]";
3787 			strcpy(iter->fmt, "%s");
3788 		} else {
3789 			strncpy(iter->fmt, p + i, j + 1);
3790 			iter->fmt[j+1] = '\0';
3791 		}
3792 		if (star)
3793 			trace_seq_printf(&iter->seq, iter->fmt, len, str);
3794 		else
3795 			trace_seq_printf(&iter->seq, iter->fmt, str);
3796 
3797 		p += i + j + 1;
3798 	}
3799  print:
3800 	if (*p)
3801 		trace_seq_vprintf(&iter->seq, p, ap);
3802 }
3803 
3804 const char *trace_event_format(struct trace_iterator *iter, const char *fmt)
3805 {
3806 	const char *p, *new_fmt;
3807 	char *q;
3808 
3809 	if (WARN_ON_ONCE(!fmt))
3810 		return fmt;
3811 
3812 	if (!iter->tr || iter->tr->trace_flags & TRACE_ITER_HASH_PTR)
3813 		return fmt;
3814 
3815 	p = fmt;
3816 	new_fmt = q = iter->fmt;
3817 	while (*p) {
3818 		if (unlikely(q - new_fmt + 3 > iter->fmt_size)) {
3819 			if (!trace_iter_expand_format(iter))
3820 				return fmt;
3821 
3822 			q += iter->fmt - new_fmt;
3823 			new_fmt = iter->fmt;
3824 		}
3825 
3826 		*q++ = *p++;
3827 
3828 		/* Replace %p with %px */
3829 		if (p[-1] == '%') {
3830 			if (p[0] == '%') {
3831 				*q++ = *p++;
3832 			} else if (p[0] == 'p' && !isalnum(p[1])) {
3833 				*q++ = *p++;
3834 				*q++ = 'x';
3835 			}
3836 		}
3837 	}
3838 	*q = '\0';
3839 
3840 	return new_fmt;
3841 }
3842 
3843 #define STATIC_TEMP_BUF_SIZE	128
3844 static char static_temp_buf[STATIC_TEMP_BUF_SIZE] __aligned(4);
3845 
3846 /* Find the next real entry, without updating the iterator itself */
3847 struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
3848 					  int *ent_cpu, u64 *ent_ts)
3849 {
3850 	/* __find_next_entry will reset ent_size */
3851 	int ent_size = iter->ent_size;
3852 	struct trace_entry *entry;
3853 
3854 	/*
3855 	 * If called from ftrace_dump(), then the iter->temp buffer
3856 	 * will be the static_temp_buf and not created from kmalloc.
3857 	 * If the entry size is greater than the buffer, we can
3858 	 * not save it. Just return NULL in that case. This is only
3859 	 * used to add markers when two consecutive events' time
3860 	 * stamps have a large delta. See trace_print_lat_context()
3861 	 */
3862 	if (iter->temp == static_temp_buf &&
3863 	    STATIC_TEMP_BUF_SIZE < ent_size)
3864 		return NULL;
3865 
3866 	/*
3867 	 * The __find_next_entry() may call peek_next_entry(), which may
3868 	 * call ring_buffer_peek() that may make the contents of iter->ent
3869 	 * undefined. Need to copy iter->ent now.
3870 	 */
3871 	if (iter->ent && iter->ent != iter->temp) {
3872 		if ((!iter->temp || iter->temp_size < iter->ent_size) &&
3873 		    !WARN_ON_ONCE(iter->temp == static_temp_buf)) {
3874 			void *temp;
3875 			temp = kmalloc(iter->ent_size, GFP_KERNEL);
3876 			if (!temp)
3877 				return NULL;
3878 			kfree(iter->temp);
3879 			iter->temp = temp;
3880 			iter->temp_size = iter->ent_size;
3881 		}
3882 		memcpy(iter->temp, iter->ent, iter->ent_size);
3883 		iter->ent = iter->temp;
3884 	}
3885 	entry = __find_next_entry(iter, ent_cpu, NULL, ent_ts);
3886 	/* Put back the original ent_size */
3887 	iter->ent_size = ent_size;
3888 
3889 	return entry;
3890 }
3891 
3892 /* Find the next real entry, and increment the iterator to the next entry */
3893 void *trace_find_next_entry_inc(struct trace_iterator *iter)
3894 {
3895 	iter->ent = __find_next_entry(iter, &iter->cpu,
3896 				      &iter->lost_events, &iter->ts);
3897 
3898 	if (iter->ent)
3899 		trace_iterator_increment(iter);
3900 
3901 	return iter->ent ? iter : NULL;
3902 }
3903 
3904 static void trace_consume(struct trace_iterator *iter)
3905 {
3906 	ring_buffer_consume(iter->array_buffer->buffer, iter->cpu, &iter->ts,
3907 			    &iter->lost_events);
3908 }
3909 
3910 static void *s_next(struct seq_file *m, void *v, loff_t *pos)
3911 {
3912 	struct trace_iterator *iter = m->private;
3913 	int i = (int)*pos;
3914 	void *ent;
3915 
3916 	WARN_ON_ONCE(iter->leftover);
3917 
3918 	(*pos)++;
3919 
3920 	/* can't go backwards */
3921 	if (iter->idx > i)
3922 		return NULL;
3923 
3924 	if (iter->idx < 0)
3925 		ent = trace_find_next_entry_inc(iter);
3926 	else
3927 		ent = iter;
3928 
3929 	while (ent && iter->idx < i)
3930 		ent = trace_find_next_entry_inc(iter);
3931 
3932 	iter->pos = *pos;
3933 
3934 	return ent;
3935 }
3936 
3937 void tracing_iter_reset(struct trace_iterator *iter, int cpu)
3938 {
3939 	struct ring_buffer_iter *buf_iter;
3940 	unsigned long entries = 0;
3941 	u64 ts;
3942 
3943 	per_cpu_ptr(iter->array_buffer->data, cpu)->skipped_entries = 0;
3944 
3945 	buf_iter = trace_buffer_iter(iter, cpu);
3946 	if (!buf_iter)
3947 		return;
3948 
3949 	ring_buffer_iter_reset(buf_iter);
3950 
3951 	/*
3952 	 * We could have the case with the max latency tracers
3953 	 * that a reset never took place on a cpu. This is evident
3954 	 * by the timestamp being before the start of the buffer.
3955 	 */
3956 	while (ring_buffer_iter_peek(buf_iter, &ts)) {
3957 		if (ts >= iter->array_buffer->time_start)
3958 			break;
3959 		entries++;
3960 		ring_buffer_iter_advance(buf_iter);
3961 	}
3962 
3963 	per_cpu_ptr(iter->array_buffer->data, cpu)->skipped_entries = entries;
3964 }
3965 
3966 /*
3967  * The current tracer is copied to avoid a global locking
3968  * all around.
3969  */
3970 static void *s_start(struct seq_file *m, loff_t *pos)
3971 {
3972 	struct trace_iterator *iter = m->private;
3973 	struct trace_array *tr = iter->tr;
3974 	int cpu_file = iter->cpu_file;
3975 	void *p = NULL;
3976 	loff_t l = 0;
3977 	int cpu;
3978 
3979 	mutex_lock(&trace_types_lock);
3980 	if (unlikely(tr->current_trace != iter->trace)) {
3981 		/* Close iter->trace before switching to the new current tracer */
3982 		if (iter->trace->close)
3983 			iter->trace->close(iter);
3984 		iter->trace = tr->current_trace;
3985 		/* Reopen the new current tracer */
3986 		if (iter->trace->open)
3987 			iter->trace->open(iter);
3988 	}
3989 	mutex_unlock(&trace_types_lock);
3990 
3991 #ifdef CONFIG_TRACER_MAX_TRACE
3992 	if (iter->snapshot && iter->trace->use_max_tr)
3993 		return ERR_PTR(-EBUSY);
3994 #endif
3995 
3996 	if (*pos != iter->pos) {
3997 		iter->ent = NULL;
3998 		iter->cpu = 0;
3999 		iter->idx = -1;
4000 
4001 		if (cpu_file == RING_BUFFER_ALL_CPUS) {
4002 			for_each_tracing_cpu(cpu)
4003 				tracing_iter_reset(iter, cpu);
4004 		} else
4005 			tracing_iter_reset(iter, cpu_file);
4006 
4007 		iter->leftover = 0;
4008 		for (p = iter; p && l < *pos; p = s_next(m, p, &l))
4009 			;
4010 
4011 	} else {
4012 		/*
4013 		 * If we overflowed the seq_file before, then we want
4014 		 * to just reuse the trace_seq buffer again.
4015 		 */
4016 		if (iter->leftover)
4017 			p = iter;
4018 		else {
4019 			l = *pos - 1;
4020 			p = s_next(m, p, &l);
4021 		}
4022 	}
4023 
4024 	trace_event_read_lock();
4025 	trace_access_lock(cpu_file);
4026 	return p;
4027 }
4028 
4029 static void s_stop(struct seq_file *m, void *p)
4030 {
4031 	struct trace_iterator *iter = m->private;
4032 
4033 #ifdef CONFIG_TRACER_MAX_TRACE
4034 	if (iter->snapshot && iter->trace->use_max_tr)
4035 		return;
4036 #endif
4037 
4038 	trace_access_unlock(iter->cpu_file);
4039 	trace_event_read_unlock();
4040 }
4041 
4042 static void
4043 get_total_entries_cpu(struct array_buffer *buf, unsigned long *total,
4044 		      unsigned long *entries, int cpu)
4045 {
4046 	unsigned long count;
4047 
4048 	count = ring_buffer_entries_cpu(buf->buffer, cpu);
4049 	/*
4050 	 * If this buffer has skipped entries, then we hold all
4051 	 * entries for the trace and we need to ignore the
4052 	 * ones before the time stamp.
4053 	 */
4054 	if (per_cpu_ptr(buf->data, cpu)->skipped_entries) {
4055 		count -= per_cpu_ptr(buf->data, cpu)->skipped_entries;
4056 		/* total is the same as the entries */
4057 		*total = count;
4058 	} else
4059 		*total = count +
4060 			ring_buffer_overrun_cpu(buf->buffer, cpu);
4061 	*entries = count;
4062 }
4063 
4064 static void
4065 get_total_entries(struct array_buffer *buf,
4066 		  unsigned long *total, unsigned long *entries)
4067 {
4068 	unsigned long t, e;
4069 	int cpu;
4070 
4071 	*total = 0;
4072 	*entries = 0;
4073 
4074 	for_each_tracing_cpu(cpu) {
4075 		get_total_entries_cpu(buf, &t, &e, cpu);
4076 		*total += t;
4077 		*entries += e;
4078 	}
4079 }
4080 
4081 unsigned long trace_total_entries_cpu(struct trace_array *tr, int cpu)
4082 {
4083 	unsigned long total, entries;
4084 
4085 	if (!tr)
4086 		tr = &global_trace;
4087 
4088 	get_total_entries_cpu(&tr->array_buffer, &total, &entries, cpu);
4089 
4090 	return entries;
4091 }
4092 
4093 unsigned long trace_total_entries(struct trace_array *tr)
4094 {
4095 	unsigned long total, entries;
4096 
4097 	if (!tr)
4098 		tr = &global_trace;
4099 
4100 	get_total_entries(&tr->array_buffer, &total, &entries);
4101 
4102 	return entries;
4103 }
4104 
4105 static void print_lat_help_header(struct seq_file *m)
4106 {
4107 	seq_puts(m, "#                    _------=> CPU#            \n"
4108 		    "#                   / _-----=> irqs-off/BH-disabled\n"
4109 		    "#                  | / _----=> need-resched    \n"
4110 		    "#                  || / _---=> hardirq/softirq \n"
4111 		    "#                  ||| / _--=> preempt-depth   \n"
4112 		    "#                  |||| / _-=> migrate-disable \n"
4113 		    "#                  ||||| /     delay           \n"
4114 		    "#  cmd     pid     |||||| time  |   caller     \n"
4115 		    "#     \\   /        ||||||  \\    |    /       \n");
4116 }
4117 
4118 static void print_event_info(struct array_buffer *buf, struct seq_file *m)
4119 {
4120 	unsigned long total;
4121 	unsigned long entries;
4122 
4123 	get_total_entries(buf, &total, &entries);
4124 	seq_printf(m, "# entries-in-buffer/entries-written: %lu/%lu   #P:%d\n",
4125 		   entries, total, num_online_cpus());
4126 	seq_puts(m, "#\n");
4127 }
4128 
4129 static void print_func_help_header(struct array_buffer *buf, struct seq_file *m,
4130 				   unsigned int flags)
4131 {
4132 	bool tgid = flags & TRACE_ITER_RECORD_TGID;
4133 
4134 	print_event_info(buf, m);
4135 
4136 	seq_printf(m, "#           TASK-PID    %s CPU#     TIMESTAMP  FUNCTION\n", tgid ? "   TGID   " : "");
4137 	seq_printf(m, "#              | |      %s   |         |         |\n",      tgid ? "     |    " : "");
4138 }
4139 
4140 static void print_func_help_header_irq(struct array_buffer *buf, struct seq_file *m,
4141 				       unsigned int flags)
4142 {
4143 	bool tgid = flags & TRACE_ITER_RECORD_TGID;
4144 	static const char space[] = "            ";
4145 	int prec = tgid ? 12 : 2;
4146 
4147 	print_event_info(buf, m);
4148 
4149 	seq_printf(m, "#                            %.*s  _-----=> irqs-off/BH-disabled\n", prec, space);
4150 	seq_printf(m, "#                            %.*s / _----=> need-resched\n", prec, space);
4151 	seq_printf(m, "#                            %.*s| / _---=> hardirq/softirq\n", prec, space);
4152 	seq_printf(m, "#                            %.*s|| / _--=> preempt-depth\n", prec, space);
4153 	seq_printf(m, "#                            %.*s||| / _-=> migrate-disable\n", prec, space);
4154 	seq_printf(m, "#                            %.*s|||| /     delay\n", prec, space);
4155 	seq_printf(m, "#           TASK-PID  %.*s CPU#  |||||  TIMESTAMP  FUNCTION\n", prec, "     TGID   ");
4156 	seq_printf(m, "#              | |    %.*s   |   |||||     |         |\n", prec, "       |    ");
4157 }
4158 
4159 void
4160 print_trace_header(struct seq_file *m, struct trace_iterator *iter)
4161 {
4162 	unsigned long sym_flags = (global_trace.trace_flags & TRACE_ITER_SYM_MASK);
4163 	struct array_buffer *buf = iter->array_buffer;
4164 	struct trace_array_cpu *data = per_cpu_ptr(buf->data, buf->cpu);
4165 	struct tracer *type = iter->trace;
4166 	unsigned long entries;
4167 	unsigned long total;
4168 	const char *name = type->name;
4169 
4170 	get_total_entries(buf, &total, &entries);
4171 
4172 	seq_printf(m, "# %s latency trace v1.1.5 on %s\n",
4173 		   name, init_utsname()->release);
4174 	seq_puts(m, "# -----------------------------------"
4175 		 "---------------------------------\n");
4176 	seq_printf(m, "# latency: %lu us, #%lu/%lu, CPU#%d |"
4177 		   " (M:%s VP:%d, KP:%d, SP:%d HP:%d",
4178 		   nsecs_to_usecs(data->saved_latency),
4179 		   entries,
4180 		   total,
4181 		   buf->cpu,
4182 		   preempt_model_none()      ? "server" :
4183 		   preempt_model_voluntary() ? "desktop" :
4184 		   preempt_model_full()      ? "preempt" :
4185 		   preempt_model_rt()        ? "preempt_rt" :
4186 		   "unknown",
4187 		   /* These are reserved for later use */
4188 		   0, 0, 0, 0);
4189 #ifdef CONFIG_SMP
4190 	seq_printf(m, " #P:%d)\n", num_online_cpus());
4191 #else
4192 	seq_puts(m, ")\n");
4193 #endif
4194 	seq_puts(m, "#    -----------------\n");
4195 	seq_printf(m, "#    | task: %.16s-%d "
4196 		   "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n",
4197 		   data->comm, data->pid,
4198 		   from_kuid_munged(seq_user_ns(m), data->uid), data->nice,
4199 		   data->policy, data->rt_priority);
4200 	seq_puts(m, "#    -----------------\n");
4201 
4202 	if (data->critical_start) {
4203 		seq_puts(m, "#  => started at: ");
4204 		seq_print_ip_sym(&iter->seq, data->critical_start, sym_flags);
4205 		trace_print_seq(m, &iter->seq);
4206 		seq_puts(m, "\n#  => ended at:   ");
4207 		seq_print_ip_sym(&iter->seq, data->critical_end, sym_flags);
4208 		trace_print_seq(m, &iter->seq);
4209 		seq_puts(m, "\n#\n");
4210 	}
4211 
4212 	seq_puts(m, "#\n");
4213 }
4214 
4215 static void test_cpu_buff_start(struct trace_iterator *iter)
4216 {
4217 	struct trace_seq *s = &iter->seq;
4218 	struct trace_array *tr = iter->tr;
4219 
4220 	if (!(tr->trace_flags & TRACE_ITER_ANNOTATE))
4221 		return;
4222 
4223 	if (!(iter->iter_flags & TRACE_FILE_ANNOTATE))
4224 		return;
4225 
4226 	if (cpumask_available(iter->started) &&
4227 	    cpumask_test_cpu(iter->cpu, iter->started))
4228 		return;
4229 
4230 	if (per_cpu_ptr(iter->array_buffer->data, iter->cpu)->skipped_entries)
4231 		return;
4232 
4233 	if (cpumask_available(iter->started))
4234 		cpumask_set_cpu(iter->cpu, iter->started);
4235 
4236 	/* Don't print started cpu buffer for the first entry of the trace */
4237 	if (iter->idx > 1)
4238 		trace_seq_printf(s, "##### CPU %u buffer started ####\n",
4239 				iter->cpu);
4240 }
4241 
4242 static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
4243 {
4244 	struct trace_array *tr = iter->tr;
4245 	struct trace_seq *s = &iter->seq;
4246 	unsigned long sym_flags = (tr->trace_flags & TRACE_ITER_SYM_MASK);
4247 	struct trace_entry *entry;
4248 	struct trace_event *event;
4249 
4250 	entry = iter->ent;
4251 
4252 	test_cpu_buff_start(iter);
4253 
4254 	event = ftrace_find_event(entry->type);
4255 
4256 	if (tr->trace_flags & TRACE_ITER_CONTEXT_INFO) {
4257 		if (iter->iter_flags & TRACE_FILE_LAT_FMT)
4258 			trace_print_lat_context(iter);
4259 		else
4260 			trace_print_context(iter);
4261 	}
4262 
4263 	if (trace_seq_has_overflowed(s))
4264 		return TRACE_TYPE_PARTIAL_LINE;
4265 
4266 	if (event) {
4267 		if (tr->trace_flags & TRACE_ITER_FIELDS)
4268 			return print_event_fields(iter, event);
4269 		return event->funcs->trace(iter, sym_flags, event);
4270 	}
4271 
4272 	trace_seq_printf(s, "Unknown type %d\n", entry->type);
4273 
4274 	return trace_handle_return(s);
4275 }
4276 
4277 static enum print_line_t print_raw_fmt(struct trace_iterator *iter)
4278 {
4279 	struct trace_array *tr = iter->tr;
4280 	struct trace_seq *s = &iter->seq;
4281 	struct trace_entry *entry;
4282 	struct trace_event *event;
4283 
4284 	entry = iter->ent;
4285 
4286 	if (tr->trace_flags & TRACE_ITER_CONTEXT_INFO)
4287 		trace_seq_printf(s, "%d %d %llu ",
4288 				 entry->pid, iter->cpu, iter->ts);
4289 
4290 	if (trace_seq_has_overflowed(s))
4291 		return TRACE_TYPE_PARTIAL_LINE;
4292 
4293 	event = ftrace_find_event(entry->type);
4294 	if (event)
4295 		return event->funcs->raw(iter, 0, event);
4296 
4297 	trace_seq_printf(s, "%d ?\n", entry->type);
4298 
4299 	return trace_handle_return(s);
4300 }
4301 
4302 static enum print_line_t print_hex_fmt(struct trace_iterator *iter)
4303 {
4304 	struct trace_array *tr = iter->tr;
4305 	struct trace_seq *s = &iter->seq;
4306 	unsigned char newline = '\n';
4307 	struct trace_entry *entry;
4308 	struct trace_event *event;
4309 
4310 	entry = iter->ent;
4311 
4312 	if (tr->trace_flags & TRACE_ITER_CONTEXT_INFO) {
4313 		SEQ_PUT_HEX_FIELD(s, entry->pid);
4314 		SEQ_PUT_HEX_FIELD(s, iter->cpu);
4315 		SEQ_PUT_HEX_FIELD(s, iter->ts);
4316 		if (trace_seq_has_overflowed(s))
4317 			return TRACE_TYPE_PARTIAL_LINE;
4318 	}
4319 
4320 	event = ftrace_find_event(entry->type);
4321 	if (event) {
4322 		enum print_line_t ret = event->funcs->hex(iter, 0, event);
4323 		if (ret != TRACE_TYPE_HANDLED)
4324 			return ret;
4325 	}
4326 
4327 	SEQ_PUT_FIELD(s, newline);
4328 
4329 	return trace_handle_return(s);
4330 }
4331 
4332 static enum print_line_t print_bin_fmt(struct trace_iterator *iter)
4333 {
4334 	struct trace_array *tr = iter->tr;
4335 	struct trace_seq *s = &iter->seq;
4336 	struct trace_entry *entry;
4337 	struct trace_event *event;
4338 
4339 	entry = iter->ent;
4340 
4341 	if (tr->trace_flags & TRACE_ITER_CONTEXT_INFO) {
4342 		SEQ_PUT_FIELD(s, entry->pid);
4343 		SEQ_PUT_FIELD(s, iter->cpu);
4344 		SEQ_PUT_FIELD(s, iter->ts);
4345 		if (trace_seq_has_overflowed(s))
4346 			return TRACE_TYPE_PARTIAL_LINE;
4347 	}
4348 
4349 	event = ftrace_find_event(entry->type);
4350 	return event ? event->funcs->binary(iter, 0, event) :
4351 		TRACE_TYPE_HANDLED;
4352 }
4353 
4354 int trace_empty(struct trace_iterator *iter)
4355 {
4356 	struct ring_buffer_iter *buf_iter;
4357 	int cpu;
4358 
4359 	/* If we are looking at one CPU buffer, only check that one */
4360 	if (iter->cpu_file != RING_BUFFER_ALL_CPUS) {
4361 		cpu = iter->cpu_file;
4362 		buf_iter = trace_buffer_iter(iter, cpu);
4363 		if (buf_iter) {
4364 			if (!ring_buffer_iter_empty(buf_iter))
4365 				return 0;
4366 		} else {
4367 			if (!ring_buffer_empty_cpu(iter->array_buffer->buffer, cpu))
4368 				return 0;
4369 		}
4370 		return 1;
4371 	}
4372 
4373 	for_each_tracing_cpu(cpu) {
4374 		buf_iter = trace_buffer_iter(iter, cpu);
4375 		if (buf_iter) {
4376 			if (!ring_buffer_iter_empty(buf_iter))
4377 				return 0;
4378 		} else {
4379 			if (!ring_buffer_empty_cpu(iter->array_buffer->buffer, cpu))
4380 				return 0;
4381 		}
4382 	}
4383 
4384 	return 1;
4385 }
4386 
4387 /*  Called with trace_event_read_lock() held. */
4388 enum print_line_t print_trace_line(struct trace_iterator *iter)
4389 {
4390 	struct trace_array *tr = iter->tr;
4391 	unsigned long trace_flags = tr->trace_flags;
4392 	enum print_line_t ret;
4393 
4394 	if (iter->lost_events) {
4395 		if (iter->lost_events == (unsigned long)-1)
4396 			trace_seq_printf(&iter->seq, "CPU:%d [LOST EVENTS]\n",
4397 					 iter->cpu);
4398 		else
4399 			trace_seq_printf(&iter->seq, "CPU:%d [LOST %lu EVENTS]\n",
4400 					 iter->cpu, iter->lost_events);
4401 		if (trace_seq_has_overflowed(&iter->seq))
4402 			return TRACE_TYPE_PARTIAL_LINE;
4403 	}
4404 
4405 	if (iter->trace && iter->trace->print_line) {
4406 		ret = iter->trace->print_line(iter);
4407 		if (ret != TRACE_TYPE_UNHANDLED)
4408 			return ret;
4409 	}
4410 
4411 	if (iter->ent->type == TRACE_BPUTS &&
4412 			trace_flags & TRACE_ITER_PRINTK &&
4413 			trace_flags & TRACE_ITER_PRINTK_MSGONLY)
4414 		return trace_print_bputs_msg_only(iter);
4415 
4416 	if (iter->ent->type == TRACE_BPRINT &&
4417 			trace_flags & TRACE_ITER_PRINTK &&
4418 			trace_flags & TRACE_ITER_PRINTK_MSGONLY)
4419 		return trace_print_bprintk_msg_only(iter);
4420 
4421 	if (iter->ent->type == TRACE_PRINT &&
4422 			trace_flags & TRACE_ITER_PRINTK &&
4423 			trace_flags & TRACE_ITER_PRINTK_MSGONLY)
4424 		return trace_print_printk_msg_only(iter);
4425 
4426 	if (trace_flags & TRACE_ITER_BIN)
4427 		return print_bin_fmt(iter);
4428 
4429 	if (trace_flags & TRACE_ITER_HEX)
4430 		return print_hex_fmt(iter);
4431 
4432 	if (trace_flags & TRACE_ITER_RAW)
4433 		return print_raw_fmt(iter);
4434 
4435 	return print_trace_fmt(iter);
4436 }
4437 
4438 void trace_latency_header(struct seq_file *m)
4439 {
4440 	struct trace_iterator *iter = m->private;
4441 	struct trace_array *tr = iter->tr;
4442 
4443 	/* print nothing if the buffers are empty */
4444 	if (trace_empty(iter))
4445 		return;
4446 
4447 	if (iter->iter_flags & TRACE_FILE_LAT_FMT)
4448 		print_trace_header(m, iter);
4449 
4450 	if (!(tr->trace_flags & TRACE_ITER_VERBOSE))
4451 		print_lat_help_header(m);
4452 }
4453 
4454 void trace_default_header(struct seq_file *m)
4455 {
4456 	struct trace_iterator *iter = m->private;
4457 	struct trace_array *tr = iter->tr;
4458 	unsigned long trace_flags = tr->trace_flags;
4459 
4460 	if (!(trace_flags & TRACE_ITER_CONTEXT_INFO))
4461 		return;
4462 
4463 	if (iter->iter_flags & TRACE_FILE_LAT_FMT) {
4464 		/* print nothing if the buffers are empty */
4465 		if (trace_empty(iter))
4466 			return;
4467 		print_trace_header(m, iter);
4468 		if (!(trace_flags & TRACE_ITER_VERBOSE))
4469 			print_lat_help_header(m);
4470 	} else {
4471 		if (!(trace_flags & TRACE_ITER_VERBOSE)) {
4472 			if (trace_flags & TRACE_ITER_IRQ_INFO)
4473 				print_func_help_header_irq(iter->array_buffer,
4474 							   m, trace_flags);
4475 			else
4476 				print_func_help_header(iter->array_buffer, m,
4477 						       trace_flags);
4478 		}
4479 	}
4480 }
4481 
4482 static void test_ftrace_alive(struct seq_file *m)
4483 {
4484 	if (!ftrace_is_dead())
4485 		return;
4486 	seq_puts(m, "# WARNING: FUNCTION TRACING IS CORRUPTED\n"
4487 		    "#          MAY BE MISSING FUNCTION EVENTS\n");
4488 }
4489 
4490 #ifdef CONFIG_TRACER_MAX_TRACE
4491 static void show_snapshot_main_help(struct seq_file *m)
4492 {
4493 	seq_puts(m, "# echo 0 > snapshot : Clears and frees snapshot buffer\n"
4494 		    "# echo 1 > snapshot : Allocates snapshot buffer, if not already allocated.\n"
4495 		    "#                      Takes a snapshot of the main buffer.\n"
4496 		    "# echo 2 > snapshot : Clears snapshot buffer (but does not allocate or free)\n"
4497 		    "#                      (Doesn't have to be '2' works with any number that\n"
4498 		    "#                       is not a '0' or '1')\n");
4499 }
4500 
4501 static void show_snapshot_percpu_help(struct seq_file *m)
4502 {
4503 	seq_puts(m, "# echo 0 > snapshot : Invalid for per_cpu snapshot file.\n");
4504 #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP
4505 	seq_puts(m, "# echo 1 > snapshot : Allocates snapshot buffer, if not already allocated.\n"
4506 		    "#                      Takes a snapshot of the main buffer for this cpu.\n");
4507 #else
4508 	seq_puts(m, "# echo 1 > snapshot : Not supported with this kernel.\n"
4509 		    "#                     Must use main snapshot file to allocate.\n");
4510 #endif
4511 	seq_puts(m, "# echo 2 > snapshot : Clears this cpu's snapshot buffer (but does not allocate)\n"
4512 		    "#                      (Doesn't have to be '2' works with any number that\n"
4513 		    "#                       is not a '0' or '1')\n");
4514 }
4515 
4516 static void print_snapshot_help(struct seq_file *m, struct trace_iterator *iter)
4517 {
4518 	if (iter->tr->allocated_snapshot)
4519 		seq_puts(m, "#\n# * Snapshot is allocated *\n#\n");
4520 	else
4521 		seq_puts(m, "#\n# * Snapshot is freed *\n#\n");
4522 
4523 	seq_puts(m, "# Snapshot commands:\n");
4524 	if (iter->cpu_file == RING_BUFFER_ALL_CPUS)
4525 		show_snapshot_main_help(m);
4526 	else
4527 		show_snapshot_percpu_help(m);
4528 }
4529 #else
4530 /* Should never be called */
4531 static inline void print_snapshot_help(struct seq_file *m, struct trace_iterator *iter) { }
4532 #endif
4533 
4534 static int s_show(struct seq_file *m, void *v)
4535 {
4536 	struct trace_iterator *iter = v;
4537 	int ret;
4538 
4539 	if (iter->ent == NULL) {
4540 		if (iter->tr) {
4541 			seq_printf(m, "# tracer: %s\n", iter->trace->name);
4542 			seq_puts(m, "#\n");
4543 			test_ftrace_alive(m);
4544 		}
4545 		if (iter->snapshot && trace_empty(iter))
4546 			print_snapshot_help(m, iter);
4547 		else if (iter->trace && iter->trace->print_header)
4548 			iter->trace->print_header(m);
4549 		else
4550 			trace_default_header(m);
4551 
4552 	} else if (iter->leftover) {
4553 		/*
4554 		 * If we filled the seq_file buffer earlier, we
4555 		 * want to just show it now.
4556 		 */
4557 		ret = trace_print_seq(m, &iter->seq);
4558 
4559 		/* ret should this time be zero, but you never know */
4560 		iter->leftover = ret;
4561 
4562 	} else {
4563 		ret = print_trace_line(iter);
4564 		if (ret == TRACE_TYPE_PARTIAL_LINE) {
4565 			iter->seq.full = 0;
4566 			trace_seq_puts(&iter->seq, "[LINE TOO BIG]\n");
4567 		}
4568 		ret = trace_print_seq(m, &iter->seq);
4569 		/*
4570 		 * If we overflow the seq_file buffer, then it will
4571 		 * ask us for this data again at start up.
4572 		 * Use that instead.
4573 		 *  ret is 0 if seq_file write succeeded.
4574 		 *        -1 otherwise.
