]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - kernel/trace/trace.c
Merge branch 'tip/tracing/ftrace' of git://git.kernel.org/pub/scm/linux/kernel/git...
[mirror_ubuntu-jammy-kernel.git] / kernel / trace / trace.c
CommitLineData
bc0c38d1
SR
1/*
2 * ring buffer based function tracer
3 *
4 * Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
5 * Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
6 *
7 * Originally taken from the RT patch by:
8 * Arnaldo Carvalho de Melo <acme@redhat.com>
9 *
10 * Based on code from the latency_tracer, that is:
11 * Copyright (C) 2004-2006 Ingo Molnar
12 * Copyright (C) 2004 William Lee Irwin III
13 */
14#include <linux/utsrelease.h>
15#include <linux/kallsyms.h>
16#include <linux/seq_file.h>
3f5a54e3 17#include <linux/notifier.h>
bc0c38d1 18#include <linux/debugfs.h>
4c11d7ae 19#include <linux/pagemap.h>
bc0c38d1
SR
20#include <linux/hardirq.h>
21#include <linux/linkage.h>
22#include <linux/uaccess.h>
23#include <linux/ftrace.h>
24#include <linux/module.h>
25#include <linux/percpu.h>
3f5a54e3 26#include <linux/kdebug.h>
bc0c38d1
SR
27#include <linux/ctype.h>
28#include <linux/init.h>
2a2cc8f7 29#include <linux/poll.h>
bc0c38d1
SR
30#include <linux/gfp.h>
31#include <linux/fs.h>
76094a2c 32#include <linux/kprobes.h>
3eefae99 33#include <linux/writeback.h>
3c56819b 34#include <linux/splice.h>
bc0c38d1 35
86387f7e 36#include <linux/stacktrace.h>
3928a8a2 37#include <linux/ring_buffer.h>
21798a84 38#include <linux/irqflags.h>
86387f7e 39
bc0c38d1 40#include "trace.h"
f0868d1e 41#include "trace_output.h"
bc0c38d1 42
3928a8a2
SR
43#define TRACE_BUFFER_FLAGS (RB_FL_OVERWRITE)
44
745b1626 45unsigned long __read_mostly tracing_max_latency;
bc0c38d1
SR
46unsigned long __read_mostly tracing_thresh;
47
8e1b82e0
FW
48/*
49 * We need to change this state when a selftest is running.
ff32504f
FW
50 * A selftest will lurk into the ring-buffer to count the
51 * entries inserted during the selftest although some concurrent
52 * insertions into the ring-buffer such as ftrace_printk could occurred
53 * at the same time, giving false positive or negative results.
54 */
8e1b82e0 55static bool __read_mostly tracing_selftest_running;
ff32504f 56
b2821ae6
SR
57/*
58 * If a tracer is running, we do not want to run SELFTEST.
59 */
60static bool __read_mostly tracing_selftest_disabled;
61
adf9f195
FW
62/* For tracers that don't implement custom flags */
63static struct tracer_opt dummy_tracer_opt[] = {
64 { }
65};
66
67static struct tracer_flags dummy_tracer_flags = {
68 .val = 0,
69 .opts = dummy_tracer_opt
70};
71
72static int dummy_set_flag(u32 old_flags, u32 bit, int set)
73{
74 return 0;
75}
0f048701
SR
76
77/*
78 * Kill all tracing for good (never come back).
79 * It is initialized to 1 but will turn to zero if the initialization
80 * of the tracer is successful. But that is the only place that sets
81 * this back to zero.
82 */
83int tracing_disabled = 1;
84
d769041f
SR
85static DEFINE_PER_CPU(local_t, ftrace_cpu_disabled);
86
87static inline void ftrace_disable_cpu(void)
88{
89 preempt_disable();
90 local_inc(&__get_cpu_var(ftrace_cpu_disabled));
91}
92
93static inline void ftrace_enable_cpu(void)
94{
95 local_dec(&__get_cpu_var(ftrace_cpu_disabled));
96 preempt_enable();
97}
98
9e01c1b7 99static cpumask_var_t __read_mostly tracing_buffer_mask;
ab46428c
SR
100
101#define for_each_tracing_cpu(cpu) \
9e01c1b7 102 for_each_cpu(cpu, tracing_buffer_mask)
ab46428c 103
944ac425
SR
104/*
105 * ftrace_dump_on_oops - variable to dump ftrace buffer on oops
106 *
107 * If there is an oops (or kernel panic) and the ftrace_dump_on_oops
108 * is set, then ftrace_dump is called. This will output the contents
109 * of the ftrace buffers to the console. This is very useful for
110 * capturing traces that lead to crashes and outputing it to a
111 * serial console.
112 *
113 * It is default off, but you can enable it with either specifying
114 * "ftrace_dump_on_oops" in the kernel command line, or setting
115 * /proc/sys/kernel/ftrace_dump_on_oops to true.
116 */
117int ftrace_dump_on_oops;
118
b2821ae6
SR
119static int tracing_set_tracer(const char *buf);
120
121#define BOOTUP_TRACER_SIZE 100
122static char bootup_tracer_buf[BOOTUP_TRACER_SIZE] __initdata;
123static char *default_bootup_tracer;
d9e54076
PZ
124
125static int __init set_ftrace(char *str)
126{
b2821ae6
SR
127 strncpy(bootup_tracer_buf, str, BOOTUP_TRACER_SIZE);
128 default_bootup_tracer = bootup_tracer_buf;
d9e54076
PZ
129 return 1;
130}
b2821ae6 131__setup("ftrace=", set_ftrace);
d9e54076 132
944ac425
SR
133static int __init set_ftrace_dump_on_oops(char *str)
134{
135 ftrace_dump_on_oops = 1;
136 return 1;
137}
138__setup("ftrace_dump_on_oops", set_ftrace_dump_on_oops);
60a11774 139
72829bc3 140long
bc0c38d1
SR
141ns2usecs(cycle_t nsec)
142{
143 nsec += 500;
144 do_div(nsec, 1000);
145 return nsec;
146}
147
e309b41d 148cycle_t ftrace_now(int cpu)
750ed1a4 149{
3928a8a2
SR
150 u64 ts = ring_buffer_time_stamp(cpu);
151 ring_buffer_normalize_time_stamp(cpu, &ts);
152 return ts;
750ed1a4
IM
153}
154
4fcdae83
SR
155/*
156 * The global_trace is the descriptor that holds the tracing
157 * buffers for the live tracing. For each CPU, it contains
158 * a link list of pages that will store trace entries. The
159 * page descriptor of the pages in the memory is used to hold
160 * the link list by linking the lru item in the page descriptor
161 * to each of the pages in the buffer per CPU.
162 *
163 * For each active CPU there is a data field that holds the
164 * pages for the buffer for that CPU. Each CPU has the same number
165 * of pages allocated for its buffer.
166 */
bc0c38d1
SR
167static struct trace_array global_trace;
168
169static DEFINE_PER_CPU(struct trace_array_cpu, global_trace_cpu);
170
4fcdae83
SR
171/*
172 * The max_tr is used to snapshot the global_trace when a maximum
173 * latency is reached. Some tracers will use this to store a maximum
174 * trace while it continues examining live traces.
175 *
176 * The buffers for the max_tr are set up the same as the global_trace.
177 * When a snapshot is taken, the link list of the max_tr is swapped
178 * with the link list of the global_trace and the buffers are reset for
179 * the global_trace so the tracing can continue.
180 */
bc0c38d1
SR
181static struct trace_array max_tr;
182
183static DEFINE_PER_CPU(struct trace_array_cpu, max_data);
184
4fcdae83 185/* tracer_enabled is used to toggle activation of a tracer */
26994ead 186static int tracer_enabled = 1;
4fcdae83 187
9036990d
SR
188/**
189 * tracing_is_enabled - return tracer_enabled status
190 *
191 * This function is used by other tracers to know the status
192 * of the tracer_enabled flag. Tracers may use this function
193 * to know if it should enable their features when starting
194 * up. See irqsoff tracer for an example (start_irqsoff_tracer).
195 */
196int tracing_is_enabled(void)
197{
198 return tracer_enabled;
199}
200
4fcdae83 201/*
3928a8a2
SR
202 * trace_buf_size is the size in bytes that is allocated
203 * for a buffer. Note, the number of bytes is always rounded
204 * to page size.
3f5a54e3
SR
205 *
206 * This number is purposely set to a low number of 16384.
207 * If the dump on oops happens, it will be much appreciated
208 * to not have to wait for all that output. Anyway this can be
209 * boot time and run time configurable.
4fcdae83 210 */
3928a8a2 211#define TRACE_BUF_SIZE_DEFAULT 1441792UL /* 16384 * 88 (sizeof(entry)) */
3f5a54e3 212
3928a8a2 213static unsigned long trace_buf_size = TRACE_BUF_SIZE_DEFAULT;
bc0c38d1 214
4fcdae83 215/* trace_types holds a link list of available tracers. */
bc0c38d1 216static struct tracer *trace_types __read_mostly;
4fcdae83
SR
217
218/* current_trace points to the tracer that is currently active */
bc0c38d1 219static struct tracer *current_trace __read_mostly;
4fcdae83
SR
220
221/*
222 * max_tracer_type_len is used to simplify the allocating of
223 * buffers to read userspace tracer names. We keep track of
224 * the longest tracer name registered.
225 */
bc0c38d1
SR
226static int max_tracer_type_len;
227
4fcdae83
SR
228/*
229 * trace_types_lock is used to protect the trace_types list.
230 * This lock is also used to keep user access serialized.
231 * Accesses from userspace will grab this lock while userspace
232 * activities happen inside the kernel.
233 */
bc0c38d1 234static DEFINE_MUTEX(trace_types_lock);
4fcdae83
SR
235
236/* trace_wait is a waitqueue for tasks blocked on trace_poll */
4e655519
IM
237static DECLARE_WAIT_QUEUE_HEAD(trace_wait);
238
ee6bce52 239/* trace_flags holds trace_options default values */
12ef7d44 240unsigned long trace_flags = TRACE_ITER_PRINT_PARENT | TRACE_ITER_PRINTK |
c4a8e8be 241 TRACE_ITER_ANNOTATE | TRACE_ITER_CONTEXT_INFO;
4e655519 242
4fcdae83
SR
243/**
244 * trace_wake_up - wake up tasks waiting for trace input
245 *
246 * Simply wakes up any task that is blocked on the trace_wait
247 * queue. These is used with trace_poll for tasks polling the trace.
248 */
4e655519
IM
249void trace_wake_up(void)
250{
017730c1
IM
251 /*
252 * The runqueue_is_locked() can fail, but this is the best we
253 * have for now:
254 */
255 if (!(trace_flags & TRACE_ITER_BLOCK) && !runqueue_is_locked())
4e655519
IM
256 wake_up(&trace_wait);
257}
bc0c38d1 258
3928a8a2 259static int __init set_buf_size(char *str)
bc0c38d1 260{
3928a8a2 261 unsigned long buf_size;
c6caeeb1
SR
262 int ret;
263
bc0c38d1
SR
264 if (!str)
265 return 0;
3928a8a2 266 ret = strict_strtoul(str, 0, &buf_size);
c6caeeb1 267 /* nr_entries can not be zero */
3928a8a2 268 if (ret < 0 || buf_size == 0)
c6caeeb1 269 return 0;
3928a8a2 270 trace_buf_size = buf_size;
bc0c38d1
SR
271 return 1;
272}
3928a8a2 273__setup("trace_buf_size=", set_buf_size);
bc0c38d1 274
57f50be1
SR
275unsigned long nsecs_to_usecs(unsigned long nsecs)
276{
277 return nsecs / 1000;
278}
279
4fcdae83 280/* These must match the bit postions in trace_iterator_flags */
bc0c38d1
SR
281static const char *trace_options[] = {
282 "print-parent",
283 "sym-offset",
284 "sym-addr",
285 "verbose",
f9896bf3 286 "raw",
5e3ca0ec 287 "hex",
cb0f12aa 288 "bin",
2a2cc8f7 289 "block",
86387f7e 290 "stacktrace",
4ac3ba41 291 "sched-tree",
f09ce573 292 "ftrace_printk",
b2a866f9 293 "ftrace_preempt",
9f029e83 294 "branch",
12ef7d44 295 "annotate",
02b67518 296 "userstacktrace",
b54d3de9 297 "sym-userobj",
66896a85 298 "printk-msg-only",
c4a8e8be 299 "context-info",
bc0c38d1
SR
300 NULL
301};
302
4fcdae83
SR
303/*
304 * ftrace_max_lock is used to protect the swapping of buffers
305 * when taking a max snapshot. The buffers themselves are
306 * protected by per_cpu spinlocks. But the action of the swap
307 * needs its own lock.
308 *
309 * This is defined as a raw_spinlock_t in order to help
310 * with performance when lockdep debugging is enabled.
311 */
92205c23
SR
312static raw_spinlock_t ftrace_max_lock =
313 (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
bc0c38d1
SR
314
315/*
316 * Copy the new maximum trace into the separate maximum-trace
317 * structure. (this way the maximum trace is permanently saved,
318 * for later retrieval via /debugfs/tracing/latency_trace)
319 */
e309b41d 320static void
bc0c38d1
SR
321__update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
322{
323 struct trace_array_cpu *data = tr->data[cpu];
324
325 max_tr.cpu = cpu;
326 max_tr.time_start = data->preempt_timestamp;
327
328 data = max_tr.data[cpu];
329 data->saved_latency = tracing_max_latency;
330
331 memcpy(data->comm, tsk->comm, TASK_COMM_LEN);
332 data->pid = tsk->pid;
b6dff3ec 333 data->uid = task_uid(tsk);
bc0c38d1
SR
334 data->nice = tsk->static_prio - 20 - MAX_RT_PRIO;
335 data->policy = tsk->policy;
336 data->rt_priority = tsk->rt_priority;
337
338 /* record this tasks comm */
339 tracing_record_cmdline(current);
340}
341
e309b41d 342static void
214023c3
SR
343trace_seq_reset(struct trace_seq *s)
344{
345 s->len = 0;
6c6c2796
PP
346 s->readpos = 0;
347}
348
349ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, size_t cnt)
350{
351 int len;
352 int ret;
353
354 if (s->len <= s->readpos)
355 return -EBUSY;
356
357 len = s->len - s->readpos;
358 if (cnt > len)
359 cnt = len;
360 ret = copy_to_user(ubuf, s->buffer + s->readpos, cnt);
361 if (ret)
362 return -EFAULT;
363
364 s->readpos += len;
365 return cnt;
214023c3
SR
366}
367
3c56819b
EGM
368ssize_t trace_seq_to_buffer(struct trace_seq *s, void *buf, size_t cnt)
369{
370 int len;
371 void *ret;
372
373 if (s->len <= s->readpos)
374 return -EBUSY;
375
376 len = s->len - s->readpos;
377 if (cnt > len)
378 cnt = len;
379 ret = memcpy(buf, s->buffer + s->readpos, cnt);
380 if (!ret)
381 return -EFAULT;
382
383 s->readpos += len;
384 return cnt;
385}
386
e309b41d 387static void
214023c3
SR
388trace_print_seq(struct seq_file *m, struct trace_seq *s)
389{
390 int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len;
391
392 s->buffer[len] = 0;
393 seq_puts(m, s->buffer);
394
395 trace_seq_reset(s);
396}
397
4fcdae83
SR
398/**
399 * update_max_tr - snapshot all trace buffers from global_trace to max_tr
400 * @tr: tracer
401 * @tsk: the task with the latency
402 * @cpu: The cpu that initiated the trace.
403 *
404 * Flip the buffers between the @tr and the max_tr and record information
405 * about which task was the cause of this latency.
406 */
e309b41d 407void
bc0c38d1
SR
408update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
409{
3928a8a2 410 struct ring_buffer *buf = tr->buffer;
bc0c38d1 411
4c11d7ae 412 WARN_ON_ONCE(!irqs_disabled());
92205c23 413 __raw_spin_lock(&ftrace_max_lock);
3928a8a2
SR
414
415 tr->buffer = max_tr.buffer;
416 max_tr.buffer = buf;
417
d769041f 418 ftrace_disable_cpu();
3928a8a2 419 ring_buffer_reset(tr->buffer);
d769041f 420 ftrace_enable_cpu();
bc0c38d1
SR
421
422 __update_max_tr(tr, tsk, cpu);
92205c23 423 __raw_spin_unlock(&ftrace_max_lock);
bc0c38d1
SR
424}
425
426/**
427 * update_max_tr_single - only copy one trace over, and reset the rest
428 * @tr - tracer
429 * @tsk - task with the latency
430 * @cpu - the cpu of the buffer to copy.
4fcdae83
SR
431 *
432 * Flip the trace of a single CPU buffer between the @tr and the max_tr.
bc0c38d1 433 */
e309b41d 434void
bc0c38d1
SR
435update_max_tr_single(struct trace_array *tr, struct task_struct *tsk, int cpu)
436{
3928a8a2 437 int ret;
bc0c38d1 438
4c11d7ae 439 WARN_ON_ONCE(!irqs_disabled());
92205c23 440 __raw_spin_lock(&ftrace_max_lock);
bc0c38d1 441
d769041f
SR
442 ftrace_disable_cpu();
443
3928a8a2
SR
444 ring_buffer_reset(max_tr.buffer);
445 ret = ring_buffer_swap_cpu(max_tr.buffer, tr->buffer, cpu);
446
d769041f
SR
447 ftrace_enable_cpu();
448
97b17efe 449 WARN_ON_ONCE(ret && ret != -EAGAIN);
bc0c38d1
SR
450
451 __update_max_tr(tr, tsk, cpu);
92205c23 452 __raw_spin_unlock(&ftrace_max_lock);
bc0c38d1
SR
453}
454
4fcdae83
SR
455/**
456 * register_tracer - register a tracer with the ftrace system.
457 * @type - the plugin for the tracer
458 *
459 * Register a new plugin tracer.
460 */
bc0c38d1
SR
461int register_tracer(struct tracer *type)
462{
463 struct tracer *t;
464 int len;
465 int ret = 0;
466
467 if (!type->name) {
468 pr_info("Tracer must have a name\n");
469 return -1;
470 }
471
86fa2f60
IM
472 /*
473 * When this gets called we hold the BKL which means that
474 * preemption is disabled. Various trace selftests however
475 * need to disable and enable preemption for successful tests.
476 * So we drop the BKL here and grab it after the tests again.
477 */
478 unlock_kernel();
bc0c38d1 479 mutex_lock(&trace_types_lock);
86fa2f60 480
8e1b82e0
FW
481 tracing_selftest_running = true;
482
bc0c38d1
SR
483 for (t = trace_types; t; t = t->next) {
484 if (strcmp(type->name, t->name) == 0) {
485 /* already found */
486 pr_info("Trace %s already registered\n",
487 type->name);
488 ret = -1;
489 goto out;
490 }
491 }
492
adf9f195
FW
493 if (!type->set_flag)
494 type->set_flag = &dummy_set_flag;
495 if (!type->flags)
496 type->flags = &dummy_tracer_flags;
497 else
498 if (!type->flags->opts)
499 type->flags->opts = dummy_tracer_opt;
500
60a11774 501#ifdef CONFIG_FTRACE_STARTUP_TEST
b2821ae6 502 if (type->selftest && !tracing_selftest_disabled) {
60a11774 503 struct tracer *saved_tracer = current_trace;
60a11774 504 struct trace_array *tr = &global_trace;
60a11774 505 int i;
ff32504f 506
60a11774
SR
507 /*
508 * Run a selftest on this tracer.
509 * Here we reset the trace buffer, and set the current
510 * tracer to be this tracer. The tracer can then run some
511 * internal tracing to verify that everything is in order.
512 * If we fail, we do not register this tracer.
513 */
86fa2f60 514 for_each_tracing_cpu(i)
3928a8a2 515 tracing_reset(tr, i);
86fa2f60 516
60a11774 517 current_trace = type;
60a11774
SR
518 /* the test is responsible for initializing and enabling */
519 pr_info("Testing tracer %s: ", type->name);
520 ret = type->selftest(type, tr);
521 /* the test is responsible for resetting too */
522 current_trace = saved_tracer;
60a11774
SR
523 if (ret) {
524 printk(KERN_CONT "FAILED!\n");
525 goto out;
526 }
1d4db00a 527 /* Only reset on passing, to avoid touching corrupted buffers */
86fa2f60 528 for_each_tracing_cpu(i)
3928a8a2 529 tracing_reset(tr, i);
86fa2f60 530
60a11774
SR
531 printk(KERN_CONT "PASSED\n");
532 }
533#endif
534
bc0c38d1
SR
535 type->next = trace_types;
536 trace_types = type;
537 len = strlen(type->name);
538 if (len > max_tracer_type_len)
539 max_tracer_type_len = len;
60a11774 540
bc0c38d1 541 out:
8e1b82e0 542 tracing_selftest_running = false;
bc0c38d1
SR
543 mutex_unlock(&trace_types_lock);
544
dac74940
SR
545 if (ret || !default_bootup_tracer)
546 goto out_unlock;
547
548 if (strncmp(default_bootup_tracer, type->name, BOOTUP_TRACER_SIZE))
549 goto out_unlock;
550
551 printk(KERN_INFO "Starting tracer '%s'\n", type->name);
552 /* Do we want this tracer to start on bootup? */
553 tracing_set_tracer(type->name);
554 default_bootup_tracer = NULL;
555 /* disable other selftests, since this will break it. */
556 tracing_selftest_disabled = 1;
b2821ae6 557#ifdef CONFIG_FTRACE_STARTUP_TEST
dac74940
SR
558 printk(KERN_INFO "Disabling FTRACE selftests due to running tracer '%s'\n",
559 type->name);
b2821ae6 560#endif
b2821ae6 561
dac74940 562 out_unlock:
b2821ae6 563 lock_kernel();
bc0c38d1
SR
564 return ret;
565}
566
567void unregister_tracer(struct tracer *type)
568{
569 struct tracer **t;
570 int len;
571
572 mutex_lock(&trace_types_lock);
573 for (t = &trace_types; *t; t = &(*t)->next) {
574 if (*t == type)
575 goto found;
576 }
577 pr_info("Trace %s not registered\n", type->name);
578 goto out;
579
580 found:
581 *t = (*t)->next;
b5db03c4
ACM
582
583 if (type == current_trace && tracer_enabled) {
584 tracer_enabled = 0;
585 tracing_stop();
586 if (current_trace->stop)
587 current_trace->stop(&global_trace);
588 current_trace = &nop_trace;
589 }
590
bc0c38d1
SR
591 if (strlen(type->name) != max_tracer_type_len)
592 goto out;
593
594 max_tracer_type_len = 0;
595 for (t = &trace_types; *t; t = &(*t)->next) {
596 len = strlen((*t)->name);
597 if (len > max_tracer_type_len)
598 max_tracer_type_len = len;
599 }
600 out:
601 mutex_unlock(&trace_types_lock);
602}
603
3928a8a2 604void tracing_reset(struct trace_array *tr, int cpu)
bc0c38d1 605{
d769041f 606 ftrace_disable_cpu();
3928a8a2 607 ring_buffer_reset_cpu(tr->buffer, cpu);
d769041f 608 ftrace_enable_cpu();
bc0c38d1
SR
609}
610
213cc060
PE
611void tracing_reset_online_cpus(struct trace_array *tr)
612{
613 int cpu;
614
615 tr->time_start = ftrace_now(tr->cpu);
616
617 for_each_online_cpu(cpu)
618 tracing_reset(tr, cpu);
619}
620
bc0c38d1
SR
621#define SAVED_CMDLINES 128
622static unsigned map_pid_to_cmdline[PID_MAX_DEFAULT+1];
623static unsigned map_cmdline_to_pid[SAVED_CMDLINES];
624static char saved_cmdlines[SAVED_CMDLINES][TASK_COMM_LEN];
625static int cmdline_idx;
626static DEFINE_SPINLOCK(trace_cmdline_lock);
25b0b44a 627
25b0b44a
SR
628/* temporary disable recording */
629atomic_t trace_record_cmdline_disabled __read_mostly;
bc0c38d1
SR
630
631static void trace_init_cmdlines(void)
632{
633 memset(&map_pid_to_cmdline, -1, sizeof(map_pid_to_cmdline));
634 memset(&map_cmdline_to_pid, -1, sizeof(map_cmdline_to_pid));
635 cmdline_idx = 0;
636}
637
0f048701
SR
638static int trace_stop_count;
639static DEFINE_SPINLOCK(tracing_start_lock);
640
69bb54ec
SR
641/**
642 * ftrace_off_permanent - disable all ftrace code permanently
643 *
644 * This should only be called when a serious anomally has
645 * been detected. This will turn off the function tracing,
646 * ring buffers, and other tracing utilites. It takes no
647 * locks and can be called from any context.
648 */
649void ftrace_off_permanent(void)
650{
651 tracing_disabled = 1;
652 ftrace_stop();
653 tracing_off_permanent();
654}
655
0f048701
SR
656/**
657 * tracing_start - quick start of the tracer
658 *
659 * If tracing is enabled but was stopped by tracing_stop,
660 * this will start the tracer back up.
661 */
662void tracing_start(void)
663{
664 struct ring_buffer *buffer;
665 unsigned long flags;
666
667 if (tracing_disabled)
668 return;
669
670 spin_lock_irqsave(&tracing_start_lock, flags);
b06a8301
SR
671 if (--trace_stop_count) {
672 if (trace_stop_count < 0) {
673 /* Someone screwed up their debugging */
674 WARN_ON_ONCE(1);
675 trace_stop_count = 0;
676 }
0f048701
SR
677 goto out;
678 }
679
680
681 buffer = global_trace.buffer;
682 if (buffer)
683 ring_buffer_record_enable(buffer);
684
685 buffer = max_tr.buffer;
686 if (buffer)
687 ring_buffer_record_enable(buffer);
688
689 ftrace_start();
690 out:
691 spin_unlock_irqrestore(&tracing_start_lock, flags);
692}
693
694/**
695 * tracing_stop - quick stop of the tracer
696 *
697 * Light weight way to stop tracing. Use in conjunction with
698 * tracing_start.
699 */
700void tracing_stop(void)
701{
702 struct ring_buffer *buffer;
703 unsigned long flags;
704
705 ftrace_stop();
706 spin_lock_irqsave(&tracing_start_lock, flags);
707 if (trace_stop_count++)
708 goto out;
709
710 buffer = global_trace.buffer;
711 if (buffer)
712 ring_buffer_record_disable(buffer);
713
714 buffer = max_tr.buffer;
715 if (buffer)
716 ring_buffer_record_disable(buffer);
717
718 out:
719 spin_unlock_irqrestore(&tracing_start_lock, flags);
720}
721
e309b41d 722void trace_stop_cmdline_recording(void);
bc0c38d1 723
e309b41d 724static void trace_save_cmdline(struct task_struct *tsk)
bc0c38d1
SR
725{
726 unsigned map;
727 unsigned idx;
728
729 if (!tsk->pid || unlikely(tsk->pid > PID_MAX_DEFAULT))
730 return;
731
732 /*
733 * It's not the end of the world if we don't get
734 * the lock, but we also don't want to spin
735 * nor do we want to disable interrupts,
736 * so if we miss here, then better luck next time.