4575 		 */
4576 		iter->leftover = ret;
4577 	}
4578 
4579 	return 0;
4580 }
4581 
4582 /*
4583  * Should be used after trace_array_get(), trace_types_lock
4584  * ensures that i_cdev was already initialized.
4585  */
4586 static inline int tracing_get_cpu(struct inode *inode)
4587 {
4588 	if (inode->i_cdev) /* See trace_create_cpu_file() */
4589 		return (long)inode->i_cdev - 1;
4590 	return RING_BUFFER_ALL_CPUS;
4591 }
4592 
4593 static const struct seq_operations tracer_seq_ops = {
4594 	.start		= s_start,
4595 	.next		= s_next,
4596 	.stop		= s_stop,
4597 	.show		= s_show,
4598 };
4599 
4600 /*
4601  * Note, as iter itself can be allocated and freed in different
4602  * ways, this function is only used to free its content, and not
4603  * the iterator itself. The only requirement to all the allocations
4604  * is that it must zero all fields (kzalloc), as freeing works with
4605  * ethier allocated content or NULL.
4606  */
4607 static void free_trace_iter_content(struct trace_iterator *iter)
4608 {
4609 	/* The fmt is either NULL, allocated or points to static_fmt_buf */
4610 	if (iter->fmt != static_fmt_buf)
4611 		kfree(iter->fmt);
4612 
4613 	kfree(iter->temp);
4614 	kfree(iter->buffer_iter);
4615 	mutex_destroy(&iter->mutex);
4616 	free_cpumask_var(iter->started);
4617 }
4618 
4619 static struct trace_iterator *
4620 __tracing_open(struct inode *inode, struct file *file, bool snapshot)
4621 {
4622 	struct trace_array *tr = inode->i_private;
4623 	struct trace_iterator *iter;
4624 	int cpu;
4625 
4626 	if (tracing_disabled)
4627 		return ERR_PTR(-ENODEV);
4628 
4629 	iter = __seq_open_private(file, &tracer_seq_ops, sizeof(*iter));
4630 	if (!iter)
4631 		return ERR_PTR(-ENOMEM);
4632 
4633 	iter->buffer_iter = kcalloc(nr_cpu_ids, sizeof(*iter->buffer_iter),
4634 				    GFP_KERNEL);
4635 	if (!iter->buffer_iter)
4636 		goto release;
4637 
4638 	/*
4639 	 * trace_find_next_entry() may need to save off iter->ent.
4640 	 * It will place it into the iter->temp buffer. As most
4641 	 * events are less than 128, allocate a buffer of that size.
4642 	 * If one is greater, then trace_find_next_entry() will
4643 	 * allocate a new buffer to adjust for the bigger iter->ent.
4644 	 * It's not critical if it fails to get allocated here.
4645 	 */
4646 	iter->temp = kmalloc(128, GFP_KERNEL);
4647 	if (iter->temp)
4648 		iter->temp_size = 128;
4649 
4650 	/*
4651 	 * trace_event_printf() may need to modify given format
4652 	 * string to replace %p with %px so that it shows real address
4653 	 * instead of hash value. However, that is only for the event
4654 	 * tracing, other tracer may not need. Defer the allocation
4655 	 * until it is needed.
4656 	 */
4657 	iter->fmt = NULL;
4658 	iter->fmt_size = 0;
4659 
4660 	mutex_lock(&trace_types_lock);
4661 	iter->trace = tr->current_trace;
4662 
4663 	if (!zalloc_cpumask_var(&iter->started, GFP_KERNEL))
4664 		goto fail;
4665 
4666 	iter->tr = tr;
4667 
4668 #ifdef CONFIG_TRACER_MAX_TRACE
4669 	/* Currently only the top directory has a snapshot */
4670 	if (tr->current_trace->print_max || snapshot)
4671 		iter->array_buffer = &tr->max_buffer;
4672 	else
4673 #endif
4674 		iter->array_buffer = &tr->array_buffer;
4675 	iter->snapshot = snapshot;
4676 	iter->pos = -1;
4677 	iter->cpu_file = tracing_get_cpu(inode);
4678 	mutex_init(&iter->mutex);
4679 
4680 	/* Notify the tracer early; before we stop tracing. */
4681 	if (iter->trace->open)
4682 		iter->trace->open(iter);
4683 
4684 	/* Annotate start of buffers if we had overruns */
4685 	if (ring_buffer_overruns(iter->array_buffer->buffer))
4686 		iter->iter_flags |= TRACE_FILE_ANNOTATE;
4687 
4688 	/* Output in nanoseconds only if we are using a clock in nanoseconds. */
4689 	if (trace_clocks[tr->clock_id].in_ns)
4690 		iter->iter_flags |= TRACE_FILE_TIME_IN_NS;
4691 
4692 	/*
4693 	 * If pause-on-trace is enabled, then stop the trace while
4694 	 * dumping, unless this is the "snapshot" file
4695 	 */
4696 	if (!iter->snapshot && (tr->trace_flags & TRACE_ITER_PAUSE_ON_TRACE))
4697 		tracing_stop_tr(tr);
4698 
4699 	if (iter->cpu_file == RING_BUFFER_ALL_CPUS) {
4700 		for_each_tracing_cpu(cpu) {
4701 			iter->buffer_iter[cpu] =
4702 				ring_buffer_read_prepare(iter->array_buffer->buffer,
4703 							 cpu, GFP_KERNEL);
4704 		}
4705 		ring_buffer_read_prepare_sync();
4706 		for_each_tracing_cpu(cpu) {
4707 			ring_buffer_read_start(iter->buffer_iter[cpu]);
4708 			tracing_iter_reset(iter, cpu);
4709 		}
4710 	} else {
4711 		cpu = iter->cpu_file;
4712 		iter->buffer_iter[cpu] =
4713 			ring_buffer_read_prepare(iter->array_buffer->buffer,
4714 						 cpu, GFP_KERNEL);
4715 		ring_buffer_read_prepare_sync();
4716 		ring_buffer_read_start(iter->buffer_iter[cpu]);
4717 		tracing_iter_reset(iter, cpu);
4718 	}
4719 
4720 	mutex_unlock(&trace_types_lock);
4721 
4722 	return iter;
4723 
4724  fail:
4725 	mutex_unlock(&trace_types_lock);
4726 	free_trace_iter_content(iter);
4727 release:
4728 	seq_release_private(inode, file);
4729 	return ERR_PTR(-ENOMEM);
4730 }
4731 
4732 int tracing_open_generic(struct inode *inode, struct file *filp)
4733 {
4734 	int ret;
4735 
4736 	ret = tracing_check_open_get_tr(NULL);
4737 	if (ret)
4738 		return ret;
4739 
4740 	filp->private_data = inode->i_private;
4741 	return 0;
4742 }
4743 
4744 bool tracing_is_disabled(void)
4745 {
4746 	return (tracing_disabled) ? true: false;
4747 }
4748 
4749 /*
4750  * Open and update trace_array ref count.
4751  * Must have the current trace_array passed to it.
4752  */
4753 int tracing_open_generic_tr(struct inode *inode, struct file *filp)
4754 {
4755 	struct trace_array *tr = inode->i_private;
4756 	int ret;
4757 
4758 	ret = tracing_check_open_get_tr(tr);
4759 	if (ret)
4760 		return ret;
4761 
4762 	filp->private_data = inode->i_private;
4763 
4764 	return 0;
4765 }
4766 
4767 /*
4768  * The private pointer of the inode is the trace_event_file.
4769  * Update the tr ref count associated to it.
4770  */
4771 int tracing_open_file_tr(struct inode *inode, struct file *filp)
4772 {
4773 	struct trace_event_file *file = inode->i_private;
4774 	int ret;
4775 
4776 	ret = tracing_check_open_get_tr(file->tr);
4777 	if (ret)
4778 		return ret;
4779 
4780 	mutex_lock(&event_mutex);
4781 
4782 	/* Fail if the file is marked for removal */
4783 	if (file->flags & EVENT_FILE_FL_FREED) {
4784 		trace_array_put(file->tr);
4785 		ret = -ENODEV;
4786 	} else {
4787 		event_file_get(file);
4788 	}
4789 
4790 	mutex_unlock(&event_mutex);
4791 	if (ret)
4792 		return ret;
4793 
4794 	filp->private_data = inode->i_private;
4795 
4796 	return 0;
4797 }
4798 
4799 int tracing_release_file_tr(struct inode *inode, struct file *filp)
4800 {
4801 	struct trace_event_file *file = inode->i_private;
4802 
4803 	trace_array_put(file->tr);
4804 	event_file_put(file);
4805 
4806 	return 0;
4807 }
4808 
4809 int tracing_single_release_file_tr(struct inode *inode, struct file *filp)
4810 {
4811 	tracing_release_file_tr(inode, filp);
4812 	return single_release(inode, filp);
4813 }
4814 
4815 static int tracing_mark_open(struct inode *inode, struct file *filp)
4816 {
4817 	stream_open(inode, filp);
4818 	return tracing_open_generic_tr(inode, filp);
4819 }
4820 
4821 static int tracing_release(struct inode *inode, struct file *file)
4822 {
4823 	struct trace_array *tr = inode->i_private;
4824 	struct seq_file *m = file->private_data;
4825 	struct trace_iterator *iter;
4826 	int cpu;
4827 
4828 	if (!(file->f_mode & FMODE_READ)) {
4829 		trace_array_put(tr);
4830 		return 0;
4831 	}
4832 
4833 	/* Writes do not use seq_file */
4834 	iter = m->private;
4835 	mutex_lock(&trace_types_lock);
4836 
4837 	for_each_tracing_cpu(cpu) {
4838 		if (iter->buffer_iter[cpu])
4839 			ring_buffer_read_finish(iter->buffer_iter[cpu]);
4840 	}
4841 
4842 	if (iter->trace && iter->trace->close)
4843 		iter->trace->close(iter);
4844 
4845 	if (!iter->snapshot && tr->stop_count)
4846 		/* reenable tracing if it was previously enabled */
4847 		tracing_start_tr(tr);
4848 
4849 	__trace_array_put(tr);
4850 
4851 	mutex_unlock(&trace_types_lock);
4852 
4853 	free_trace_iter_content(iter);
4854 	seq_release_private(inode, file);
4855 
4856 	return 0;
4857 }
4858 
4859 int tracing_release_generic_tr(struct inode *inode, struct file *file)
4860 {
4861 	struct trace_array *tr = inode->i_private;
4862 
4863 	trace_array_put(tr);
4864 	return 0;
4865 }
4866 
4867 static int tracing_single_release_tr(struct inode *inode, struct file *file)
4868 {
4869 	struct trace_array *tr = inode->i_private;
4870 
4871 	trace_array_put(tr);
4872 
4873 	return single_release(inode, file);
4874 }
4875 
4876 static int tracing_open(struct inode *inode, struct file *file)
4877 {
4878 	struct trace_array *tr = inode->i_private;
4879 	struct trace_iterator *iter;
4880 	int ret;
4881 
4882 	ret = tracing_check_open_get_tr(tr);
4883 	if (ret)
4884 		return ret;
4885 
4886 	/* If this file was open for write, then erase contents */
4887 	if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC)) {
4888 		int cpu = tracing_get_cpu(inode);
4889 		struct array_buffer *trace_buf = &tr->array_buffer;
4890 
4891 #ifdef CONFIG_TRACER_MAX_TRACE
4892 		if (tr->current_trace->print_max)
4893 			trace_buf = &tr->max_buffer;
4894 #endif
4895 
4896 		if (cpu == RING_BUFFER_ALL_CPUS)
4897 			tracing_reset_online_cpus(trace_buf);
4898 		else
4899 			tracing_reset_cpu(trace_buf, cpu);
4900 	}
4901 
4902 	if (file->f_mode & FMODE_READ) {
4903 		iter = __tracing_open(inode, file, false);
4904 		if (IS_ERR(iter))
4905 			ret = PTR_ERR(iter);
4906 		else if (tr->trace_flags & TRACE_ITER_LATENCY_FMT)
4907 			iter->iter_flags |= TRACE_FILE_LAT_FMT;
4908 	}
4909 
4910 	if (ret < 0)
4911 		trace_array_put(tr);
4912 
4913 	return ret;
4914 }
4915 
4916 /*
4917  * Some tracers are not suitable for instance buffers.
4918  * A tracer is always available for the global array (toplevel)
4919  * or if it explicitly states that it is.
4920  */
4921 static bool
4922 trace_ok_for_array(struct tracer *t, struct trace_array *tr)
4923 {
4924 	return (tr->flags & TRACE_ARRAY_FL_GLOBAL) || t->allow_instances;
4925 }
4926 
4927 /* Find the next tracer that this trace array may use */
4928 static struct tracer *
4929 get_tracer_for_array(struct trace_array *tr, struct tracer *t)
4930 {
4931 	while (t && !trace_ok_for_array(t, tr))
4932 		t = t->next;
4933 
4934 	return t;
4935 }
4936 
4937 static void *
4938 t_next(struct seq_file *m, void *v, loff_t *pos)
4939 {
4940 	struct trace_array *tr = m->private;
4941 	struct tracer *t = v;
4942 
4943 	(*pos)++;
4944 
4945 	if (t)
4946 		t = get_tracer_for_array(tr, t->next);
4947 
4948 	return t;
4949 }
4950 
4951 static void *t_start(struct seq_file *m, loff_t *pos)
4952 {
4953 	struct trace_array *tr = m->private;
4954 	struct tracer *t;
4955 	loff_t l = 0;
4956 
4957 	mutex_lock(&trace_types_lock);
4958 
4959 	t = get_tracer_for_array(tr, trace_types);
4960 	for (; t && l < *pos; t = t_next(m, t, &l))
4961 			;
4962 
4963 	return t;
4964 }
4965 
4966 static void t_stop(struct seq_file *m, void *p)
4967 {
4968 	mutex_unlock(&trace_types_lock);
4969 }
4970 
4971 static int t_show(struct seq_file *m, void *v)
4972 {
4973 	struct tracer *t = v;
4974 
4975 	if (!t)
4976 		return 0;
4977 
4978 	seq_puts(m, t->name);
4979 	if (t->next)
4980 		seq_putc(m, ' ');
4981 	else
4982 		seq_putc(m, '\n');
4983 
4984 	return 0;
4985 }
4986 
4987 static const struct seq_operations show_traces_seq_ops = {
4988 	.start		= t_start,
4989 	.next		= t_next,
4990 	.stop		= t_stop,
4991 	.show		= t_show,
4992 };
4993 
4994 static int show_traces_open(struct inode *inode, struct file *file)
4995 {
4996 	struct trace_array *tr = inode->i_private;
4997 	struct seq_file *m;
4998 	int ret;
4999 
5000 	ret = tracing_check_open_get_tr(tr);
5001 	if (ret)
5002 		return ret;
5003 
5004 	ret = seq_open(file, &show_traces_seq_ops);
5005 	if (ret) {
5006 		trace_array_put(tr);
5007 		return ret;
5008 	}
5009 
5010 	m = file->private_data;
5011 	m->private = tr;
5012 
5013 	return 0;
5014 }
5015 
5016 static int show_traces_release(struct inode *inode, struct file *file)
5017 {
5018 	struct trace_array *tr = inode->i_private;
5019 
5020 	trace_array_put(tr);
5021 	return seq_release(inode, file);
5022 }
5023 
5024 static ssize_t
5025 tracing_write_stub(struct file *filp, const char __user *ubuf,
5026 		   size_t count, loff_t *ppos)
5027 {
5028 	return count;
5029 }
5030 
5031 loff_t tracing_lseek(struct file *file, loff_t offset, int whence)
5032 {
5033 	int ret;
5034 
5035 	if (file->f_mode & FMODE_READ)
5036 		ret = seq_lseek(file, offset, whence);
5037 	else
5038 		file->f_pos = ret = 0;
5039 
5040 	return ret;
5041 }
5042 
5043 static const struct file_operations tracing_fops = {
5044 	.open		= tracing_open,
5045 	.read		= seq_read,
5046 	.read_iter	= seq_read_iter,
5047 	.splice_read	= copy_splice_read,
5048 	.write		= tracing_write_stub,
5049 	.llseek		= tracing_lseek,
5050 	.release	= tracing_release,
5051 };
5052 
5053 static const struct file_operations show_traces_fops = {
5054 	.open		= show_traces_open,
5055 	.read		= seq_read,
5056 	.llseek		= seq_lseek,
5057 	.release	= show_traces_release,
5058 };
5059 
5060 static ssize_t
5061 tracing_cpumask_read(struct file *filp, char __user *ubuf,
5062 		     size_t count, loff_t *ppos)
5063 {
5064 	struct trace_array *tr = file_inode(filp)->i_private;
5065 	char *mask_str;
5066 	int len;
5067 
5068 	len = snprintf(NULL, 0, "%*pb\n",
5069 		       cpumask_pr_args(tr->tracing_cpumask)) + 1;
5070 	mask_str = kmalloc(len, GFP_KERNEL);
5071 	if (!mask_str)
5072 		return -ENOMEM;
5073 
5074 	len = snprintf(mask_str, len, "%*pb\n",
5075 		       cpumask_pr_args(tr->tracing_cpumask));
5076 	if (len >= count) {
5077 		count = -EINVAL;
5078 		goto out_err;
5079 	}
5080 	count = simple_read_from_buffer(ubuf, count, ppos, mask_str, len);
5081 
5082 out_err:
5083 	kfree(mask_str);
5084 
5085 	return count;
5086 }
5087 
5088 int tracing_set_cpumask(struct trace_array *tr,
5089 			cpumask_var_t tracing_cpumask_new)
5090 {
5091 	int cpu;
5092 
5093 	if (!tr)
5094 		return -EINVAL;
5095 
5096 	local_irq_disable();
5097 	arch_spin_lock(&tr->max_lock);
5098 	for_each_tracing_cpu(cpu) {
5099 		/*
5100 		 * Increase/decrease the disabled counter if we are
5101 		 * about to flip a bit in the cpumask:
5102 		 */
5103 		if (cpumask_test_cpu(cpu, tr->tracing_cpumask) &&
5104 				!cpumask_test_cpu(cpu, tracing_cpumask_new)) {
5105 			atomic_inc(&per_cpu_ptr(tr->array_buffer.data, cpu)->disabled);
5106 			ring_buffer_record_disable_cpu(tr->array_buffer.buffer, cpu);
5107 #ifdef CONFIG_TRACER_MAX_TRACE
5108 			ring_buffer_record_disable_cpu(tr->max_buffer.buffer, cpu);
5109 #endif
5110 		}
5111 		if (!cpumask_test_cpu(cpu, tr->tracing_cpumask) &&
5112 				cpumask_test_cpu(cpu, tracing_cpumask_new)) {
5113 			atomic_dec(&per_cpu_ptr(tr->array_buffer.data, cpu)->disabled);
5114 			ring_buffer_record_enable_cpu(tr->array_buffer.buffer, cpu);
5115 #ifdef CONFIG_TRACER_MAX_TRACE
5116 			ring_buffer_record_enable_cpu(tr->max_buffer.buffer, cpu);
5117 #endif
5118 		}
5119 	}
5120 	arch_spin_unlock(&tr->max_lock);
5121 	local_irq_enable();
5122 
5123 	cpumask_copy(tr->tracing_cpumask, tracing_cpumask_new);
5124 
5125 	return 0;
5126 }
5127 
5128 static ssize_t
5129 tracing_cpumask_write(struct file *filp, const char __user *ubuf,
5130 		      size_t count, loff_t *ppos)
5131 {
5132 	struct trace_array *tr = file_inode(filp)->i_private;
5133 	cpumask_var_t tracing_cpumask_new;
5134 	int err;
5135 
5136 	if (!zalloc_cpumask_var(&tracing_cpumask_new, GFP_KERNEL))
5137 		return -ENOMEM;
5138 
5139 	err = cpumask_parse_user(ubuf, count, tracing_cpumask_new);
5140 	if (err)
5141 		goto err_free;
5142 
5143 	err = tracing_set_cpumask(tr, tracing_cpumask_new);
5144 	if (err)
5145 		goto err_free;
5146 
5147 	free_cpumask_var(tracing_cpumask_new);
5148 
5149 	return count;
5150 
5151 err_free:
5152 	free_cpumask_var(tracing_cpumask_new);
5153 
5154 	return err;
5155 }
5156 
5157 static const struct file_operations tracing_cpumask_fops = {
5158 	.open		= tracing_open_generic_tr,
5159 	.read		= tracing_cpumask_read,
5160 	.write		= tracing_cpumask_write,
5161 	.release	= tracing_release_generic_tr,
5162 	.llseek		= generic_file_llseek,
5163 };
5164 
5165 static int tracing_trace_options_show(struct seq_file *m, void *v)
5166 {
5167 	struct tracer_opt *trace_opts;
5168 	struct trace_array *tr = m->private;
5169 	u32 tracer_flags;
5170 	int i;
5171 
5172 	mutex_lock(&trace_types_lock);
5173 	tracer_flags = tr->current_trace->flags->val;
5174 	trace_opts = tr->current_trace->flags->opts;
5175 
5176 	for (i = 0; trace_options[i]; i++) {
5177 		if (tr->trace_flags & (1 << i))
5178 			seq_printf(m, "%s\n", trace_options[i]);
5179 		else
5180 			seq_printf(m, "no%s\n", trace_options[i]);
5181 	}
5182 
5183 	for (i = 0; trace_opts[i].name; i++) {
5184 		if (tracer_flags & trace_opts[i].bit)
5185 			seq_printf(m, "%s\n", trace_opts[i].name);
5186 		else
5187 			seq_printf(m, "no%s\n", trace_opts[i].name);
5188 	}
5189 	mutex_unlock(&trace_types_lock);
5190 
5191 	return 0;
5192 }
5193 
5194 static int __set_tracer_option(struct trace_array *tr,
5195 			       struct tracer_flags *tracer_flags,
5196 			       struct tracer_opt *opts, int neg)
5197 {
5198 	struct tracer *trace = tracer_flags->trace;
5199 	int ret;
5200 
5201 	ret = trace->set_flag(tr, tracer_flags->val, opts->bit, !neg);
5202 	if (ret)
5203 		return ret;
5204 
5205 	if (neg)
5206 		tracer_flags->val &= ~opts->bit;
5207 	else
5208 		tracer_flags->val |= opts->bit;
5209 	return 0;
5210 }
5211 
5212 /* Try to assign a tracer specific option */
5213 static int set_tracer_option(struct trace_array *tr, char *cmp, int neg)
5214 {
5215 	struct tracer *trace = tr->current_trace;
5216 	struct tracer_flags *tracer_flags = trace->flags;
5217 	struct tracer_opt *opts = NULL;
5218 	int i;
5219 
5220 	for (i = 0; tracer_flags->opts[i].name; i++) {
5221 		opts = &tracer_flags->opts[i];
5222 
5223 		if (strcmp(cmp, opts->name) == 0)
5224 			return __set_tracer_option(tr, trace->flags, opts, neg);
5225 	}
5226 
5227 	return -EINVAL;
5228 }
5229 
5230 /* Some tracers require overwrite to stay enabled */
5231 int trace_keep_overwrite(struct tracer *tracer, u32 mask, int set)
5232 {
5233 	if (tracer->enabled && (mask & TRACE_ITER_OVERWRITE) && !set)
5234 		return -1;
5235 
5236 	return 0;
5237 }
5238 
5239 int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled)
5240 {
5241 	if ((mask == TRACE_ITER_RECORD_TGID) ||
5242 	    (mask == TRACE_ITER_RECORD_CMD))
5243 		lockdep_assert_held(&event_mutex);
5244 
5245 	/* do nothing if flag is already set */
5246 	if (!!(tr->trace_flags & mask) == !!enabled)
5247 		return 0;
5248 
5249 	/* Give the tracer a chance to approve the change */
5250 	if (tr->current_trace->flag_changed)
5251 		if (tr->current_trace->flag_changed(tr, mask, !!enabled))
5252 			return -EINVAL;
5253 
5254 	if (enabled)
5255 		tr->trace_flags |= mask;
5256 	else
5257 		tr->trace_flags &= ~mask;
5258 
5259 	if (mask == TRACE_ITER_RECORD_CMD)
5260 		trace_event_enable_cmd_record(enabled);
5261 
5262 	if (mask == TRACE_ITER_RECORD_TGID) {
5263 
5264 		if (trace_alloc_tgid_map() < 0) {
5265 			tr->trace_flags &= ~TRACE_ITER_RECORD_TGID;
5266 			return -ENOMEM;
5267 		}
5268 
5269 		trace_event_enable_tgid_record(enabled);
5270 	}
5271 
5272 	if (mask == TRACE_ITER_EVENT_FORK)
5273 		trace_event_follow_fork(tr, enabled);
5274 
5275 	if (mask == TRACE_ITER_FUNC_FORK)
5276 		ftrace_pid_follow_fork(tr, enabled);
5277 
5278 	if (mask == TRACE_ITER_OVERWRITE) {
5279 		ring_buffer_change_overwrite(tr->array_buffer.buffer, enabled);
5280 #ifdef CONFIG_TRACER_MAX_TRACE
5281 		ring_buffer_change_overwrite(tr->max_buffer.buffer, enabled);
5282 #endif
5283 	}
5284 
5285 	if (mask == TRACE_ITER_PRINTK) {
5286 		trace_printk_start_stop_comm(enabled);
5287 		trace_printk_control(enabled);
5288 	}
5289 
5290 	return 0;
5291 }
5292 
5293 int trace_set_options(struct trace_array *tr, char *option)
5294 {
5295 	char *cmp;
5296 	int neg = 0;
5297 	int ret;
5298 	size_t orig_len = strlen(option);
5299 	int len;
5300 
5301 	cmp = strstrip(option);
5302 
5303 	len = str_has_prefix(cmp, "no");
5304 	if (len)
5305 		neg = 1;
5306 
5307 	cmp += len;
5308 
5309 	mutex_lock(&event_mutex);
5310 	mutex_lock(&trace_types_lock);
5311 
5312 	ret = match_string(trace_options, -1, cmp);
5313 	/* If no option could be set, test the specific tracer options */
5314 	if (ret < 0)
5315 		ret = set_tracer_option(tr, cmp, neg);
5316 	else
5317 		ret = set_tracer_flag(tr, 1 << ret, !neg);
5318 
5319 	mutex_unlock(&trace_types_lock);
5320 	mutex_unlock(&event_mutex);
5321 
5322 	/*
5323 	 * If the first trailing whitespace is replaced with '\0' by strstrip,
5324 	 * turn it back into a space.