737 */
738 if (!spin_trylock(&trace_cmdline_lock))
739 return;
740
741 idx = map_pid_to_cmdline[tsk->pid];
742 if (idx >= SAVED_CMDLINES) {
743 idx = (cmdline_idx + 1) % SAVED_CMDLINES;
744
745 map = map_cmdline_to_pid[idx];
746 if (map <= PID_MAX_DEFAULT)
747 map_pid_to_cmdline[map] = (unsigned)-1;
748
749 map_pid_to_cmdline[tsk->pid] = idx;
750
751 cmdline_idx = idx;
752 }
753
754 memcpy(&saved_cmdlines[idx], tsk->comm, TASK_COMM_LEN);
755
756 spin_unlock(&trace_cmdline_lock);
757}
758
660c7f9b 759char *trace_find_cmdline(int pid)
bc0c38d1
SR
760{
761 char *cmdline = "<...>";
762 unsigned map;
763
764 if (!pid)
765 return "<idle>";
766
767 if (pid > PID_MAX_DEFAULT)
768 goto out;
769
770 map = map_pid_to_cmdline[pid];
771 if (map >= SAVED_CMDLINES)
772 goto out;
773
774 cmdline = saved_cmdlines[map];
775
776 out:
777 return cmdline;
778}
779
e309b41d 780void tracing_record_cmdline(struct task_struct *tsk)
bc0c38d1
SR
781{
782 if (atomic_read(&trace_record_cmdline_disabled))
783 return;
784
785 trace_save_cmdline(tsk);
786}
787
45dcd8b8 788void
38697053
SR
789tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags,
790 int pc)
bc0c38d1
SR
791{
792 struct task_struct *tsk = current;
bc0c38d1 793
777e208d
SR
794 entry->preempt_count = pc & 0xff;
795 entry->pid = (tsk) ? tsk->pid : 0;
b54d3de9 796 entry->tgid = (tsk) ? tsk->tgid : 0;
777e208d 797 entry->flags =
9244489a 798#ifdef CONFIG_TRACE_IRQFLAGS_SUPPORT
2e2ca155 799 (irqs_disabled_flags(flags) ? TRACE_FLAG_IRQS_OFF : 0) |
9244489a
SR
800#else
801 TRACE_FLAG_IRQS_NOSUPPORT |
802#endif
bc0c38d1
SR
803 ((pc & HARDIRQ_MASK) ? TRACE_FLAG_HARDIRQ : 0) |
804 ((pc & SOFTIRQ_MASK) ? TRACE_FLAG_SOFTIRQ : 0) |
805 (need_resched() ? TRACE_FLAG_NEED_RESCHED : 0);
806}
807
51a763dd
ACM
808struct ring_buffer_event *trace_buffer_lock_reserve(struct trace_array *tr,
809 unsigned char type,
810 unsigned long len,
811 unsigned long flags, int pc)
812{
813 struct ring_buffer_event *event;
814
815 event = ring_buffer_lock_reserve(tr->buffer, len);
816 if (event != NULL) {
817 struct trace_entry *ent = ring_buffer_event_data(event);
818
819 tracing_generic_entry_update(ent, flags, pc);
820 ent->type = type;
821 }
822
823 return event;
824}
825static void ftrace_trace_stack(struct trace_array *tr,
826 unsigned long flags, int skip, int pc);
827static void ftrace_trace_userstack(struct trace_array *tr,
828 unsigned long flags, int pc);
829
830void trace_buffer_unlock_commit(struct trace_array *tr,
831 struct ring_buffer_event *event,
832 unsigned long flags, int pc)
833{
834 ring_buffer_unlock_commit(tr->buffer, event);
835
836 ftrace_trace_stack(tr, flags, 6, pc);
837 ftrace_trace_userstack(tr, flags, pc);
838 trace_wake_up();
839}
840
e309b41d 841void
7be42151 842trace_function(struct trace_array *tr,
38697053
SR
843 unsigned long ip, unsigned long parent_ip, unsigned long flags,
844 int pc)
bc0c38d1 845{
3928a8a2 846 struct ring_buffer_event *event;
777e208d 847 struct ftrace_entry *entry;
bc0c38d1 848
d769041f
SR
849 /* If we are reading the ring buffer, don't trace */
850 if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled))))
851 return;
852
51a763dd
ACM
853 event = trace_buffer_lock_reserve(tr, TRACE_FN, sizeof(*entry),
854 flags, pc);
3928a8a2
SR
855 if (!event)
856 return;
857 entry = ring_buffer_event_data(event);
777e208d
SR
858 entry->ip = ip;
859 entry->parent_ip = parent_ip;
0a987751 860 ring_buffer_unlock_commit(tr->buffer, event);
bc0c38d1
SR
861}
862
fb52607a 863#ifdef CONFIG_FUNCTION_GRAPH_TRACER
287b6e68 864static void __trace_graph_entry(struct trace_array *tr,
287b6e68
FW
865 struct ftrace_graph_ent *trace,
866 unsigned long flags,
867 int pc)
868{
869 struct ring_buffer_event *event;
870 struct ftrace_graph_ent_entry *entry;
287b6e68
FW
871
872 if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled))))
873 return;
874
51a763dd
ACM
875 event = trace_buffer_lock_reserve(&global_trace, TRACE_GRAPH_ENT,
876 sizeof(*entry), flags, pc);
287b6e68
FW
877 if (!event)
878 return;
879 entry = ring_buffer_event_data(event);
287b6e68 880 entry->graph_ent = *trace;
0a987751 881 ring_buffer_unlock_commit(global_trace.buffer, event);
287b6e68
FW
882}
883
884static void __trace_graph_return(struct trace_array *tr,
fb52607a 885 struct ftrace_graph_ret *trace,
15e6cb36
FW
886 unsigned long flags,
887 int pc)
888{
889 struct ring_buffer_event *event;
287b6e68 890 struct ftrace_graph_ret_entry *entry;
15e6cb36
FW
891
892 if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled))))
893 return;
894
51a763dd
ACM
895 event = trace_buffer_lock_reserve(&global_trace, TRACE_GRAPH_RET,
896 sizeof(*entry), flags, pc);
15e6cb36
FW
897 if (!event)
898 return;
899 entry = ring_buffer_event_data(event);
287b6e68 900 entry->ret = *trace;
0a987751 901 ring_buffer_unlock_commit(global_trace.buffer, event);
15e6cb36
FW
902}
903#endif
904
e309b41d 905void
2e0f5761 906ftrace(struct trace_array *tr, struct trace_array_cpu *data,
38697053
SR
907 unsigned long ip, unsigned long parent_ip, unsigned long flags,
908 int pc)
2e0f5761
IM
909{
910 if (likely(!atomic_read(&data->disabled)))
7be42151 911 trace_function(tr, ip, parent_ip, flags, pc);
2e0f5761
IM
912}
913
53614991 914static void __ftrace_trace_stack(struct trace_array *tr,
53614991
SR
915 unsigned long flags,
916 int skip, int pc)
86387f7e 917{
c2c80529 918#ifdef CONFIG_STACKTRACE
3928a8a2 919 struct ring_buffer_event *event;
777e208d 920 struct stack_entry *entry;
86387f7e
IM
921 struct stack_trace trace;
922
51a763dd
ACM
923 event = trace_buffer_lock_reserve(tr, TRACE_STACK,
924 sizeof(*entry), flags, pc);
3928a8a2
SR
925 if (!event)
926 return;
927 entry = ring_buffer_event_data(event);
777e208d 928 memset(&entry->caller, 0, sizeof(entry->caller));
86387f7e
IM
929
930 trace.nr_entries = 0;
931 trace.max_entries = FTRACE_STACK_ENTRIES;
932 trace.skip = skip;
777e208d 933 trace.entries = entry->caller;
86387f7e
IM
934
935 save_stack_trace(&trace);
0a987751 936 ring_buffer_unlock_commit(tr->buffer, event);
c2c80529 937#endif
f0a920d5
IM
938}
939
53614991 940static void ftrace_trace_stack(struct trace_array *tr,
53614991
SR
941 unsigned long flags,
942 int skip, int pc)
943{
944 if (!(trace_flags & TRACE_ITER_STACKTRACE))
945 return;
946
7be42151 947 __ftrace_trace_stack(tr, flags, skip, pc);
53614991
SR
948}
949
38697053 950void __trace_stack(struct trace_array *tr,
38697053 951 unsigned long flags,
53614991 952 int skip, int pc)
38697053 953{
7be42151 954 __ftrace_trace_stack(tr, flags, skip, pc);
38697053
SR
955}
956
02b67518 957static void ftrace_trace_userstack(struct trace_array *tr,
7be42151 958 unsigned long flags, int pc)
02b67518 959{
c7425acb 960#ifdef CONFIG_STACKTRACE
8d7c6a96 961 struct ring_buffer_event *event;
02b67518
TE
962 struct userstack_entry *entry;
963 struct stack_trace trace;
02b67518
TE
964
965 if (!(trace_flags & TRACE_ITER_USERSTACKTRACE))
966 return;
967
51a763dd
ACM
968 event = trace_buffer_lock_reserve(tr, TRACE_USER_STACK,
969 sizeof(*entry), flags, pc);
02b67518
TE
970 if (!event)
971 return;
972 entry = ring_buffer_event_data(event);
02b67518
TE
973
974 memset(&entry->caller, 0, sizeof(entry->caller));
975
976 trace.nr_entries = 0;
977 trace.max_entries = FTRACE_STACK_ENTRIES;
978 trace.skip = 0;
979 trace.entries = entry->caller;
980
981 save_stack_trace_user(&trace);
0a987751 982 ring_buffer_unlock_commit(tr->buffer, event);
c7425acb 983#endif
02b67518
TE
984}
985
7be42151 986void __trace_userstack(struct trace_array *tr, unsigned long flags)
02b67518 987{
7be42151 988 ftrace_trace_userstack(tr, flags, preempt_count());
02b67518
TE
989}
990
38697053 991static void
7be42151 992ftrace_trace_special(void *__tr,
38697053
SR
993 unsigned long arg1, unsigned long arg2, unsigned long arg3,
994 int pc)
a4feb834 995{
3928a8a2 996 struct ring_buffer_event *event;
a4feb834 997 struct trace_array *tr = __tr;
777e208d 998 struct special_entry *entry;
a4feb834 999
51a763dd
ACM
1000 event = trace_buffer_lock_reserve(tr, TRACE_SPECIAL,
1001 sizeof(*entry), 0, pc);
3928a8a2
SR
1002 if (!event)
1003 return;
1004 entry = ring_buffer_event_data(event);
777e208d
SR
1005 entry->arg1 = arg1;
1006 entry->arg2 = arg2;
1007 entry->arg3 = arg3;
51a763dd 1008 trace_buffer_unlock_commit(tr, event, 0, pc);
a4feb834
IM
1009}
1010
38697053
SR
1011void
1012__trace_special(void *__tr, void *__data,
1013 unsigned long arg1, unsigned long arg2, unsigned long arg3)
1014{
7be42151 1015 ftrace_trace_special(__tr, arg1, arg2, arg3, preempt_count());
38697053
SR
1016}
1017
e309b41d 1018void
bc0c38d1 1019tracing_sched_switch_trace(struct trace_array *tr,
86387f7e
IM
1020 struct task_struct *prev,
1021 struct task_struct *next,
38697053 1022 unsigned long flags, int pc)
bc0c38d1 1023{
3928a8a2 1024 struct ring_buffer_event *event;
777e208d 1025 struct ctx_switch_entry *entry;
bc0c38d1 1026
51a763dd
ACM
1027 event = trace_buffer_lock_reserve(tr, TRACE_CTX,
1028 sizeof(*entry), flags, pc);
3928a8a2
SR
1029 if (!event)
1030 return;
1031 entry = ring_buffer_event_data(event);
777e208d
SR
1032 entry->prev_pid = prev->pid;
1033 entry->prev_prio = prev->prio;
1034 entry->prev_state = prev->state;
1035 entry->next_pid = next->pid;
1036 entry->next_prio = next->prio;
1037 entry->next_state = next->state;
1038 entry->next_cpu = task_cpu(next);
51a763dd 1039 trace_buffer_unlock_commit(tr, event, flags, pc);
bc0c38d1
SR
1040}
1041
57422797
IM
1042void
1043tracing_sched_wakeup_trace(struct trace_array *tr,
86387f7e
IM
1044 struct task_struct *wakee,
1045 struct task_struct *curr,
38697053 1046 unsigned long flags, int pc)
57422797 1047{
3928a8a2 1048 struct ring_buffer_event *event;
777e208d 1049 struct ctx_switch_entry *entry;
57422797 1050
51a763dd
ACM
1051 event = trace_buffer_lock_reserve(tr, TRACE_WAKE,
1052 sizeof(*entry), flags, pc);
3928a8a2
SR
1053 if (!event)
1054 return;
1055 entry = ring_buffer_event_data(event);
777e208d
SR
1056 entry->prev_pid = curr->pid;
1057 entry->prev_prio = curr->prio;
1058 entry->prev_state = curr->state;
1059 entry->next_pid = wakee->pid;
1060 entry->next_prio = wakee->prio;
1061 entry->next_state = wakee->state;
1062 entry->next_cpu = task_cpu(wakee);
51a763dd 1063 trace_buffer_unlock_commit(tr, event, flags, pc);
57422797
IM
1064}
1065
4902f884
SR
1066void
1067ftrace_special(unsigned long arg1, unsigned long arg2, unsigned long arg3)
1068{
1069 struct trace_array *tr = &global_trace;
1070 struct trace_array_cpu *data;
5aa1ba6a 1071 unsigned long flags;
4902f884 1072 int cpu;
38697053 1073 int pc;
4902f884 1074
c76f0694 1075 if (tracing_disabled)
4902f884
SR
1076 return;
1077
38697053 1078 pc = preempt_count();
5aa1ba6a 1079 local_irq_save(flags);
4902f884
SR
1080 cpu = raw_smp_processor_id();
1081 data = tr->data[cpu];
4902f884 1082
5aa1ba6a 1083 if (likely(atomic_inc_return(&data->disabled) == 1))
7be42151 1084 ftrace_trace_special(tr, arg1, arg2, arg3, pc);
4902f884 1085
5aa1ba6a
SR
1086 atomic_dec(&data->disabled);
1087 local_irq_restore(flags);
4902f884
SR
1088}
1089
fb52607a 1090#ifdef CONFIG_FUNCTION_GRAPH_TRACER
e49dc19c 1091int trace_graph_entry(struct ftrace_graph_ent *trace)
15e6cb36
FW
1092{
1093 struct trace_array *tr = &global_trace;
1094 struct trace_array_cpu *data;
1095 unsigned long flags;
1096 long disabled;
1097 int cpu;
1098 int pc;
1099
804a6851
SR
1100 if (!ftrace_trace_task(current))
1101 return 0;
1102
ea4e2bc4
SR
1103 if (!ftrace_graph_addr(trace->func))
1104 return 0;
1105
a5e25883 1106 local_irq_save(flags);
15e6cb36
FW
1107 cpu = raw_smp_processor_id();
1108 data = tr->data[cpu];
1109 disabled = atomic_inc_return(&data->disabled);
1110 if (likely(disabled == 1)) {
1111 pc = preempt_count();
7be42151 1112 __trace_graph_entry(tr, trace, flags, pc);
287b6e68 1113 }
ea4e2bc4
SR
1114 /* Only do the atomic if it is not already set */
1115 if (!test_tsk_trace_graph(current))
1116 set_tsk_trace_graph(current);
287b6e68 1117 atomic_dec(&data->disabled);
a5e25883 1118 local_irq_restore(flags);
e49dc19c
SR
1119
1120 return 1;
287b6e68
FW
1121}
1122
1123void trace_graph_return(struct ftrace_graph_ret *trace)
1124{
1125 struct trace_array *tr = &global_trace;
1126 struct trace_array_cpu *data;
1127 unsigned long flags;
1128 long disabled;
1129 int cpu;
1130 int pc;
1131
a5e25883 1132 local_irq_save(flags);
287b6e68
FW
1133 cpu = raw_smp_processor_id();
1134 data = tr->data[cpu];