5325 	 */
5326 	if (orig_len > strlen(option))
5327 		option[strlen(option)] = ' ';
5328 
5329 	return ret;
5330 }
5331 
5332 static void __init apply_trace_boot_options(void)
5333 {
5334 	char *buf = trace_boot_options_buf;
5335 	char *option;
5336 
5337 	while (true) {
5338 		option = strsep(&buf, ",");
5339 
5340 		if (!option)
5341 			break;
5342 
5343 		if (*option)
5344 			trace_set_options(&global_trace, option);
5345 
5346 		/* Put back the comma to allow this to be called again */
5347 		if (buf)
5348 			*(buf - 1) = ',';
5349 	}
5350 }
5351 
5352 static ssize_t
5353 tracing_trace_options_write(struct file *filp, const char __user *ubuf,
5354 			size_t cnt, loff_t *ppos)
5355 {
5356 	struct seq_file *m = filp->private_data;
5357 	struct trace_array *tr = m->private;
5358 	char buf[64];
5359 	int ret;
5360 
5361 	if (cnt >= sizeof(buf))
5362 		return -EINVAL;
5363 
5364 	if (copy_from_user(buf, ubuf, cnt))
5365 		return -EFAULT;
5366 
5367 	buf[cnt] = 0;
5368 
5369 	ret = trace_set_options(tr, buf);
5370 	if (ret < 0)
5371 		return ret;
5372 
5373 	*ppos += cnt;
5374 
5375 	return cnt;
5376 }
5377 
5378 static int tracing_trace_options_open(struct inode *inode, struct file *file)
5379 {
5380 	struct trace_array *tr = inode->i_private;
5381 	int ret;
5382 
5383 	ret = tracing_check_open_get_tr(tr);
5384 	if (ret)
5385 		return ret;
5386 
5387 	ret = single_open(file, tracing_trace_options_show, inode->i_private);
5388 	if (ret < 0)
5389 		trace_array_put(tr);
5390 
5391 	return ret;
5392 }
5393 
5394 static const struct file_operations tracing_iter_fops = {
5395 	.open		= tracing_trace_options_open,
5396 	.read		= seq_read,
5397 	.llseek		= seq_lseek,
5398 	.release	= tracing_single_release_tr,
5399 	.write		= tracing_trace_options_write,
5400 };
5401 
5402 static const char readme_msg[] =
5403 	"tracing mini-HOWTO:\n\n"
5404 	"# echo 0 > tracing_on : quick way to disable tracing\n"
5405 	"# echo 1 > tracing_on : quick way to re-enable tracing\n\n"
5406 	" Important files:\n"
5407 	"  trace\t\t\t- The static contents of the buffer\n"
5408 	"\t\t\t  To clear the buffer write into this file: echo > trace\n"
5409 	"  trace_pipe\t\t- A consuming read to see the contents of the buffer\n"
5410 	"  current_tracer\t- function and latency tracers\n"
5411 	"  available_tracers\t- list of configured tracers for current_tracer\n"
5412 	"  error_log\t- error log for failed commands (that support it)\n"
5413 	"  buffer_size_kb\t- view and modify size of per cpu buffer\n"
5414 	"  buffer_total_size_kb  - view total size of all cpu buffers\n\n"
5415 	"  trace_clock\t\t- change the clock used to order events\n"
5416 	"       local:   Per cpu clock but may not be synced across CPUs\n"
5417 	"      global:   Synced across CPUs but slows tracing down.\n"
5418 	"     counter:   Not a clock, but just an increment\n"
5419 	"      uptime:   Jiffy counter from time of boot\n"
5420 	"        perf:   Same clock that perf events use\n"
5421 #ifdef CONFIG_X86_64
5422 	"     x86-tsc:   TSC cycle counter\n"
5423 #endif
5424 	"\n  timestamp_mode\t- view the mode used to timestamp events\n"
5425 	"       delta:   Delta difference against a buffer-wide timestamp\n"
5426 	"    absolute:   Absolute (standalone) timestamp\n"
5427 	"\n  trace_marker\t\t- Writes into this file writes into the kernel buffer\n"
5428 	"\n  trace_marker_raw\t\t- Writes into this file writes binary data into the kernel buffer\n"
5429 	"  tracing_cpumask\t- Limit which CPUs to trace\n"
5430 	"  instances\t\t- Make sub-buffers with: mkdir instances/foo\n"
5431 	"\t\t\t  Remove sub-buffer with rmdir\n"
5432 	"  trace_options\t\t- Set format or modify how tracing happens\n"
5433 	"\t\t\t  Disable an option by prefixing 'no' to the\n"
5434 	"\t\t\t  option name\n"
5435 	"  saved_cmdlines_size\t- echo command number in here to store comm-pid list\n"
5436 #ifdef CONFIG_DYNAMIC_FTRACE
5437 	"\n  available_filter_functions - list of functions that can be filtered on\n"
5438 	"  set_ftrace_filter\t- echo function name in here to only trace these\n"
5439 	"\t\t\t  functions\n"
5440 	"\t     accepts: func_full_name or glob-matching-pattern\n"
5441 	"\t     modules: Can select a group via module\n"
5442 	"\t      Format: :mod:<module-name>\n"
5443 	"\t     example: echo :mod:ext3 > set_ftrace_filter\n"
5444 	"\t    triggers: a command to perform when function is hit\n"
5445 	"\t      Format: <function>:<trigger>[:count]\n"
5446 	"\t     trigger: traceon, traceoff\n"
5447 	"\t\t      enable_event:<system>:<event>\n"
5448 	"\t\t      disable_event:<system>:<event>\n"
5449 #ifdef CONFIG_STACKTRACE
5450 	"\t\t      stacktrace\n"
5451 #endif
5452 #ifdef CONFIG_TRACER_SNAPSHOT
5453 	"\t\t      snapshot\n"
5454 #endif
5455 	"\t\t      dump\n"
5456 	"\t\t      cpudump\n"
5457 	"\t     example: echo do_fault:traceoff > set_ftrace_filter\n"
5458 	"\t              echo do_trap:traceoff:3 > set_ftrace_filter\n"
5459 	"\t     The first one will disable tracing every time do_fault is hit\n"
5460 	"\t     The second will disable tracing at most 3 times when do_trap is hit\n"
5461 	"\t       The first time do trap is hit and it disables tracing, the\n"
5462 	"\t       counter will decrement to 2. If tracing is already disabled,\n"
5463 	"\t       the counter will not decrement. It only decrements when the\n"
5464 	"\t       trigger did work\n"
5465 	"\t     To remove trigger without count:\n"
5466 	"\t       echo '!<function>:<trigger> > set_ftrace_filter\n"
5467 	"\t     To remove trigger with a count:\n"
5468 	"\t       echo '!<function>:<trigger>:0 > set_ftrace_filter\n"
5469 	"  set_ftrace_notrace\t- echo function name in here to never trace.\n"
5470 	"\t    accepts: func_full_name, *func_end, func_begin*, *func_middle*\n"
5471 	"\t    modules: Can select a group via module command :mod:\n"
5472 	"\t    Does not accept triggers\n"
5473 #endif /* CONFIG_DYNAMIC_FTRACE */
5474 #ifdef CONFIG_FUNCTION_TRACER
5475 	"  set_ftrace_pid\t- Write pid(s) to only function trace those pids\n"
5476 	"\t\t    (function)\n"
5477 	"  set_ftrace_notrace_pid\t- Write pid(s) to not function trace those pids\n"
5478 	"\t\t    (function)\n"
5479 #endif
5480 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
5481 	"  set_graph_function\t- Trace the nested calls of a function (function_graph)\n"
5482 	"  set_graph_notrace\t- Do not trace the nested calls of a function (function_graph)\n"
5483 	"  max_graph_depth\t- Trace a limited depth of nested calls (0 is unlimited)\n"
5484 #endif
5485 #ifdef CONFIG_TRACER_SNAPSHOT
5486 	"\n  snapshot\t\t- Like 'trace' but shows the content of the static\n"
5487 	"\t\t\t  snapshot buffer. Read the contents for more\n"
5488 	"\t\t\t  information\n"
5489 #endif
5490 #ifdef CONFIG_STACK_TRACER
5491 	"  stack_trace\t\t- Shows the max stack trace when active\n"
5492 	"  stack_max_size\t- Shows current max stack size that was traced\n"
5493 	"\t\t\t  Write into this file to reset the max size (trigger a\n"
5494 	"\t\t\t  new trace)\n"
5495 #ifdef CONFIG_DYNAMIC_FTRACE
5496 	"  stack_trace_filter\t- Like set_ftrace_filter but limits what stack_trace\n"
5497 	"\t\t\t  traces\n"
5498 #endif
5499 #endif /* CONFIG_STACK_TRACER */
5500 #ifdef CONFIG_DYNAMIC_EVENTS
5501 	"  dynamic_events\t\t- Create/append/remove/show the generic dynamic events\n"
5502 	"\t\t\t  Write into this file to define/undefine new trace events.\n"
5503 #endif
5504 #ifdef CONFIG_KPROBE_EVENTS
5505 	"  kprobe_events\t\t- Create/append/remove/show the kernel dynamic events\n"
5506 	"\t\t\t  Write into this file to define/undefine new trace events.\n"
5507 #endif
5508 #ifdef CONFIG_UPROBE_EVENTS
5509 	"  uprobe_events\t\t- Create/append/remove/show the userspace dynamic events\n"
5510 	"\t\t\t  Write into this file to define/undefine new trace events.\n"
5511 #endif
5512 #if defined(CONFIG_KPROBE_EVENTS) || defined(CONFIG_UPROBE_EVENTS) || \
5513     defined(CONFIG_FPROBE_EVENTS)
5514 	"\t  accepts: event-definitions (one definition per line)\n"
5515 #if defined(CONFIG_KPROBE_EVENTS) || defined(CONFIG_UPROBE_EVENTS)
5516 	"\t   Format: p[:[<group>/][<event>]] <place> [<args>]\n"
5517 	"\t           r[maxactive][:[<group>/][<event>]] <place> [<args>]\n"
5518 #endif
5519 #ifdef CONFIG_FPROBE_EVENTS
5520 	"\t           f[:[<group>/][<event>]] <func-name>[%return] [<args>]\n"
5521 	"\t           t[:[<group>/][<event>]] <tracepoint> [<args>]\n"
5522 #endif
5523 #ifdef CONFIG_HIST_TRIGGERS
5524 	"\t           s:[synthetic/]<event> <field> [<field>]\n"
5525 #endif
5526 	"\t           e[:[<group>/][<event>]] <attached-group>.<attached-event> [<args>] [if <filter>]\n"
5527 	"\t           -:[<group>/][<event>]\n"
5528 #ifdef CONFIG_KPROBE_EVENTS
5529 	"\t    place: [<module>:]<symbol>[+<offset>]|<memaddr>\n"
5530   "place (kretprobe): [<module>:]<symbol>[+<offset>]%return|<memaddr>\n"
5531 #endif
5532 #ifdef CONFIG_UPROBE_EVENTS
5533   "   place (uprobe): <path>:<offset>[%return][(ref_ctr_offset)]\n"
5534 #endif
5535 	"\t     args: <name>=fetcharg[:type]\n"
5536 	"\t fetcharg: (%<register>|$<efield>), @<address>, @<symbol>[+|-<offset>],\n"
5537 #ifdef CONFIG_HAVE_FUNCTION_ARG_ACCESS_API
5538 	"\t           $stack<index>, $stack, $retval, $comm, $arg<N>,\n"
5539 #ifdef CONFIG_PROBE_EVENTS_BTF_ARGS
5540 	"\t           <argname>[->field[->field|.field...]],\n"
5541 #endif
5542 #else
5543 	"\t           $stack<index>, $stack, $retval, $comm,\n"
5544 #endif
5545 	"\t           +|-[u]<offset>(<fetcharg>), \\imm-value, \\\"imm-string\"\n"
5546 	"\t     kernel return probes support: $retval, $arg<N>, $comm\n"
5547 	"\t     type: s8/16/32/64, u8/16/32/64, x8/16/32/64, char, string, symbol,\n"
5548 	"\t           b<bit-width>@<bit-offset>/<container-size>, ustring,\n"
5549 	"\t           symstr, %pd/%pD, <type>\\[<array-size>\\]\n"
5550 #ifdef CONFIG_HIST_TRIGGERS
5551 	"\t    field: <stype> <name>;\n"
5552 	"\t    stype: u8/u16/u32/u64, s8/s16/s32/s64, pid_t,\n"
5553 	"\t           [unsigned] char/int/long\n"
5554 #endif
5555 	"\t    efield: For event probes ('e' types), the field is on of the fields\n"
5556 	"\t            of the <attached-group>/<attached-event>.\n"
5557 #endif
5558 	"  events/\t\t- Directory containing all trace event subsystems:\n"
5559 	"      enable\t\t- Write 0/1 to enable/disable tracing of all events\n"
5560 	"  events/<system>/\t- Directory containing all trace events for <system>:\n"
5561 	"      enable\t\t- Write 0/1 to enable/disable tracing of all <system>\n"
5562 	"\t\t\t  events\n"
5563 	"      filter\t\t- If set, only events passing filter are traced\n"
5564 	"  events/<system>/<event>/\t- Directory containing control files for\n"
5565 	"\t\t\t  <event>:\n"
5566 	"      enable\t\t- Write 0/1 to enable/disable tracing of <event>\n"
5567 	"      filter\t\t- If set, only events passing filter are traced\n"
5568 	"      trigger\t\t- If set, a command to perform when event is hit\n"
5569 	"\t    Format: <trigger>[:count][if <filter>]\n"
5570 	"\t   trigger: traceon, traceoff\n"
5571 	"\t            enable_event:<system>:<event>\n"
5572 	"\t            disable_event:<system>:<event>\n"
5573 #ifdef CONFIG_HIST_TRIGGERS
5574 	"\t            enable_hist:<system>:<event>\n"
5575 	"\t            disable_hist:<system>:<event>\n"
5576 #endif
5577 #ifdef CONFIG_STACKTRACE
5578 	"\t\t    stacktrace\n"
5579 #endif
5580 #ifdef CONFIG_TRACER_SNAPSHOT
5581 	"\t\t    snapshot\n"
5582 #endif
5583 #ifdef CONFIG_HIST_TRIGGERS
5584 	"\t\t    hist (see below)\n"
5585 #endif
5586 	"\t   example: echo traceoff > events/block/block_unplug/trigger\n"
5587 	"\t            echo traceoff:3 > events/block/block_unplug/trigger\n"
5588 	"\t            echo 'enable_event:kmem:kmalloc:3 if nr_rq > 1' > \\\n"
5589 	"\t                  events/block/block_unplug/trigger\n"
5590 	"\t   The first disables tracing every time block_unplug is hit.\n"
5591 	"\t   The second disables tracing the first 3 times block_unplug is hit.\n"
5592 	"\t   The third enables the kmalloc event the first 3 times block_unplug\n"
5593 	"\t     is hit and has value of greater than 1 for the 'nr_rq' event field.\n"
5594 	"\t   Like function triggers, the counter is only decremented if it\n"
5595 	"\t    enabled or disabled tracing.\n"
5596 	"\t   To remove a trigger without a count:\n"
5597 	"\t     echo '!<trigger> > <system>/<event>/trigger\n"
5598 	"\t   To remove a trigger with a count:\n"
5599 	"\t     echo '!<trigger>:0 > <system>/<event>/trigger\n"
5600 	"\t   Filters can be ignored when removing a trigger.\n"
5601 #ifdef CONFIG_HIST_TRIGGERS
5602 	"      hist trigger\t- If set, event hits are aggregated into a hash table\n"
5603 	"\t    Format: hist:keys=<field1[,field2,...]>\n"
5604 	"\t            [:<var1>=<field|var_ref|numeric_literal>[,<var2>=...]]\n"
5605 	"\t            [:values=<field1[,field2,...]>]\n"
5606 	"\t            [:sort=<field1[,field2,...]>]\n"
5607 	"\t            [:size=#entries]\n"
5608 	"\t            [:pause][:continue][:clear]\n"
5609 	"\t            [:name=histname1]\n"
5610 	"\t            [:nohitcount]\n"
5611 	"\t            [:<handler>.<action>]\n"
5612 	"\t            [if <filter>]\n\n"
5613 	"\t    Note, special fields can be used as well:\n"
5614 	"\t            common_timestamp - to record current timestamp\n"
5615 	"\t            common_cpu - to record the CPU the event happened on\n"
5616 	"\n"
5617 	"\t    A hist trigger variable can be:\n"
5618 	"\t        - a reference to a field e.g. x=current_timestamp,\n"
5619 	"\t        - a reference to another variable e.g. y=$x,\n"
5620 	"\t        - a numeric literal: e.g. ms_per_sec=1000,\n"
5621 	"\t        - an arithmetic expression: e.g. time_secs=current_timestamp/1000\n"
5622 	"\n"
5623 	"\t    hist trigger arithmetic expressions support addition(+), subtraction(-),\n"
5624 	"\t    multiplication(*) and division(/) operators. An operand can be either a\n"
5625 	"\t    variable reference, field or numeric literal.\n"
5626 	"\n"
5627 	"\t    When a matching event is hit, an entry is added to a hash\n"
5628 	"\t    table using the key(s) and value(s) named, and the value of a\n"
5629 	"\t    sum called 'hitcount' is incremented.  Keys and values\n"
5630 	"\t    correspond to fields in the event's format description.  Keys\n"
5631 	"\t    can be any field, or the special string 'common_stacktrace'.\n"
5632 	"\t    Compound keys consisting of up to two fields can be specified\n"
5633 	"\t    by the 'keys' keyword.  Values must correspond to numeric\n"
5634 	"\t    fields.  Sort keys consisting of up to two fields can be\n"
5635 	"\t    specified using the 'sort' keyword.  The sort direction can\n"
5636 	"\t    be modified by appending '.descending' or '.ascending' to a\n"
5637 	"\t    sort field.  The 'size' parameter can be used to specify more\n"
5638 	"\t    or fewer than the default 2048 entries for the hashtable size.\n"
5639 	"\t    If a hist trigger is given a name using the 'name' parameter,\n"
5640 	"\t    its histogram data will be shared with other triggers of the\n"
5641 	"\t    same name, and trigger hits will update this common data.\n\n"
5642 	"\t    Reading the 'hist' file for the event will dump the hash\n"
5643 	"\t    table in its entirety to stdout.  If there are multiple hist\n"
5644 	"\t    triggers attached to an event, there will be a table for each\n"
5645 	"\t    trigger in the output.  The table displayed for a named\n"
5646 	"\t    trigger will be the same as any other instance having the\n"
5647 	"\t    same name.  The default format used to display a given field\n"
5648 	"\t    can be modified by appending any of the following modifiers\n"
5649 	"\t    to the field name, as applicable:\n\n"
5650 	"\t            .hex        display a number as a hex value\n"
5651 	"\t            .sym        display an address as a symbol\n"
5652 	"\t            .sym-offset display an address as a symbol and offset\n"
5653 	"\t            .execname   display a common_pid as a program name\n"
5654 	"\t            .syscall    display a syscall id as a syscall name\n"
5655 	"\t            .log2       display log2 value rather than raw number\n"
5656 	"\t            .buckets=size  display values in groups of size rather than raw number\n"
5657 	"\t            .usecs      display a common_timestamp in microseconds\n"
5658 	"\t            .percent    display a number of percentage value\n"
5659 	"\t            .graph      display a bar-graph of a value\n\n"
5660 	"\t    The 'pause' parameter can be used to pause an existing hist\n"
5661 	"\t    trigger or to start a hist trigger but not log any events\n"
5662 	"\t    until told to do so.  'continue' can be used to start or\n"
5663 	"\t    restart a paused hist trigger.\n\n"
5664 	"\t    The 'clear' parameter will clear the contents of a running\n"
5665 	"\t    hist trigger and leave its current paused/active state\n"
5666 	"\t    unchanged.\n\n"
5667 	"\t    The 'nohitcount' (or NOHC) parameter will suppress display of\n"
5668 	"\t    raw hitcount in the histogram.\n\n"
5669 	"\t    The enable_hist and disable_hist triggers can be used to\n"
5670 	"\t    have one event conditionally start and stop another event's\n"
5671 	"\t    already-attached hist trigger.  The syntax is analogous to\n"
5672 	"\t    the enable_event and disable_event triggers.\n\n"
5673 	"\t    Hist trigger handlers and actions are executed whenever a\n"
5674 	"\t    a histogram entry is added or updated.  They take the form:\n\n"
5675 	"\t        <handler>.<action>\n\n"
5676 	"\t    The available handlers are:\n\n"
5677 	"\t        onmatch(matching.event)  - invoke on addition or update\n"
5678 	"\t        onmax(var)               - invoke if var exceeds current max\n"
5679 	"\t        onchange(var)            - invoke action if var changes\n\n"
5680 	"\t    The available actions are:\n\n"
5681 	"\t        trace(<synthetic_event>,param list)  - generate synthetic event\n"
5682 	"\t        save(field,...)                      - save current event fields\n"
5683 #ifdef CONFIG_TRACER_SNAPSHOT
5684 	"\t        snapshot()                           - snapshot the trace buffer\n\n"
5685 #endif
5686 #ifdef CONFIG_SYNTH_EVENTS
5687 	"  events/synthetic_events\t- Create/append/remove/show synthetic events\n"
5688 	"\t  Write into this file to define/undefine new synthetic events.\n"
5689 	"\t     example: echo 'myevent u64 lat; char name[]; long[] stack' >> synthetic_events\n"
5690 #endif
5691 #endif
5692 ;
5693 
5694 static ssize_t
5695 tracing_readme_read(struct file *filp, char __user *ubuf,
5696 		       size_t cnt, loff_t *ppos)
5697 {
5698 	return simple_read_from_buffer(ubuf, cnt, ppos,
5699 					readme_msg, strlen(readme_msg));
5700 }
5701 
5702 static const struct file_operations tracing_readme_fops = {
5703 	.open		= tracing_open_generic,
5704 	.read		= tracing_readme_read,
5705 	.llseek		= generic_file_llseek,
5706 };
5707 
5708 #ifdef CONFIG_TRACE_EVAL_MAP_FILE
5709 static union trace_eval_map_item *
5710 update_eval_map(union trace_eval_map_item *ptr)
5711 {
5712 	if (!ptr->map.eval_string) {
5713 		if (ptr->tail.next) {
5714 			ptr = ptr->tail.next;
5715 			/* Set ptr to the next real item (skip head) */
5716 			ptr++;
5717 		} else
5718 			return NULL;
5719 	}
5720 	return ptr;
5721 }
5722 
5723 static void *eval_map_next(struct seq_file *m, void *v, loff_t *pos)
5724 {
5725 	union trace_eval_map_item *ptr = v;
5726 
5727 	/*
5728 	 * Paranoid! If ptr points to end, we don't want to increment past it.
5729 	 * This really should never happen.
5730 	 */
5731 	(*pos)++;
5732 	ptr = update_eval_map(ptr);
5733 	if (WARN_ON_ONCE(!ptr))
5734 		return NULL;
5735 
5736 	ptr++;
5737 	ptr = update_eval_map(ptr);
5738 
5739 	return ptr;
5740 }
5741 
5742 static void *eval_map_start(struct seq_file *m, loff_t *pos)
5743 {
5744 	union trace_eval_map_item *v;
5745 	loff_t l = 0;
5746 
5747 	mutex_lock(&trace_eval_mutex);
5748 
5749 	v = trace_eval_maps;
5750 	if (v)
5751 		v++;
5752 
5753 	while (v && l < *pos) {
5754 		v = eval_map_next(m, v, &l);
5755 	}
5756 
5757 	return v;
5758 }
5759 
5760 static void eval_map_stop(struct seq_file *m, void *v)
5761 {
5762 	mutex_unlock(&trace_eval_mutex);
5763 }
5764 
5765 static int eval_map_show(struct seq_file *m, void *v)
5766 {
5767 	union trace_eval_map_item *ptr = v;
5768 
5769 	seq_printf(m, "%s %ld (%s)\n",
5770 		   ptr->map.eval_string, ptr->map.eval_value,
5771 		   ptr->map.system);
5772 
5773 	return 0;
5774 }
5775 
5776 static const struct seq_operations tracing_eval_map_seq_ops = {
5777 	.start		= eval_map_start,
5778 	.next		= eval_map_next,
5779 	.stop		= eval_map_stop,
5780 	.show		= eval_map_show,
5781 };
5782 
5783 static int tracing_eval_map_open(struct inode *inode, struct file *filp)
5784 {
5785 	int ret;
5786 
5787 	ret = tracing_check_open_get_tr(NULL);
5788 	if (ret)
5789 		return ret;
5790 
5791 	return seq_open(filp, &tracing_eval_map_seq_ops);
5792 }
5793 
5794 static const struct file_operations tracing_eval_map_fops = {
5795 	.open		= tracing_eval_map_open,
5796 	.read		= seq_read,
5797 	.llseek		= seq_lseek,
5798 	.release	= seq_release,
5799 };
5800 
5801 static inline union trace_eval_map_item *
5802 trace_eval_jmp_to_tail(union trace_eval_map_item *ptr)
5803 {
5804 	/* Return tail of array given the head */
5805 	return ptr + ptr->head.length + 1;
5806 }
5807 
5808 static void
5809 trace_insert_eval_map_file(struct module *mod, struct trace_eval_map **start,
5810 			   int len)
5811 {
5812 	struct trace_eval_map **stop;
5813 	struct trace_eval_map **map;
5814 	union trace_eval_map_item *map_array;
5815 	union trace_eval_map_item *ptr;
5816 
5817 	stop = start + len;
5818 
5819 	/*
5820 	 * The trace_eval_maps contains the map plus a head and tail item,
5821 	 * where the head holds the module and length of array, and the
5822 	 * tail holds a pointer to the next list.
5823 	 */
5824 	map_array = kmalloc_array(len + 2, sizeof(*map_array), GFP_KERNEL);
5825 	if (!map_array) {
5826 		pr_warn("Unable to allocate trace eval mapping\n");
5827 		return;
5828 	}
5829 
5830 	mutex_lock(&trace_eval_mutex);
5831 
5832 	if (!trace_eval_maps)
5833 		trace_eval_maps = map_array;
5834 	else {
5835 		ptr = trace_eval_maps;
5836 		for (;;) {
5837 			ptr = trace_eval_jmp_to_tail(ptr);
5838 			if (!ptr->tail.next)
5839 				break;
5840 			ptr = ptr->tail.next;
5841 
5842 		}
5843 		ptr->tail.next = map_array;
5844 	}
5845 	map_array->head.mod = mod;
5846 	map_array->head.length = len;
5847 	map_array++;
5848 
5849 	for (map = start; (unsigned long)map < (unsigned long)stop; map++) {
5850 		map_array->map = **map;
5851 		map_array++;
5852 	}
5853 	memset(map_array, 0, sizeof(*map_array));
5854 
5855 	mutex_unlock(&trace_eval_mutex);
5856 }
5857 
5858 static void trace_create_eval_file(struct dentry *d_tracer)
5859 {
5860 	trace_create_file("eval_map", TRACE_MODE_READ, d_tracer,
5861 			  NULL, &tracing_eval_map_fops);
5862 }
5863 
5864 #else /* CONFIG_TRACE_EVAL_MAP_FILE */
5865 static inline void trace_create_eval_file(struct dentry *d_tracer) { }
5866 static inline void trace_insert_eval_map_file(struct module *mod,
5867 			      struct trace_eval_map **start, int len) { }
5868 #endif /* !CONFIG_TRACE_EVAL_MAP_FILE */
5869 
5870 static void trace_insert_eval_map(struct module *mod,
5871 				  struct trace_eval_map **start, int len)
5872 {
5873 	struct trace_eval_map **map;
5874 
5875 	if (len <= 0)
5876 		return;
5877 
5878 	map = start;
5879 
5880 	trace_event_eval_update(map, len);
5881 
5882 	trace_insert_eval_map_file(mod, start, len);
5883 }
5884 
5885 static ssize_t
5886 tracing_set_trace_read(struct file *filp, char __user *ubuf,
5887 		       size_t cnt, loff_t *ppos)
5888 {
5889 	struct trace_array *tr = filp->private_data;
5890 	char buf[MAX_TRACER_SIZE+2];
5891 	int r;
5892 
5893 	mutex_lock(&trace_types_lock);
5894 	r = sprintf(buf, "%s\n", tr->current_trace->name);
5895 	mutex_unlock(&trace_types_lock);
5896 
5897 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
5898 }
5899 
5900 int tracer_init(struct tracer *t, struct trace_array *tr)
5901 {
5902 	tracing_reset_online_cpus(&tr->array_buffer);
5903 	return t->init(tr);
5904 }
5905 
5906 static void set_buffer_entries(struct array_buffer *buf, unsigned long val)
5907 {
5908 	int cpu;
5909 
5910 	for_each_tracing_cpu(cpu)
5911 		per_cpu_ptr(buf->data, cpu)->entries = val;
5912 }
5913 
5914 static void update_buffer_entries(struct array_buffer *buf, int cpu)
5915 {
5916 	if (cpu == RING_BUFFER_ALL_CPUS) {
5917 		set_buffer_entries(buf, ring_buffer_size(buf->buffer, 0));
5918 	} else {
5919 		per_cpu_ptr(buf->data, cpu)->entries = ring_buffer_size(buf->buffer, cpu);
5920 	}
5921 }
5922 
5923 #ifdef CONFIG_TRACER_MAX_TRACE
5924 /* resize @tr's buffer to the size of @size_tr's entries */
5925 static int resize_buffer_duplicate_size(struct array_buffer *trace_buf,
5926 					struct array_buffer *size_buf, int cpu_id)
5927 {
5928 	int cpu, ret = 0;
5929 
5930 	if (cpu_id == RING_BUFFER_ALL_CPUS) {
5931 		for_each_tracing_cpu(cpu) {
5932 			ret = ring_buffer_resize(trace_buf->buffer,
5933 				 per_cpu_ptr(size_buf->data, cpu)->entries, cpu);
5934 			if (ret < 0)
5935 				break;
5936 			per_cpu_ptr(trace_buf->data, cpu)->entries =
5937 				per_cpu_ptr(size_buf->data, cpu)->entries;
5938 		}
5939 	} else {
5940 		ret = ring_buffer_resize(trace_buf->buffer,
5941 				 per_cpu_ptr(size_buf->data, cpu_id)->entries, cpu_id);
5942 		if (ret == 0)
5943 			per_cpu_ptr(trace_buf->data, cpu_id)->entries =
5944 				per_cpu_ptr(size_buf->data, cpu_id)->entries;
5945 	}
5946 
5947 	return ret;
5948 }
5949 #endif /* CONFIG_TRACER_MAX_TRACE */
5950 
5951 static int __tracing_resize_ring_buffer(struct trace_array *tr,
5952 					unsigned long size, int cpu)
5953 {
5954 	int ret;
5955 
5956 	/*
5957 	 * If kernel or user changes the size of the ring buffer
5958 	 * we use the size that was given, and we can forget about
5959 	 * expanding it later.
5960 	 */
5961 	trace_set_ring_buffer_expanded(tr);
5962 
5963 	/* May be called before buffers are initialized */
5964 	if (!tr->array_buffer.buffer)
5965 		return 0;
5966 
5967 	/* Do not allow tracing while resizing ring buffer */
5968 	tracing_stop_tr(tr);
5969 
5970 	ret = ring_buffer_resize(tr->array_buffer.buffer, size, cpu);
5971 	if (ret < 0)
5972 		goto out_start;
5973 
5974 #ifdef CONFIG_TRACER_MAX_TRACE
5975 	if (!tr->allocated_snapshot)
5976 		goto out;
5977 
5978 	ret = ring_buffer_resize(tr->max_buffer.buffer, size, cpu);
5979 	if (ret < 0) {
5980 		int r = resize_buffer_duplicate_size(&tr->array_buffer,
5981 						     &tr->array_buffer, cpu);
5982 		if (r < 0) {
5983 			/*
5984 			 * AARGH! We are left with different
5985 			 * size max buffer!!!!
5986 			 * The max buffer is our "snapshot" buffer.
5987 			 * When a tracer needs a snapshot (one of the
5988 			 * latency tracers), it swaps the max buffer
5989 			 * with the saved snap shot. We succeeded to
5990 			 * update the size of the main buffer, but failed to
5991 			 * update the size of the max buffer. But when we tried
5992 			 * to reset the main buffer to the original size, we
5993 			 * failed there too. This is very unlikely to
5994 			 * happen, but if it does, warn and kill all
5995 			 * tracing.
5996 			 */
5997 			WARN_ON(1);
5998 			tracing_disabled = 1;
5999 		}
6000 		goto out_start;
6001 	}
6002 
6003 	update_buffer_entries(&tr->max_buffer, cpu);
6004 
6005  out:
6006 #endif /* CONFIG_TRACER_MAX_TRACE */
6007 
6008 	update_buffer_entries(&tr->array_buffer, cpu);
6009  out_start:
6010 	tracing_start_tr(tr);
6011 	return ret;
6012 }
6013 
6014 ssize_t tracing_resize_ring_buffer(struct trace_array *tr,
6015 				  unsigned long size, int cpu_id)
6016 {
6017 	int ret;
6018 
6019 	mutex_lock(&trace_types_lock);
6020 
6021 	if (cpu_id != RING_BUFFER_ALL_CPUS) {
6022 		/* make sure, this cpu is enabled in the mask */
6023 		if (!cpumask_test_cpu(cpu_id, tracing_buffer_mask)) {
6024 			ret = -EINVAL;
6025 			goto out;
6026 		}
6027 	}
6028 
6029 	ret = __tracing_resize_ring_buffer(tr, size, cpu_id);
6030 	if (ret < 0)
6031 		ret = -ENOMEM;
6032 
6033 out:
6034 	mutex_unlock(&trace_types_lock);
6035 
6036 	return ret;
6037 }
6038 
6039 
6040 /**
6041  * tracing_update_buffers - used by tracing facility to expand ring buffers
6042  * @tr: The tracing instance
6043  *
6044  * To save on memory when the tracing is never used on a system with it
6045  * configured in. The ring buffers are set to a minimum size. But once
6046  * a user starts to use the tracing facility, then they need to grow
6047  * to their default size.
6048  *
6049  * This function is to be called when a tracer is about to be used.
6050  */
6051 int tracing_update_buffers(struct trace_array *tr)
6052 {
6053 	int ret = 0;
6054 
6055 	mutex_lock(&trace_types_lock);
6056 	if (!tr->ring_buffer_expanded)
6057 		ret = __tracing_resize_ring_buffer(tr, trace_buf_size,
6058 						RING_BUFFER_ALL_CPUS);
6059 	mutex_unlock(&trace_types_lock);
6060 
6061 	return ret;
6062 }
6063 
6064 struct trace_option_dentry;
6065 
6066 static void
6067 create_trace_option_files(struct trace_array *tr, struct tracer *tracer);
6068 
6069 /*
6070  * Used to clear out the tracer before deletion of an instance.
6071  * Must have trace_types_lock held.
6072  */
6073 static void tracing_set_nop(struct trace_array *tr)
6074 {
6075 	if (tr->current_trace == &nop_trace)
6076 		return;
6077 
6078 	tr->current_trace->enabled--;
6079 
6080 	if (tr->current_trace->reset)
6081 		tr->current_trace->reset(tr);
6082 
6083 	tr->current_trace = &nop_trace;
6084 }
6085 
6086 static bool tracer_options_updated;
6087 
6088 static void add_tracer_options(struct trace_array *tr, struct tracer *t)
6089 {
6090 	/* Only enable if the directory has been created already. */
6091 	if (!tr->dir)
6092 		return;
6093 
6094 	/* Only create trace option files after update_tracer_options finish */
6095 	if (!tracer_options_updated)
6096 		return;
6097 
6098 	create_trace_option_files(tr, t);
6099 }
6100 
6101 int tracing_set_tracer(struct trace_array *tr, const char *buf)
6102 {
6103 	struct tracer *t;
6104 #ifdef CONFIG_TRACER_MAX_TRACE
6105 	bool had_max_tr;
6106 #endif
6107 	int ret = 0;
6108 
6109 	mutex_lock(&trace_types_lock);
6110 
6111 	if (!tr->ring_buffer_expanded) {
6112 		ret = __tracing_resize_ring_buffer(tr, trace_buf_size,
6113 						RING_BUFFER_ALL_CPUS);
6114 		if (ret < 0)
6115 			goto out;
6116 		ret = 0;
6117 	}
6118 
6119 	for (t = trace_types; t; t = t->next) {
6120 		if (strcmp(t->name, buf) == 0)
6121 			break;
6122 	}
6123 	if (!t) {
6124 		ret = -EINVAL;
6125 		goto out;
6126 	}
6127 	if (t == tr->current_trace)
6128 		goto out;
6129 
6130 #ifdef CONFIG_TRACER_SNAPSHOT
6131 	if (t->use_max_tr) {
6132 		local_irq_disable();
6133 		arch_spin_lock(&tr->max_lock);
6134 		if (tr->cond_snapshot)
6135 			ret = -EBUSY;
6136 		arch_spin_unlock(&tr->max_lock);
6137 		local_irq_enable();
6138 		if (ret)
6139 			goto out;
6140 	}
6141 #endif
6142 	/* Some tracers won't work on kernel command line */
6143 	if (system_state < SYSTEM_RUNNING && t->noboot) {
6144 		pr_warn("Tracer '%s' is not allowed on command line, ignored\n",
6145 			t->name);
6146 		goto out;
6147 	}
6148 
6149 	/* Some tracers are only allowed for the top level buffer */
6150 	if (!trace_ok_for_array(t, tr)) {
6151 		ret = -EINVAL;
6152 		goto out;
6153 	}
6154 
6155 	/* If trace pipe files are being read, we can't change the tracer */
6156 	if (tr->trace_ref) {
6157 		ret = -EBUSY;
6158 		goto out;
6159 	}
6160 
6161 	trace_branch_disable();
6162 
6163 	tr->current_trace->enabled--;
6164 
6165 	if (tr->current_trace->reset)
6166 		tr->current_trace->reset(tr);
6167 
6168 #ifdef CONFIG_TRACER_MAX_TRACE
6169 	had_max_tr = tr->current_trace->use_max_tr;
6170 
6171 	/* Current trace needs to be nop_trace before synchronize_rcu */
6172 	tr->current_trace = &nop_trace;
6173 
6174 	if (had_max_tr && !t->use_max_tr) {
6175 		/*
6176 		 * We need to make sure that the update_max_tr sees that
6177 		 * current_trace changed to nop_trace to keep it from
6178 		 * swapping the buffers after we resize it.