1135 disabled = atomic_inc_return(&data->disabled);
1136 if (likely(disabled == 1)) {
1137 pc = preempt_count();
7be42151 1138 __trace_graph_return(tr, trace, flags, pc);
15e6cb36 1139 }
ea4e2bc4
SR
1140 if (!trace->depth)
1141 clear_tsk_trace_graph(current);
15e6cb36 1142 atomic_dec(&data->disabled);
a5e25883 1143 local_irq_restore(flags);
15e6cb36 1144}
fb52607a 1145#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
15e6cb36 1146
bc0c38d1
SR
1147enum trace_file_type {
1148 TRACE_FILE_LAT_FMT = 1,
12ef7d44 1149 TRACE_FILE_ANNOTATE = 2,
bc0c38d1
SR
1150};
1151
e2ac8ef5 1152static void trace_iterator_increment(struct trace_iterator *iter)
5a90f577 1153{
d769041f
SR
1154 /* Don't allow ftrace to trace into the ring buffers */
1155 ftrace_disable_cpu();
1156
5a90f577 1157 iter->idx++;
d769041f
SR
1158 if (iter->buffer_iter[iter->cpu])
1159 ring_buffer_read(iter->buffer_iter[iter->cpu], NULL);
1160
1161 ftrace_enable_cpu();
5a90f577
SR
1162}
1163
e309b41d 1164static struct trace_entry *
3928a8a2 1165peek_next_entry(struct trace_iterator *iter, int cpu, u64 *ts)
dd0e545f 1166{
3928a8a2
SR
1167 struct ring_buffer_event *event;
1168 struct ring_buffer_iter *buf_iter = iter->buffer_iter[cpu];
dd0e545f 1169
d769041f
SR
1170 /* Don't allow ftrace to trace into the ring buffers */
1171 ftrace_disable_cpu();
1172
1173 if (buf_iter)
1174 event = ring_buffer_iter_peek(buf_iter, ts);
1175 else
1176 event = ring_buffer_peek(iter->tr->buffer, cpu, ts);
1177
1178 ftrace_enable_cpu();
1179
3928a8a2 1180 return event ? ring_buffer_event_data(event) : NULL;
dd0e545f 1181}
d769041f 1182
dd0e545f 1183static struct trace_entry *
3928a8a2 1184__find_next_entry(struct trace_iterator *iter, int *ent_cpu, u64 *ent_ts)
bc0c38d1 1185{
3928a8a2 1186 struct ring_buffer *buffer = iter->tr->buffer;
bc0c38d1 1187 struct trace_entry *ent, *next = NULL;
3928a8a2 1188 u64 next_ts = 0, ts;
bc0c38d1
SR
1189 int next_cpu = -1;
1190 int cpu;
1191
ab46428c 1192 for_each_tracing_cpu(cpu) {
dd0e545f 1193
3928a8a2
SR
1194 if (ring_buffer_empty_cpu(buffer, cpu))
1195 continue;
dd0e545f 1196
3928a8a2 1197 ent = peek_next_entry(iter, cpu, &ts);
dd0e545f 1198
cdd31cd2
IM
1199 /*
1200 * Pick the entry with the smallest timestamp:
1201 */
3928a8a2 1202 if (ent && (!next || ts < next_ts)) {
bc0c38d1
SR
1203 next = ent;
1204 next_cpu = cpu;
3928a8a2 1205 next_ts = ts;
bc0c38d1
SR
1206 }
1207 }
1208
1209 if (ent_cpu)
1210 *ent_cpu = next_cpu;
1211
3928a8a2
SR
1212 if (ent_ts)
1213 *ent_ts = next_ts;
1214
bc0c38d1
SR
1215 return next;
1216}
1217
dd0e545f 1218/* Find the next real entry, without updating the iterator itself */
c4a8e8be
FW
1219struct trace_entry *trace_find_next_entry(struct trace_iterator *iter,
1220 int *ent_cpu, u64 *ent_ts)
bc0c38d1 1221{
3928a8a2 1222 return __find_next_entry(iter, ent_cpu, ent_ts);
dd0e545f
SR
1223}
1224
1225/* Find the next real entry, and increment the iterator to the next entry */
1226static void *find_next_entry_inc(struct trace_iterator *iter)
1227{
3928a8a2 1228 iter->ent = __find_next_entry(iter, &iter->cpu, &iter->ts);
dd0e545f 1229
3928a8a2 1230 if (iter->ent)
e2ac8ef5 1231 trace_iterator_increment(iter);
dd0e545f 1232
3928a8a2 1233 return iter->ent ? iter : NULL;
b3806b43 1234}
bc0c38d1 1235
e309b41d 1236static void trace_consume(struct trace_iterator *iter)
b3806b43 1237{
d769041f
SR
1238 /* Don't allow ftrace to trace into the ring buffers */
1239 ftrace_disable_cpu();
3928a8a2 1240 ring_buffer_consume(iter->tr->buffer, iter->cpu, &iter->ts);
d769041f 1241 ftrace_enable_cpu();
bc0c38d1
SR
1242}
1243
e309b41d 1244static void *s_next(struct seq_file *m, void *v, loff_t *pos)
bc0c38d1
SR
1245{
1246 struct trace_iterator *iter = m->private;
bc0c38d1 1247 int i = (int)*pos;
4e3c3333 1248 void *ent;
bc0c38d1
SR
1249
1250 (*pos)++;
1251
1252 /* can't go backwards */
1253 if (iter->idx > i)
1254 return NULL;
1255
1256 if (iter->idx < 0)
1257 ent = find_next_entry_inc(iter);
1258 else
1259 ent = iter;
1260
1261 while (ent && iter->idx < i)
1262 ent = find_next_entry_inc(iter);
1263
1264 iter->pos = *pos;
1265
bc0c38d1
SR
1266 return ent;
1267}
1268
1269static void *s_start(struct seq_file *m, loff_t *pos)
1270{
1271 struct trace_iterator *iter = m->private;
1272 void *p = NULL;
1273 loff_t l = 0;
3928a8a2 1274 int cpu;
bc0c38d1
SR
1275
1276 mutex_lock(&trace_types_lock);
1277
d15f57f2
SR
1278 if (!current_trace || current_trace != iter->trace) {
1279 mutex_unlock(&trace_types_lock);
bc0c38d1 1280 return NULL;
d15f57f2 1281 }
bc0c38d1
SR
1282
1283 atomic_inc(&trace_record_cmdline_disabled);
1284
bc0c38d1
SR
1285 if (*pos != iter->pos) {
1286 iter->ent = NULL;
1287 iter->cpu = 0;
1288 iter->idx = -1;
1289
d769041f
SR
1290 ftrace_disable_cpu();
1291
3928a8a2
SR
1292 for_each_tracing_cpu(cpu) {
1293 ring_buffer_iter_reset(iter->buffer_iter[cpu]);
4c11d7ae 1294 }
bc0c38d1 1295
d769041f
SR
1296 ftrace_enable_cpu();
1297
bc0c38d1
SR
1298 for (p = iter; p && l < *pos; p = s_next(m, p, &l))
1299 ;
1300
1301 } else {
4c11d7ae 1302 l = *pos - 1;
bc0c38d1
SR
1303 p = s_next(m, p, &l);
1304 }
1305
1306 return p;
1307}
1308
1309static void s_stop(struct seq_file *m, void *p)
1310{
bc0c38d1 1311 atomic_dec(&trace_record_cmdline_disabled);
bc0c38d1
SR
1312 mutex_unlock(&trace_types_lock);
1313}
1314
e309b41d 1315static void print_lat_help_header(struct seq_file *m)
bc0c38d1 1316{
a6168353
ME
1317 seq_puts(m, "# _------=> CPU# \n");
1318 seq_puts(m, "# / _-----=> irqs-off \n");
1319 seq_puts(m, "# | / _----=> need-resched \n");
1320 seq_puts(m, "# || / _---=> hardirq/softirq \n");
1321 seq_puts(m, "# ||| / _--=> preempt-depth \n");
1322 seq_puts(m, "# |||| / \n");
1323 seq_puts(m, "# ||||| delay \n");
1324 seq_puts(m, "# cmd pid ||||| time | caller \n");
1325 seq_puts(m, "# \\ / ||||| \\ | / \n");
bc0c38d1
SR
1326}
1327
e309b41d 1328static void print_func_help_header(struct seq_file *m)
bc0c38d1 1329{
a6168353
ME
1330 seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n");
1331 seq_puts(m, "# | | | | |\n");
bc0c38d1
SR
1332}
1333
1334
e309b41d 1335static void
bc0c38d1
SR
1336print_trace_header(struct seq_file *m, struct trace_iterator *iter)
1337{
1338 unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
1339 struct trace_array *tr = iter->tr;
1340 struct trace_array_cpu *data = tr->data[tr->cpu];
1341 struct tracer *type = current_trace;
3928a8a2
SR
1342 unsigned long total;
1343 unsigned long entries;
bc0c38d1
SR
1344 const char *name = "preemption";
1345
1346 if (type)
1347 name = type->name;
1348
3928a8a2
SR
1349 entries = ring_buffer_entries(iter->tr->buffer);
1350 total = entries +
1351 ring_buffer_overruns(iter->tr->buffer);
bc0c38d1
SR
1352
1353 seq_printf(m, "%s latency trace v1.1.5 on %s\n",
1354 name, UTS_RELEASE);
1355 seq_puts(m, "-----------------------------------"
1356 "---------------------------------\n");
1357 seq_printf(m, " latency: %lu us, #%lu/%lu, CPU#%d |"
1358 " (M:%s VP:%d, KP:%d, SP:%d HP:%d",
57f50be1 1359 nsecs_to_usecs(data->saved_latency),
bc0c38d1 1360 entries,
4c11d7ae 1361 total,
bc0c38d1
SR
1362 tr->cpu,
1363#if defined(CONFIG_PREEMPT_NONE)
1364 "server",
1365#elif defined(CONFIG_PREEMPT_VOLUNTARY)
1366 "desktop",
b5c21b45 1367#elif defined(CONFIG_PREEMPT)
bc0c38d1
SR
1368 "preempt",
1369#else
1370 "unknown",
1371#endif
1372 /* These are reserved for later use */
1373 0, 0, 0, 0);
1374#ifdef CONFIG_SMP
1375 seq_printf(m, " #P:%d)\n", num_online_cpus());
1376#else
1377 seq_puts(m, ")\n");
1378#endif
1379 seq_puts(m, " -----------------\n");
1380 seq_printf(m, " | task: %.16s-%d "
1381 "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n",
1382 data->comm, data->pid, data->uid, data->nice,
1383 data->policy, data->rt_priority);
1384 seq_puts(m, " -----------------\n");
1385
1386 if (data->critical_start) {
1387 seq_puts(m, " => started at: ");
214023c3
SR
1388 seq_print_ip_sym(&iter->seq, data->critical_start, sym_flags);
1389 trace_print_seq(m, &iter->seq);
bc0c38d1 1390 seq_puts(m, "\n => ended at: ");
214023c3
SR
1391 seq_print_ip_sym(&iter->seq, data->critical_end, sym_flags);
1392 trace_print_seq(m, &iter->seq);
bc0c38d1
SR
1393 seq_puts(m, "\n");
1394 }
1395
1396 seq_puts(m, "\n");
1397}
1398
a309720c
SR
1399static void test_cpu_buff_start(struct trace_iterator *iter)
1400{
1401 struct trace_seq *s = &iter->seq;
1402
12ef7d44
SR
1403 if (!(trace_flags & TRACE_ITER_ANNOTATE))
1404 return;
1405
1406 if (!(iter->iter_flags & TRACE_FILE_ANNOTATE))
1407 return;
1408
4462344e 1409 if (cpumask_test_cpu(iter->cpu, iter->started))
a309720c
SR
1410 return;
1411
4462344e 1412 cpumask_set_cpu(iter->cpu, iter->started);
a309720c
SR
1413 trace_seq_printf(s, "##### CPU %u buffer started ####\n", iter->cpu);
1414}
1415
c4a8e8be 1416static enum print_line_t print_lat_fmt(struct trace_iterator *iter)
bc0c38d1 1417{
214023c3 1418 struct trace_seq *s = &iter->seq;
bc0c38d1 1419 unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
f633cef0 1420 struct trace_event *event;
bc0c38d1 1421 struct trace_entry *entry = iter->ent;
dd0e545f 1422
a309720c
SR
1423 test_cpu_buff_start(iter);
1424
c4a8e8be
FW
1425 event = ftrace_find_event(entry->type);
1426
1427 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
d9793bd8
ACM
1428 if (!trace_print_lat_context(iter))
1429 goto partial;
bc0c38d1 1430 }
777e208d 1431
268ccda0 1432 if (event)
d9793bd8
ACM
1433 return event->latency_trace(iter, sym_flags);
1434
1435 if (!trace_seq_printf(s, "Unknown type %d\n", entry->type))
1436 goto partial;
02b67518 1437
2c4f035f 1438 return TRACE_TYPE_HANDLED;
d9793bd8
ACM
1439partial:
1440 return TRACE_TYPE_PARTIAL_LINE;
bc0c38d1
SR
1441}
1442
2c4f035f 1443static enum print_line_t print_trace_fmt(struct trace_iterator *iter)
bc0c38d1 1444{
214023c3 1445 struct trace_seq *s = &iter->seq;
bc0c38d1 1446 unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
4e3c3333 1447 struct trace_entry *entry;
f633cef0 1448 struct trace_event *event;
bc0c38d1 1449
4e3c3333 1450 entry = iter->ent;
dd0e545f 1451
a309720c
SR
1452 test_cpu_buff_start(iter);
1453
c4a8e8be 1454 event = ftrace_find_event(entry->type);
bc0c38d1 1455
c4a8e8be 1456 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
d9793bd8
ACM
1457 if (!trace_print_context(iter))
1458 goto partial;
c4a8e8be 1459 }
bc0c38d1 1460
268ccda0 1461 if (event)
d9793bd8
ACM
1462 return event->trace(iter, sym_flags);
1463
1464 if (!trace_seq_printf(s, "Unknown type %d\n", entry->type))
1465 goto partial;
02b67518 1466
2c4f035f 1467 return TRACE_TYPE_HANDLED;
d9793bd8
ACM
1468partial:
1469 return TRACE_TYPE_PARTIAL_LINE;
bc0c38d1
SR
1470}
1471
2c4f035f 1472static enum print_line_t print_raw_fmt(struct trace_iterator *iter)
f9896bf3
IM
1473{
1474 struct trace_seq *s = &iter->seq;
1475 struct trace_entry *entry;
f633cef0 1476 struct trace_event *event;
f9896bf3
IM
1477
1478 entry = iter->ent;
dd0e545f 1479
c4a8e8be 1480 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
d9793bd8
ACM
1481 if (!trace_seq_printf(s, "%d %d %llu ",
1482 entry->pid, iter->cpu, iter->ts))
1483 goto partial;
c4a8e8be 1484 }
f9896bf3 1485
f633cef0 1486 event = ftrace_find_event(entry->type);
268ccda0 1487 if (event)
d9793bd8
ACM
1488 return event->raw(iter, 0);
1489
1490 if (!trace_seq_printf(s, "%d ?\n", entry->type))
1491 goto partial;
777e208d 1492
2c4f035f 1493 return TRACE_TYPE_HANDLED;
d9793bd8
ACM
1494partial:
1495 return TRACE_TYPE_PARTIAL_LINE;
f9896bf3
IM
1496}
1497
2c4f035f 1498static enum print_line_t print_hex_fmt(struct trace_iterator *iter)
5e3ca0ec
IM
1499{
1500 struct trace_seq *s = &iter->seq;
1501 unsigned char newline = '\n';
1502 struct trace_entry *entry;
f633cef0 1503 struct trace_event *event;
5e3ca0ec
IM
1504
1505 entry = iter->ent;
dd0e545f 1506
c4a8e8be
FW
1507 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
1508 SEQ_PUT_HEX_FIELD_RET(s, entry->pid);
1509 SEQ_PUT_HEX_FIELD_RET(s, iter->cpu);
1510 SEQ_PUT_HEX_FIELD_RET(s, iter->ts);
1511 }
5e3ca0ec 1512
f633cef0 1513 event = ftrace_find_event(entry->type);
268ccda0 1514 if (event) {
ae7462b4 1515 enum print_line_t ret = event->hex(iter, 0);
d9793bd8
ACM
1516 if (ret != TRACE_TYPE_HANDLED)
1517 return ret;
1518 }
7104f300 1519
5e3ca0ec
IM
1520 SEQ_PUT_FIELD_RET(s, newline);
1521
2c4f035f 1522 return TRACE_TYPE_HANDLED;
5e3ca0ec
IM
1523}
1524
66896a85
FW
1525static enum print_line_t print_printk_msg_only(struct trace_iterator *iter)
1526{
1527 struct trace_seq *s = &iter->seq;
1528 struct trace_entry *entry = iter->ent;
1529 struct print_entry *field;
1530 int ret;
1531
1532 trace_assign_type(field, entry);
1533
c4a8e8be 1534 ret = trace_seq_printf(s, "%s", field->buf);
66896a85
FW
1535 if (!ret)
1536 return TRACE_TYPE_PARTIAL_LINE;
1537
66896a85
FW
1538 return TRACE_TYPE_HANDLED;
1539}
1540
2c4f035f 1541static enum print_line_t print_bin_fmt(struct trace_iterator *iter)
cb0f12aa
IM
1542{
1543 struct trace_seq *s = &iter->seq;
1544 struct trace_entry *entry;
f633cef0 1545 struct trace_event *event;
cb0f12aa
IM
1546
1547 entry = iter->ent;
dd0e545f 1548
c4a8e8be
FW
1549 if (trace_flags & TRACE_ITER_CONTEXT_INFO) {
1550 SEQ_PUT_FIELD_RET(s, entry->pid);
1830b52d 1551 SEQ_PUT_FIELD_RET(s, iter->cpu);
c4a8e8be
FW
1552 SEQ_PUT_FIELD_RET(s, iter->ts);
1553 }
cb0f12aa 1554
f633cef0 1555 event = ftrace_find_event(entry->type);
268ccda0 1556 return event ? event->binary(iter, 0) : TRACE_TYPE_HANDLED;
cb0f12aa
IM
1557}
1558
bc0c38d1
SR
1559static int trace_empty(struct trace_iterator *iter)
1560{
bc0c38d1
SR
1561 int cpu;
1562
ab46428c 1563 for_each_tracing_cpu(cpu) {
d769041f
SR
1564 if (iter->buffer_iter[cpu]) {
1565 if (!ring_buffer_iter_empty(iter->buffer_iter[cpu]))
1566 return 0;
1567 } else {
1568 if (!ring_buffer_empty_cpu(iter->tr->buffer, cpu))
1569 return 0;
1570 }
bc0c38d1 1571 }
d769041f 1572
797d3712 1573 return 1;
bc0c38d1
SR
1574}
1575
2c4f035f 1576static enum print_line_t print_trace_line(struct trace_iterator *iter)
f9896bf3 1577{
2c4f035f
FW
1578 enum print_line_t ret;
1579
1580 if (iter->trace && iter->trace->print_line) {
1581 ret = iter->trace->print_line(iter);
1582 if (ret != TRACE_TYPE_UNHANDLED)
1583 return ret;
1584 }
72829bc3 1585
66896a85
FW
1586 if (iter->ent->type == TRACE_PRINT &&
1587 trace_flags & TRACE_ITER_PRINTK &&
1588 trace_flags & TRACE_ITER_PRINTK_MSGONLY)
1589 return print_printk_msg_only(iter);
1590
cb0f12aa
IM
1591 if (trace_flags & TRACE_ITER_BIN)
1592 return print_bin_fmt(iter);
1593
5e3ca0ec
IM
1594 if (trace_flags & TRACE_ITER_HEX)
1595 return print_hex_fmt(iter);
1596
f9896bf3
IM
1597 if (trace_flags & TRACE_ITER_RAW)
1598 return print_raw_fmt(iter);
1599
1600 if (iter->iter_flags & TRACE_FILE_LAT_FMT)
c4a8e8be 1601 return print_lat_fmt(iter);
f9896bf3
IM
1602
1603 return print_trace_fmt(iter);
1604}
1605
bc0c38d1
SR
1606static int s_show(struct seq_file *m, void *v)
1607{
1608 struct trace_iterator *iter = v;
1609
1610 if (iter->ent == NULL) {
1611 if (iter->tr) {
1612 seq_printf(m, "# tracer: %s\n", iter->trace->name);
1613 seq_puts(m, "#\n");
1614 }
8bba1bf5
MM
1615 if (iter->trace && iter->trace->print_header)
1616 iter->trace->print_header(m);
1617 else if (iter->iter_flags & TRACE_FILE_LAT_FMT) {
bc0c38d1
SR
1618 /* print nothing if the buffers are empty */
1619 if (trace_empty(iter))
1620 return 0;
1621 print_trace_header(m, iter);
1622 if (!(trace_flags & TRACE_ITER_VERBOSE))
1623 print_lat_help_header(m);
1624 } else {
1625 if (!(trace_flags & TRACE_ITER_VERBOSE))
1626 print_func_help_header(m);
1627 }
1628 } else {
f9896bf3 1629 print_trace_line(iter);
214023c3 1630 trace_print_seq(m, &iter->seq);
bc0c38d1
SR
1631 }
1632
1633 return 0;
1634}
1635
1636static struct seq_operations tracer_seq_ops = {
4bf39a94
IM
1637 .start = s_start,
1638 .next = s_next,
1639 .stop = s_stop,
1640 .show = s_show,
bc0c38d1
SR
1641};
1642
e309b41d 1643static struct trace_iterator *
bc0c38d1
SR
1644__tracing_open(struct inode *inode, struct file *file, int *ret)
1645{
1646 struct trace_iterator *iter;
3928a8a2
SR
1647 struct seq_file *m;
1648 int cpu;
bc0c38d1 1649
60a11774
SR
1650 if (tracing_disabled) {
1651 *ret = -ENODEV;
1652 return NULL;
1653 }
1654
bc0c38d1
SR
1655 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
1656 if (!iter) {
1657 *ret = -ENOMEM;
1658 goto out;
1659 }
1660
1661 mutex_lock(&trace_types_lock);
1662 if (current_trace && current_trace->print_max)
1663 iter->tr = &max_tr;
1664 else
1665 iter->tr = inode->i_private;
1666 iter->trace = current_trace;
1667 iter->pos = -1;
1668
8bba1bf5
MM
1669 /* Notify the tracer early; before we stop tracing. */
1670 if (iter->trace && iter->trace->open)
a93751ca 1671 iter->trace->open(iter);
8bba1bf5 1672
12ef7d44
SR
1673 /* Annotate start of buffers if we had overruns */
1674 if (ring_buffer_overruns(iter->tr->buffer))
1675 iter->iter_flags |= TRACE_FILE_ANNOTATE;
1676
1677
3928a8a2 1678 for_each_tracing_cpu(cpu) {
d769041f 1679
3928a8a2
SR
1680 iter->buffer_iter[cpu] =
1681 ring_buffer_read_start(iter->tr->buffer, cpu);
d769041f 1682
3928a8a2
SR
1683 if (!iter->buffer_iter[cpu])
1684 goto fail_buffer;
1685 }
1686
bc0c38d1
SR
1687 /* TODO stop tracer */
1688 *ret = seq_open(file, &tracer_seq_ops);
3928a8a2
SR
1689 if (*ret)
1690 goto fail_buffer;
bc0c38d1 1691
3928a8a2
SR
1692 m = file->private_data;
1693 m->private = iter;
bc0c38d1 1694
3928a8a2 1695 /* stop the trace while dumping */
9036990d 1696 tracing_stop();
3928a8a2 1697
bc0c38d1
SR
1698 mutex_unlock(&trace_types_lock);
1699
1700 out:
1701 return iter;
3928a8a2
SR
1702
1703 fail_buffer:
1704 for_each_tracing_cpu(cpu) {
1705 if (iter->buffer_iter[cpu])
1706 ring_buffer_read_finish(iter->buffer_iter[cpu]);
1707 }
1708 mutex_unlock(&trace_types_lock);
0bb943c7 1709 kfree(iter);
3928a8a2
SR
1710
1711 return ERR_PTR(-ENOMEM);
bc0c38d1
SR
1712}
1713
1714int tracing_open_generic(struct inode *inode, struct file *filp)
1715{
60a11774
SR
1716 if (tracing_disabled)
1717 return -ENODEV;
1718
bc0c38d1
SR
1719 filp->private_data = inode->i_private;
1720 return 0;
1721}
1722
1723int tracing_release(struct inode *inode, struct file *file)
1724{
1725 struct seq_file *m = (struct seq_file *)file->private_data;
1726 struct trace_iterator *iter = m->private;
3928a8a2 1727 int cpu;
bc0c38d1
SR
1728
1729 mutex_lock(&trace_types_lock);
3928a8a2
SR
1730 for_each_tracing_cpu(cpu) {
1731 if (iter->buffer_iter[cpu])
1732 ring_buffer_read_finish(iter->buffer_iter[cpu]);
1733 }
1734
bc0c38d1
SR
1735 if (iter->trace && iter->trace->close)
1736 iter->trace->close(iter);
1737
1738 /* reenable tracing if it was previously enabled */
9036990d 1739 tracing_start();
bc0c38d1
SR
1740 mutex_unlock(&trace_types_lock);
1741
1742 seq_release(inode, file);
1743 kfree(iter);
1744 return 0;
1745}
1746
1747static int tracing_open(struct inode *inode, struct file *file)
1748{
1749 int ret;
1750
1751 __tracing_open(inode, file, &ret);
1752
1753 return ret;
1754}
1755
1756static int tracing_lt_open(struct inode *inode, struct file *file)
1757{
1758 struct trace_iterator *iter;
1759 int ret;
1760
1761 iter = __tracing_open(inode, file, &ret);
1762
1763 if (!ret)
1764 iter->iter_flags |= TRACE_FILE_LAT_FMT;
1765
1766 return ret;
1767}
1768
1769
e309b41d 1770static void *
bc0c38d1
SR
1771t_next(struct seq_file *m, void *v, loff_t *pos)
1772{
1773 struct tracer *t = m->private;
1774
1775 (*pos)++;
1776
1777 if (t)
1778 t = t->next;
1779
1780 m->private = t;
1781
1782 return t;
1783}
1784
1785static void *t_start(struct seq_file *m, loff_t *pos)
1786{
1787 struct tracer *t = m->private;
1788 loff_t l = 0;
1789
1790 mutex_lock(&trace_types_lock);
1791 for (; t && l < *pos; t = t_next(m, t, &l))
1792 ;
1793
1794 return t;
1795}
1796
1797static void t_stop(struct seq_file *m, void *p)
1798{
1799 mutex_unlock(&trace_types_lock);
1800}
1801
1802static int t_show(struct seq_file *m, void *v)
1803{
1804 struct tracer *t = v;
1805
1806 if (!t)
1807 return 0;
1808
1809 seq_printf(m, "%s", t->name);
1810 if (t->next)
1811 seq_putc(m, ' ');
1812 else
1813 seq_putc(m, '\n');
1814
1815 return 0;
1816}
1817
1818static struct seq_operations show_traces_seq_ops = {
4bf39a94
IM
1819 .start = t_start,
1820 .next = t_next,
1821 .stop = t_stop,
1822 .show = t_show,
bc0c38d1
SR
1823};
1824
1825static int show_traces_open(struct inode *inode, struct file *file)
1826{
1827 int ret;
1828
60a11774
SR
1829 if (tracing_disabled)
1830 return -ENODEV;
1831
bc0c38d1
SR
1832 ret = seq_open(file, &show_traces_seq_ops);
1833 if (!ret) {
1834 struct seq_file *m = file->private_data;
1835 m->private = trace_types;
1836 }
1837
1838 return ret;
1839}
1840
1841static struct file_operations tracing_fops = {
4bf39a94
IM
1842 .open = tracing_open,
1843 .read = seq_read,
1844 .llseek = seq_lseek,
1845 .release = tracing_release,
bc0c38d1
SR
1846};
1847
1848static struct file_operations tracing_lt_fops = {
4bf39a94
IM
1849 .open = tracing_lt_open,
1850 .read = seq_read,
1851 .llseek = seq_lseek,
1852 .release = tracing_release,
bc0c38d1
SR
1853};
1854
1855static struct file_operations show_traces_fops = {
c7078de1
IM
1856 .open = show_traces_open,
1857 .read = seq_read,
1858 .release = seq_release,
1859};
1860
36dfe925
IM
1861/*
1862 * Only trace on a CPU if the bitmask is set:
1863 */
9e01c1b7 1864static cpumask_var_t tracing_cpumask;
36dfe925
IM
1865
1866/*
1867 * The tracer itself will not take this lock, but still we want
1868 * to provide a consistent cpumask to user-space:
1869 */
1870static DEFINE_MUTEX(tracing_cpumask_update_lock);
1871
1872/*
1873 * Temporary storage for the character representation of the
1874 * CPU bitmask (and one more byte for the newline):
1875 */
1876static char mask_str[NR_CPUS + 1];
1877
c7078de1
IM
1878static ssize_t
1879tracing_cpumask_read(struct file *filp, char __user *ubuf,
1880 size_t count, loff_t *ppos)
1881{
36dfe925 1882 int len;
c7078de1
IM
1883
1884 mutex_lock(&tracing_cpumask_update_lock);
36dfe925 1885
9e01c1b7 1886 len = cpumask_scnprintf(mask_str, count, tracing_cpumask);
36dfe925
IM
1887 if (count - len < 2) {
1888 count = -EINVAL;
1889 goto out_err;
1890 }
1891 len += sprintf(mask_str + len, "\n");
1892 count = simple_read_from_buffer(ubuf, count, ppos, mask_str, NR_CPUS+1);
1893
1894out_err:
c7078de1
IM
1895 mutex_unlock(&tracing_cpumask_update_lock);
1896
1897 return count;
1898}
1899
1900static ssize_t
1901tracing_cpumask_write(struct file *filp, const char __user *ubuf,
1902 size_t count, loff_t *ppos)
1903{
36dfe925 1904 int err, cpu;
9e01c1b7
RR
1905 cpumask_var_t tracing_cpumask_new;
1906
1907 if (!alloc_cpumask_var(&tracing_cpumask_new, GFP_KERNEL))
1908 return -ENOMEM;
c7078de1
IM
1909
1910 mutex_lock(&tracing_cpumask_update_lock);
9e01c1b7 1911 err = cpumask_parse_user(ubuf, count, tracing_cpumask_new);
c7078de1 1912 if (err)
36dfe925
IM
1913 goto err_unlock;
1914
a5e25883 1915 local_irq_disable();
92205c23 1916 __raw_spin_lock(&ftrace_max_lock);
ab46428c 1917 for_each_tracing_cpu(cpu) {
36dfe925
IM
1918 /*
1919 * Increase/decrease the disabled counter if we are
1920 * about to flip a bit in the cpumask:
1921 */
9e01c1b7
RR
1922 if (cpumask_test_cpu(cpu, tracing_cpumask) &&
1923 !cpumask_test_cpu(cpu, tracing_cpumask_new)) {
36dfe925
IM
1924 atomic_inc(&global_trace.data[cpu]->disabled);
1925 }
9e01c1b7
RR
1926 if (!cpumask_test_cpu(cpu, tracing_cpumask) &&
1927 cpumask_test_cpu(cpu, tracing_cpumask_new)) {
36dfe925
IM
1928 atomic_dec(&global_trace.data[cpu]->disabled);
1929 }
1930 }
92205c23 1931 __raw_spin_unlock(&ftrace_max_lock);
a5e25883 1932 local_irq_enable();
36dfe925 1933
9e01c1b7 1934 cpumask_copy(tracing_cpumask, tracing_cpumask_new);
36dfe925
IM
1935
1936 mutex_unlock(&tracing_cpumask_update_lock);
9e01c1b7 1937 free_cpumask_var(tracing_cpumask_new);
c7078de1
IM
1938
1939 return count;
36dfe925
IM
1940
1941err_unlock:
1942 mutex_unlock(&tracing_cpumask_update_lock);
9e01c1b7 1943 free_cpumask_var(tracing_cpumask);
36dfe925
IM
1944
1945 return err;
c7078de1
IM
1946}
1947
1948static struct file_operations tracing_cpumask_fops = {
1949 .open = tracing_open_generic,
1950 .read = tracing_cpumask_read,
1951 .write = tracing_cpumask_write,
bc0c38d1
SR
1952};
1953
1954static ssize_t
ee6bce52 1955tracing_trace_options_read(struct file *filp, char __user *ubuf,
bc0c38d1
SR
1956 size_t cnt, loff_t *ppos)
1957{
adf9f195 1958 int i;
bc0c38d1
SR
1959 char *buf;
1960 int r = 0;
1961 int len = 0;
adf9f195
FW
1962 u32 tracer_flags = current_trace->flags->val;
1963 struct tracer_opt *trace_opts = current_trace->flags->opts;
1964
bc0c38d1 1965
c3706f00 1966 /* calculate max size */
bc0c38d1
SR
1967 for (i = 0; trace_options[i]; i++) {
1968 len += strlen(trace_options[i]);
1969 len += 3; /* "no" and space */
1970 }
1971
adf9f195
FW
1972 /*
1973 * Increase the size with names of options specific
1974 * of the current tracer.