6179 		 * The update_max_tr is called from interrupts disabled
6180 		 * so a synchronized_sched() is sufficient.
6181 		 */
6182 		synchronize_rcu();
6183 		free_snapshot(tr);
6184 		tracing_disarm_snapshot(tr);
6185 	}
6186 
6187 	if (!had_max_tr && t->use_max_tr) {
6188 		ret = tracing_arm_snapshot_locked(tr);
6189 		if (ret)
6190 			goto out;
6191 	}
6192 #else
6193 	tr->current_trace = &nop_trace;
6194 #endif
6195 
6196 	if (t->init) {
6197 		ret = tracer_init(t, tr);
6198 		if (ret) {
6199 #ifdef CONFIG_TRACER_MAX_TRACE
6200 			if (t->use_max_tr)
6201 				tracing_disarm_snapshot(tr);
6202 #endif
6203 			goto out;
6204 		}
6205 	}
6206 
6207 	tr->current_trace = t;
6208 	tr->current_trace->enabled++;
6209 	trace_branch_enable(tr);
6210  out:
6211 	mutex_unlock(&trace_types_lock);
6212 
6213 	return ret;
6214 }
6215 
6216 static ssize_t
6217 tracing_set_trace_write(struct file *filp, const char __user *ubuf,
6218 			size_t cnt, loff_t *ppos)
6219 {
6220 	struct trace_array *tr = filp->private_data;
6221 	char buf[MAX_TRACER_SIZE+1];
6222 	char *name;
6223 	size_t ret;
6224 	int err;
6225 
6226 	ret = cnt;
6227 
6228 	if (cnt > MAX_TRACER_SIZE)
6229 		cnt = MAX_TRACER_SIZE;
6230 
6231 	if (copy_from_user(buf, ubuf, cnt))
6232 		return -EFAULT;
6233 
6234 	buf[cnt] = 0;
6235 
6236 	name = strim(buf);
6237 
6238 	err = tracing_set_tracer(tr, name);
6239 	if (err)
6240 		return err;
6241 
6242 	*ppos += ret;
6243 
6244 	return ret;
6245 }
6246 
6247 static ssize_t
6248 tracing_nsecs_read(unsigned long *ptr, char __user *ubuf,
6249 		   size_t cnt, loff_t *ppos)
6250 {
6251 	char buf[64];
6252 	int r;
6253 
6254 	r = snprintf(buf, sizeof(buf), "%ld\n",
6255 		     *ptr == (unsigned long)-1 ? -1 : nsecs_to_usecs(*ptr));
6256 	if (r > sizeof(buf))
6257 		r = sizeof(buf);
6258 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
6259 }
6260 
6261 static ssize_t
6262 tracing_nsecs_write(unsigned long *ptr, const char __user *ubuf,
6263 		    size_t cnt, loff_t *ppos)
6264 {
6265 	unsigned long val;
6266 	int ret;
6267 
6268 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
6269 	if (ret)
6270 		return ret;
6271 
6272 	*ptr = val * 1000;
6273 
6274 	return cnt;
6275 }
6276 
6277 static ssize_t
6278 tracing_thresh_read(struct file *filp, char __user *ubuf,
6279 		    size_t cnt, loff_t *ppos)
6280 {
6281 	return tracing_nsecs_read(&tracing_thresh, ubuf, cnt, ppos);
6282 }
6283 
6284 static ssize_t
6285 tracing_thresh_write(struct file *filp, const char __user *ubuf,
6286 		     size_t cnt, loff_t *ppos)
6287 {
6288 	struct trace_array *tr = filp->private_data;
6289 	int ret;
6290 
6291 	mutex_lock(&trace_types_lock);
6292 	ret = tracing_nsecs_write(&tracing_thresh, ubuf, cnt, ppos);
6293 	if (ret < 0)
6294 		goto out;
6295 
6296 	if (tr->current_trace->update_thresh) {
6297 		ret = tr->current_trace->update_thresh(tr);
6298 		if (ret < 0)
6299 			goto out;
6300 	}
6301 
6302 	ret = cnt;
6303 out:
6304 	mutex_unlock(&trace_types_lock);
6305 
6306 	return ret;
6307 }
6308 
6309 #ifdef CONFIG_TRACER_MAX_TRACE
6310 
6311 static ssize_t
6312 tracing_max_lat_read(struct file *filp, char __user *ubuf,
6313 		     size_t cnt, loff_t *ppos)
6314 {
6315 	struct trace_array *tr = filp->private_data;
6316 
6317 	return tracing_nsecs_read(&tr->max_latency, ubuf, cnt, ppos);
6318 }
6319 
6320 static ssize_t
6321 tracing_max_lat_write(struct file *filp, const char __user *ubuf,
6322 		      size_t cnt, loff_t *ppos)
6323 {
6324 	struct trace_array *tr = filp->private_data;
6325 
6326 	return tracing_nsecs_write(&tr->max_latency, ubuf, cnt, ppos);
6327 }
6328 
6329 #endif
6330 
6331 static int open_pipe_on_cpu(struct trace_array *tr, int cpu)
6332 {
6333 	if (cpu == RING_BUFFER_ALL_CPUS) {
6334 		if (cpumask_empty(tr->pipe_cpumask)) {
6335 			cpumask_setall(tr->pipe_cpumask);
6336 			return 0;
6337 		}
6338 	} else if (!cpumask_test_cpu(cpu, tr->pipe_cpumask)) {
6339 		cpumask_set_cpu(cpu, tr->pipe_cpumask);
6340 		return 0;
6341 	}
6342 	return -EBUSY;
6343 }
6344 
6345 static void close_pipe_on_cpu(struct trace_array *tr, int cpu)
6346 {
6347 	if (cpu == RING_BUFFER_ALL_CPUS) {
6348 		WARN_ON(!cpumask_full(tr->pipe_cpumask));
6349 		cpumask_clear(tr->pipe_cpumask);
6350 	} else {
6351 		WARN_ON(!cpumask_test_cpu(cpu, tr->pipe_cpumask));
6352 		cpumask_clear_cpu(cpu, tr->pipe_cpumask);
6353 	}
6354 }
6355 
6356 static int tracing_open_pipe(struct inode *inode, struct file *filp)
6357 {
6358 	struct trace_array *tr = inode->i_private;
6359 	struct trace_iterator *iter;
6360 	int cpu;
6361 	int ret;
6362 
6363 	ret = tracing_check_open_get_tr(tr);
6364 	if (ret)
6365 		return ret;
6366 
6367 	mutex_lock(&trace_types_lock);
6368 	cpu = tracing_get_cpu(inode);
6369 	ret = open_pipe_on_cpu(tr, cpu);
6370 	if (ret)
6371 		goto fail_pipe_on_cpu;
6372 
6373 	/* create a buffer to store the information to pass to userspace */
6374 	iter = kzalloc(sizeof(*iter), GFP_KERNEL);
6375 	if (!iter) {
6376 		ret = -ENOMEM;
6377 		goto fail_alloc_iter;
6378 	}
6379 
6380 	trace_seq_init(&iter->seq);
6381 	iter->trace = tr->current_trace;
6382 
6383 	if (!alloc_cpumask_var(&iter->started, GFP_KERNEL)) {
6384 		ret = -ENOMEM;
6385 		goto fail;
6386 	}
6387 
6388 	/* trace pipe does not show start of buffer */
6389 	cpumask_setall(iter->started);
6390 
6391 	if (tr->trace_flags & TRACE_ITER_LATENCY_FMT)
6392 		iter->iter_flags |= TRACE_FILE_LAT_FMT;
6393 
6394 	/* Output in nanoseconds only if we are using a clock in nanoseconds. */
6395 	if (trace_clocks[tr->clock_id].in_ns)
6396 		iter->iter_flags |= TRACE_FILE_TIME_IN_NS;
6397 
6398 	iter->tr = tr;
6399 	iter->array_buffer = &tr->array_buffer;
6400 	iter->cpu_file = cpu;
6401 	mutex_init(&iter->mutex);
6402 	filp->private_data = iter;
6403 
6404 	if (iter->trace->pipe_open)
6405 		iter->trace->pipe_open(iter);
6406 
6407 	nonseekable_open(inode, filp);
6408 
6409 	tr->trace_ref++;
6410 
6411 	mutex_unlock(&trace_types_lock);
6412 	return ret;
6413 
6414 fail:
6415 	kfree(iter);
6416 fail_alloc_iter:
6417 	close_pipe_on_cpu(tr, cpu);
6418 fail_pipe_on_cpu:
6419 	__trace_array_put(tr);
6420 	mutex_unlock(&trace_types_lock);
6421 	return ret;
6422 }
6423 
6424 static int tracing_release_pipe(struct inode *inode, struct file *file)
6425 {
6426 	struct trace_iterator *iter = file->private_data;
6427 	struct trace_array *tr = inode->i_private;
6428 
6429 	mutex_lock(&trace_types_lock);
6430 
6431 	tr->trace_ref--;
6432 
6433 	if (iter->trace->pipe_close)
6434 		iter->trace->pipe_close(iter);
6435 	close_pipe_on_cpu(tr, iter->cpu_file);
6436 	mutex_unlock(&trace_types_lock);
6437 
6438 	free_trace_iter_content(iter);
6439 	kfree(iter);
6440 
6441 	trace_array_put(tr);
6442 
6443 	return 0;
6444 }
6445 
6446 static __poll_t
6447 trace_poll(struct trace_iterator *iter, struct file *filp, poll_table *poll_table)
6448 {
6449 	struct trace_array *tr = iter->tr;
6450 
6451 	/* Iterators are static, they should be filled or empty */
6452 	if (trace_buffer_iter(iter, iter->cpu_file))
6453 		return EPOLLIN | EPOLLRDNORM;
6454 
6455 	if (tr->trace_flags & TRACE_ITER_BLOCK)
6456 		/*
6457 		 * Always select as readable when in blocking mode
6458 		 */
6459 		return EPOLLIN | EPOLLRDNORM;
6460 	else
6461 		return ring_buffer_poll_wait(iter->array_buffer->buffer, iter->cpu_file,
6462 					     filp, poll_table, iter->tr->buffer_percent);
6463 }
6464 
6465 static __poll_t
6466 tracing_poll_pipe(struct file *filp, poll_table *poll_table)
6467 {
6468 	struct trace_iterator *iter = filp->private_data;
6469 
6470 	return trace_poll(iter, filp, poll_table);
6471 }
6472 
6473 /* Must be called with iter->mutex held. */
6474 static int tracing_wait_pipe(struct file *filp)
6475 {
6476 	struct trace_iterator *iter = filp->private_data;
6477 	int ret;
6478 
6479 	while (trace_empty(iter)) {
6480 
6481 		if ((filp->f_flags & O_NONBLOCK)) {
6482 			return -EAGAIN;
6483 		}
6484 
6485 		/*
6486 		 * We block until we read something and tracing is disabled.
6487 		 * We still block if tracing is disabled, but we have never
6488 		 * read anything. This allows a user to cat this file, and
6489 		 * then enable tracing. But after we have read something,
6490 		 * we give an EOF when tracing is again disabled.
6491 		 *
6492 		 * iter->pos will be 0 if we haven't read anything.
6493 		 */
6494 		if (!tracer_tracing_is_on(iter->tr) && iter->pos)
6495 			break;
6496 
6497 		mutex_unlock(&iter->mutex);
6498 
6499 		ret = wait_on_pipe(iter, 0);
6500 
6501 		mutex_lock(&iter->mutex);
6502 
6503 		if (ret)
6504 			return ret;
6505 	}
6506 
6507 	return 1;
6508 }
6509 
6510 /*
6511  * Consumer reader.
6512  */
6513 static ssize_t
6514 tracing_read_pipe(struct file *filp, char __user *ubuf,
6515 		  size_t cnt, loff_t *ppos)
6516 {
6517 	struct trace_iterator *iter = filp->private_data;
6518 	ssize_t sret;
6519 
6520 	/*
6521 	 * Avoid more than one consumer on a single file descriptor
6522 	 * This is just a matter of traces coherency, the ring buffer itself
6523 	 * is protected.
6524 	 */
6525 	mutex_lock(&iter->mutex);
6526 
6527 	/* return any leftover data */
6528 	sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
6529 	if (sret != -EBUSY)
6530 		goto out;
6531 
6532 	trace_seq_init(&iter->seq);
6533 
6534 	if (iter->trace->read) {
6535 		sret = iter->trace->read(iter, filp, ubuf, cnt, ppos);
6536 		if (sret)
6537 			goto out;
6538 	}
6539 
6540 waitagain:
6541 	sret = tracing_wait_pipe(filp);
6542 	if (sret <= 0)
6543 		goto out;
6544 
6545 	/* stop when tracing is finished */
6546 	if (trace_empty(iter)) {
6547 		sret = 0;
6548 		goto out;
6549 	}
6550 
6551 	if (cnt >= TRACE_SEQ_BUFFER_SIZE)
6552 		cnt = TRACE_SEQ_BUFFER_SIZE - 1;
6553 
6554 	/* reset all but tr, trace, and overruns */
6555 	trace_iterator_reset(iter);
6556 	cpumask_clear(iter->started);
6557 	trace_seq_init(&iter->seq);
6558 
6559 	trace_event_read_lock();
6560 	trace_access_lock(iter->cpu_file);
6561 	while (trace_find_next_entry_inc(iter) != NULL) {
6562 		enum print_line_t ret;
6563 		int save_len = iter->seq.seq.len;
6564 
6565 		ret = print_trace_line(iter);
6566 		if (ret == TRACE_TYPE_PARTIAL_LINE) {
6567 			/*
6568 			 * If one print_trace_line() fills entire trace_seq in one shot,
6569 			 * trace_seq_to_user() will returns -EBUSY because save_len == 0,
6570 			 * In this case, we need to consume it, otherwise, loop will peek
6571 			 * this event next time, resulting in an infinite loop.
6572 			 */
6573 			if (save_len == 0) {
6574 				iter->seq.full = 0;
6575 				trace_seq_puts(&iter->seq, "[LINE TOO BIG]\n");
6576 				trace_consume(iter);
6577 				break;
6578 			}
6579 
6580 			/* In other cases, don't print partial lines */
6581 			iter->seq.seq.len = save_len;
6582 			break;
6583 		}
6584 		if (ret != TRACE_TYPE_NO_CONSUME)
6585 			trace_consume(iter);
6586 
6587 		if (trace_seq_used(&iter->seq) >= cnt)
6588 			break;
6589 
6590 		/*
6591 		 * Setting the full flag means we reached the trace_seq buffer
6592 		 * size and we should leave by partial output condition above.
6593 		 * One of the trace_seq_* functions is not used properly.
6594 		 */
6595 		WARN_ONCE(iter->seq.full, "full flag set for trace type %d",
6596 			  iter->ent->type);
6597 	}
6598 	trace_access_unlock(iter->cpu_file);
6599 	trace_event_read_unlock();
6600 
6601 	/* Now copy what we have to the user */
6602 	sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
6603 	if (iter->seq.readpos >= trace_seq_used(&iter->seq))
6604 		trace_seq_init(&iter->seq);
6605 
6606 	/*
6607 	 * If there was nothing to send to user, in spite of consuming trace
6608 	 * entries, go back to wait for more entries.
6609 	 */
6610 	if (sret == -EBUSY)
6611 		goto waitagain;
6612 
6613 out:
6614 	mutex_unlock(&iter->mutex);
6615 
6616 	return sret;
6617 }
6618 
6619 static void tracing_spd_release_pipe(struct splice_pipe_desc *spd,
6620 				     unsigned int idx)
6621 {
6622 	__free_page(spd->pages[idx]);
6623 }
6624 
6625 static size_t
6626 tracing_fill_pipe_page(size_t rem, struct trace_iterator *iter)
6627 {
6628 	size_t count;
6629 	int save_len;
6630 	int ret;
6631 
6632 	/* Seq buffer is page-sized, exactly what we need. */
6633 	for (;;) {
6634 		save_len = iter->seq.seq.len;
6635 		ret = print_trace_line(iter);
6636 
6637 		if (trace_seq_has_overflowed(&iter->seq)) {
6638 			iter->seq.seq.len = save_len;
6639 			break;
6640 		}
6641 
6642 		/*
6643 		 * This should not be hit, because it should only
6644 		 * be set if the iter->seq overflowed. But check it
6645 		 * anyway to be safe.
6646 		 */
6647 		if (ret == TRACE_TYPE_PARTIAL_LINE) {
6648 			iter->seq.seq.len = save_len;
6649 			break;
6650 		}
6651 
6652 		count = trace_seq_used(&iter->seq) - save_len;
6653 		if (rem < count) {
6654 			rem = 0;
6655 			iter->seq.seq.len = save_len;
6656 			break;
6657 		}
6658 
6659 		if (ret != TRACE_TYPE_NO_CONSUME)
6660 			trace_consume(iter);
6661 		rem -= count;
6662 		if (!trace_find_next_entry_inc(iter))	{
6663 			rem = 0;
6664 			iter->ent = NULL;
6665 			break;
6666 		}
6667 	}
6668 
6669 	return rem;
6670 }
6671 
6672 static ssize_t tracing_splice_read_pipe(struct file *filp,
6673 					loff_t *ppos,
6674 					struct pipe_inode_info *pipe,
6675 					size_t len,
6676 					unsigned int flags)
6677 {
6678 	struct page *pages_def[PIPE_DEF_BUFFERS];
6679 	struct partial_page partial_def[PIPE_DEF_BUFFERS];
6680 	struct trace_iterator *iter = filp->private_data;
6681 	struct splice_pipe_desc spd = {
6682 		.pages		= pages_def,
6683 		.partial	= partial_def,
6684 		.nr_pages	= 0, /* This gets updated below. */
6685 		.nr_pages_max	= PIPE_DEF_BUFFERS,
6686 		.ops		= &default_pipe_buf_ops,
6687 		.spd_release	= tracing_spd_release_pipe,
6688 	};
6689 	ssize_t ret;
6690 	size_t rem;
6691 	unsigned int i;
6692 
6693 	if (splice_grow_spd(pipe, &spd))
6694 		return -ENOMEM;
6695 
6696 	mutex_lock(&iter->mutex);
6697 
6698 	if (iter->trace->splice_read) {
6699 		ret = iter->trace->splice_read(iter, filp,
6700 					       ppos, pipe, len, flags);
6701 		if (ret)
6702 			goto out_err;
6703 	}
6704 
6705 	ret = tracing_wait_pipe(filp);
6706 	if (ret <= 0)
6707 		goto out_err;
6708 
6709 	if (!iter->ent && !trace_find_next_entry_inc(iter)) {
6710 		ret = -EFAULT;
6711 		goto out_err;
6712 	}
6713 
6714 	trace_event_read_lock();
6715 	trace_access_lock(iter->cpu_file);
6716 
6717 	/* Fill as many pages as possible. */
6718 	for (i = 0, rem = len; i < spd.nr_pages_max && rem; i++) {
6719 		spd.pages[i] = alloc_page(GFP_KERNEL);
6720 		if (!spd.pages[i])
6721 			break;
6722 
6723 		rem = tracing_fill_pipe_page(rem, iter);
6724 
6725 		/* Copy the data into the page, so we can start over. */
6726 		ret = trace_seq_to_buffer(&iter->seq,
6727 					  page_address(spd.pages[i]),
6728 					  trace_seq_used(&iter->seq));
6729 		if (ret < 0) {
6730 			__free_page(spd.pages[i]);
6731 			break;
6732 		}
6733 		spd.partial[i].offset = 0;
6734 		spd.partial[i].len = trace_seq_used(&iter->seq);
6735 
6736 		trace_seq_init(&iter->seq);
6737 	}
6738 
6739 	trace_access_unlock(iter->cpu_file);
6740 	trace_event_read_unlock();
6741 	mutex_unlock(&iter->mutex);
6742 
6743 	spd.nr_pages = i;
6744 
6745 	if (i)
6746 		ret = splice_to_pipe(pipe, &spd);
6747 	else
6748 		ret = 0;
6749 out:
6750 	splice_shrink_spd(&spd);
6751 	return ret;
6752 
6753 out_err:
6754 	mutex_unlock(&iter->mutex);
6755 	goto out;
6756 }
6757 
6758 static ssize_t
6759 tracing_entries_read(struct file *filp, char __user *ubuf,
6760 		     size_t cnt, loff_t *ppos)
6761 {
6762 	struct inode *inode = file_inode(filp);
6763 	struct trace_array *tr = inode->i_private;
6764 	int cpu = tracing_get_cpu(inode);
6765 	char buf[64];
6766 	int r = 0;
6767 	ssize_t ret;
6768 
6769 	mutex_lock(&trace_types_lock);
6770 
6771 	if (cpu == RING_BUFFER_ALL_CPUS) {
6772 		int cpu, buf_size_same;
6773 		unsigned long size;
6774 
6775 		size = 0;
6776 		buf_size_same = 1;
6777 		/* check if all cpu sizes are same */
6778 		for_each_tracing_cpu(cpu) {
6779 			/* fill in the size from first enabled cpu */
6780 			if (size == 0)
6781 				size = per_cpu_ptr(tr->array_buffer.data, cpu)->entries;
6782 			if (size != per_cpu_ptr(tr->array_buffer.data, cpu)->entries) {
6783 				buf_size_same = 0;
6784 				break;
6785 			}
6786 		}
6787 
6788 		if (buf_size_same) {
6789 			if (!tr->ring_buffer_expanded)
6790 				r = sprintf(buf, "%lu (expanded: %lu)\n",
6791 					    size >> 10,
6792 					    trace_buf_size >> 10);
6793 			else
6794 				r = sprintf(buf, "%lu\n", size >> 10);
6795 		} else
6796 			r = sprintf(buf, "X\n");
6797 	} else
6798 		r = sprintf(buf, "%lu\n", per_cpu_ptr(tr->array_buffer.data, cpu)->entries >> 10);
6799 
6800 	mutex_unlock(&trace_types_lock);
6801 
6802 	ret = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
6803 	return ret;
6804 }
6805 
6806 static ssize_t
6807 tracing_entries_write(struct file *filp, const char __user *ubuf,
6808 		      size_t cnt, loff_t *ppos)
6809 {
6810 	struct inode *inode = file_inode(filp);
6811 	struct trace_array *tr = inode->i_private;
6812 	unsigned long val;
6813 	int ret;
6814 
6815 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
6816 	if (ret)
6817 		return ret;
6818 
6819 	/* must have at least 1 entry */
6820 	if (!val)
6821 		return -EINVAL;
6822 
6823 	/* value is in KB */
6824 	val <<= 10;
6825 	ret = tracing_resize_ring_buffer(tr, val, tracing_get_cpu(inode));
6826 	if (ret < 0)
6827 		return ret;
6828 
6829 	*ppos += cnt;
6830 
6831 	return cnt;
6832 }
6833 
6834 static ssize_t
6835 tracing_total_entries_read(struct file *filp, char __user *ubuf,
6836 				size_t cnt, loff_t *ppos)
6837 {
6838 	struct trace_array *tr = filp->private_data;
6839 	char buf[64];
6840 	int r, cpu;
6841 	unsigned long size = 0, expanded_size = 0;
6842 
6843 	mutex_lock(&trace_types_lock);
6844 	for_each_tracing_cpu(cpu) {
6845 		size += per_cpu_ptr(tr->array_buffer.data, cpu)->entries >> 10;
6846 		if (!tr->ring_buffer_expanded)
6847 			expanded_size += trace_buf_size >> 10;
6848 	}
6849 	if (tr->ring_buffer_expanded)
6850 		r = sprintf(buf, "%lu\n", size);
6851 	else
6852 		r = sprintf(buf, "%lu (expanded: %lu)\n", size, expanded_size);
6853 	mutex_unlock(&trace_types_lock);
6854 
6855 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
6856 }
6857 
6858 static ssize_t
6859 tracing_free_buffer_write(struct file *filp, const char __user *ubuf,
6860 			  size_t cnt, loff_t *ppos)
6861 {
6862 	/*
6863 	 * There is no need to read what the user has written, this function
6864 	 * is just to make sure that there is no error when "echo" is used
6865 	 */
6866 
6867 	*ppos += cnt;
6868 
6869 	return cnt;
6870 }
6871 
6872 static int
6873 tracing_free_buffer_release(struct inode *inode, struct file *filp)
6874 {
6875 	struct trace_array *tr = inode->i_private;
6876 
6877 	/* disable tracing ? */
6878 	if (tr->trace_flags & TRACE_ITER_STOP_ON_FREE)
6879 		tracer_tracing_off(tr);
6880 	/* resize the ring buffer to 0 */
6881 	tracing_resize_ring_buffer(tr, 0, RING_BUFFER_ALL_CPUS);
6882 
6883 	trace_array_put(tr);
6884 
6885 	return 0;
6886 }
6887 
6888 #define TRACE_MARKER_MAX_SIZE		4096
6889 
6890 static ssize_t
6891 tracing_mark_write(struct file *filp, const char __user *ubuf,
6892 					size_t cnt, loff_t *fpos)
6893 {
6894 	struct trace_array *tr = filp->private_data;
6895 	struct ring_buffer_event *event;
6896 	enum event_trigger_type tt = ETT_NONE;
6897 	struct trace_buffer *buffer;
6898 	struct print_entry *entry;
6899 	int meta_size;
6900 	ssize_t written;
6901 	size_t size;
6902 	int len;
6903 
6904 /* Used in tracing_mark_raw_write() as well */
6905 #define FAULTED_STR "<faulted>"
6906 #define FAULTED_SIZE (sizeof(FAULTED_STR) - 1) /* '\0' is already accounted for */
6907 
6908 	if (tracing_disabled)
6909 		return -EINVAL;
6910 
6911 	if (!(tr->trace_flags & TRACE_ITER_MARKERS))
6912 		return -EINVAL;
6913 
6914 	if ((ssize_t)cnt < 0)
6915 		return -EINVAL;
6916 
6917 	if (cnt > TRACE_MARKER_MAX_SIZE)
6918 		cnt = TRACE_MARKER_MAX_SIZE;
6919 
6920 	meta_size = sizeof(*entry) + 2;  /* add '\0' and possible '\n' */
6921  again:
6922 	size = cnt + meta_size;
6923 
6924 	/* If less than "<faulted>", then make sure we can still add that */
6925 	if (cnt < FAULTED_SIZE)
6926 		size += FAULTED_SIZE - cnt;
6927 
6928 	buffer = tr->array_buffer.buffer;
6929 	event = __trace_buffer_lock_reserve(buffer, TRACE_PRINT, size,
6930 					    tracing_gen_ctx());
6931 	if (unlikely(!event)) {
6932 		/*
6933 		 * If the size was greater than what was allowed, then
6934 		 * make it smaller and try again.
6935 		 */
6936 		if (size > ring_buffer_max_event_size(buffer)) {
6937 			/* cnt < FAULTED size should never be bigger than max */
6938 			if (WARN_ON_ONCE(cnt < FAULTED_SIZE))
6939 				return -EBADF;
6940 			cnt = ring_buffer_max_event_size(buffer) - meta_size;
6941 			/* The above should only happen once */
6942 			if (WARN_ON_ONCE(cnt + meta_size == size))
6943 				return -EBADF;
6944 			goto again;
6945 		}
6946 
6947 		/* Ring buffer disabled, return as if not open for write */
6948 		return -EBADF;
6949 	}
6950 
6951 	entry = ring_buffer_event_data(event);
6952 	entry->ip = _THIS_IP_;
6953 
6954 	len = __copy_from_user_inatomic(&entry->buf, ubuf, cnt);
6955 	if (len) {
6956 		memcpy(&entry->buf, FAULTED_STR, FAULTED_SIZE);
6957 		cnt = FAULTED_SIZE;
6958 		written = -EFAULT;
6959 	} else
6960 		written = cnt;
6961 
6962 	if (tr->trace_marker_file && !list_empty(&tr->trace_marker_file->triggers)) {
6963 		/* do not add \n before testing triggers, but add \0 */
6964 		entry->buf[cnt] = '\0';
6965 		tt = event_triggers_call(tr->trace_marker_file, buffer, entry, event);
6966 	}
6967 
6968 	if (entry->buf[cnt - 1] != '\n') {
6969 		entry->buf[cnt] = '\n';
6970 		entry->buf[cnt + 1] = '\0';
6971 	} else
6972 		entry->buf[cnt] = '\0';
6973 
6974 	if (static_branch_unlikely(&trace_marker_exports_enabled))
6975 		ftrace_exports(event, TRACE_EXPORT_MARKER);
6976 	__buffer_unlock_commit(buffer, event);
6977 
6978 	if (tt)
6979 		event_triggers_post_call(tr->trace_marker_file, tt);
6980 
6981 	return written;
6982 }
6983 
6984 static ssize_t
6985 tracing_mark_raw_write(struct file *filp, const char __user *ubuf,
6986 					size_t cnt, loff_t *fpos)
6987 {
6988 	struct trace_array *tr = filp->private_data;
6989 	struct ring_buffer_event *event;
6990 	struct trace_buffer *buffer;
6991 	struct raw_data_entry *entry;
6992 	ssize_t written;
6993 	int size;
6994 	int len;
6995 
6996 #define FAULT_SIZE_ID (FAULTED_SIZE + sizeof(int))
6997 
6998 	if (tracing_disabled)
6999 		return -EINVAL;
7000 
7001 	if (!(tr->trace_flags & TRACE_ITER_MARKERS))
7002 		return -EINVAL;
7003 
7004 	/* The marker must at least have a tag id */
7005 	if (cnt < sizeof(unsigned int))
7006 		return -EINVAL;
7007 
7008 	size = sizeof(*entry) + cnt;
7009 	if (cnt < FAULT_SIZE_ID)
7010 		size += FAULT_SIZE_ID - cnt;
7011 
7012 	buffer = tr->array_buffer.buffer;
7013 
7014 	if (size > ring_buffer_max_event_size(buffer))
7015 		return -EINVAL;
7016 
7017 	event = __trace_buffer_lock_reserve(buffer, TRACE_RAW_DATA, size,
7018 					    tracing_gen_ctx());
7019 	if (!event)
7020 		/* Ring buffer disabled, return as if not open for write */
7021 		return -EBADF;
7022 
7023 	entry = ring_buffer_event_data(event);
7024 
7025 	len = __copy_from_user_inatomic(&entry->id, ubuf, cnt);
7026 	if (len) {
7027 		entry->id = -1;
7028 		memcpy(&entry->buf, FAULTED_STR, FAULTED_SIZE);
7029 		written = -EFAULT;
7030 	} else
7031 		written = cnt;
7032 
7033 	__buffer_unlock_commit(buffer, event);
7034 
7035 	return written;
7036 }
7037 
7038 static int tracing_clock_show(struct seq_file *m, void *v)
7039 {
7040 	struct trace_array *tr = m->private;
7041 	int i;
7042 
7043 	for (i = 0; i < ARRAY_SIZE(trace_clocks); i++)
7044 		seq_printf(m,
7045 			"%s%s%s%s", i ? " " : "",
7046 			i == tr->clock_id ? "[" : "", trace_clocks[i].name,
7047 			i == tr->clock_id ? "]" : "");
7048 	seq_putc(m, '\n');
7049 
7050 	return 0;
7051 }
7052 
7053 int tracing_set_clock(struct trace_array *tr, const char *clockstr)
7054 {
7055 	int i;
7056 
7057 	for (i = 0; i < ARRAY_SIZE(trace_clocks); i++) {
7058 		if (strcmp(trace_clocks[i].name, clockstr) == 0)
7059 			break;
7060 	}
7061 	if (i == ARRAY_SIZE(trace_clocks))
7062 		return -EINVAL;
7063 
7064 	mutex_lock(&trace_types_lock);
7065 
7066 	tr->clock_id = i;
7067 
7068 	ring_buffer_set_clock(tr->array_buffer.buffer, trace_clocks[i].func);
7069 
7070 	/*
7071 	 * New clock may not be consistent with the previous clock.