1975 */
1976 for (i = 0; trace_opts[i].name; i++) {
1977 len += strlen(trace_opts[i].name);
1978 len += 3; /* "no" and space */
1979 }
1980
bc0c38d1
SR
1981 /* +2 for \n and \0 */
1982 buf = kmalloc(len + 2, GFP_KERNEL);
1983 if (!buf)
1984 return -ENOMEM;
1985
1986 for (i = 0; trace_options[i]; i++) {
1987 if (trace_flags & (1 << i))
1988 r += sprintf(buf + r, "%s ", trace_options[i]);
1989 else
1990 r += sprintf(buf + r, "no%s ", trace_options[i]);
1991 }
1992
adf9f195
FW
1993 for (i = 0; trace_opts[i].name; i++) {
1994 if (tracer_flags & trace_opts[i].bit)
1995 r += sprintf(buf + r, "%s ",
1996 trace_opts[i].name);
1997 else
1998 r += sprintf(buf + r, "no%s ",
1999 trace_opts[i].name);
2000 }
2001
bc0c38d1
SR
2002 r += sprintf(buf + r, "\n");
2003 WARN_ON(r >= len + 2);
2004
36dfe925 2005 r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2006
2007 kfree(buf);
2008
2009 return r;
2010}
2011
adf9f195
FW
2012/* Try to assign a tracer specific option */
2013static int set_tracer_option(struct tracer *trace, char *cmp, int neg)
2014{
2015 struct tracer_flags *trace_flags = trace->flags;
2016 struct tracer_opt *opts = NULL;
2017 int ret = 0, i = 0;
2018 int len;
2019
2020 for (i = 0; trace_flags->opts[i].name; i++) {
2021 opts = &trace_flags->opts[i];
2022 len = strlen(opts->name);
2023
2024 if (strncmp(cmp, opts->name, len) == 0) {
2025 ret = trace->set_flag(trace_flags->val,
2026 opts->bit, !neg);
2027 break;
2028 }
2029 }
2030 /* Not found */
2031 if (!trace_flags->opts[i].name)
2032 return -EINVAL;
2033
2034 /* Refused to handle */
2035 if (ret)
2036 return ret;
2037
2038 if (neg)
2039 trace_flags->val &= ~opts->bit;
2040 else
2041 trace_flags->val |= opts->bit;
2042
2043 return 0;
2044}
2045
bc0c38d1 2046static ssize_t
ee6bce52 2047tracing_trace_options_write(struct file *filp, const char __user *ubuf,
bc0c38d1
SR
2048 size_t cnt, loff_t *ppos)
2049{
2050 char buf[64];
2051 char *cmp = buf;
2052 int neg = 0;
adf9f195 2053 int ret;
bc0c38d1
SR
2054 int i;
2055
cffae437
SR
2056 if (cnt >= sizeof(buf))
2057 return -EINVAL;
bc0c38d1
SR
2058
2059 if (copy_from_user(&buf, ubuf, cnt))
2060 return -EFAULT;
2061
2062 buf[cnt] = 0;
2063
2064 if (strncmp(buf, "no", 2) == 0) {
2065 neg = 1;
2066 cmp += 2;
2067 }
2068
2069 for (i = 0; trace_options[i]; i++) {
2070 int len = strlen(trace_options[i]);
2071
2072 if (strncmp(cmp, trace_options[i], len) == 0) {
2073 if (neg)
2074 trace_flags &= ~(1 << i);
2075 else
2076 trace_flags |= (1 << i);
2077 break;
2078 }
2079 }
adf9f195
FW
2080
2081 /* If no option could be set, test the specific tracer options */
2082 if (!trace_options[i]) {
2083 ret = set_tracer_option(current_trace, cmp, neg);
2084 if (ret)
2085 return ret;
2086 }
bc0c38d1
SR
2087
2088 filp->f_pos += cnt;
2089
2090 return cnt;
2091}
2092
2093static struct file_operations tracing_iter_fops = {
c7078de1 2094 .open = tracing_open_generic,
ee6bce52
SR
2095 .read = tracing_trace_options_read,
2096 .write = tracing_trace_options_write,
bc0c38d1
SR
2097};
2098
7bd2f24c
IM
2099static const char readme_msg[] =
2100 "tracing mini-HOWTO:\n\n"
2101 "# mkdir /debug\n"
2102 "# mount -t debugfs nodev /debug\n\n"
2103 "# cat /debug/tracing/available_tracers\n"
2104 "wakeup preemptirqsoff preemptoff irqsoff ftrace sched_switch none\n\n"
2105 "# cat /debug/tracing/current_tracer\n"
2106 "none\n"
2107 "# echo sched_switch > /debug/tracing/current_tracer\n"
2108 "# cat /debug/tracing/current_tracer\n"
2109 "sched_switch\n"
ee6bce52 2110 "# cat /debug/tracing/trace_options\n"
7bd2f24c 2111 "noprint-parent nosym-offset nosym-addr noverbose\n"
ee6bce52 2112 "# echo print-parent > /debug/tracing/trace_options\n"
7bd2f24c
IM
2113 "# echo 1 > /debug/tracing/tracing_enabled\n"
2114 "# cat /debug/tracing/trace > /tmp/trace.txt\n"
2115 "echo 0 > /debug/tracing/tracing_enabled\n"
2116;
2117
2118static ssize_t
2119tracing_readme_read(struct file *filp, char __user *ubuf,
2120 size_t cnt, loff_t *ppos)
2121{
2122 return simple_read_from_buffer(ubuf, cnt, ppos,
2123 readme_msg, strlen(readme_msg));
2124}
2125
2126static struct file_operations tracing_readme_fops = {
c7078de1
IM
2127 .open = tracing_open_generic,
2128 .read = tracing_readme_read,
7bd2f24c
IM
2129};
2130
bc0c38d1
SR
2131static ssize_t
2132tracing_ctrl_read(struct file *filp, char __user *ubuf,
2133 size_t cnt, loff_t *ppos)
2134{
bc0c38d1
SR
2135 char buf[64];
2136 int r;
2137
9036990d 2138 r = sprintf(buf, "%u\n", tracer_enabled);
4e3c3333 2139 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2140}
2141
2142static ssize_t
2143tracing_ctrl_write(struct file *filp, const char __user *ubuf,
2144 size_t cnt, loff_t *ppos)
2145{
2146 struct trace_array *tr = filp->private_data;
bc0c38d1 2147 char buf[64];
c6caeeb1
SR
2148 long val;
2149 int ret;
bc0c38d1 2150
cffae437
SR
2151 if (cnt >= sizeof(buf))
2152 return -EINVAL;
bc0c38d1
SR
2153
2154 if (copy_from_user(&buf, ubuf, cnt))
2155 return -EFAULT;
2156
2157 buf[cnt] = 0;
2158
c6caeeb1
SR
2159 ret = strict_strtoul(buf, 10, &val);
2160 if (ret < 0)
2161 return ret;
bc0c38d1
SR
2162
2163 val = !!val;
2164
2165 mutex_lock(&trace_types_lock);
9036990d
SR
2166 if (tracer_enabled ^ val) {
2167 if (val) {
bc0c38d1 2168 tracer_enabled = 1;
9036990d
SR
2169 if (current_trace->start)
2170 current_trace->start(tr);
2171 tracing_start();
2172 } else {
bc0c38d1 2173 tracer_enabled = 0;
9036990d
SR
2174 tracing_stop();
2175 if (current_trace->stop)
2176 current_trace->stop(tr);
2177 }
bc0c38d1
SR
2178 }
2179 mutex_unlock(&trace_types_lock);
2180
2181 filp->f_pos += cnt;
2182
2183 return cnt;
2184}
2185
2186static ssize_t
2187tracing_set_trace_read(struct file *filp, char __user *ubuf,
2188 size_t cnt, loff_t *ppos)
2189{
2190 char buf[max_tracer_type_len+2];
2191 int r;
2192
2193 mutex_lock(&trace_types_lock);
2194 if (current_trace)
2195 r = sprintf(buf, "%s\n", current_trace->name);
2196 else
2197 r = sprintf(buf, "\n");
2198 mutex_unlock(&trace_types_lock);
2199
4bf39a94 2200 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2201}
2202
b6f11df2
ACM
2203int tracer_init(struct tracer *t, struct trace_array *tr)
2204{
2205 tracing_reset_online_cpus(tr);
2206 return t->init(tr);
2207}
2208
b2821ae6 2209static int tracing_set_tracer(const char *buf)
bc0c38d1
SR
2210{
2211 struct trace_array *tr = &global_trace;
2212 struct tracer *t;
d9e54076 2213 int ret = 0;
bc0c38d1
SR
2214
2215 mutex_lock(&trace_types_lock);
2216 for (t = trace_types; t; t = t->next) {
2217 if (strcmp(t->name, buf) == 0)
2218 break;
2219 }
c2931e05
FW
2220 if (!t) {
2221 ret = -EINVAL;
2222 goto out;
2223 }
2224 if (t == current_trace)
bc0c38d1
SR
2225 goto out;
2226
9f029e83 2227 trace_branch_disable();
bc0c38d1
SR
2228 if (current_trace && current_trace->reset)
2229 current_trace->reset(tr);
2230
2231 current_trace = t;
1c80025a 2232 if (t->init) {
b6f11df2 2233 ret = tracer_init(t, tr);
1c80025a
FW
2234 if (ret)
2235 goto out;
2236 }
bc0c38d1 2237
9f029e83 2238 trace_branch_enable(tr);
bc0c38d1
SR
2239 out:
2240 mutex_unlock(&trace_types_lock);
2241
d9e54076
PZ
2242 return ret;
2243}
2244
2245static ssize_t
2246tracing_set_trace_write(struct file *filp, const char __user *ubuf,
2247 size_t cnt, loff_t *ppos)
2248{
2249 char buf[max_tracer_type_len+1];
2250 int i;
2251 size_t ret;
e6e7a65a
FW
2252 int err;
2253
2254 ret = cnt;
d9e54076
PZ
2255
2256 if (cnt > max_tracer_type_len)
2257 cnt = max_tracer_type_len;
2258
2259 if (copy_from_user(&buf, ubuf, cnt))
2260 return -EFAULT;
2261
2262 buf[cnt] = 0;
2263
2264 /* strip ending whitespace. */
2265 for (i = cnt - 1; i > 0 && isspace(buf[i]); i--)
2266 buf[i] = 0;
2267
e6e7a65a
FW
2268 err = tracing_set_tracer(buf);
2269 if (err)
2270 return err;
d9e54076 2271
e6e7a65a 2272 filp->f_pos += ret;
bc0c38d1 2273
c2931e05 2274 return ret;
bc0c38d1
SR
2275}
2276
2277static ssize_t
2278tracing_max_lat_read(struct file *filp, char __user *ubuf,
2279 size_t cnt, loff_t *ppos)
2280{
2281 unsigned long *ptr = filp->private_data;
2282 char buf[64];
2283 int r;
2284
cffae437 2285 r = snprintf(buf, sizeof(buf), "%ld\n",
bc0c38d1 2286 *ptr == (unsigned long)-1 ? -1 : nsecs_to_usecs(*ptr));
cffae437
SR
2287 if (r > sizeof(buf))
2288 r = sizeof(buf);
4bf39a94 2289 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
bc0c38d1
SR
2290}
2291
2292static ssize_t
2293tracing_max_lat_write(struct file *filp, const char __user *ubuf,
2294 size_t cnt, loff_t *ppos)
2295{
2296 long *ptr = filp->private_data;
bc0c38d1 2297 char buf[64];
c6caeeb1
SR
2298 long val;
2299 int ret;
bc0c38d1 2300
cffae437
SR
2301 if (cnt >= sizeof(buf))
2302 return -EINVAL;
bc0c38d1
SR
2303
2304 if (copy_from_user(&buf, ubuf, cnt))
2305 return -EFAULT;
2306
2307 buf[cnt] = 0;
2308
c6caeeb1
SR
2309 ret = strict_strtoul(buf, 10, &val);
2310 if (ret < 0)
2311 return ret;
bc0c38d1
SR
2312
2313 *ptr = val * 1000;
2314
2315 return cnt;
2316}
2317
b3806b43
SR
2318static atomic_t tracing_reader;
2319
2320static int tracing_open_pipe(struct inode *inode, struct file *filp)
2321{
2322 struct trace_iterator *iter;
2323
2324 if (tracing_disabled)
2325 return -ENODEV;
2326
2327 /* We only allow for reader of the pipe */
2328 if (atomic_inc_return(&tracing_reader) != 1) {
2329 atomic_dec(&tracing_reader);
2330 return -EBUSY;
2331 }
2332
2333 /* create a buffer to store the information to pass to userspace */
2334 iter = kzalloc(sizeof(*iter), GFP_KERNEL);
2335 if (!iter)
2336 return -ENOMEM;
2337
4462344e
RR
2338 if (!alloc_cpumask_var(&iter->started, GFP_KERNEL)) {
2339 kfree(iter);
2340 return -ENOMEM;
2341 }
2342
107bad8b 2343 mutex_lock(&trace_types_lock);
a309720c
SR
2344
2345 /* trace pipe does not show start of buffer */
4462344e 2346 cpumask_setall(iter->started);
a309720c 2347
b3806b43 2348 iter->tr = &global_trace;
72829bc3 2349 iter->trace = current_trace;
b3806b43
SR
2350 filp->private_data = iter;
2351
107bad8b
SR
2352 if (iter->trace->pipe_open)
2353 iter->trace->pipe_open(iter);
2354 mutex_unlock(&trace_types_lock);
2355
b3806b43
SR
2356 return 0;
2357}
2358
2359static int tracing_release_pipe(struct inode *inode, struct file *file)
2360{
2361 struct trace_iterator *iter = file->private_data;
2362
4462344e 2363 free_cpumask_var(iter->started);
b3806b43
SR
2364 kfree(iter);
2365 atomic_dec(&tracing_reader);
2366
2367 return 0;
2368}
2369
2a2cc8f7
SSP
2370static unsigned int
2371tracing_poll_pipe(struct file *filp, poll_table *poll_table)
2372{
2373 struct trace_iterator *iter = filp->private_data;
2374
2375 if (trace_flags & TRACE_ITER_BLOCK) {
2376 /*
2377 * Always select as readable when in blocking mode
2378 */
2379 return POLLIN | POLLRDNORM;
afc2abc0 2380 } else {
2a2cc8f7
SSP
2381 if (!trace_empty(iter))
2382 return POLLIN | POLLRDNORM;
2383 poll_wait(filp, &trace_wait, poll_table);
2384 if (!trace_empty(iter))
2385 return POLLIN | POLLRDNORM;
2386
2387 return 0;
2388 }
2389}
2390
ff98781b
EGM
2391/* Must be called with trace_types_lock mutex held. */
2392static int tracing_wait_pipe(struct file *filp)
b3806b43
SR
2393{
2394 struct trace_iterator *iter = filp->private_data;
b3806b43 2395
b3806b43 2396 while (trace_empty(iter)) {
2dc8f095 2397
107bad8b 2398 if ((filp->f_flags & O_NONBLOCK)) {
ff98781b 2399 return -EAGAIN;
107bad8b 2400 }
2dc8f095 2401
b3806b43
SR
2402 /*
2403 * This is a make-shift waitqueue. The reason we don't use
2404 * an actual wait queue is because:
2405 * 1) we only ever have one waiter
2406 * 2) the tracing, traces all functions, we don't want
2407 * the overhead of calling wake_up and friends
2408 * (and tracing them too)
2409 * Anyway, this is really very primitive wakeup.