7072 	 * Reset the buffer so that it doesn't have incomparable timestamps.
7073 	 */
7074 	tracing_reset_online_cpus(&tr->array_buffer);
7075 
7076 #ifdef CONFIG_TRACER_MAX_TRACE
7077 	if (tr->max_buffer.buffer)
7078 		ring_buffer_set_clock(tr->max_buffer.buffer, trace_clocks[i].func);
7079 	tracing_reset_online_cpus(&tr->max_buffer);
7080 #endif
7081 
7082 	mutex_unlock(&trace_types_lock);
7083 
7084 	return 0;
7085 }
7086 
7087 static ssize_t tracing_clock_write(struct file *filp, const char __user *ubuf,
7088 				   size_t cnt, loff_t *fpos)
7089 {
7090 	struct seq_file *m = filp->private_data;
7091 	struct trace_array *tr = m->private;
7092 	char buf[64];
7093 	const char *clockstr;
7094 	int ret;
7095 
7096 	if (cnt >= sizeof(buf))
7097 		return -EINVAL;
7098 
7099 	if (copy_from_user(buf, ubuf, cnt))
7100 		return -EFAULT;
7101 
7102 	buf[cnt] = 0;
7103 
7104 	clockstr = strstrip(buf);
7105 
7106 	ret = tracing_set_clock(tr, clockstr);
7107 	if (ret)
7108 		return ret;
7109 
7110 	*fpos += cnt;
7111 
7112 	return cnt;
7113 }
7114 
7115 static int tracing_clock_open(struct inode *inode, struct file *file)
7116 {
7117 	struct trace_array *tr = inode->i_private;
7118 	int ret;
7119 
7120 	ret = tracing_check_open_get_tr(tr);
7121 	if (ret)
7122 		return ret;
7123 
7124 	ret = single_open(file, tracing_clock_show, inode->i_private);
7125 	if (ret < 0)
7126 		trace_array_put(tr);
7127 
7128 	return ret;
7129 }
7130 
7131 static int tracing_time_stamp_mode_show(struct seq_file *m, void *v)
7132 {
7133 	struct trace_array *tr = m->private;
7134 
7135 	mutex_lock(&trace_types_lock);
7136 
7137 	if (ring_buffer_time_stamp_abs(tr->array_buffer.buffer))
7138 		seq_puts(m, "delta [absolute]\n");
7139 	else
7140 		seq_puts(m, "[delta] absolute\n");
7141 
7142 	mutex_unlock(&trace_types_lock);
7143 
7144 	return 0;
7145 }
7146 
7147 static int tracing_time_stamp_mode_open(struct inode *inode, struct file *file)
7148 {
7149 	struct trace_array *tr = inode->i_private;
7150 	int ret;
7151 
7152 	ret = tracing_check_open_get_tr(tr);
7153 	if (ret)
7154 		return ret;
7155 
7156 	ret = single_open(file, tracing_time_stamp_mode_show, inode->i_private);
7157 	if (ret < 0)
7158 		trace_array_put(tr);
7159 
7160 	return ret;
7161 }
7162 
7163 u64 tracing_event_time_stamp(struct trace_buffer *buffer, struct ring_buffer_event *rbe)
7164 {
7165 	if (rbe == this_cpu_read(trace_buffered_event))
7166 		return ring_buffer_time_stamp(buffer);
7167 
7168 	return ring_buffer_event_time_stamp(buffer, rbe);
7169 }
7170 
7171 /*
7172  * Set or disable using the per CPU trace_buffer_event when possible.
7173  */
7174 int tracing_set_filter_buffering(struct trace_array *tr, bool set)
7175 {
7176 	int ret = 0;
7177 
7178 	mutex_lock(&trace_types_lock);
7179 
7180 	if (set && tr->no_filter_buffering_ref++)
7181 		goto out;
7182 
7183 	if (!set) {
7184 		if (WARN_ON_ONCE(!tr->no_filter_buffering_ref)) {
7185 			ret = -EINVAL;
7186 			goto out;
7187 		}
7188 
7189 		--tr->no_filter_buffering_ref;
7190 	}
7191  out:
7192 	mutex_unlock(&trace_types_lock);
7193 
7194 	return ret;
7195 }
7196 
7197 struct ftrace_buffer_info {
7198 	struct trace_iterator	iter;
7199 	void			*spare;
7200 	unsigned int		spare_cpu;
7201 	unsigned int		spare_size;
7202 	unsigned int		read;
7203 };
7204 
7205 #ifdef CONFIG_TRACER_SNAPSHOT
7206 static int tracing_snapshot_open(struct inode *inode, struct file *file)
7207 {
7208 	struct trace_array *tr = inode->i_private;
7209 	struct trace_iterator *iter;
7210 	struct seq_file *m;
7211 	int ret;
7212 
7213 	ret = tracing_check_open_get_tr(tr);
7214 	if (ret)
7215 		return ret;
7216 
7217 	if (file->f_mode & FMODE_READ) {
7218 		iter = __tracing_open(inode, file, true);
7219 		if (IS_ERR(iter))
7220 			ret = PTR_ERR(iter);
7221 	} else {
7222 		/* Writes still need the seq_file to hold the private data */
7223 		ret = -ENOMEM;
7224 		m = kzalloc(sizeof(*m), GFP_KERNEL);
7225 		if (!m)
7226 			goto out;
7227 		iter = kzalloc(sizeof(*iter), GFP_KERNEL);
7228 		if (!iter) {
7229 			kfree(m);
7230 			goto out;
7231 		}
7232 		ret = 0;
7233 
7234 		iter->tr = tr;
7235 		iter->array_buffer = &tr->max_buffer;
7236 		iter->cpu_file = tracing_get_cpu(inode);
7237 		m->private = iter;
7238 		file->private_data = m;
7239 	}
7240 out:
7241 	if (ret < 0)
7242 		trace_array_put(tr);
7243 
7244 	return ret;
7245 }
7246 
7247 static void tracing_swap_cpu_buffer(void *tr)
7248 {
7249 	update_max_tr_single((struct trace_array *)tr, current, smp_processor_id());
7250 }
7251 
7252 static ssize_t
7253 tracing_snapshot_write(struct file *filp, const char __user *ubuf, size_t cnt,
7254 		       loff_t *ppos)
7255 {
7256 	struct seq_file *m = filp->private_data;
7257 	struct trace_iterator *iter = m->private;
7258 	struct trace_array *tr = iter->tr;
7259 	unsigned long val;
7260 	int ret;
7261 
7262 	ret = tracing_update_buffers(tr);
7263 	if (ret < 0)
7264 		return ret;
7265 
7266 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
7267 	if (ret)
7268 		return ret;
7269 
7270 	mutex_lock(&trace_types_lock);
7271 
7272 	if (tr->current_trace->use_max_tr) {
7273 		ret = -EBUSY;
7274 		goto out;
7275 	}
7276 
7277 	local_irq_disable();
7278 	arch_spin_lock(&tr->max_lock);
7279 	if (tr->cond_snapshot)
7280 		ret = -EBUSY;
7281 	arch_spin_unlock(&tr->max_lock);
7282 	local_irq_enable();
7283 	if (ret)
7284 		goto out;
7285 
7286 	switch (val) {
7287 	case 0:
7288 		if (iter->cpu_file != RING_BUFFER_ALL_CPUS) {
7289 			ret = -EINVAL;
7290 			break;
7291 		}
7292 		if (tr->allocated_snapshot)
7293 			free_snapshot(tr);
7294 		break;
7295 	case 1:
7296 /* Only allow per-cpu swap if the ring buffer supports it */
7297 #ifndef CONFIG_RING_BUFFER_ALLOW_SWAP
7298 		if (iter->cpu_file != RING_BUFFER_ALL_CPUS) {
7299 			ret = -EINVAL;
7300 			break;
7301 		}
7302 #endif
7303 		if (tr->allocated_snapshot)
7304 			ret = resize_buffer_duplicate_size(&tr->max_buffer,
7305 					&tr->array_buffer, iter->cpu_file);
7306 
7307 		ret = tracing_arm_snapshot_locked(tr);
7308 		if (ret)
7309 			break;
7310 
7311 		/* Now, we're going to swap */
7312 		if (iter->cpu_file == RING_BUFFER_ALL_CPUS) {
7313 			local_irq_disable();
7314 			update_max_tr(tr, current, smp_processor_id(), NULL);
7315 			local_irq_enable();
7316 		} else {
7317 			smp_call_function_single(iter->cpu_file, tracing_swap_cpu_buffer,
7318 						 (void *)tr, 1);
7319 		}
7320 		tracing_disarm_snapshot(tr);
7321 		break;
7322 	default:
7323 		if (tr->allocated_snapshot) {
7324 			if (iter->cpu_file == RING_BUFFER_ALL_CPUS)
7325 				tracing_reset_online_cpus(&tr->max_buffer);
7326 			else
7327 				tracing_reset_cpu(&tr->max_buffer, iter->cpu_file);
7328 		}
7329 		break;
7330 	}
7331 
7332 	if (ret >= 0) {
7333 		*ppos += cnt;
7334 		ret = cnt;
7335 	}
7336 out:
7337 	mutex_unlock(&trace_types_lock);
7338 	return ret;
7339 }
7340 
7341 static int tracing_snapshot_release(struct inode *inode, struct file *file)
7342 {
7343 	struct seq_file *m = file->private_data;
7344 	int ret;
7345 
7346 	ret = tracing_release(inode, file);
7347 
7348 	if (file->f_mode & FMODE_READ)
7349 		return ret;
7350 
7351 	/* If write only, the seq_file is just a stub */
7352 	if (m)
7353 		kfree(m->private);
7354 	kfree(m);
7355 
7356 	return 0;
7357 }
7358 
7359 static int tracing_buffers_open(struct inode *inode, struct file *filp);
7360 static ssize_t tracing_buffers_read(struct file *filp, char __user *ubuf,
7361 				    size_t count, loff_t *ppos);
7362 static int tracing_buffers_release(struct inode *inode, struct file *file);
7363 static ssize_t tracing_buffers_splice_read(struct file *file, loff_t *ppos,
7364 		   struct pipe_inode_info *pipe, size_t len, unsigned int flags);
7365 
7366 static int snapshot_raw_open(struct inode *inode, struct file *filp)
7367 {
7368 	struct ftrace_buffer_info *info;
7369 	int ret;
7370 
7371 	/* The following checks for tracefs lockdown */
7372 	ret = tracing_buffers_open(inode, filp);
7373 	if (ret < 0)
7374 		return ret;
7375 
7376 	info = filp->private_data;
7377 
7378 	if (info->iter.trace->use_max_tr) {
7379 		tracing_buffers_release(inode, filp);
7380 		return -EBUSY;
7381 	}
7382 
7383 	info->iter.snapshot = true;
7384 	info->iter.array_buffer = &info->iter.tr->max_buffer;
7385 
7386 	return ret;
7387 }
7388 
7389 #endif /* CONFIG_TRACER_SNAPSHOT */
7390 
7391 
7392 static const struct file_operations tracing_thresh_fops = {
7393 	.open		= tracing_open_generic,
7394 	.read		= tracing_thresh_read,
7395 	.write		= tracing_thresh_write,
7396 	.llseek		= generic_file_llseek,
7397 };
7398 
7399 #ifdef CONFIG_TRACER_MAX_TRACE
7400 static const struct file_operations tracing_max_lat_fops = {
7401 	.open		= tracing_open_generic_tr,
7402 	.read		= tracing_max_lat_read,
7403 	.write		= tracing_max_lat_write,
7404 	.llseek		= generic_file_llseek,
7405 	.release	= tracing_release_generic_tr,
7406 };
7407 #endif
7408 
7409 static const struct file_operations set_tracer_fops = {
7410 	.open		= tracing_open_generic_tr,
7411 	.read		= tracing_set_trace_read,
7412 	.write		= tracing_set_trace_write,
7413 	.llseek		= generic_file_llseek,
7414 	.release	= tracing_release_generic_tr,
7415 };
7416 
7417 static const struct file_operations tracing_pipe_fops = {
7418 	.open		= tracing_open_pipe,
7419 	.poll		= tracing_poll_pipe,
7420 	.read		= tracing_read_pipe,
7421 	.splice_read	= tracing_splice_read_pipe,
7422 	.release	= tracing_release_pipe,
7423 	.llseek		= no_llseek,
7424 };
7425 
7426 static const struct file_operations tracing_entries_fops = {
7427 	.open		= tracing_open_generic_tr,
7428 	.read		= tracing_entries_read,
7429 	.write		= tracing_entries_write,
7430 	.llseek		= generic_file_llseek,
7431 	.release	= tracing_release_generic_tr,
7432 };
7433 
7434 static const struct file_operations tracing_total_entries_fops = {
7435 	.open		= tracing_open_generic_tr,
7436 	.read		= tracing_total_entries_read,
7437 	.llseek		= generic_file_llseek,
7438 	.release	= tracing_release_generic_tr,
7439 };
7440 
7441 static const struct file_operations tracing_free_buffer_fops = {
7442 	.open		= tracing_open_generic_tr,
7443 	.write		= tracing_free_buffer_write,
7444 	.release	= tracing_free_buffer_release,
7445 };
7446 
7447 static const struct file_operations tracing_mark_fops = {
7448 	.open		= tracing_mark_open,
7449 	.write		= tracing_mark_write,
7450 	.release	= tracing_release_generic_tr,
7451 };
7452 
7453 static const struct file_operations tracing_mark_raw_fops = {
7454 	.open		= tracing_mark_open,
7455 	.write		= tracing_mark_raw_write,
7456 	.release	= tracing_release_generic_tr,
7457 };
7458 
7459 static const struct file_operations trace_clock_fops = {
7460 	.open		= tracing_clock_open,
7461 	.read		= seq_read,
7462 	.llseek		= seq_lseek,
7463 	.release	= tracing_single_release_tr,
7464 	.write		= tracing_clock_write,
7465 };
7466 
7467 static const struct file_operations trace_time_stamp_mode_fops = {
7468 	.open		= tracing_time_stamp_mode_open,
7469 	.read		= seq_read,
7470 	.llseek		= seq_lseek,
7471 	.release	= tracing_single_release_tr,
7472 };
7473 
7474 #ifdef CONFIG_TRACER_SNAPSHOT
7475 static const struct file_operations snapshot_fops = {
7476 	.open		= tracing_snapshot_open,
7477 	.read		= seq_read,
7478 	.write		= tracing_snapshot_write,
7479 	.llseek		= tracing_lseek,
7480 	.release	= tracing_snapshot_release,
7481 };
7482 
7483 static const struct file_operations snapshot_raw_fops = {
7484 	.open		= snapshot_raw_open,
7485 	.read		= tracing_buffers_read,
7486 	.release	= tracing_buffers_release,
7487 	.splice_read	= tracing_buffers_splice_read,
7488 	.llseek		= no_llseek,
7489 };
7490 
7491 #endif /* CONFIG_TRACER_SNAPSHOT */
7492 
7493 /*
7494  * trace_min_max_write - Write a u64 value to a trace_min_max_param struct
7495  * @filp: The active open file structure
7496  * @ubuf: The userspace provided buffer to read value into
7497  * @cnt: The maximum number of bytes to read
7498  * @ppos: The current "file" position
7499  *
7500  * This function implements the write interface for a struct trace_min_max_param.
7501  * The filp->private_data must point to a trace_min_max_param structure that
7502  * defines where to write the value, the min and the max acceptable values,
7503  * and a lock to protect the write.
7504  */
7505 static ssize_t
7506 trace_min_max_write(struct file *filp, const char __user *ubuf, size_t cnt, loff_t *ppos)
7507 {
7508 	struct trace_min_max_param *param = filp->private_data;
7509 	u64 val;
7510 	int err;
7511 
7512 	if (!param)
7513 		return -EFAULT;
7514 
7515 	err = kstrtoull_from_user(ubuf, cnt, 10, &val);
7516 	if (err)
7517 		return err;
7518 
7519 	if (param->lock)
7520 		mutex_lock(param->lock);
7521 
7522 	if (param->min && val < *param->min)
7523 		err = -EINVAL;
7524 
7525 	if (param->max && val > *param->max)
7526 		err = -EINVAL;
7527 
7528 	if (!err)
7529 		*param->val = val;
7530 
7531 	if (param->lock)
7532 		mutex_unlock(param->lock);
7533 
7534 	if (err)
7535 		return err;
7536 
7537 	return cnt;
7538 }
7539 
7540 /*
7541  * trace_min_max_read - Read a u64 value from a trace_min_max_param struct
7542  * @filp: The active open file structure
7543  * @ubuf: The userspace provided buffer to read value into
7544  * @cnt: The maximum number of bytes to read
7545  * @ppos: The current "file" position
7546  *
7547  * This function implements the read interface for a struct trace_min_max_param.
7548  * The filp->private_data must point to a trace_min_max_param struct with valid
7549  * data.
7550  */
7551 static ssize_t
7552 trace_min_max_read(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos)
7553 {
7554 	struct trace_min_max_param *param = filp->private_data;
7555 	char buf[U64_STR_SIZE];
7556 	int len;
7557 	u64 val;
7558 
7559 	if (!param)
7560 		return -EFAULT;
7561 
7562 	val = *param->val;
7563 
7564 	if (cnt > sizeof(buf))
7565 		cnt = sizeof(buf);
7566 
7567 	len = snprintf(buf, sizeof(buf), "%llu\n", val);
7568 
7569 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, len);
7570 }
7571 
7572 const struct file_operations trace_min_max_fops = {
7573 	.open		= tracing_open_generic,
7574 	.read		= trace_min_max_read,
7575 	.write		= trace_min_max_write,
7576 };
7577 
7578 #define TRACING_LOG_ERRS_MAX	8
7579 #define TRACING_LOG_LOC_MAX	128
7580 
7581 #define CMD_PREFIX "  Command: "
7582 
7583 struct err_info {
7584 	const char	**errs;	/* ptr to loc-specific array of err strings */
7585 	u8		type;	/* index into errs -> specific err string */
7586 	u16		pos;	/* caret position */
7587 	u64		ts;
7588 };
7589 
7590 struct tracing_log_err {
7591 	struct list_head	list;
7592 	struct err_info		info;
7593 	char			loc[TRACING_LOG_LOC_MAX]; /* err location */
7594 	char			*cmd;                     /* what caused err */
7595 };
7596 
7597 static DEFINE_MUTEX(tracing_err_log_lock);
7598 
7599 static struct tracing_log_err *alloc_tracing_log_err(int len)
7600 {
7601 	struct tracing_log_err *err;
7602 
7603 	err = kzalloc(sizeof(*err), GFP_KERNEL);
7604 	if (!err)
7605 		return ERR_PTR(-ENOMEM);
7606 
7607 	err->cmd = kzalloc(len, GFP_KERNEL);
7608 	if (!err->cmd) {
7609 		kfree(err);
7610 		return ERR_PTR(-ENOMEM);
7611 	}
7612 
7613 	return err;
7614 }
7615 
7616 static void free_tracing_log_err(struct tracing_log_err *err)
7617 {
7618 	kfree(err->cmd);
7619 	kfree(err);
7620 }
7621 
7622 static struct tracing_log_err *get_tracing_log_err(struct trace_array *tr,
7623 						   int len)
7624 {
7625 	struct tracing_log_err *err;
7626 	char *cmd;
7627 
7628 	if (tr->n_err_log_entries < TRACING_LOG_ERRS_MAX) {
7629 		err = alloc_tracing_log_err(len);
7630 		if (PTR_ERR(err) != -ENOMEM)
7631 			tr->n_err_log_entries++;
7632 
7633 		return err;
7634 	}
7635 	cmd = kzalloc(len, GFP_KERNEL);
7636 	if (!cmd)
7637 		return ERR_PTR(-ENOMEM);
7638 	err = list_first_entry(&tr->err_log, struct tracing_log_err, list);
7639 	kfree(err->cmd);
7640 	err->cmd = cmd;
7641 	list_del(&err->list);
7642 
7643 	return err;
7644 }
7645 
7646 /**
7647  * err_pos - find the position of a string within a command for error careting
7648  * @cmd: The tracing command that caused the error
7649  * @str: The string to position the caret at within @cmd
7650  *
7651  * Finds the position of the first occurrence of @str within @cmd.  The
7652  * return value can be passed to tracing_log_err() for caret placement
7653  * within @cmd.
7654  *
7655  * Returns the index within @cmd of the first occurrence of @str or 0
7656  * if @str was not found.
7657  */
7658 unsigned int err_pos(char *cmd, const char *str)
7659 {
7660 	char *found;
7661 
7662 	if (WARN_ON(!strlen(cmd)))
7663 		return 0;
7664 
7665 	found = strstr(cmd, str);
7666 	if (found)
7667 		return found - cmd;
7668 
7669 	return 0;
7670 }
7671 
7672 /**
7673  * tracing_log_err - write an error to the tracing error log
7674  * @tr: The associated trace array for the error (NULL for top level array)
7675  * @loc: A string describing where the error occurred
7676  * @cmd: The tracing command that caused the error
7677  * @errs: The array of loc-specific static error strings
7678  * @type: The index into errs[], which produces the specific static err string
7679  * @pos: The position the caret should be placed in the cmd
7680  *
7681  * Writes an error into tracing/error_log of the form:
7682  *
7683  * <loc>: error: <text>
7684  *   Command: <cmd>
7685  *              ^
7686  *
7687  * tracing/error_log is a small log file containing the last
7688  * TRACING_LOG_ERRS_MAX errors (8).  Memory for errors isn't allocated
7689  * unless there has been a tracing error, and the error log can be
7690  * cleared and have its memory freed by writing the empty string in
7691  * truncation mode to it i.e. echo > tracing/error_log.
7692  *
7693  * NOTE: the @errs array along with the @type param are used to
7694  * produce a static error string - this string is not copied and saved
7695  * when the error is logged - only a pointer to it is saved.  See
7696  * existing callers for examples of how static strings are typically
7697  * defined for use with tracing_log_err().
7698  */
7699 void tracing_log_err(struct trace_array *tr,
7700 		     const char *loc, const char *cmd,
7701 		     const char **errs, u8 type, u16 pos)
7702 {
7703 	struct tracing_log_err *err;
7704 	int len = 0;
7705 
7706 	if (!tr)
7707 		tr = &global_trace;
7708 
7709 	len += sizeof(CMD_PREFIX) + 2 * sizeof("\n") + strlen(cmd) + 1;
7710 
7711 	mutex_lock(&tracing_err_log_lock);
7712 	err = get_tracing_log_err(tr, len);
7713 	if (PTR_ERR(err) == -ENOMEM) {
7714 		mutex_unlock(&tracing_err_log_lock);
7715 		return;
7716 	}
7717 
7718 	snprintf(err->loc, TRACING_LOG_LOC_MAX, "%s: error: ", loc);
7719 	snprintf(err->cmd, len, "\n" CMD_PREFIX "%s\n", cmd);
7720 
7721 	err->info.errs = errs;
7722 	err->info.type = type;
7723 	err->info.pos = pos;
7724 	err->info.ts = local_clock();
7725 
7726 	list_add_tail(&err->list, &tr->err_log);
7727 	mutex_unlock(&tracing_err_log_lock);
7728 }
7729 
7730 static void clear_tracing_err_log(struct trace_array *tr)
7731 {
7732 	struct tracing_log_err *err, *next;
7733 
7734 	mutex_lock(&tracing_err_log_lock);
7735 	list_for_each_entry_safe(err, next, &tr->err_log, list) {
7736 		list_del(&err->list);
7737 		free_tracing_log_err(err);
7738 	}
7739 
7740 	tr->n_err_log_entries = 0;
7741 	mutex_unlock(&tracing_err_log_lock);
7742 }
7743 
7744 static void *tracing_err_log_seq_start(struct seq_file *m, loff_t *pos)
7745 {
7746 	struct trace_array *tr = m->private;
7747 
7748 	mutex_lock(&tracing_err_log_lock);
7749 
7750 	return seq_list_start(&tr->err_log, *pos);
7751 }
7752 
7753 static void *tracing_err_log_seq_next(struct seq_file *m, void *v, loff_t *pos)
7754 {
7755 	struct trace_array *tr = m->private;
7756 
7757 	return seq_list_next(v, &tr->err_log, pos);
7758 }
7759 
7760 static void tracing_err_log_seq_stop(struct seq_file *m, void *v)
7761 {
7762 	mutex_unlock(&tracing_err_log_lock);
7763 }
7764 
7765 static void tracing_err_log_show_pos(struct seq_file *m, u16 pos)
7766 {
7767 	u16 i;
7768 
7769 	for (i = 0; i < sizeof(CMD_PREFIX) - 1; i++)
7770 		seq_putc(m, ' ');
7771 	for (i = 0; i < pos; i++)
7772 		seq_putc(m, ' ');
7773 	seq_puts(m, "^\n");
7774 }
7775 
7776 static int tracing_err_log_seq_show(struct seq_file *m, void *v)
7777 {
7778 	struct tracing_log_err *err = v;
7779 
7780 	if (err) {
7781 		const char *err_text = err->info.errs[err->info.type];
7782 		u64 sec = err->info.ts;
7783 		u32 nsec;
7784 
7785 		nsec = do_div(sec, NSEC_PER_SEC);
7786 		seq_printf(m, "[%5llu.%06u] %s%s", sec, nsec / 1000,
7787 			   err->loc, err_text);
7788 		seq_printf(m, "%s", err->cmd);
7789 		tracing_err_log_show_pos(m, err->info.pos);
7790 	}
7791 
7792 	return 0;
7793 }
7794 
7795 static const struct seq_operations tracing_err_log_seq_ops = {
7796 	.start  = tracing_err_log_seq_start,
7797 	.next   = tracing_err_log_seq_next,
7798 	.stop   = tracing_err_log_seq_stop,
7799 	.show   = tracing_err_log_seq_show
7800 };
7801 
7802 static int tracing_err_log_open(struct inode *inode, struct file *file)
7803 {
7804 	struct trace_array *tr = inode->i_private;
7805 	int ret = 0;
7806 
7807 	ret = tracing_check_open_get_tr(tr);
7808 	if (ret)
7809 		return ret;
7810 
7811 	/* If this file was opened for write, then erase contents */
7812 	if ((file->f_mode & FMODE_WRITE) && (file->f_flags & O_TRUNC))
7813 		clear_tracing_err_log(tr);
7814 
7815 	if (file->f_mode & FMODE_READ) {
7816 		ret = seq_open(file, &tracing_err_log_seq_ops);
7817 		if (!ret) {
7818 			struct seq_file *m = file->private_data;
7819 			m->private = tr;
7820 		} else {
7821 			trace_array_put(tr);
7822 		}
7823 	}
7824 	return ret;
7825 }
7826 
7827 static ssize_t tracing_err_log_write(struct file *file,
7828 				     const char __user *buffer,
7829 				     size_t count, loff_t *ppos)
7830 {
7831 	return count;
7832 }
7833 
7834 static int tracing_err_log_release(struct inode *inode, struct file *file)
7835 {
7836 	struct trace_array *tr = inode->i_private;
7837 
7838 	trace_array_put(tr);
7839 
7840 	if (file->f_mode & FMODE_READ)
7841 		seq_release(inode, file);
7842 
7843 	return 0;
7844 }
7845 
7846 static const struct file_operations tracing_err_log_fops = {
7847 	.open           = tracing_err_log_open,
7848 	.write		= tracing_err_log_write,
7849 	.read           = seq_read,
7850 	.llseek         = tracing_lseek,
7851 	.release        = tracing_err_log_release,
7852 };
7853 
7854 static int tracing_buffers_open(struct inode *inode, struct file *filp)
7855 {
7856 	struct trace_array *tr = inode->i_private;
7857 	struct ftrace_buffer_info *info;
7858 	int ret;
7859 
7860 	ret = tracing_check_open_get_tr(tr);
7861 	if (ret)
7862 		return ret;
7863 
7864 	info = kvzalloc(sizeof(*info), GFP_KERNEL);
7865 	if (!info) {
7866 		trace_array_put(tr);
7867 		return -ENOMEM;
7868 	}
7869 
7870 	mutex_lock(&trace_types_lock);
7871 
7872 	info->iter.tr		= tr;
7873 	info->iter.cpu_file	= tracing_get_cpu(inode);
7874 	info->iter.trace	= tr->current_trace;
7875 	info->iter.array_buffer = &tr->array_buffer;
7876 	info->spare		= NULL;
7877 	/* Force reading ring buffer for first read */
7878 	info->read		= (unsigned int)-1;
7879 
7880 	filp->private_data = info;
7881 
7882 	tr->trace_ref++;
7883 
7884 	mutex_unlock(&trace_types_lock);
7885 
7886 	ret = nonseekable_open(inode, filp);
7887 	if (ret < 0)
7888 		trace_array_put(tr);
7889 
7890 	return ret;
7891 }
7892 
7893 static __poll_t
7894 tracing_buffers_poll(struct file *filp, poll_table *poll_table)
7895 {
7896 	struct ftrace_buffer_info *info = filp->private_data;
7897 	struct trace_iterator *iter = &info->iter;
7898 
7899 	return trace_poll(iter, filp, poll_table);
7900 }
7901 
7902 static ssize_t
7903 tracing_buffers_read(struct file *filp, char __user *ubuf,
7904 		     size_t count, loff_t *ppos)
7905 {
7906 	struct ftrace_buffer_info *info = filp->private_data;
7907 	struct trace_iterator *iter = &info->iter;
7908 	void *trace_data;
7909 	int page_size;
7910 	ssize_t ret = 0;
7911 	ssize_t size;
7912 
7913 	if (!count)
7914 		return 0;
7915 
7916 #ifdef CONFIG_TRACER_MAX_TRACE
7917 	if (iter->snapshot && iter->tr->current_trace->use_max_tr)
7918 		return -EBUSY;
7919 #endif
7920 
7921 	page_size = ring_buffer_subbuf_size_get(iter->array_buffer->buffer);
7922 
7923 	/* Make sure the spare matches the current sub buffer size */
7924 	if (info->spare) {
7925 		if (page_size != info->spare_size) {
7926 			ring_buffer_free_read_page(iter->array_buffer->buffer,
7927 						   info->spare_cpu, info->spare);
7928 			info->spare = NULL;
7929 		}
7930 	}
7931 
7932 	if (!info->spare) {
7933 		info->spare = ring_buffer_alloc_read_page(iter->array_buffer->buffer,
7934 							  iter->cpu_file);
7935 		if (IS_ERR(info->spare)) {
7936 			ret = PTR_ERR(info->spare);
7937 			info->spare = NULL;
7938 		} else {
7939 			info->spare_cpu = iter->cpu_file;
7940 			info->spare_size = page_size;
7941 		}
7942 	}
7943 	if (!info->spare)
7944 		return ret;
7945 
7946 	/* Do we have previous read data to read? */
7947 	if (info->read < page_size)
7948 		goto read;
7949 
7950  again:
7951 	trace_access_lock(iter->cpu_file);
7952 	ret = ring_buffer_read_page(iter->array_buffer->buffer,
7953 				    info->spare,
7954 				    count,
7955 				    iter->cpu_file, 0);
7956 	trace_access_unlock(iter->cpu_file);
7957 
7958 	if (ret < 0) {
7959 		if (trace_empty(iter)) {
7960 			if ((filp->f_flags & O_NONBLOCK))
7961 				return -EAGAIN;
7962 
7963 			ret = wait_on_pipe(iter, 0);
7964 			if (ret)
7965 				return ret;
7966 
7967 			goto again;
7968 		}
7969 		return 0;
7970 	}
7971 
7972 	info->read = 0;
7973  read:
7974 	size = page_size - info->read;
7975 	if (size > count)
7976 		size = count;
7977 	trace_data = ring_buffer_read_page_data(info->spare);
7978 	ret = copy_to_user(ubuf, trace_data + info->read, size);
7979 	if (ret == size)
7980 		return -EFAULT;
7981 
7982 	size -= ret;
7983 
7984 	*ppos += size;
7985 	info->read += size;
7986 
7987 	return size;
7988 }
7989 
7990 static int tracing_buffers_flush(struct file *file, fl_owner_t id)
7991 {
7992 	struct ftrace_buffer_info *info = file->private_data;
7993 	struct trace_iterator *iter = &info->iter;
7994 
7995 	iter->closed = true;
7996 	/* Make sure the waiters see the new wait_index */
7997 	(void)atomic_fetch_inc_release(&iter->wait_index);
7998 
7999 	ring_buffer_wake_waiters(iter->array_buffer->buffer, iter->cpu_file);
8000 
8001 	return 0;
8002 }
8003 
8004 static int tracing_buffers_release(struct inode *inode, struct file *file)
8005 {
8006 	struct ftrace_buffer_info *info = file->private_data;
8007 	struct trace_iterator *iter = &info->iter;
8008 
8009 	mutex_lock(&trace_types_lock);
8010 
8011 	iter->tr->trace_ref--;
8012 
8013 	__trace_array_put(iter->tr);
8014 
8015 	if (info->spare)
8016 		ring_buffer_free_read_page(iter->array_buffer->buffer,
8017 					   info->spare_cpu, info->spare);
8018 	kvfree(info);
8019 
8020 	mutex_unlock(&trace_types_lock);
8021 
8022 	return 0;
8023 }
8024 
8025 struct buffer_ref {
8026 	struct trace_buffer	*buffer;
8027 	void			*page;
8028 	int			cpu;
8029 	refcount_t		refcount;
8030 };
8031 
8032 static void buffer_ref_release(struct buffer_ref *ref)
8033 {
8034 	if (!refcount_dec_and_test(&ref->refcount))
8035 		return;
8036 	ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page);
8037 	kfree(ref);
8038 }
8039 
8040 static void buffer_pipe_buf_release(struct pipe_inode_info *pipe,
8041 				    struct pipe_buffer *buf)
8042 {
8043 	struct buffer_ref *ref = (struct buffer_ref *)buf->private;
8044 
8045 	buffer_ref_release(ref);
8046 	buf->private = 0;
8047 }
8048 
8049 static bool buffer_pipe_buf_get(struct pipe_inode_info *pipe,
8050 				struct pipe_buffer *buf)
8051 {
8052 	struct buffer_ref *ref = (struct buffer_ref *)buf->private;
8053 
8054 	if (refcount_read(&ref->refcount) > INT_MAX/2)
8055 		return false;
8056 
8057 	refcount_inc(&ref->refcount);
8058 	return true;
8059 }
8060 
8061 /* Pipe buffer operations for a buffer. */
8062 static const struct pipe_buf_operations buffer_pipe_buf_ops = {
8063 	.release		= buffer_pipe_buf_release,
8064 	.get			= buffer_pipe_buf_get,
8065 };
8066 
8067 /*
8068  * Callback from splice_to_pipe(), if we need to release some pages
8069  * at the end of the spd in case we error'ed out in filling the pipe.
8070  */
8071 static void buffer_spd_release(struct splice_pipe_desc *spd, unsigned int i)
8072 {
8073 	struct buffer_ref *ref =
8074 		(struct buffer_ref *)spd->partial[i].private;
8075 
8076 	buffer_ref_release(ref);
8077 	spd->partial[i].private = 0;
8078 }
8079 
8080 static ssize_t
8081 tracing_buffers_splice_read(struct file *file, loff_t *ppos,
8082 			    struct pipe_inode_info *pipe, size_t len,
8083 			    unsigned int flags)
8084 {
8085 	struct ftrace_buffer_info *info = file->private_data;
8086 	struct trace_iterator *iter = &info->iter;
8087 	struct partial_page partial_def[PIPE_DEF_BUFFERS];
8088 	struct page *pages_def[PIPE_DEF_BUFFERS];
8089 	struct splice_pipe_desc spd = {
8090 		.pages		= pages_def,
8091 		.partial	= partial_def,
8092 		.nr_pages_max	= PIPE_DEF_BUFFERS,
8093 		.ops		= &buffer_pipe_buf_ops,
8094 		.spd_release	= buffer_spd_release,
8095 	};
8096 	struct buffer_ref *ref;
8097 	bool woken = false;
8098 	int page_size;
8099 	int entries, i;
8100 	ssize_t ret = 0;
8101 
8102 #ifdef CONFIG_TRACER_MAX_TRACE
8103 	if (iter->snapshot && iter->tr->current_trace->use_max_tr)
8104 		return -EBUSY;
8105 #endif
8106 
8107 	page_size = ring_buffer_subbuf_size_get(iter->array_buffer->buffer);
8108 	if (*ppos & (page_size - 1))
8109 		return -EINVAL;
8110 
8111 	if (len & (page_size - 1)) {
8112 		if (len < page_size)
8113 			return -EINVAL;
8114 		len &= (~(page_size - 1));
8115 	}
8116 
8117 	if (splice_grow_spd(pipe, &spd))
8118 		return -ENOMEM;
8119 
8120  again:
8121 	trace_access_lock(iter->cpu_file);
8122 	entries = ring_buffer_entries_cpu(iter->array_buffer->buffer, iter->cpu_file);
8123 
8124 	for (i = 0; i < spd.nr_pages_max && len && entries; i++, len -= page_size) {
8125 		struct page *page;
8126 		int r;
8127 
8128 		ref = kzalloc(sizeof(*ref), GFP_KERNEL);
8129 		if (!ref) {
8130 			ret = -ENOMEM;
8131 			break;
8132 		}
8133 
8134 		refcount_set(&ref->refcount, 1);
8135 		ref->buffer = iter->array_buffer->buffer;
8136 		ref->page = ring_buffer_alloc_read_page(ref->buffer, iter->cpu_file);
8137 		if (IS_ERR(ref->page)) {
8138 			ret = PTR_ERR(ref->page);
8139 			ref->page = NULL;
8140 			kfree(ref);
8141 			break;
8142 		}
8143 		ref->cpu = iter->cpu_file;
8144 
8145 		r = ring_buffer_read_page(ref->buffer, ref->page,
8146 					  len, iter->cpu_file, 1);
8147 		if (r < 0) {
8148 			ring_buffer_free_read_page(ref->buffer, ref->cpu,
8149 						   ref->page);
8150 			kfree(ref);
8151 			break;
8152 		}
8153 
8154 		page = virt_to_page(ring_buffer_read_page_data(ref->page));
8155 
8156 		spd.pages[i] = page;
8157 		spd.partial[i].len = page_size;
8158 		spd.partial[i].offset = 0;
8159 		spd.partial[i].private = (unsigned long)ref;
8160 		spd.nr_pages++;
8161 		*ppos += page_size;
8162 
8163 		entries = ring_buffer_entries_cpu(iter->array_buffer->buffer, iter->cpu_file);
8164 	}
8165 
8166 	trace_access_unlock(iter->cpu_file);
8167 	spd.nr_pages = i;
8168 
8169 	/* did we read anything? */
8170 	if (!spd.nr_pages) {
8171 
8172 		if (ret)
8173 			goto out;
8174 
8175 		if (woken)
8176 			goto out;
8177 
8178 		ret = -EAGAIN;
8179 		if ((file->f_flags & O_NONBLOCK) || (flags & SPLICE_F_NONBLOCK))
8180 			goto out;
8181 
8182 		ret = wait_on_pipe(iter, iter->snapshot ? 0 : iter->tr->buffer_percent);
8183 		if (ret)
8184 			goto out;
8185 
8186 		/* No need to wait after waking up when tracing is off */
8187 		if (!tracer_tracing_is_on(iter->tr))
8188 			goto out;
8189 
8190 		/* Iterate one more time to collect any new data then exit */
8191 		woken = true;
8192 
8193 		goto again;
8194 	}
8195 
8196 	ret = splice_to_pipe(pipe, &spd);
8197 out:
8198 	splice_shrink_spd(&spd);
8199 
8200 	return ret;
8201 }
8202 
8203 static long tracing_buffers_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
8204 {
8205 	struct ftrace_buffer_info *info = file->private_data;
8206 	struct trace_iterator *iter = &info->iter;
8207 	int err;
8208 
8209 	if (cmd == TRACE_MMAP_IOCTL_GET_READER) {
8210 		if (!(file->f_flags & O_NONBLOCK)) {
8211 			err = ring_buffer_wait(iter->array_buffer->buffer,
8212 					       iter->cpu_file,
8213 					       iter->tr->buffer_percent,
8214 					       NULL, NULL);
8215 			if (err)
8216 				return err;
8217 		}
8218 
8219 		return ring_buffer_map_get_reader(iter->array_buffer->buffer,
8220 						  iter->cpu_file);
8221 	} else if (cmd) {
8222 		return -ENOTTY;
8223 	}
8224 
8225 	/*
8226 	 * An ioctl call with cmd 0 to the ring buffer file will wake up all
8227 	 * waiters
8228 	 */
8229 	mutex_lock(&trace_types_lock);
8230 
8231 	/* Make sure the waiters see the new wait_index */
8232 	(void)atomic_fetch_inc_release(&iter->wait_index);
8233 
8234 	ring_buffer_wake_waiters(iter->array_buffer->buffer, iter->cpu_file);
8235 
8236 	mutex_unlock(&trace_types_lock);
8237 	return 0;
8238 }
8239 
8240 #ifdef CONFIG_TRACER_MAX_TRACE
8241 static int get_snapshot_map(struct trace_array *tr)
8242 {
8243 	int err = 0;
8244 
8245 	/*
8246 	 * Called with mmap_lock held. lockdep would be unhappy if we would now
8247 	 * take trace_types_lock. Instead use the specific
8248 	 * snapshot_trigger_lock.
8249 	 */
8250 	spin_lock(&tr->snapshot_trigger_lock);
8251 
8252 	if (tr->snapshot || tr->mapped == UINT_MAX)
8253 		err = -EBUSY;
8254 	else
8255 		tr->mapped++;
8256 
8257 	spin_unlock(&tr->snapshot_trigger_lock);
8258 
8259 	/* Wait for update_max_tr() to observe iter->tr->mapped */
8260 	if (tr->mapped == 1)
8261 		synchronize_rcu();
8262 
8263 	return err;
8264 
8265 }
8266 static void put_snapshot_map(struct trace_array *tr)
8267 {
8268 	spin_lock(&tr->snapshot_trigger_lock);
8269 	if (!WARN_ON(!tr->mapped))
8270 		tr->mapped--;
8271 	spin_unlock(&tr->snapshot_trigger_lock);
8272 }
8273 #else
8274 static inline int get_snapshot_map(struct trace_array *tr) { return 0; }
8275 static inline void put_snapshot_map(struct trace_array *tr) { }
8276 #endif
8277 
8278 static void tracing_buffers_mmap_close(struct vm_area_struct *vma)
8279 {
8280 	struct ftrace_buffer_info *info = vma->vm_file->private_data;
8281 	struct trace_iterator *iter = &info->iter;
8282 
8283 	WARN_ON(ring_buffer_unmap(iter->array_buffer->buffer, iter->cpu_file));
8284 	put_snapshot_map(iter->tr);
8285 }
8286 
8287 static const struct vm_operations_struct tracing_buffers_vmops = {
8288 	.close		= tracing_buffers_mmap_close,
8289 };
8290 
8291 static int tracing_buffers_mmap(struct file *filp, struct vm_area_struct *vma)
8292 {
8293 	struct ftrace_buffer_info *info = filp->private_data;
8294 	struct trace_iterator *iter = &info->iter;
8295 	int ret = 0;
8296 
8297 	ret = get_snapshot_map(iter->tr);
8298 	if (ret)
8299 		return ret;
8300 
8301 	ret = ring_buffer_map(iter->array_buffer->buffer, iter->cpu_file, vma);
8302 	if (ret)
8303 		put_snapshot_map(iter->tr);
8304 
8305 	vma->vm_ops = &tracing_buffers_vmops;
8306 
8307 	return ret;
8308 }
8309 
8310 static const struct file_operations tracing_buffers_fops = {
8311 	.open		= tracing_buffers_open,
8312 	.read		= tracing_buffers_read,
8313 	.poll		= tracing_buffers_poll,
8314 	.release	= tracing_buffers_release,
8315 	.flush		= tracing_buffers_flush,
8316 	.splice_read	= tracing_buffers_splice_read,
8317 	.unlocked_ioctl = tracing_buffers_ioctl,
8318 	.llseek		= no_llseek,
8319 	.mmap		= tracing_buffers_mmap,
8320 };
8321 
8322 static ssize_t
8323 tracing_stats_read(struct file *filp, char __user *ubuf,
8324 		   size_t count, loff_t *ppos)
8325 {
8326 	struct inode *inode = file_inode(filp);
8327 	struct trace_array *tr = inode->i_private;
8328 	struct array_buffer *trace_buf = &tr->array_buffer;
8329 	int cpu = tracing_get_cpu(inode);
8330 	struct trace_seq *s;
8331 	unsigned long cnt;
8332 	unsigned long long t;
8333 	unsigned long usec_rem;
8334 
8335 	s = kmalloc(sizeof(*s), GFP_KERNEL);
8336 	if (!s)
8337 		return -ENOMEM;
8338 
8339 	trace_seq_init(s);
8340 
8341 	cnt = ring_buffer_entries_cpu(trace_buf->buffer, cpu);
8342 	trace_seq_printf(s, "entries: %ld\n", cnt);
8343 
8344 	cnt = ring_buffer_overrun_cpu(trace_buf->buffer, cpu);
8345 	trace_seq_printf(s, "overrun: %ld\n", cnt);
8346 
8347 	cnt = ring_buffer_commit_overrun_cpu(trace_buf->buffer, cpu);
8348 	trace_seq_printf(s, "commit overrun: %ld\n", cnt);
8349 
8350 	cnt = ring_buffer_bytes_cpu(trace_buf->buffer, cpu);
8351 	trace_seq_printf(s, "bytes: %ld\n", cnt);
8352 
8353 	if (trace_clocks[tr->clock_id].in_ns) {
8354 		/* local or global for trace_clock */
8355 		t = ns2usecs(ring_buffer_oldest_event_ts(trace_buf->buffer, cpu));
8356 		usec_rem = do_div(t, USEC_PER_SEC);
8357 		trace_seq_printf(s, "oldest event ts: %5llu.%06lu\n",
8358 								t, usec_rem);
8359 
8360 		t = ns2usecs(ring_buffer_time_stamp(trace_buf->buffer));
8361 		usec_rem = do_div(t, USEC_PER_SEC);
8362 		trace_seq_printf(s, "now ts: %5llu.%06lu\n", t, usec_rem);
8363 	} else {
8364 		/* counter or tsc mode for trace_clock */
8365 		trace_seq_printf(s, "oldest event ts: %llu\n",
8366 				ring_buffer_oldest_event_ts(trace_buf->buffer, cpu));
8367 
8368 		trace_seq_printf(s, "now ts: %llu\n",
8369 				ring_buffer_time_stamp(trace_buf->buffer));
8370 	}
8371 
8372 	cnt = ring_buffer_dropped_events_cpu(trace_buf->buffer, cpu);
8373 	trace_seq_printf(s, "dropped events: %ld\n", cnt);
8374 
8375 	cnt = ring_buffer_read_events_cpu(trace_buf->buffer, cpu);
8376 	trace_seq_printf(s, "read events: %ld\n", cnt);
8377 
8378 	count = simple_read_from_buffer(ubuf, count, ppos,
8379 					s->buffer, trace_seq_used(s));
8380 
8381 	kfree(s);
8382 
8383 	return count;
8384 }
8385 
8386 static const struct file_operations tracing_stats_fops = {
8387 	.open		= tracing_open_generic_tr,
8388 	.read		= tracing_stats_read,
8389 	.llseek		= generic_file_llseek,
8390 	.release	= tracing_release_generic_tr,
8391 };
8392 
8393 #ifdef CONFIG_DYNAMIC_FTRACE
8394 
8395 static ssize_t
8396 tracing_read_dyn_info(struct file *filp, char __user *ubuf,
8397 		  size_t cnt, loff_t *ppos)
8398 {
8399 	ssize_t ret;
8400 	char *buf;
8401 	int r;
8402 
8403 	/* 256 should be plenty to hold the amount needed */
8404 	buf = kmalloc(256, GFP_KERNEL);
8405 	if (!buf)
8406 		return -ENOMEM;
8407 
8408 	r = scnprintf(buf, 256, "%ld pages:%ld groups: %ld\n",
8409 		      ftrace_update_tot_cnt,
8410 		      ftrace_number_of_pages,
8411 		      ftrace_number_of_groups);
8412 
8413 	ret = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
8414 	kfree(buf);
8415 	return ret;
8416 }
8417 
8418 static const struct file_operations tracing_dyn_info_fops = {
8419 	.open		= tracing_open_generic,
8420 	.read		= tracing_read_dyn_info,
8421 	.llseek		= generic_file_llseek,
8422 };
8423 #endif /* CONFIG_DYNAMIC_FTRACE */
8424 
8425 #if defined(CONFIG_TRACER_SNAPSHOT) && defined(CONFIG_DYNAMIC_FTRACE)
8426 static void
8427 ftrace_snapshot(unsigned long ip, unsigned long parent_ip,
8428 		struct trace_array *tr, struct ftrace_probe_ops *ops,
8429 		void *data)
8430 {
8431 	tracing_snapshot_instance(tr);
8432 }
8433 
8434 static void
8435 ftrace_count_snapshot(unsigned long ip, unsigned long parent_ip,
8436 		      struct trace_array *tr, struct ftrace_probe_ops *ops,
8437 		      void *data)
8438 {
8439 	struct ftrace_func_mapper *mapper = data;
8440 	long *count = NULL;
8441 
8442 	if (mapper)
8443 		count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
8444 
8445 	if (count) {
8446 
8447 		if (*count <= 0)
8448 			return;
8449 
8450 		(*count)--;
8451 	}
8452 
8453 	tracing_snapshot_instance(tr);
8454 }
8455 
8456 static int
8457 ftrace_snapshot_print(struct seq_file *m, unsigned long ip,
8458 		      struct ftrace_probe_ops *ops, void *data)
8459 {
8460 	struct ftrace_func_mapper *mapper = data;
8461 	long *count = NULL;
8462 
8463 	seq_printf(m, "%ps:", (void *)ip);
8464 
8465 	seq_puts(m, "snapshot");
8466 
8467 	if (mapper)
8468 		count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
8469 
8470 	if (count)
8471 		seq_printf(m, ":count=%ld\n", *count);
8472 	else
8473 		seq_puts(m, ":unlimited\n");
8474 
8475 	return 0;
8476 }
8477 
8478 static int
8479 ftrace_snapshot_init(struct ftrace_probe_ops *ops, struct trace_array *tr,
8480 		     unsigned long ip, void *init_data, void **data)
8481 {
8482 	struct ftrace_func_mapper *mapper = *data;
8483 
8484 	if (!mapper) {
8485 		mapper = allocate_ftrace_func_mapper();
8486 		if (!mapper)
8487 			return -ENOMEM;
8488 		*data = mapper;
8489 	}
8490 
8491 	return ftrace_func_mapper_add_ip(mapper, ip, init_data);
8492 }
8493 
8494 static void
8495 ftrace_snapshot_free(struct ftrace_probe_ops *ops, struct trace_array *tr,
8496 		     unsigned long ip, void *data)
8497 {
8498 	struct ftrace_func_mapper *mapper = data;
8499 
8500 	if (!ip) {
8501 		if (!mapper)
8502 			return;
8503 		free_ftrace_func_mapper(mapper, NULL);
8504 		return;
8505 	}
8506 
8507 	ftrace_func_mapper_remove_ip(mapper, ip);
8508 }
8509 
8510 static struct ftrace_probe_ops snapshot_probe_ops = {
8511 	.func			= ftrace_snapshot,
8512 	.print			= ftrace_snapshot_print,
8513 };
8514 
8515 static struct ftrace_probe_ops snapshot_count_probe_ops = {
8516 	.func			= ftrace_count_snapshot,
8517 	.print			= ftrace_snapshot_print,
8518 	.init			= ftrace_snapshot_init,
8519 	.free			= ftrace_snapshot_free,
8520 };
8521 
8522 static int
8523 ftrace_trace_snapshot_callback(struct trace_array *tr, struct ftrace_hash *hash,
8524 			       char *glob, char *cmd, char *param, int enable)
8525 {
8526 	struct ftrace_probe_ops *ops;
8527 	void *count = (void *)-1;
8528 	char *number;
8529 	int ret;
8530 
8531 	if (!tr)
8532 		return -ENODEV;
8533 
8534 	/* hash funcs only work with set_ftrace_filter */
8535 	if (!enable)
8536 		return -EINVAL;
8537 
8538 	ops = param ? &snapshot_count_probe_ops :  &snapshot_probe_ops;
8539 
8540 	if (glob[0] == '!') {
8541 		ret = unregister_ftrace_function_probe_func(glob+1, tr, ops);
8542 		if (!ret)
8543 			tracing_disarm_snapshot(tr);
8544 
8545 		return ret;
8546 	}
8547 
8548 	if (!param)
8549 		goto out_reg;
8550 
8551 	number = strsep(&param, ":");
8552 
8553 	if (!strlen(number))
8554 		goto out_reg;
8555 
8556 	/*
8557 	 * We use the callback data field (which is a pointer)
8558 	 * as our counter.
8559 	 */
8560 	ret = kstrtoul(number, 0, (unsigned long *)&count);
8561 	if (ret)
8562 		return ret;
8563 
8564  out_reg:
8565 	ret = tracing_arm_snapshot(tr);
8566 	if (ret < 0)
8567 		goto out;
8568 
8569 	ret = register_ftrace_function_probe(glob, tr, ops, count);
8570 	if (ret < 0)
8571 		tracing_disarm_snapshot(tr);
8572  out:
8573 	return ret < 0 ? ret : 0;
8574 }
8575 
8576 static struct ftrace_func_command ftrace_snapshot_cmd = {
8577 	.name			= "snapshot",
8578 	.func			= ftrace_trace_snapshot_callback,
8579 };
8580 
8581 static __init int register_snapshot_cmd(void)
8582 {
8583 	return register_ftrace_command(&ftrace_snapshot_cmd);
8584 }
8585 #else
8586 static inline __init int register_snapshot_cmd(void) { return 0; }
8587 #endif /* defined(CONFIG_TRACER_SNAPSHOT) && defined(CONFIG_DYNAMIC_FTRACE) */
8588 
8589 static struct dentry *tracing_get_dentry(struct trace_array *tr)
8590 {
8591 	if (WARN_ON(!tr->dir))
8592 		return ERR_PTR(-ENODEV);
8593 
8594 	/* Top directory uses NULL as the parent */
8595 	if (tr->flags & TRACE_ARRAY_FL_GLOBAL)
8596 		return NULL;
8597 
8598 	/* All sub buffers have a descriptor */
8599 	return tr->dir;
8600 }
8601 
8602 static struct dentry *tracing_dentry_percpu(struct trace_array *tr, int cpu)
8603 {
8604 	struct dentry *d_tracer;
8605 
8606 	if (tr->percpu_dir)
8607 		return tr->percpu_dir;
8608 
8609 	d_tracer = tracing_get_dentry(tr);
8610 	if (IS_ERR(d_tracer))
8611 		return NULL;
8612 
8613 	tr->percpu_dir = tracefs_create_dir("per_cpu", d_tracer);
8614 
8615 	MEM_FAIL(!tr->percpu_dir,
8616 		  "Could not create tracefs directory 'per_cpu/%d'\n", cpu);
8617 
8618 	return tr->percpu_dir;
8619 }
8620 
8621 static struct dentry *
8622 trace_create_cpu_file(const char *name, umode_t mode, struct dentry *parent,
8623 		      void *data, long cpu, const struct file_operations *fops)
8624 {
8625 	struct dentry *ret = trace_create_file(name, mode, parent, data, fops);
8626 
8627 	if (ret) /* See tracing_get_cpu() */
8628 		d_inode(ret)->i_cdev = (void *)(cpu + 1);
8629 	return ret;
8630 }
8631 
8632 static void
8633 tracing_init_tracefs_percpu(struct trace_array *tr, long cpu)
8634 {
8635 	struct dentry *d_percpu = tracing_dentry_percpu(tr, cpu);
8636 	struct dentry *d_cpu;
8637 	char cpu_dir[30]; /* 30 characters should be more than enough */
8638 
8639 	if (!d_percpu)
8640 		return;
8641 
8642 	snprintf(cpu_dir, 30, "cpu%ld", cpu);
8643 	d_cpu = tracefs_create_dir(cpu_dir, d_percpu);
8644 	if (!d_cpu) {
8645 		pr_warn("Could not create tracefs '%s' entry\n", cpu_dir);
8646 		return;
8647 	}
8648 
8649 	/* per cpu trace_pipe */
8650 	trace_create_cpu_file("trace_pipe", TRACE_MODE_READ, d_cpu,
8651 				tr, cpu, &tracing_pipe_fops);
8652 
8653 	/* per cpu trace */
8654 	trace_create_cpu_file("trace", TRACE_MODE_WRITE, d_cpu,
8655 				tr, cpu, &tracing_fops);
8656 
8657 	trace_create_cpu_file("trace_pipe_raw", TRACE_MODE_READ, d_cpu,
8658 				tr, cpu, &tracing_buffers_fops);
8659 
8660 	trace_create_cpu_file("stats", TRACE_MODE_READ, d_cpu,
8661 				tr, cpu, &tracing_stats_fops);
8662 
8663 	trace_create_cpu_file("buffer_size_kb", TRACE_MODE_READ, d_cpu,
8664 				tr, cpu, &tracing_entries_fops);
8665 
8666 #ifdef CONFIG_TRACER_SNAPSHOT
8667 	trace_create_cpu_file("snapshot", TRACE_MODE_WRITE, d_cpu,
8668 				tr, cpu, &snapshot_fops);
8669 
8670 	trace_create_cpu_file("snapshot_raw", TRACE_MODE_READ, d_cpu,
8671 				tr, cpu, &snapshot_raw_fops);
8672 #endif
8673 }
8674 
8675 #ifdef CONFIG_FTRACE_SELFTEST
8676 /* Let selftest have access to static functions in this file */
8677 #include "trace_selftest.c"
8678 #endif
8679 
8680 static ssize_t
8681 trace_options_read(struct file *filp, char __user *ubuf, size_t cnt,
8682 			loff_t *ppos)
8683 {
8684 	struct trace_option_dentry *topt = filp->private_data;
8685 	char *buf;
8686 
8687 	if (topt->flags->val & topt->opt->bit)
8688 		buf = "1\n";
8689 	else
8690 		buf = "0\n";
8691 
8692 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, 2);
8693 }
8694 
8695 static ssize_t
8696 trace_options_write(struct file *filp, const char __user *ubuf, size_t cnt,
8697 			 loff_t *ppos)
8698 {
8699 	struct trace_option_dentry *topt = filp->private_data;
8700 	unsigned long val;
8701 	int ret;
8702 
8703 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
8704 	if (ret)
8705 		return ret;
8706 
8707 	if (val != 0 && val != 1)
8708 		return -EINVAL;
8709 
8710 	if (!!(topt->flags->val & topt->opt->bit) != val) {
8711 		mutex_lock(&trace_types_lock);
8712 		ret = __set_tracer_option(topt->tr, topt->flags,
8713 					  topt->opt, !val);
8714 		mutex_unlock(&trace_types_lock);
8715 		if (ret)
8716 			return ret;
8717 	}
8718 
8719 	*ppos += cnt;
8720 
8721 	return cnt;
8722 }
8723 
8724 static int tracing_open_options(struct inode *inode, struct file *filp)
8725 {
8726 	struct trace_option_dentry *topt = inode->i_private;
8727 	int ret;
8728 
8729 	ret = tracing_check_open_get_tr(topt->tr);
8730 	if (ret)
8731 		return ret;
8732 
8733 	filp->private_data = inode->i_private;
8734 	return 0;
8735 }
8736 
8737 static int tracing_release_options(struct inode *inode, struct file *file)
8738 {
8739 	struct trace_option_dentry *topt = file->private_data;
8740 
8741 	trace_array_put(topt->tr);
8742 	return 0;
8743 }
8744 
8745 static const struct file_operations trace_options_fops = {
8746 	.open = tracing_open_options,
8747 	.read = trace_options_read,
8748 	.write = trace_options_write,
8749 	.llseek	= generic_file_llseek,
8750 	.release = tracing_release_options,
8751 };
8752 
8753 /*
8754  * In order to pass in both the trace_array descriptor as well as the index
8755  * to the flag that the trace option file represents, the trace_array
8756  * has a character array of trace_flags_index[], which holds the index
8757  * of the bit for the flag it represents. index[0] == 0, index[1] == 1, etc.
8758  * The address of this character array is passed to the flag option file
8759  * read/write callbacks.
8760  *
8761  * In order to extract both the index and the trace_array descriptor,
8762  * get_tr_index() uses the following algorithm.
8763  *
8764  *   idx = *ptr;
8765  *
8766  * As the pointer itself contains the address of the index (remember
8767  * index[1] == 1).
8768  *
8769  * Then to get the trace_array descriptor, by subtracting that index
8770  * from the ptr, we get to the start of the index itself.
8771  *
8772  *   ptr - idx == &index[0]
8773  *
8774  * Then a simple container_of() from that pointer gets us to the
8775  * trace_array descriptor.