2410 */
2411 set_current_state(TASK_INTERRUPTIBLE);
2412 iter->tr->waiter = current;
2413
107bad8b
SR
2414 mutex_unlock(&trace_types_lock);
2415
9fe068e9
IM
2416 /* sleep for 100 msecs, and try again. */
2417 schedule_timeout(HZ/10);
b3806b43 2418
107bad8b
SR
2419 mutex_lock(&trace_types_lock);
2420
b3806b43
SR
2421 iter->tr->waiter = NULL;
2422
107bad8b 2423 if (signal_pending(current)) {
ff98781b 2424 return -EINTR;
107bad8b 2425 }
b3806b43 2426
84527997 2427 if (iter->trace != current_trace)
ff98781b 2428 return 0;
84527997 2429
b3806b43
SR
2430 /*
2431 * We block until we read something and tracing is disabled.
2432 * We still block if tracing is disabled, but we have never
2433 * read anything. This allows a user to cat this file, and
2434 * then enable tracing. But after we have read something,
2435 * we give an EOF when tracing is again disabled.
2436 *
2437 * iter->pos will be 0 if we haven't read anything.
2438 */
2439 if (!tracer_enabled && iter->pos)
2440 break;
2441
2442 continue;
2443 }
2444
ff98781b
EGM
2445 return 1;
2446}
2447
2448/*
2449 * Consumer reader.
2450 */
2451static ssize_t
2452tracing_read_pipe(struct file *filp, char __user *ubuf,
2453 size_t cnt, loff_t *ppos)
2454{
2455 struct trace_iterator *iter = filp->private_data;
2456 ssize_t sret;
2457
2458 /* return any leftover data */
2459 sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
2460 if (sret != -EBUSY)
2461 return sret;
2462
2463 trace_seq_reset(&iter->seq);
2464
2465 mutex_lock(&trace_types_lock);
2466 if (iter->trace->read) {
2467 sret = iter->trace->read(iter, filp, ubuf, cnt, ppos);
2468 if (sret)
2469 goto out;
2470 }
2471
2472waitagain:
2473 sret = tracing_wait_pipe(filp);
2474 if (sret <= 0)
2475 goto out;
2476
b3806b43 2477 /* stop when tracing is finished */
ff98781b
EGM
2478 if (trace_empty(iter)) {
2479 sret = 0;
107bad8b 2480 goto out;
ff98781b 2481 }
b3806b43
SR
2482
2483 if (cnt >= PAGE_SIZE)
2484 cnt = PAGE_SIZE - 1;
2485
53d0aa77 2486 /* reset all but tr, trace, and overruns */
53d0aa77
SR
2487 memset(&iter->seq, 0,
2488 sizeof(struct trace_iterator) -
2489 offsetof(struct trace_iterator, seq));
4823ed7e 2490 iter->pos = -1;
b3806b43 2491
088b1e42 2492 while (find_next_entry_inc(iter) != NULL) {
2c4f035f 2493 enum print_line_t ret;
088b1e42
SR
2494 int len = iter->seq.len;
2495
f9896bf3 2496 ret = print_trace_line(iter);
2c4f035f 2497 if (ret == TRACE_TYPE_PARTIAL_LINE) {
088b1e42
SR
2498 /* don't print partial lines */
2499 iter->seq.len = len;
b3806b43 2500 break;
088b1e42 2501 }
b91facc3
FW
2502 if (ret != TRACE_TYPE_NO_CONSUME)
2503 trace_consume(iter);
b3806b43
SR
2504
2505 if (iter->seq.len >= cnt)
2506 break;
b3806b43
SR
2507 }
2508
b3806b43 2509 /* Now copy what we have to the user */
6c6c2796
PP
2510 sret = trace_seq_to_user(&iter->seq, ubuf, cnt);
2511 if (iter->seq.readpos >= iter->seq.len)
b3806b43 2512 trace_seq_reset(&iter->seq);
9ff4b974
PP
2513
2514 /*
2515 * If there was nothing to send to user, inspite of consuming trace
2516 * entries, go back to wait for more entries.
2517 */
6c6c2796 2518 if (sret == -EBUSY)
9ff4b974 2519 goto waitagain;
b3806b43 2520
107bad8b
SR
2521out:
2522 mutex_unlock(&trace_types_lock);
2523
6c6c2796 2524 return sret;
b3806b43
SR
2525}
2526
3c56819b
EGM
2527static void tracing_pipe_buf_release(struct pipe_inode_info *pipe,
2528 struct pipe_buffer *buf)
2529{
2530 __free_page(buf->page);
2531}
2532
2533static void tracing_spd_release_pipe(struct splice_pipe_desc *spd,
2534 unsigned int idx)
2535{
2536 __free_page(spd->pages[idx]);
2537}
2538
2539static struct pipe_buf_operations tracing_pipe_buf_ops = {
34cd4998
SR
2540 .can_merge = 0,
2541 .map = generic_pipe_buf_map,
2542 .unmap = generic_pipe_buf_unmap,
2543 .confirm = generic_pipe_buf_confirm,
2544 .release = tracing_pipe_buf_release,
2545 .steal = generic_pipe_buf_steal,
2546 .get = generic_pipe_buf_get,
3c56819b
EGM
2547};
2548
34cd4998
SR
2549static size_t
2550tracing_fill_pipe_page(struct page *pages, size_t rem,
2551 struct trace_iterator *iter)
2552{
2553 size_t count;
2554 int ret;
2555
2556 /* Seq buffer is page-sized, exactly what we need. */
2557 for (;;) {
2558 count = iter->seq.len;
2559 ret = print_trace_line(iter);
2560 count = iter->seq.len - count;
2561 if (rem < count) {
2562 rem = 0;
2563 iter->seq.len -= count;
2564 break;
2565 }
2566 if (ret == TRACE_TYPE_PARTIAL_LINE) {
2567 iter->seq.len -= count;
2568 break;
2569 }
2570
2571 trace_consume(iter);
2572 rem -= count;
2573 if (!find_next_entry_inc(iter)) {
2574 rem = 0;
2575 iter->ent = NULL;
2576 break;
2577 }
2578 }
2579
2580 return rem;
2581}
2582
3c56819b
EGM
2583static ssize_t tracing_splice_read_pipe(struct file *filp,
2584 loff_t *ppos,
2585 struct pipe_inode_info *pipe,
2586 size_t len,
2587 unsigned int flags)
2588{
2589 struct page *pages[PIPE_BUFFERS];
2590 struct partial_page partial[PIPE_BUFFERS];
2591 struct trace_iterator *iter = filp->private_data;
2592 struct splice_pipe_desc spd = {
34cd4998
SR
2593 .pages = pages,
2594 .partial = partial,
2595 .nr_pages = 0, /* This gets updated below. */
2596 .flags = flags,
2597 .ops = &tracing_pipe_buf_ops,
2598 .spd_release = tracing_spd_release_pipe,
3c56819b
EGM
2599 };
2600 ssize_t ret;
34cd4998 2601 size_t rem;
3c56819b
EGM
2602 unsigned int i;
2603
2604 mutex_lock(&trace_types_lock);
2605
2606 if (iter->trace->splice_read) {
2607 ret = iter->trace->splice_read(iter, filp,
2608 ppos, pipe, len, flags);
2609 if (ret)
34cd4998 2610 goto out_err;
3c56819b
EGM
2611 }
2612
2613 ret = tracing_wait_pipe(filp);
2614 if (ret <= 0)
34cd4998 2615 goto out_err;
3c56819b
EGM
2616
2617 if (!iter->ent && !find_next_entry_inc(iter)) {
2618 ret = -EFAULT;
34cd4998 2619 goto out_err;
3c56819b
EGM
2620 }
2621
2622 /* Fill as many pages as possible. */
2623 for (i = 0, rem = len; i < PIPE_BUFFERS && rem; i++) {
2624 pages[i] = alloc_page(GFP_KERNEL);
34cd4998
SR
2625 if (!pages[i])
2626 break;
3c56819b 2627
34cd4998 2628 rem = tracing_fill_pipe_page(pages[i], rem, iter);
3c56819b
EGM
2629
2630 /* Copy the data into the page, so we can start over. */
2631 ret = trace_seq_to_buffer(&iter->seq,
2632 page_address(pages[i]),
2633 iter->seq.len);
2634 if (ret < 0) {
2635 __free_page(pages[i]);
2636 break;
2637 }
2638 partial[i].offset = 0;
2639 partial[i].len = iter->seq.len;
2640
2641 trace_seq_reset(&iter->seq);
2642 }
2643
2644 mutex_unlock(&trace_types_lock);
2645
2646 spd.nr_pages = i;
2647
2648 return splice_to_pipe(pipe, &spd);
2649
34cd4998 2650out_err:
3c56819b
EGM
2651 mutex_unlock(&trace_types_lock);
2652
2653 return ret;
2654}
2655
a98a3c3f
SR
2656static ssize_t
2657tracing_entries_read(struct file *filp, char __user *ubuf,
2658 size_t cnt, loff_t *ppos)
2659{
2660 struct trace_array *tr = filp->private_data;
2661 char buf[64];
2662 int r;
2663
1696b2b0 2664 r = sprintf(buf, "%lu\n", tr->entries >> 10);
a98a3c3f
SR
2665 return simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
2666}
2667
2668static ssize_t
2669tracing_entries_write(struct file *filp, const char __user *ubuf,
2670 size_t cnt, loff_t *ppos)
2671{
2672 unsigned long val;
2673 char buf[64];
bf5e6519 2674 int ret, cpu;
a98a3c3f 2675
cffae437
SR
2676 if (cnt >= sizeof(buf))
2677 return -EINVAL;
a98a3c3f
SR
2678
2679 if (copy_from_user(&buf, ubuf, cnt))
2680 return -EFAULT;
2681
2682 buf[cnt] = 0;
2683
c6caeeb1
SR
2684 ret = strict_strtoul(buf, 10, &val);
2685 if (ret < 0)
2686 return ret;
a98a3c3f
SR
2687
2688 /* must have at least 1 entry */
2689 if (!val)
2690 return -EINVAL;
2691
2692 mutex_lock(&trace_types_lock);
2693
c76f0694 2694 tracing_stop();
a98a3c3f 2695
bf5e6519
SR
2696 /* disable all cpu buffers */
2697 for_each_tracing_cpu(cpu) {
2698 if (global_trace.data[cpu])
2699 atomic_inc(&global_trace.data[cpu]->disabled);
2700 if (max_tr.data[cpu])
2701 atomic_inc(&max_tr.data[cpu]->disabled);
2702 }
2703
1696b2b0
SR
2704 /* value is in KB */
2705 val <<= 10;
2706
3928a8a2
SR
2707 if (val != global_trace.entries) {
2708 ret = ring_buffer_resize(global_trace.buffer, val);
2709 if (ret < 0) {
2710 cnt = ret;
3eefae99
SR
2711 goto out;
2712 }
2713
3928a8a2
SR
2714 ret = ring_buffer_resize(max_tr.buffer, val);
2715 if (ret < 0) {
2716 int r;
2717 cnt = ret;
2718 r = ring_buffer_resize(global_trace.buffer,
2719 global_trace.entries);
2720 if (r < 0) {
2721 /* AARGH! We are left with different
2722 * size max buffer!!!! */
2723 WARN_ON(1);
2724 tracing_disabled = 1;
a98a3c3f 2725 }
3928a8a2 2726 goto out;
a98a3c3f 2727 }
3eefae99 2728
3928a8a2 2729 global_trace.entries = val;
a98a3c3f
SR
2730 }
2731
2732 filp->f_pos += cnt;
2733
19384c03
SR
2734 /* If check pages failed, return ENOMEM */
2735 if (tracing_disabled)
2736 cnt = -ENOMEM;
a98a3c3f 2737 out:
bf5e6519
SR
2738 for_each_tracing_cpu(cpu) {
2739 if (global_trace.data[cpu])
2740 atomic_dec(&global_trace.data[cpu]->disabled);
2741 if (max_tr.data[cpu])
2742 atomic_dec(&max_tr.data[cpu]->disabled);
2743 }
2744
c76f0694 2745 tracing_start();
a98a3c3f
SR
2746 max_tr.entries = global_trace.entries;
2747 mutex_unlock(&trace_types_lock);
2748
2749 return cnt;
2750}
2751
5bf9a1ee
PP
2752static int mark_printk(const char *fmt, ...)