8776  */
8777 static void get_tr_index(void *data, struct trace_array **ptr,
8778 			 unsigned int *pindex)
8779 {
8780 	*pindex = *(unsigned char *)data;
8781 
8782 	*ptr = container_of(data - *pindex, struct trace_array,
8783 			    trace_flags_index);
8784 }
8785 
8786 static ssize_t
8787 trace_options_core_read(struct file *filp, char __user *ubuf, size_t cnt,
8788 			loff_t *ppos)
8789 {
8790 	void *tr_index = filp->private_data;
8791 	struct trace_array *tr;
8792 	unsigned int index;
8793 	char *buf;
8794 
8795 	get_tr_index(tr_index, &tr, &index);
8796 
8797 	if (tr->trace_flags & (1 << index))
8798 		buf = "1\n";
8799 	else
8800 		buf = "0\n";
8801 
8802 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, 2);
8803 }
8804 
8805 static ssize_t
8806 trace_options_core_write(struct file *filp, const char __user *ubuf, size_t cnt,
8807 			 loff_t *ppos)
8808 {
8809 	void *tr_index = filp->private_data;
8810 	struct trace_array *tr;
8811 	unsigned int index;
8812 	unsigned long val;
8813 	int ret;
8814 
8815 	get_tr_index(tr_index, &tr, &index);
8816 
8817 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
8818 	if (ret)
8819 		return ret;
8820 
8821 	if (val != 0 && val != 1)
8822 		return -EINVAL;
8823 
8824 	mutex_lock(&event_mutex);
8825 	mutex_lock(&trace_types_lock);
8826 	ret = set_tracer_flag(tr, 1 << index, val);
8827 	mutex_unlock(&trace_types_lock);
8828 	mutex_unlock(&event_mutex);
8829 
8830 	if (ret < 0)
8831 		return ret;
8832 
8833 	*ppos += cnt;
8834 
8835 	return cnt;
8836 }
8837 
8838 static const struct file_operations trace_options_core_fops = {
8839 	.open = tracing_open_generic,
8840 	.read = trace_options_core_read,
8841 	.write = trace_options_core_write,
8842 	.llseek = generic_file_llseek,
8843 };
8844 
8845 struct dentry *trace_create_file(const char *name,
8846 				 umode_t mode,
8847 				 struct dentry *parent,
8848 				 void *data,
8849 				 const struct file_operations *fops)
8850 {
8851 	struct dentry *ret;
8852 
8853 	ret = tracefs_create_file(name, mode, parent, data, fops);
8854 	if (!ret)
8855 		pr_warn("Could not create tracefs '%s' entry\n", name);
8856 
8857 	return ret;
8858 }
8859 
8860 
8861 static struct dentry *trace_options_init_dentry(struct trace_array *tr)
8862 {
8863 	struct dentry *d_tracer;
8864 
8865 	if (tr->options)
8866 		return tr->options;
8867 
8868 	d_tracer = tracing_get_dentry(tr);
8869 	if (IS_ERR(d_tracer))
8870 		return NULL;
8871 
8872 	tr->options = tracefs_create_dir("options", d_tracer);
8873 	if (!tr->options) {
8874 		pr_warn("Could not create tracefs directory 'options'\n");
8875 		return NULL;
8876 	}
8877 
8878 	return tr->options;
8879 }
8880 
8881 static void
8882 create_trace_option_file(struct trace_array *tr,
8883 			 struct trace_option_dentry *topt,
8884 			 struct tracer_flags *flags,
8885 			 struct tracer_opt *opt)
8886 {
8887 	struct dentry *t_options;
8888 
8889 	t_options = trace_options_init_dentry(tr);
8890 	if (!t_options)
8891 		return;
8892 
8893 	topt->flags = flags;
8894 	topt->opt = opt;
8895 	topt->tr = tr;
8896 
8897 	topt->entry = trace_create_file(opt->name, TRACE_MODE_WRITE,
8898 					t_options, topt, &trace_options_fops);
8899 
8900 }
8901 
8902 static void
8903 create_trace_option_files(struct trace_array *tr, struct tracer *tracer)
8904 {
8905 	struct trace_option_dentry *topts;
8906 	struct trace_options *tr_topts;
8907 	struct tracer_flags *flags;
8908 	struct tracer_opt *opts;
8909 	int cnt;
8910 	int i;
8911 
8912 	if (!tracer)
8913 		return;
8914 
8915 	flags = tracer->flags;
8916 
8917 	if (!flags || !flags->opts)
8918 		return;
8919 
8920 	/*
8921 	 * If this is an instance, only create flags for tracers
8922 	 * the instance may have.
8923 	 */
8924 	if (!trace_ok_for_array(tracer, tr))
8925 		return;
8926 
8927 	for (i = 0; i < tr->nr_topts; i++) {
8928 		/* Make sure there's no duplicate flags. */
8929 		if (WARN_ON_ONCE(tr->topts[i].tracer->flags == tracer->flags))
8930 			return;
8931 	}
8932 
8933 	opts = flags->opts;
8934 
8935 	for (cnt = 0; opts[cnt].name; cnt++)
8936 		;
8937 
8938 	topts = kcalloc(cnt + 1, sizeof(*topts), GFP_KERNEL);
8939 	if (!topts)
8940 		return;
8941 
8942 	tr_topts = krealloc(tr->topts, sizeof(*tr->topts) * (tr->nr_topts + 1),
8943 			    GFP_KERNEL);
8944 	if (!tr_topts) {
8945 		kfree(topts);
8946 		return;
8947 	}
8948 
8949 	tr->topts = tr_topts;
8950 	tr->topts[tr->nr_topts].tracer = tracer;
8951 	tr->topts[tr->nr_topts].topts = topts;
8952 	tr->nr_topts++;
8953 
8954 	for (cnt = 0; opts[cnt].name; cnt++) {
8955 		create_trace_option_file(tr, &topts[cnt], flags,
8956 					 &opts[cnt]);
8957 		MEM_FAIL(topts[cnt].entry == NULL,
8958 			  "Failed to create trace option: %s",
8959 			  opts[cnt].name);
8960 	}
8961 }
8962 
8963 static struct dentry *
8964 create_trace_option_core_file(struct trace_array *tr,
8965 			      const char *option, long index)
8966 {
8967 	struct dentry *t_options;
8968 
8969 	t_options = trace_options_init_dentry(tr);
8970 	if (!t_options)
8971 		return NULL;
8972 
8973 	return trace_create_file(option, TRACE_MODE_WRITE, t_options,
8974 				 (void *)&tr->trace_flags_index[index],
8975 				 &trace_options_core_fops);
8976 }
8977 
8978 static void create_trace_options_dir(struct trace_array *tr)
8979 {
8980 	struct dentry *t_options;
8981 	bool top_level = tr == &global_trace;
8982 	int i;
8983 
8984 	t_options = trace_options_init_dentry(tr);
8985 	if (!t_options)
8986 		return;
8987 
8988 	for (i = 0; trace_options[i]; i++) {
8989 		if (top_level ||
8990 		    !((1 << i) & TOP_LEVEL_TRACE_FLAGS))
8991 			create_trace_option_core_file(tr, trace_options[i], i);
8992 	}
8993 }
8994 
8995 static ssize_t
8996 rb_simple_read(struct file *filp, char __user *ubuf,
8997 	       size_t cnt, loff_t *ppos)
8998 {
8999 	struct trace_array *tr = filp->private_data;
9000 	char buf[64];
9001 	int r;
9002 
9003 	r = tracer_tracing_is_on(tr);
9004 	r = sprintf(buf, "%d\n", r);
9005 
9006 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
9007 }
9008 
9009 static ssize_t
9010 rb_simple_write(struct file *filp, const char __user *ubuf,
9011 		size_t cnt, loff_t *ppos)
9012 {
9013 	struct trace_array *tr = filp->private_data;
9014 	struct trace_buffer *buffer = tr->array_buffer.buffer;
9015 	unsigned long val;
9016 	int ret;
9017 
9018 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
9019 	if (ret)
9020 		return ret;
9021 
9022 	if (buffer) {
9023 		mutex_lock(&trace_types_lock);
9024 		if (!!val == tracer_tracing_is_on(tr)) {
9025 			val = 0; /* do nothing */
9026 		} else if (val) {
9027 			tracer_tracing_on(tr);
9028 			if (tr->current_trace->start)
9029 				tr->current_trace->start(tr);
9030 		} else {
9031 			tracer_tracing_off(tr);
9032 			if (tr->current_trace->stop)
9033 				tr->current_trace->stop(tr);
9034 			/* Wake up any waiters */
9035 			ring_buffer_wake_waiters(buffer, RING_BUFFER_ALL_CPUS);
9036 		}
9037 		mutex_unlock(&trace_types_lock);
9038 	}
9039 
9040 	(*ppos)++;
9041 
9042 	return cnt;
9043 }
9044 
9045 static const struct file_operations rb_simple_fops = {
9046 	.open		= tracing_open_generic_tr,
9047 	.read		= rb_simple_read,
9048 	.write		= rb_simple_write,
9049 	.release	= tracing_release_generic_tr,
9050 	.llseek		= default_llseek,
9051 };
9052 
9053 static ssize_t
9054 buffer_percent_read(struct file *filp, char __user *ubuf,
9055 		    size_t cnt, loff_t *ppos)
9056 {
9057 	struct trace_array *tr = filp->private_data;
9058 	char buf[64];
9059 	int r;
9060 
9061 	r = tr->buffer_percent;
9062 	r = sprintf(buf, "%d\n", r);
9063 
9064 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
9065 }
9066 
9067 static ssize_t
9068 buffer_percent_write(struct file *filp, const char __user *ubuf,
9069 		     size_t cnt, loff_t *ppos)
9070 {
9071 	struct trace_array *tr = filp->private_data;
9072 	unsigned long val;
9073 	int ret;
9074 
9075 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
9076 	if (ret)
9077 		return ret;
9078 
9079 	if (val > 100)
9080 		return -EINVAL;
9081 
9082 	tr->buffer_percent = val;
9083 
9084 	(*ppos)++;
9085 
9086 	return cnt;
9087 }
9088 
9089 static const struct file_operations buffer_percent_fops = {
9090 	.open		= tracing_open_generic_tr,
9091 	.read		= buffer_percent_read,
9092 	.write		= buffer_percent_write,
9093 	.release	= tracing_release_generic_tr,
9094 	.llseek		= default_llseek,
9095 };
9096 
9097 static ssize_t
9098 buffer_subbuf_size_read(struct file *filp, char __user *ubuf, size_t cnt, loff_t *ppos)
9099 {
9100 	struct trace_array *tr = filp->private_data;
9101 	size_t size;
9102 	char buf[64];
9103 	int order;
9104 	int r;
9105 
9106 	order = ring_buffer_subbuf_order_get(tr->array_buffer.buffer);
9107 	size = (PAGE_SIZE << order) / 1024;
9108 
9109 	r = sprintf(buf, "%zd\n", size);
9110 
9111 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
9112 }
9113 
9114 static ssize_t
9115 buffer_subbuf_size_write(struct file *filp, const char __user *ubuf,
9116 			 size_t cnt, loff_t *ppos)
9117 {
9118 	struct trace_array *tr = filp->private_data;
9119 	unsigned long val;
9120 	int old_order;
9121 	int order;
9122 	int pages;
9123 	int ret;
9124 
9125 	ret = kstrtoul_from_user(ubuf, cnt, 10, &val);
9126 	if (ret)
9127 		return ret;
9128 
9129 	val *= 1024; /* value passed in is in KB */
9130 
9131 	pages = DIV_ROUND_UP(val, PAGE_SIZE);
9132 	order = fls(pages - 1);
9133 
9134 	/* limit between 1 and 128 system pages */
9135 	if (order < 0 || order > 7)
9136 		return -EINVAL;
9137 
9138 	/* Do not allow tracing while changing the order of the ring buffer */
9139 	tracing_stop_tr(tr);
9140 
9141 	old_order = ring_buffer_subbuf_order_get(tr->array_buffer.buffer);
9142 	if (old_order == order)
9143 		goto out;
9144 
9145 	ret = ring_buffer_subbuf_order_set(tr->array_buffer.buffer, order);
9146 	if (ret)
9147 		goto out;
9148 
9149 #ifdef CONFIG_TRACER_MAX_TRACE
9150 
9151 	if (!tr->allocated_snapshot)
9152 		goto out_max;
9153 
9154 	ret = ring_buffer_subbuf_order_set(tr->max_buffer.buffer, order);
9155 	if (ret) {
9156 		/* Put back the old order */
9157 		cnt = ring_buffer_subbuf_order_set(tr->array_buffer.buffer, old_order);
9158 		if (WARN_ON_ONCE(cnt)) {
9159 			/*
9160 			 * AARGH! We are left with different orders!
9161 			 * The max buffer is our "snapshot" buffer.
9162 			 * When a tracer needs a snapshot (one of the
9163 			 * latency tracers), it swaps the max buffer
9164 			 * with the saved snap shot. We succeeded to
9165 			 * update the order of the main buffer, but failed to
9166 			 * update the order of the max buffer. But when we tried
9167 			 * to reset the main buffer to the original size, we
9168 			 * failed there too. This is very unlikely to
9169 			 * happen, but if it does, warn and kill all
9170 			 * tracing.
9171 			 */
9172 			tracing_disabled = 1;
9173 		}
9174 		goto out;
9175 	}
9176  out_max:
9177 #endif
9178 	(*ppos)++;
9179  out:
9180 	if (ret)
9181 		cnt = ret;
9182 	tracing_start_tr(tr);
9183 	return cnt;
9184 }
9185 
9186 static const struct file_operations buffer_subbuf_size_fops = {
9187 	.open		= tracing_open_generic_tr,
9188 	.read		= buffer_subbuf_size_read,
9189 	.write		= buffer_subbuf_size_write,
9190 	.release	= tracing_release_generic_tr,
9191 	.llseek		= default_llseek,
9192 };
9193 
9194 static struct dentry *trace_instance_dir;
9195 
9196 static void
9197 init_tracer_tracefs(struct trace_array *tr, struct dentry *d_tracer);
9198 
9199 static int
9200 allocate_trace_buffer(struct trace_array *tr, struct array_buffer *buf, int size)
9201 {
9202 	enum ring_buffer_flags rb_flags;
9203 
9204 	rb_flags = tr->trace_flags & TRACE_ITER_OVERWRITE ? RB_FL_OVERWRITE : 0;
9205 
9206 	buf->tr = tr;
9207 
9208 	buf->buffer = ring_buffer_alloc(size, rb_flags);
9209 	if (!buf->buffer)
9210 		return -ENOMEM;
9211 
9212 	buf->data = alloc_percpu(struct trace_array_cpu);
9213 	if (!buf->data) {
9214 		ring_buffer_free(buf->buffer);
9215 		buf->buffer = NULL;
9216 		return -ENOMEM;
9217 	}
9218 
9219 	/* Allocate the first page for all buffers */
9220 	set_buffer_entries(&tr->array_buffer,
9221 			   ring_buffer_size(tr->array_buffer.buffer, 0));
9222 
9223 	return 0;
9224 }
9225 
9226 static void free_trace_buffer(struct array_buffer *buf)
9227 {
9228 	if (buf->buffer) {
9229 		ring_buffer_free(buf->buffer);
9230 		buf->buffer = NULL;
9231 		free_percpu(buf->data);
9232 		buf->data = NULL;
9233 	}
9234 }
9235 
9236 static int allocate_trace_buffers(struct trace_array *tr, int size)
9237 {
9238 	int ret;
9239 
9240 	ret = allocate_trace_buffer(tr, &tr->array_buffer, size);
9241 	if (ret)
9242 		return ret;
9243 
9244 #ifdef CONFIG_TRACER_MAX_TRACE
9245 	ret = allocate_trace_buffer(tr, &tr->max_buffer,
9246 				    allocate_snapshot ? size : 1);
9247 	if (MEM_FAIL(ret, "Failed to allocate trace buffer\n")) {
9248 		free_trace_buffer(&tr->array_buffer);
9249 		return -ENOMEM;
9250 	}
9251 	tr->allocated_snapshot = allocate_snapshot;
9252 
9253 	allocate_snapshot = false;
9254 #endif
9255 
9256 	return 0;
9257 }
9258 
9259 static void free_trace_buffers(struct trace_array *tr)
9260 {
9261 	if (!tr)
9262 		return;
9263 
9264 	free_trace_buffer(&tr->array_buffer);
9265 
9266 #ifdef CONFIG_TRACER_MAX_TRACE
9267 	free_trace_buffer(&tr->max_buffer);
9268 #endif
9269 }
9270 
9271 static void init_trace_flags_index(struct trace_array *tr)
9272 {
9273 	int i;
9274 
9275 	/* Used by the trace options files */
9276 	for (i = 0; i < TRACE_FLAGS_MAX_SIZE; i++)
9277 		tr->trace_flags_index[i] = i;
9278 }
9279 
9280 static void __update_tracer_options(struct trace_array *tr)
9281 {
9282 	struct tracer *t;
9283 
9284 	for (t = trace_types; t; t = t->next)
9285 		add_tracer_options(tr, t);
9286 }
9287 
9288 static void update_tracer_options(struct trace_array *tr)
9289 {
9290 	mutex_lock(&trace_types_lock);
9291 	tracer_options_updated = true;
9292 	__update_tracer_options(tr);
9293 	mutex_unlock(&trace_types_lock);
9294 }
9295 
9296 /* Must have trace_types_lock held */
9297 struct trace_array *trace_array_find(const char *instance)
9298 {
9299 	struct trace_array *tr, *found = NULL;
9300 
9301 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
9302 		if (tr->name && strcmp(tr->name, instance) == 0) {
9303 			found = tr;
9304 			break;
9305 		}
9306 	}
9307 
9308 	return found;
9309 }
9310 
9311 struct trace_array *trace_array_find_get(const char *instance)
9312 {
9313 	struct trace_array *tr;
9314 
9315 	mutex_lock(&trace_types_lock);
9316 	tr = trace_array_find(instance);
9317 	if (tr)
9318 		tr->ref++;
9319 	mutex_unlock(&trace_types_lock);
9320 
9321 	return tr;
9322 }
9323 
9324 static int trace_array_create_dir(struct trace_array *tr)
9325 {
9326 	int ret;
9327 
9328 	tr->dir = tracefs_create_dir(tr->name, trace_instance_dir);
9329 	if (!tr->dir)
9330 		return -EINVAL;
9331 
9332 	ret = event_trace_add_tracer(tr->dir, tr);
9333 	if (ret) {
9334 		tracefs_remove(tr->dir);
9335 		return ret;
9336 	}
9337 
9338 	init_tracer_tracefs(tr, tr->dir);
9339 	__update_tracer_options(tr);
9340 
9341 	return ret;
9342 }
9343 
9344 static struct trace_array *
9345 trace_array_create_systems(const char *name, const char *systems)
9346 {
9347 	struct trace_array *tr;
9348 	int ret;
9349 
9350 	ret = -ENOMEM;
9351 	tr = kzalloc(sizeof(*tr), GFP_KERNEL);
9352 	if (!tr)
9353 		return ERR_PTR(ret);
9354 
9355 	tr->name = kstrdup(name, GFP_KERNEL);
9356 	if (!tr->name)
9357 		goto out_free_tr;
9358 
9359 	if (!alloc_cpumask_var(&tr->tracing_cpumask, GFP_KERNEL))
9360 		goto out_free_tr;
9361 
9362 	if (!zalloc_cpumask_var(&tr->pipe_cpumask, GFP_KERNEL))
9363 		goto out_free_tr;
9364 
9365 	if (systems) {
9366 		tr->system_names = kstrdup_const(systems, GFP_KERNEL);
9367 		if (!tr->system_names)
9368 			goto out_free_tr;
9369 	}
9370 
9371 	tr->trace_flags = global_trace.trace_flags & ~ZEROED_TRACE_FLAGS;
9372 
9373 	cpumask_copy(tr->tracing_cpumask, cpu_all_mask);
9374 
9375 	raw_spin_lock_init(&tr->start_lock);
9376 
9377 	tr->max_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
9378 #ifdef CONFIG_TRACER_MAX_TRACE
9379 	spin_lock_init(&tr->snapshot_trigger_lock);
9380 #endif
9381 	tr->current_trace = &nop_trace;
9382 
9383 	INIT_LIST_HEAD(&tr->systems);
9384 	INIT_LIST_HEAD(&tr->events);
9385 	INIT_LIST_HEAD(&tr->hist_vars);
9386 	INIT_LIST_HEAD(&tr->err_log);
9387 
9388 	if (allocate_trace_buffers(tr, trace_buf_size) < 0)
9389 		goto out_free_tr;
9390 
9391 	/* The ring buffer is defaultly expanded */
9392 	trace_set_ring_buffer_expanded(tr);
9393 
9394 	if (ftrace_allocate_ftrace_ops(tr) < 0)
9395 		goto out_free_tr;
9396 
9397 	ftrace_init_trace_array(tr);
9398 
9399 	init_trace_flags_index(tr);
9400 
9401 	if (trace_instance_dir) {
9402 		ret = trace_array_create_dir(tr);
9403 		if (ret)
9404 			goto out_free_tr;
9405 	} else
9406 		__trace_early_add_events(tr);
9407 
9408 	list_add(&tr->list, &ftrace_trace_arrays);
9409 
9410 	tr->ref++;
9411 
9412 	return tr;
9413 
9414  out_free_tr:
9415 	ftrace_free_ftrace_ops(tr);
9416 	free_trace_buffers(tr);
9417 	free_cpumask_var(tr->pipe_cpumask);
9418 	free_cpumask_var(tr->tracing_cpumask);
9419 	kfree_const(tr->system_names);
9420 	kfree(tr->name);
9421 	kfree(tr);
9422 
9423 	return ERR_PTR(ret);
9424 }
9425 
9426 static struct trace_array *trace_array_create(const char *name)
9427 {
9428 	return trace_array_create_systems(name, NULL);
9429 }
9430 
9431 static int instance_mkdir(const char *name)
9432 {
9433 	struct trace_array *tr;
9434 	int ret;
9435 
9436 	mutex_lock(&event_mutex);
9437 	mutex_lock(&trace_types_lock);
9438 
9439 	ret = -EEXIST;
9440 	if (trace_array_find(name))
9441 		goto out_unlock;
9442 
9443 	tr = trace_array_create(name);
9444 
9445 	ret = PTR_ERR_OR_ZERO(tr);
9446 
9447 out_unlock:
9448 	mutex_unlock(&trace_types_lock);
9449 	mutex_unlock(&event_mutex);
9450 	return ret;
9451 }
9452 
9453 /**
9454  * trace_array_get_by_name - Create/Lookup a trace array, given its name.
9455  * @name: The name of the trace array to be looked up/created.
9456  * @systems: A list of systems to create event directories for (NULL for all)
9457  *
9458  * Returns pointer to trace array with given name.
9459  * NULL, if it cannot be created.
9460  *
9461  * NOTE: This function increments the reference counter associated with the
9462  * trace array returned. This makes sure it cannot be freed while in use.
9463  * Use trace_array_put() once the trace array is no longer needed.
9464  * If the trace_array is to be freed, trace_array_destroy() needs to
9465  * be called after the trace_array_put(), or simply let user space delete
9466  * it from the tracefs instances directory. But until the
9467  * trace_array_put() is called, user space can not delete it.
9468  *
9469  */
9470 struct trace_array *trace_array_get_by_name(const char *name, const char *systems)
9471 {
9472 	struct trace_array *tr;
9473 
9474 	mutex_lock(&event_mutex);
9475 	mutex_lock(&trace_types_lock);
9476 
9477 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
9478 		if (tr->name && strcmp(tr->name, name) == 0)
9479 			goto out_unlock;
9480 	}
9481 
9482 	tr = trace_array_create_systems(name, systems);
9483 
9484 	if (IS_ERR(tr))
9485 		tr = NULL;
9486 out_unlock:
9487 	if (tr)
9488 		tr->ref++;
9489 
9490 	mutex_unlock(&trace_types_lock);
9491 	mutex_unlock(&event_mutex);
9492 	return tr;
9493 }
9494 EXPORT_SYMBOL_GPL(trace_array_get_by_name);
9495 
9496 static int __remove_instance(struct trace_array *tr)
9497 {
9498 	int i;
9499 
9500 	/* Reference counter for a newly created trace array = 1. */
9501 	if (tr->ref > 1 || (tr->current_trace && tr->trace_ref))
9502 		return -EBUSY;
9503 
9504 	list_del(&tr->list);
9505 
9506 	/* Disable all the flags that were enabled coming in */
9507 	for (i = 0; i < TRACE_FLAGS_MAX_SIZE; i++) {
9508 		if ((1 << i) & ZEROED_TRACE_FLAGS)
9509 			set_tracer_flag(tr, 1 << i, 0);
9510 	}
9511 
9512 	tracing_set_nop(tr);
9513 	clear_ftrace_function_probes(tr);
9514 	event_trace_del_tracer(tr);
9515 	ftrace_clear_pids(tr);
9516 	ftrace_destroy_function_files(tr);
9517 	tracefs_remove(tr->dir);
9518 	free_percpu(tr->last_func_repeats);
9519 	free_trace_buffers(tr);
9520 	clear_tracing_err_log(tr);
9521 
9522 	for (i = 0; i < tr->nr_topts; i++) {
9523 		kfree(tr->topts[i].topts);
9524 	}
9525 	kfree(tr->topts);
9526 
9527 	free_cpumask_var(tr->pipe_cpumask);
9528 	free_cpumask_var(tr->tracing_cpumask);
9529 	kfree_const(tr->system_names);
9530 	kfree(tr->name);
9531 	kfree(tr);
9532 
9533 	return 0;
9534 }
9535 
9536 int trace_array_destroy(struct trace_array *this_tr)
9537 {
9538 	struct trace_array *tr;
9539 	int ret;
9540 
9541 	if (!this_tr)
9542 		return -EINVAL;
9543 
9544 	mutex_lock(&event_mutex);
9545 	mutex_lock(&trace_types_lock);
9546 
9547 	ret = -ENODEV;
9548 
9549 	/* Making sure trace array exists before destroying it. */
9550 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
9551 		if (tr == this_tr) {
9552 			ret = __remove_instance(tr);
9553 			break;
9554 		}
9555 	}
9556 
9557 	mutex_unlock(&trace_types_lock);
9558 	mutex_unlock(&event_mutex);
9559 
9560 	return ret;
9561 }
9562 EXPORT_SYMBOL_GPL(trace_array_destroy);
9563 
9564 static int instance_rmdir(const char *name)
9565 {
9566 	struct trace_array *tr;
9567 	int ret;
9568 
9569 	mutex_lock(&event_mutex);
9570 	mutex_lock(&trace_types_lock);
9571 
9572 	ret = -ENODEV;
9573 	tr = trace_array_find(name);
9574 	if (tr)
9575 		ret = __remove_instance(tr);
9576 
9577 	mutex_unlock(&trace_types_lock);
9578 	mutex_unlock(&event_mutex);
9579 
9580 	return ret;
9581 }
9582 
9583 static __init void create_trace_instances(struct dentry *d_tracer)
9584 {
9585 	struct trace_array *tr;
9586 
9587 	trace_instance_dir = tracefs_create_instance_dir("instances", d_tracer,
9588 							 instance_mkdir,
9589 							 instance_rmdir);
9590 	if (MEM_FAIL(!trace_instance_dir, "Failed to create instances directory\n"))
9591 		return;
9592 
9593 	mutex_lock(&event_mutex);
9594 	mutex_lock(&trace_types_lock);
9595 
9596 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
9597 		if (!tr->name)
9598 			continue;
9599 		if (MEM_FAIL(trace_array_create_dir(tr) < 0,
9600 			     "Failed to create instance directory\n"))
9601 			break;
9602 	}
9603 
9604 	mutex_unlock(&trace_types_lock);
9605 	mutex_unlock(&event_mutex);
9606 }
9607 
9608 static void
9609 init_tracer_tracefs(struct trace_array *tr, struct dentry *d_tracer)
9610 {
9611 	int cpu;
9612 
9613 	trace_create_file("available_tracers", TRACE_MODE_READ, d_tracer,
9614 			tr, &show_traces_fops);
9615 
9616 	trace_create_file("current_tracer", TRACE_MODE_WRITE, d_tracer,
9617 			tr, &set_tracer_fops);
9618 
9619 	trace_create_file("tracing_cpumask", TRACE_MODE_WRITE, d_tracer,
9620 			  tr, &tracing_cpumask_fops);
9621 
9622 	trace_create_file("trace_options", TRACE_MODE_WRITE, d_tracer,
9623 			  tr, &tracing_iter_fops);
9624 
9625 	trace_create_file("trace", TRACE_MODE_WRITE, d_tracer,
9626 			  tr, &tracing_fops);
9627 
9628 	trace_create_file("trace_pipe", TRACE_MODE_READ, d_tracer,
9629 			  tr, &tracing_pipe_fops);
9630 
9631 	trace_create_file("buffer_size_kb", TRACE_MODE_WRITE, d_tracer,
9632 			  tr, &tracing_entries_fops);
9633 
9634 	trace_create_file("buffer_total_size_kb", TRACE_MODE_READ, d_tracer,
9635 			  tr, &tracing_total_entries_fops);
9636 
9637 	trace_create_file("free_buffer", 0200, d_tracer,
9638 			  tr, &tracing_free_buffer_fops);
9639 
9640 	trace_create_file("trace_marker", 0220, d_tracer,
9641 			  tr, &tracing_mark_fops);
9642 
9643 	tr->trace_marker_file = __find_event_file(tr, "ftrace", "print");
9644 
9645 	trace_create_file("trace_marker_raw", 0220, d_tracer,
9646 			  tr, &tracing_mark_raw_fops);
9647 
9648 	trace_create_file("trace_clock", TRACE_MODE_WRITE, d_tracer, tr,
9649 			  &trace_clock_fops);
9650 
9651 	trace_create_file("tracing_on", TRACE_MODE_WRITE, d_tracer,
9652 			  tr, &rb_simple_fops);
9653 
9654 	trace_create_file("timestamp_mode", TRACE_MODE_READ, d_tracer, tr,
9655 			  &trace_time_stamp_mode_fops);
9656 
9657 	tr->buffer_percent = 50;
9658 
9659 	trace_create_file("buffer_percent", TRACE_MODE_WRITE, d_tracer,
9660 			tr, &buffer_percent_fops);
9661 
9662 	trace_create_file("buffer_subbuf_size_kb", TRACE_MODE_WRITE, d_tracer,
9663 			  tr, &buffer_subbuf_size_fops);
9664 
9665 	create_trace_options_dir(tr);
9666 
9667 #ifdef CONFIG_TRACER_MAX_TRACE
9668 	trace_create_maxlat_file(tr, d_tracer);
9669 #endif
9670 
9671 	if (ftrace_create_function_files(tr, d_tracer))
9672 		MEM_FAIL(1, "Could not allocate function filter files");
9673 
9674 #ifdef CONFIG_TRACER_SNAPSHOT
9675 	trace_create_file("snapshot", TRACE_MODE_WRITE, d_tracer,
9676 			  tr, &snapshot_fops);
9677 #endif
9678 
9679 	trace_create_file("error_log", TRACE_MODE_WRITE, d_tracer,
9680 			  tr, &tracing_err_log_fops);
9681 
9682 	for_each_tracing_cpu(cpu)
9683 		tracing_init_tracefs_percpu(tr, cpu);
9684 
9685 	ftrace_init_tracefs(tr, d_tracer);
9686 }
9687 
9688 static struct vfsmount *trace_automount(struct dentry *mntpt, void *ingore)
9689 {
9690 	struct vfsmount *mnt;
9691 	struct file_system_type *type;
9692 
9693 	/*
9694 	 * To maintain backward compatibility for tools that mount
9695 	 * debugfs to get to the tracing facility, tracefs is automatically
9696 	 * mounted to the debugfs/tracing directory.
9697 	 */
9698 	type = get_fs_type("tracefs");
9699 	if (!type)
9700 		return NULL;
9701 	mnt = vfs_submount(mntpt, type, "tracefs", NULL);
9702 	put_filesystem(type);
9703 	if (IS_ERR(mnt))
9704 		return NULL;
9705 	mntget(mnt);
9706 
9707 	return mnt;
9708 }
9709 
9710 /**
9711  * tracing_init_dentry - initialize top level trace array
9712  *
9713  * This is called when creating files or directories in the tracing
9714  * directory. It is called via fs_initcall() by any of the boot up code
9715  * and expects to return the dentry of the top level tracing directory.
9716  */
9717 int tracing_init_dentry(void)
9718 {
9719 	struct trace_array *tr = &global_trace;
9720 
9721 	if (security_locked_down(LOCKDOWN_TRACEFS)) {
9722 		pr_warn("Tracing disabled due to lockdown\n");
9723 		return -EPERM;
9724 	}
9725 
9726 	/* The top level trace array uses  NULL as parent */
9727 	if (tr->dir)
9728 		return 0;
9729 
9730 	if (WARN_ON(!tracefs_initialized()))
9731 		return -ENODEV;
9732 
9733 	/*
9734 	 * As there may still be users that expect the tracing
9735 	 * files to exist in debugfs/tracing, we must automount
9736 	 * the tracefs file system there, so older tools still
9737 	 * work with the newer kernel.
9738 	 */
9739 	tr->dir = debugfs_create_automount("tracing", NULL,
9740 					   trace_automount, NULL);
9741 
9742 	return 0;
9743 }
9744 
9745 extern struct trace_eval_map *__start_ftrace_eval_maps[];
9746 extern struct trace_eval_map *__stop_ftrace_eval_maps[];
9747 
9748 static struct workqueue_struct *eval_map_wq __initdata;
9749 static struct work_struct eval_map_work __initdata;
9750 static struct work_struct tracerfs_init_work __initdata;
9751 
9752 static void __init eval_map_work_func(struct work_struct *work)
9753 {
9754 	int len;
9755 
9756 	len = __stop_ftrace_eval_maps - __start_ftrace_eval_maps;
9757 	trace_insert_eval_map(NULL, __start_ftrace_eval_maps, len);
9758 }
9759 
9760 static int __init trace_eval_init(void)
9761 {
9762 	INIT_WORK(&eval_map_work, eval_map_work_func);
9763 
9764 	eval_map_wq = alloc_workqueue("eval_map_wq", WQ_UNBOUND, 0);
9765 	if (!eval_map_wq) {
9766 		pr_err("Unable to allocate eval_map_wq\n");
9767 		/* Do work here */
9768 		eval_map_work_func(&eval_map_work);
9769 		return -ENOMEM;
9770 	}
9771 
9772 	queue_work(eval_map_wq, &eval_map_work);
9773 	return 0;
9774 }
9775 
9776 subsys_initcall(trace_eval_init);
9777 
9778 static int __init trace_eval_sync(void)
9779 {
9780 	/* Make sure the eval map updates are finished */
9781 	if (eval_map_wq)
9782 		destroy_workqueue(eval_map_wq);
9783 	return 0;
9784 }
9785 
9786 late_initcall_sync(trace_eval_sync);
9787 
9788 
9789 #ifdef CONFIG_MODULES
9790 static void trace_module_add_evals(struct module *mod)
9791 {
9792 	if (!mod->num_trace_evals)
9793 		return;
9794 
9795 	/*
9796 	 * Modules with bad taint do not have events created, do
9797 	 * not bother with enums either.
9798 	 */
9799 	if (trace_module_has_bad_taint(mod))
9800 		return;
9801 
9802 	trace_insert_eval_map(mod, mod->trace_evals, mod->num_trace_evals);
9803 }
9804 
9805 #ifdef CONFIG_TRACE_EVAL_MAP_FILE
9806 static void trace_module_remove_evals(struct module *mod)
9807 {
9808 	union trace_eval_map_item *map;
9809 	union trace_eval_map_item **last = &trace_eval_maps;
9810 
9811 	if (!mod->num_trace_evals)
9812 		return;
9813 
9814 	mutex_lock(&trace_eval_mutex);
9815 
9816 	map = trace_eval_maps;
9817 
9818 	while (map) {
9819 		if (map->head.mod == mod)
9820 			break;
9821 		map = trace_eval_jmp_to_tail(map);
9822 		last = &map->tail.next;
9823 		map = map->tail.next;
9824 	}
9825 	if (!map)
9826 		goto out;
9827 
9828 	*last = trace_eval_jmp_to_tail(map)->tail.next;
9829 	kfree(map);
9830  out:
9831 	mutex_unlock(&trace_eval_mutex);
9832 }
9833 #else
9834 static inline void trace_module_remove_evals(struct module *mod) { }
9835 #endif /* CONFIG_TRACE_EVAL_MAP_FILE */
9836 
9837 static int trace_module_notify(struct notifier_block *self,
9838 			       unsigned long val, void *data)
9839 {
9840 	struct module *mod = data;
9841 
9842 	switch (val) {
9843 	case MODULE_STATE_COMING:
9844 		trace_module_add_evals(mod);
9845 		break;
9846 	case MODULE_STATE_GOING:
9847 		trace_module_remove_evals(mod);
9848 		break;
9849 	}
9850 
9851 	return NOTIFY_OK;
9852 }
9853 
9854 static struct notifier_block trace_module_nb = {
9855 	.notifier_call = trace_module_notify,
9856 	.priority = 0,
9857 };
9858 #endif /* CONFIG_MODULES */
9859 
9860 static __init void tracer_init_tracefs_work_func(struct work_struct *work)
9861 {
9862 
9863 	event_trace_init();
9864 
9865 	init_tracer_tracefs(&global_trace, NULL);
9866 	ftrace_init_tracefs_toplevel(&global_trace, NULL);
9867 
9868 	trace_create_file("tracing_thresh", TRACE_MODE_WRITE, NULL,
9869 			&global_trace, &tracing_thresh_fops);
9870 
9871 	trace_create_file("README", TRACE_MODE_READ, NULL,
9872 			NULL, &tracing_readme_fops);
9873 
9874 	trace_create_file("saved_cmdlines", TRACE_MODE_READ, NULL,
9875 			NULL, &tracing_saved_cmdlines_fops);
9876 
9877 	trace_create_file("saved_cmdlines_size", TRACE_MODE_WRITE, NULL,
9878 			  NULL, &tracing_saved_cmdlines_size_fops);
9879 
9880 	trace_create_file("saved_tgids", TRACE_MODE_READ, NULL,
9881 			NULL, &tracing_saved_tgids_fops);
9882 
9883 	trace_create_eval_file(NULL);
9884 
9885 #ifdef CONFIG_MODULES
9886 	register_module_notifier(&trace_module_nb);
9887 #endif
9888 
9889 #ifdef CONFIG_DYNAMIC_FTRACE
9890 	trace_create_file("dyn_ftrace_total_info", TRACE_MODE_READ, NULL,
9891 			NULL, &tracing_dyn_info_fops);
9892 #endif
9893 
9894 	create_trace_instances(NULL);
9895 
9896 	update_tracer_options(&global_trace);
9897 }
9898 
9899 static __init int tracer_init_tracefs(void)
9900 {
9901 	int ret;
9902 
9903 	trace_access_lock_init();
9904 
9905 	ret = tracing_init_dentry();
9906 	if (ret)
9907 		return 0;
9908 
9909 	if (eval_map_wq) {
9910 		INIT_WORK(&tracerfs_init_work, tracer_init_tracefs_work_func);
9911 		queue_work(eval_map_wq, &tracerfs_init_work);
9912 	} else {
9913 		tracer_init_tracefs_work_func(NULL);
9914 	}
9915 
9916 	rv_init_interface();
9917 
9918 	return 0;
9919 }
9920 
9921 fs_initcall(tracer_init_tracefs);
9922 
9923 static int trace_die_panic_handler(struct notifier_block *self,
9924 				unsigned long ev, void *unused);
9925 
9926 static struct notifier_block trace_panic_notifier = {
9927 	.notifier_call = trace_die_panic_handler,
9928 	.priority = INT_MAX - 1,
9929 };
9930 
9931 static struct notifier_block trace_die_notifier = {
9932 	.notifier_call = trace_die_panic_handler,
9933 	.priority = INT_MAX - 1,
9934 };
9935 
9936 /*
9937  * The idea is to execute the following die/panic callback early, in order
9938  * to avoid showing irrelevant information in the trace (like other panic
9939  * notifier functions); we are the 2nd to run, after hung_task/rcu_stall
9940  * warnings get disabled (to prevent potential log flooding).
9941  */
9942 static int trace_die_panic_handler(struct notifier_block *self,
9943 				unsigned long ev, void *unused)
9944 {
9945 	if (!ftrace_dump_on_oops_enabled())
9946 		return NOTIFY_DONE;
9947 
9948 	/* The die notifier requires DIE_OOPS to trigger */
9949 	if (self == &trace_die_notifier && ev != DIE_OOPS)
9950 		return NOTIFY_DONE;
9951 
9952 	ftrace_dump(DUMP_PARAM);
9953 
9954 	return NOTIFY_DONE;
9955 }
9956 
9957 /*
9958  * printk is set to max of 1024, we really don't need it that big.
9959  * Nothing should be printing 1000 characters anyway.
9960  */
9961 #define TRACE_MAX_PRINT		1000
9962 
9963 /*
9964  * Define here KERN_TRACE so that we have one place to modify
9965  * it if we decide to change what log level the ftrace dump
9966  * should be at.
9967  */
9968 #define KERN_TRACE		KERN_EMERG
9969 
9970 void
9971 trace_printk_seq(struct trace_seq *s)
9972 {
9973 	/* Probably should print a warning here. */
9974 	if (s->seq.len >= TRACE_MAX_PRINT)
9975 		s->seq.len = TRACE_MAX_PRINT;
9976 
9977 	/*
9978 	 * More paranoid code. Although the buffer size is set to
9979 	 * PAGE_SIZE, and TRACE_MAX_PRINT is 1000, this is just
9980 	 * an extra layer of protection.
9981 	 */
9982 	if (WARN_ON_ONCE(s->seq.len >= s->seq.size))
9983 		s->seq.len = s->seq.size - 1;
9984 
9985 	/* should be zero ended, but we are paranoid. */
9986 	s->buffer[s->seq.len] = 0;
9987 
9988 	printk(KERN_TRACE "%s", s->buffer);
9989 
9990 	trace_seq_init(s);
9991 }
9992 
9993 static void trace_init_iter(struct trace_iterator *iter, struct trace_array *tr)
9994 {
9995 	iter->tr = tr;
9996 	iter->trace = iter->tr->current_trace;
9997 	iter->cpu_file = RING_BUFFER_ALL_CPUS;
9998 	iter->array_buffer = &tr->array_buffer;
9999 
10000 	if (iter->trace && iter->trace->open)
10001 		iter->trace->open(iter);
10002 
10003 	/* Annotate start of buffers if we had overruns */
10004 	if (ring_buffer_overruns(iter->array_buffer->buffer))
10005 		iter->iter_flags |= TRACE_FILE_ANNOTATE;
10006 
10007 	/* Output in nanoseconds only if we are using a clock in nanoseconds. */
10008 	if (trace_clocks[iter->tr->clock_id].in_ns)
10009 		iter->iter_flags |= TRACE_FILE_TIME_IN_NS;
10010 
10011 	/* Can not use kmalloc for iter.temp and iter.fmt */
10012 	iter->temp = static_temp_buf;
10013 	iter->temp_size = STATIC_TEMP_BUF_SIZE;
10014 	iter->fmt = static_fmt_buf;
10015 	iter->fmt_size = STATIC_FMT_BUF_SIZE;
10016 }
10017 
10018 void trace_init_global_iter(struct trace_iterator *iter)
10019 {
10020 	trace_init_iter(iter, &global_trace);
10021 }
10022 
10023 static void ftrace_dump_one(struct trace_array *tr, enum ftrace_dump_mode dump_mode)
10024 {
10025 	/* use static because iter can be a bit big for the stack */
10026 	static struct trace_iterator iter;
10027 	unsigned int old_userobj;
10028 	unsigned long flags;
10029 	int cnt = 0, cpu;
10030 
10031 	/*
10032 	 * Always turn off tracing when we dump.
10033 	 * We don't need to show trace output of what happens
10034 	 * between multiple crashes.
10035 	 *
10036 	 * If the user does a sysrq-z, then they can re-enable
10037 	 * tracing with echo 1 > tracing_on.
10038 	 */
10039 	tracer_tracing_off(tr);
10040 
10041 	local_irq_save(flags);
10042 
10043 	/* Simulate the iterator */
10044 	trace_init_iter(&iter, tr);
10045 
10046 	for_each_tracing_cpu(cpu) {
10047 		atomic_inc(&per_cpu_ptr(iter.array_buffer->data, cpu)->disabled);
10048 	}
10049 
10050 	old_userobj = tr->trace_flags & TRACE_ITER_SYM_USEROBJ;
10051 
10052 	/* don't look at user memory in panic mode */
10053 	tr->trace_flags &= ~TRACE_ITER_SYM_USEROBJ;
10054 
10055 	if (dump_mode == DUMP_ORIG)
10056 		iter.cpu_file = raw_smp_processor_id();
10057 	else
10058 		iter.cpu_file = RING_BUFFER_ALL_CPUS;
10059 
10060 	if (tr == &global_trace)
10061 		printk(KERN_TRACE "Dumping ftrace buffer:\n");
10062 	else
10063 		printk(KERN_TRACE "Dumping ftrace instance %s buffer:\n", tr->name);
10064 
10065 	/* Did function tracer already get disabled? */
10066 	if (ftrace_is_dead()) {
10067 		printk("# WARNING: FUNCTION TRACING IS CORRUPTED\n");
10068 		printk("#          MAY BE MISSING FUNCTION EVENTS\n");
10069 	}
10070 
10071 	/*
10072 	 * We need to stop all tracing on all CPUS to read
10073 	 * the next buffer. This is a bit expensive, but is
10074 	 * not done often. We fill all what we can read,
10075 	 * and then release the locks again.
10076 	 */
10077 
10078 	while (!trace_empty(&iter)) {
10079 
10080 		if (!cnt)
10081 			printk(KERN_TRACE "---------------------------------\n");
10082 
10083 		cnt++;
10084 
10085 		trace_iterator_reset(&iter);
10086 		iter.iter_flags |= TRACE_FILE_LAT_FMT;
10087 
10088 		if (trace_find_next_entry_inc(&iter) != NULL) {
10089 			int ret;
10090 
10091 			ret = print_trace_line(&iter);
10092 			if (ret != TRACE_TYPE_NO_CONSUME)
10093 				trace_consume(&iter);
10094 		}
10095 		touch_nmi_watchdog();
10096 
10097 		trace_printk_seq(&iter.seq);
10098 	}
10099 
10100 	if (!cnt)
10101 		printk(KERN_TRACE "   (ftrace buffer empty)\n");
10102 	else
10103 		printk(KERN_TRACE "---------------------------------\n");
10104 
10105 	tr->trace_flags |= old_userobj;
10106 
10107 	for_each_tracing_cpu(cpu) {
10108 		atomic_dec(&per_cpu_ptr(iter.array_buffer->data, cpu)->disabled);
10109 	}
10110 	local_irq_restore(flags);
10111 }
10112 
10113 static void ftrace_dump_by_param(void)
10114 {
10115 	bool first_param = true;
10116 	char dump_param[MAX_TRACER_SIZE];
10117 	char *buf, *token, *inst_name;
10118 	struct trace_array *tr;
10119 
10120 	strscpy(dump_param, ftrace_dump_on_oops, MAX_TRACER_SIZE);
10121 	buf = dump_param;
10122 
10123 	while ((token = strsep(&buf, ",")) != NULL) {
10124 		if (first_param) {
10125 			first_param = false;
10126 			if (!strcmp("0", token))
10127 				continue;
10128 			else if (!strcmp("1", token)) {
10129 				ftrace_dump_one(&global_trace, DUMP_ALL);
10130 				continue;
10131 			}
10132 			else if (!strcmp("2", token) ||
10133 			  !strcmp("orig_cpu", token)) {
10134 				ftrace_dump_one(&global_trace, DUMP_ORIG);
10135 				continue;
10136 			}
10137 		}
10138 
10139 		inst_name = strsep(&token, "=");
10140 		tr = trace_array_find(inst_name);
10141 		if (!tr) {
10142 			printk(KERN_TRACE "Instance %s not found\n", inst_name);
10143 			continue;
10144 		}
10145 
10146 		if (token && (!strcmp("2", token) ||
10147 			  !strcmp("orig_cpu", token)))
10148 			ftrace_dump_one(tr, DUMP_ORIG);
10149 		else
10150 			ftrace_dump_one(tr, DUMP_ALL);
10151 	}
10152 }
10153 
10154 void ftrace_dump(enum ftrace_dump_mode oops_dump_mode)
10155 {
10156 	static atomic_t dump_running;
10157 
10158 	/* Only allow one dump user at a time. */
10159 	if (atomic_inc_return(&dump_running) != 1) {
10160 		atomic_dec(&dump_running);
10161 		return;
10162 	}
10163 
10164 	switch (oops_dump_mode) {
10165 	case DUMP_ALL:
10166 		ftrace_dump_one(&global_trace, DUMP_ALL);
10167 		break;
10168 	case DUMP_ORIG:
10169 		ftrace_dump_one(&global_trace, DUMP_ORIG);
10170 		break;
10171 	case DUMP_PARAM:
10172 		ftrace_dump_by_param();
10173 		break;
10174 	case DUMP_NONE:
10175 		break;
10176 	default:
10177 		printk(KERN_TRACE "Bad dumping mode, switching to all CPUs dump\n");
10178 		ftrace_dump_one(&global_trace, DUMP_ALL);
10179 	}
10180 
10181 	atomic_dec(&dump_running);
10182 }
10183 EXPORT_SYMBOL_GPL(ftrace_dump);
10184 
10185 #define WRITE_BUFSIZE  4096
10186 
10187 ssize_t trace_parse_run_command(struct file *file, const char __user *buffer,
10188 				size_t count, loff_t *ppos,
10189 				int (*createfn)(const char *))
10190 {
10191 	char *kbuf, *buf, *tmp;
10192 	int ret = 0;
10193 	size_t done = 0;
10194 	size_t size;
10195 
10196 	kbuf = kmalloc(WRITE_BUFSIZE, GFP_KERNEL);
10197 	if (!kbuf)
10198 		return -ENOMEM;
10199 
10200 	while (done < count) {
10201 		size = count - done;
10202 
10203 		if (size >= WRITE_BUFSIZE)
10204 			size = WRITE_BUFSIZE - 1;
10205 
10206 		if (copy_from_user(kbuf, buffer + done, size)) {
10207 			ret = -EFAULT;
10208 			goto out;
10209 		}
10210 		kbuf[size] = '\0';
10211 		buf = kbuf;
10212 		do {
10213 			tmp = strchr(buf, '\n');
10214 			if (tmp) {
10215 				*tmp = '\0';
10216 				size = tmp - buf + 1;
10217 			} else {
10218 				size = strlen(buf);
10219 				if (done + size < count) {
10220 					if (buf != kbuf)
10221 						break;
10222 					/* This can accept WRITE_BUFSIZE - 2 ('\n' + '\0') */
10223 					pr_warn("Line length is too long: Should be less than %d\n",
10224 						WRITE_BUFSIZE - 2);
10225 					ret = -EINVAL;
10226 					goto out;
10227 				}
10228 			}
10229 			done += size;
10230 
10231 			/* Remove comments */
10232 			tmp = strchr(buf, '#');
10233 
10234 			if (tmp)
10235 				*tmp = '\0';
10236 
10237 			ret = createfn(buf);
10238 			if (ret)
10239 				goto out;
10240 			buf += size;
10241 
10242 		} while (done < count);
10243 	}
10244 	ret = done;
10245 
10246 out:
10247 	kfree(kbuf);
10248 
10249 	return ret;
10250 }
10251 
10252 #ifdef CONFIG_TRACER_MAX_TRACE
10253 __init static bool tr_needs_alloc_snapshot(const char *name)
10254 {
10255 	char *test;
10256 	int len = strlen(name);
10257 	bool ret;
10258 
10259 	if (!boot_snapshot_index)
10260 		return false;
10261 
10262 	if (strncmp(name, boot_snapshot_info, len) == 0 &&
10263 	    boot_snapshot_info[len] == '\t')
10264 		return true;
10265 
10266 	test = kmalloc(strlen(name) + 3, GFP_KERNEL);
10267 	if (!test)
10268 		return false;
10269 
10270 	sprintf(test, "\t%s\t", name);
10271 	ret = strstr(boot_snapshot_info, test) == NULL;
10272 	kfree(test);
10273 	return ret;
10274 }
10275 
10276 __init static void do_allocate_snapshot(const char *name)
10277 {
10278 	if (!tr_needs_alloc_snapshot(name))
10279 		return;
10280 
10281 	/*
10282 	 * When allocate_snapshot is set, the next call to
10283 	 * allocate_trace_buffers() (called by trace_array_get_by_name())
10284 	 * will allocate the snapshot buffer. That will alse clear
10285 	 * this flag.
10286 	 */
10287 	allocate_snapshot = true;
10288 }
10289 #else
10290 static inline void do_allocate_snapshot(const char *name) { }
10291 #endif
10292 
10293 __init static void enable_instances(void)
10294 {
10295 	struct trace_array *tr;
10296 	char *curr_str;
10297 	char *str;
10298 	char *tok;
10299 
10300 	/* A tab is always appended */
10301 	boot_instance_info[boot_instance_index - 1] = '\0';
10302 	str = boot_instance_info;
10303 
10304 	while ((curr_str = strsep(&str, "\t"))) {
10305 
10306 		tok = strsep(&curr_str, ",");
10307 
10308 		if (IS_ENABLED(CONFIG_TRACER_MAX_TRACE))
10309 			do_allocate_snapshot(tok);
10310 
10311 		tr = trace_array_get_by_name(tok, NULL);
10312 		if (!tr) {
10313 			pr_warn("Failed to create instance buffer %s\n", curr_str);
10314 			continue;
10315 		}
10316 		/* Allow user space to delete it */
10317 		trace_array_put(tr);
10318 
10319 		while ((tok = strsep(&curr_str, ","))) {
10320 			early_enable_events(tr, tok, true);
10321 		}
10322 	}
10323 }
10324 
10325 __init static int tracer_alloc_buffers(void)
10326 {
10327 	int ring_buf_size;
10328 	int ret = -ENOMEM;
10329 
10330 
10331 	if (security_locked_down(LOCKDOWN_TRACEFS)) {
10332 		pr_warn("Tracing disabled due to lockdown\n");
10333 		return -EPERM;
10334 	}
10335 
10336 	/*
10337 	 * Make sure we don't accidentally add more trace options
10338 	 * than we have bits for.
10339 	 */
10340 	BUILD_BUG_ON(TRACE_ITER_LAST_BIT > TRACE_FLAGS_MAX_SIZE);
10341 
10342 	if (!alloc_cpumask_var(&tracing_buffer_mask, GFP_KERNEL))
10343 		goto out;
10344 
10345 	if (!alloc_cpumask_var(&global_trace.tracing_cpumask, GFP_KERNEL))
10346 		goto out_free_buffer_mask;
10347 
10348 	/* Only allocate trace_printk buffers if a trace_printk exists */
10349 	if (&__stop___trace_bprintk_fmt != &__start___trace_bprintk_fmt)
10350 		/* Must be called before global_trace.buffer is allocated */
10351 		trace_printk_init_buffers();
10352 
10353 	/* To save memory, keep the ring buffer size to its minimum */
10354 	if (global_trace.ring_buffer_expanded)
10355 		ring_buf_size = trace_buf_size;
10356 	else
10357 		ring_buf_size = 1;
10358 
10359 	cpumask_copy(tracing_buffer_mask, cpu_possible_mask);
10360 	cpumask_copy(global_trace.tracing_cpumask, cpu_all_mask);
10361 
10362 	raw_spin_lock_init(&global_trace.start_lock);
10363 
10364 	/*
10365 	 * The prepare callbacks allocates some memory for the ring buffer. We
10366 	 * don't free the buffer if the CPU goes down. If we were to free
10367 	 * the buffer, then the user would lose any trace that was in the
10368 	 * buffer. The memory will be removed once the "instance" is removed.
10369 	 */
10370 	ret = cpuhp_setup_state_multi(CPUHP_TRACE_RB_PREPARE,
10371 				      "trace/RB:prepare", trace_rb_cpu_prepare,
10372 				      NULL);
10373 	if (ret < 0)
10374 		goto out_free_cpumask;
10375 	/* Used for event triggers */
10376 	ret = -ENOMEM;
10377 	temp_buffer = ring_buffer_alloc(PAGE_SIZE, RB_FL_OVERWRITE);
10378 	if (!temp_buffer)
10379 		goto out_rm_hp_state;
10380 
10381 	if (trace_create_savedcmd() < 0)
10382 		goto out_free_temp_buffer;
10383 
10384 	if (!zalloc_cpumask_var(&global_trace.pipe_cpumask, GFP_KERNEL))
10385 		goto out_free_savedcmd;
10386 
10387 	/* TODO: make the number of buffers hot pluggable with CPUS */
10388 	if (allocate_trace_buffers(&global_trace, ring_buf_size) < 0) {
10389 		MEM_FAIL(1, "tracer: failed to allocate ring buffer!\n");
10390 		goto out_free_pipe_cpumask;
10391 	}
10392 	if (global_trace.buffer_disabled)
10393 		tracing_off();
10394 
10395 	if (trace_boot_clock) {
10396 		ret = tracing_set_clock(&global_trace, trace_boot_clock);
10397 		if (ret < 0)
10398 			pr_warn("Trace clock %s not defined, going back to default\n",
10399 				trace_boot_clock);
10400 	}
10401 
10402 	/*
10403 	 * register_tracer() might reference current_trace, so it
10404 	 * needs to be set before we register anything. This is
10405 	 * just a bootstrap of current_trace anyway.
10406 	 */
10407 	global_trace.current_trace = &nop_trace;
10408 
10409 	global_trace.max_lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
10410 #ifdef CONFIG_TRACER_MAX_TRACE
10411 	spin_lock_init(&global_trace.snapshot_trigger_lock);
10412 #endif
10413 	ftrace_init_global_array_ops(&global_trace);
10414 
10415 	init_trace_flags_index(&global_trace);
10416 
10417 	register_tracer(&nop_trace);
10418 
10419 	/* Function tracing may start here (via kernel command line) */
10420 	init_function_trace();
10421 
10422 	/* All seems OK, enable tracing */
10423 	tracing_disabled = 0;
10424 
10425 	atomic_notifier_chain_register(&panic_notifier_list,
10426 				       &trace_panic_notifier);
10427 
10428 	register_die_notifier(&trace_die_notifier);
10429 
10430 	global_trace.flags = TRACE_ARRAY_FL_GLOBAL;
10431 
10432 	INIT_LIST_HEAD(&global_trace.systems);
10433 	INIT_LIST_HEAD(&global_trace.events);
10434 	INIT_LIST_HEAD(&global_trace.hist_vars);
10435 	INIT_LIST_HEAD(&global_trace.err_log);
10436 	list_add(&global_trace.list, &ftrace_trace_arrays);
10437 
10438 	apply_trace_boot_options();
10439 
10440 	register_snapshot_cmd();
10441 
10442 	test_can_verify();
10443 
10444 	return 0;
10445 
10446 out_free_pipe_cpumask:
10447 	free_cpumask_var(global_trace.pipe_cpumask);
10448 out_free_savedcmd:
10449 	trace_free_saved_cmdlines_buffer();
10450 out_free_temp_buffer:
10451 	ring_buffer_free(temp_buffer);
10452 out_rm_hp_state:
10453 	cpuhp_remove_multi_state(CPUHP_TRACE_RB_PREPARE);
10454 out_free_cpumask:
10455 	free_cpumask_var(global_trace.tracing_cpumask);
10456 out_free_buffer_mask:
10457 	free_cpumask_var(tracing_buffer_mask);
10458 out:
10459 	return ret;
10460 }
10461 
10462 void __init ftrace_boot_snapshot(void)
10463 {
10464 #ifdef CONFIG_TRACER_MAX_TRACE
10465 	struct trace_array *tr;
10466 
10467 	if (!snapshot_at_boot)
10468 		return;
10469 
10470 	list_for_each_entry(tr, &ftrace_trace_arrays, list) {
10471 		if (!tr->allocated_snapshot)
10472 			continue;
10473 
10474 		tracing_snapshot_instance(tr);
10475 		trace_array_puts(tr, "** Boot snapshot taken **\n");
10476 	}
10477 #endif
10478 }
10479 
10480 void __init early_trace_init(void)
10481 {
10482 	if (tracepoint_printk) {
10483 		tracepoint_print_iter =
10484 			kzalloc(sizeof(*tracepoint_print_iter), GFP_KERNEL);
10485 		if (MEM_FAIL(!tracepoint_print_iter,
10486 			     "Failed to allocate trace iterator\n"))
10487 			tracepoint_printk = 0;
10488 		else
10489 			static_key_enable(&tracepoint_printk_key.key);
10490 	}
10491 	tracer_alloc_buffers();
10492 
10493 	init_events();
10494 }
10495 
10496 void __init trace_init(void)
10497 {
10498 	trace_event_init();
10499 
10500 	if (boot_instance_index)
10501 		enable_instances();
10502 }
10503 
10504 __init static void clear_boot_tracer(void)
10505 {
10506 	/*
10507 	 * The default tracer at boot buffer is an init section.
10508 	 * This function is called in lateinit. If we did not
10509 	 * find the boot tracer, then clear it out, to prevent
10510 	 * later registration from accessing the buffer that is
10511 	 * about to be freed.
10512 	 */
10513 	if (!default_bootup_tracer)
10514 		return;
10515 
10516 	printk(KERN_INFO "ftrace bootup tracer '%s' not registered.\n",
10517 	       default_bootup_tracer);
10518 	default_bootup_tracer = NULL;
10519 }
10520 
10521 #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
10522 __init static void tracing_set_default_clock(void)
10523 {
10524 	/* sched_clock_stable() is determined in late_initcall */
10525 	if (!trace_boot_clock && !sched_clock_stable()) {
10526 		if (security_locked_down(LOCKDOWN_TRACEFS)) {
10527 			pr_warn("Can not set tracing clock due to lockdown\n");
10528 			return;
10529 		}
10530 
10531 		printk(KERN_WARNING
10532 		       "Unstable clock detected, switching default tracing clock to \"global\"\n"
10533 		       "If you want to keep using the local clock, then add:\n"
10534 		       "  \"trace_clock=local\"\n"
10535 		       "on the kernel command line\n");
10536 		tracing_set_clock(&global_trace, "global");
10537 	}
10538 }
10539 #else
10540 static inline void tracing_set_default_clock(void) { }
10541 #endif
10542 
10543 __init static int late_trace_init(void)
10544 {
10545 	if (tracepoint_printk && tracepoint_printk_stop_on_boot) {
10546 		static_key_disable(&tracepoint_printk_key.key);
10547 		tracepoint_printk = 0;
10548 	}
10549 
10550 	tracing_set_default_clock();
10551 	clear_boot_tracer();
10552 	return 0;
10553 }
10554 
10555 late_initcall_sync(late_trace_init);
10556