2753{
2754 int ret;
2755 va_list args;
2756 va_start(args, fmt);
1fd8f2a3 2757 ret = trace_vprintk(0, -1, fmt, args);
5bf9a1ee
PP
2758 va_end(args);
2759 return ret;
2760}
2761
2762static ssize_t
2763tracing_mark_write(struct file *filp, const char __user *ubuf,
2764 size_t cnt, loff_t *fpos)
2765{
2766 char *buf;
2767 char *end;
5bf9a1ee 2768
c76f0694 2769 if (tracing_disabled)
5bf9a1ee
PP
2770 return -EINVAL;
2771
2772 if (cnt > TRACE_BUF_SIZE)
2773 cnt = TRACE_BUF_SIZE;
2774
2775 buf = kmalloc(cnt + 1, GFP_KERNEL);
2776 if (buf == NULL)
2777 return -ENOMEM;
2778
2779 if (copy_from_user(buf, ubuf, cnt)) {
2780 kfree(buf);
2781 return -EFAULT;
2782 }
2783
2784 /* Cut from the first nil or newline. */
2785 buf[cnt] = '\0';
2786 end = strchr(buf, '\n');
2787 if (end)
2788 *end = '\0';
2789
2790 cnt = mark_printk("%s\n", buf);
2791 kfree(buf);
2792 *fpos += cnt;
2793
2794 return cnt;
2795}
2796
bc0c38d1 2797static struct file_operations tracing_max_lat_fops = {
4bf39a94
IM
2798 .open = tracing_open_generic,
2799 .read = tracing_max_lat_read,
2800 .write = tracing_max_lat_write,
bc0c38d1
SR
2801};
2802
2803static struct file_operations tracing_ctrl_fops = {
4bf39a94
IM
2804 .open = tracing_open_generic,
2805 .read = tracing_ctrl_read,
2806 .write = tracing_ctrl_write,
bc0c38d1
SR
2807};
2808
2809static struct file_operations set_tracer_fops = {
4bf39a94
IM
2810 .open = tracing_open_generic,
2811 .read = tracing_set_trace_read,
2812 .write = tracing_set_trace_write,
bc0c38d1
SR
2813};
2814
b3806b43 2815static struct file_operations tracing_pipe_fops = {
4bf39a94 2816 .open = tracing_open_pipe,
2a2cc8f7 2817 .poll = tracing_poll_pipe,
4bf39a94 2818 .read = tracing_read_pipe,
3c56819b 2819 .splice_read = tracing_splice_read_pipe,
4bf39a94 2820 .release = tracing_release_pipe,
b3806b43
SR
2821};
2822
a98a3c3f
SR
2823static struct file_operations tracing_entries_fops = {
2824 .open = tracing_open_generic,
2825 .read = tracing_entries_read,
2826 .write = tracing_entries_write,
2827};
2828
5bf9a1ee 2829static struct file_operations tracing_mark_fops = {
43a15386 2830 .open = tracing_open_generic,
5bf9a1ee
PP
2831 .write = tracing_mark_write,
2832};
2833
bc0c38d1
SR
2834#ifdef CONFIG_DYNAMIC_FTRACE
2835
b807c3d0
SR
2836int __weak ftrace_arch_read_dyn_info(char *buf, int size)
2837{
2838 return 0;
2839}
2840
bc0c38d1 2841static ssize_t
b807c3d0 2842tracing_read_dyn_info(struct file *filp, char __user *ubuf,
bc0c38d1
SR
2843 size_t cnt, loff_t *ppos)
2844{
a26a2a27
SR
2845 static char ftrace_dyn_info_buffer[1024];
2846 static DEFINE_MUTEX(dyn_info_mutex);
bc0c38d1 2847 unsigned long *p = filp->private_data;
b807c3d0 2848 char *buf = ftrace_dyn_info_buffer;
a26a2a27 2849 int size = ARRAY_SIZE(ftrace_dyn_info_buffer);
bc0c38d1
SR
2850 int r;
2851
b807c3d0
SR
2852 mutex_lock(&dyn_info_mutex);
2853 r = sprintf(buf, "%ld ", *p);
4bf39a94 2854
a26a2a27 2855 r += ftrace_arch_read_dyn_info(buf+r, (size-1)-r);
b807c3d0
SR
2856 buf[r++] = '\n';
2857
2858 r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r);
2859
2860 mutex_unlock(&dyn_info_mutex);
2861
2862 return r;
bc0c38d1
SR
2863}
2864
b807c3d0 2865static struct file_operations tracing_dyn_info_fops = {
4bf39a94 2866 .open = tracing_open_generic,
b807c3d0 2867 .read = tracing_read_dyn_info,
bc0c38d1
SR
2868};
2869#endif
2870
2871static struct dentry *d_tracer;
2872
2873struct dentry *tracing_init_dentry(void)
2874{
2875 static int once;
2876
2877 if (d_tracer)
2878 return d_tracer;
2879
2880 d_tracer = debugfs_create_dir("tracing", NULL);
2881
2882 if (!d_tracer && !once) {
2883 once = 1;
2884 pr_warning("Could not create debugfs directory 'tracing'\n");
2885 return NULL;
2886 }
2887
2888 return d_tracer;
2889}
2890
60a11774
SR
2891#ifdef CONFIG_FTRACE_SELFTEST
2892/* Let selftest have access to static functions in this file */
2893#include "trace_selftest.c"
2894#endif
2895
b5ad384e 2896static __init int tracer_init_debugfs(void)
bc0c38d1
SR
2897{
2898 struct dentry *d_tracer;
2899 struct dentry *entry;
2900
2901 d_tracer = tracing_init_dentry();
2902
2903 entry = debugfs_create_file("tracing_enabled", 0644, d_tracer,
2904 &global_trace, &tracing_ctrl_fops);
2905 if (!entry)
2906 pr_warning("Could not create debugfs 'tracing_enabled' entry\n");
2907
ee6bce52 2908 entry = debugfs_create_file("trace_options", 0644, d_tracer,
bc0c38d1
SR
2909 NULL, &tracing_iter_fops);
2910 if (!entry)
ee6bce52 2911 pr_warning("Could not create debugfs 'trace_options' entry\n");
bc0c38d1 2912
c7078de1
IM
2913 entry = debugfs_create_file("tracing_cpumask", 0644, d_tracer,
2914 NULL, &tracing_cpumask_fops);
2915 if (!entry)
2916 pr_warning("Could not create debugfs 'tracing_cpumask' entry\n");
2917
bc0c38d1
SR
2918 entry = debugfs_create_file("latency_trace", 0444, d_tracer,
2919 &global_trace, &tracing_lt_fops);
2920 if (!entry)
2921 pr_warning("Could not create debugfs 'latency_trace' entry\n");
2922
2923 entry = debugfs_create_file("trace", 0444, d_tracer,
2924 &global_trace, &tracing_fops);
2925 if (!entry)
2926 pr_warning("Could not create debugfs 'trace' entry\n");
2927
2928 entry = debugfs_create_file("available_tracers", 0444, d_tracer,
2929 &global_trace, &show_traces_fops);
2930 if (!entry)
98a983aa 2931 pr_warning("Could not create debugfs 'available_tracers' entry\n");
bc0c38d1
SR
2932
2933 entry = debugfs_create_file("current_tracer", 0444, d_tracer,
2934 &global_trace, &set_tracer_fops);
2935 if (!entry)
98a983aa 2936 pr_warning("Could not create debugfs 'current_tracer' entry\n");
bc0c38d1
SR
2937
2938 entry = debugfs_create_file("tracing_max_latency", 0644, d_tracer,
2939 &tracing_max_latency,
2940 &tracing_max_lat_fops);
2941 if (!entry)
2942 pr_warning("Could not create debugfs "
2943 "'tracing_max_latency' entry\n");
2944
2945 entry = debugfs_create_file("tracing_thresh", 0644, d_tracer,
2946 &tracing_thresh, &tracing_max_lat_fops);
2947 if (!entry)
2948 pr_warning("Could not create debugfs "
98a983aa 2949 "'tracing_thresh' entry\n");
7bd2f24c
IM
2950 entry = debugfs_create_file("README", 0644, d_tracer,
2951 NULL, &tracing_readme_fops);
2952 if (!entry)
2953 pr_warning("Could not create debugfs 'README' entry\n");
2954
b3806b43
SR
2955 entry = debugfs_create_file("trace_pipe", 0644, d_tracer,
2956 NULL, &tracing_pipe_fops);
2957 if (!entry)
2958 pr_warning("Could not create debugfs "
98a983aa 2959 "'trace_pipe' entry\n");
bc0c38d1 2960
a94c80e7 2961 entry = debugfs_create_file("buffer_size_kb", 0644, d_tracer,
a98a3c3f
SR
2962 &global_trace, &tracing_entries_fops);
2963 if (!entry)
2964 pr_warning("Could not create debugfs "
a94c80e7 2965 "'buffer_size_kb' entry\n");
a98a3c3f 2966
5bf9a1ee
PP
2967 entry = debugfs_create_file("trace_marker", 0220, d_tracer,
2968 NULL, &tracing_mark_fops);
2969 if (!entry)
2970 pr_warning("Could not create debugfs "
2971 "'trace_marker' entry\n");
2972
bc0c38d1
SR
2973#ifdef CONFIG_DYNAMIC_FTRACE
2974 entry = debugfs_create_file("dyn_ftrace_total_info", 0444, d_tracer,
2975 &ftrace_update_tot_cnt,
b807c3d0 2976 &tracing_dyn_info_fops);
bc0c38d1
SR
2977 if (!entry)
2978 pr_warning("Could not create debugfs "
2979 "'dyn_ftrace_total_info' entry\n");
2980#endif
d618b3e6
IM
2981#ifdef CONFIG_SYSPROF_TRACER
2982 init_tracer_sysprof_debugfs(d_tracer);
2983#endif
b5ad384e 2984 return 0;
bc0c38d1
SR
2985}
2986
1fd8f2a3 2987int trace_vprintk(unsigned long ip, int depth, const char *fmt, va_list args)
dd0e545f 2988{
dd0e545f
SR
2989 static DEFINE_SPINLOCK(trace_buf_lock);
2990 static char trace_buf[TRACE_BUF_SIZE];
f09ce573 2991
3928a8a2 2992 struct ring_buffer_event *event;
f09ce573 2993 struct trace_array *tr = &global_trace;
dd0e545f 2994 struct trace_array_cpu *data;
38697053 2995 int cpu, len = 0, size, pc;
e726f5f9
IM
2996 struct print_entry *entry;
2997 unsigned long irq_flags;
dd0e545f 2998
8e1b82e0 2999 if (tracing_disabled || tracing_selftest_running)
dd0e545f
SR
3000 return 0;
3001
38697053
SR
3002 pc = preempt_count();
3003 preempt_disable_notrace();
dd0e545f
SR
3004 cpu = raw_smp_processor_id();
3005 data = tr->data[cpu];
dd0e545f 3006
3ea2e6d7 3007 if (unlikely(atomic_read(&data->disabled)))
dd0e545f
SR
3008 goto out;
3009
380c4b14
FW
3010 pause_graph_tracing();
3011 spin_lock_irqsave(&trace_buf_lock, irq_flags);
801fe400 3012 len = vsnprintf(trace_buf, TRACE_BUF_SIZE, fmt, args);
dd0e545f
SR
3013
3014 len = min(len, TRACE_BUF_SIZE-1);
3015 trace_buf[len] = 0;
3016
777e208d 3017 size = sizeof(*entry) + len + 1;
51a763dd 3018 event = trace_buffer_lock_reserve(tr, TRACE_PRINT, size, irq_flags, pc);
3928a8a2
SR
3019 if (!event)
3020 goto out_unlock;
777e208d 3021 entry = ring_buffer_event_data(event);
777e208d 3022 entry->ip = ip;
1fd8f2a3 3023 entry->depth = depth;
dd0e545f 3024
777e208d
SR
3025 memcpy(&entry->buf, trace_buf, len);
3026 entry->buf[len] = 0;
0a987751 3027 ring_buffer_unlock_commit(tr->buffer, event);
dd0e545f 3028
3928a8a2 3029 out_unlock:
380c4b14
FW
3030 spin_unlock_irqrestore(&trace_buf_lock, irq_flags);
3031 unpause_graph_tracing();
dd0e545f 3032 out:
38697053 3033 preempt_enable_notrace();
dd0e545f
SR
3034
3035 return len;
3036}
801fe400
PP
3037EXPORT_SYMBOL_GPL(trace_vprintk);
3038
3039int __ftrace_printk(unsigned long ip, const char *fmt, ...)
3040{
3041 int ret;
3042 va_list ap;
3043
3044 if (!(trace_flags & TRACE_ITER_PRINTK))
3045 return 0;
3046
3047 va_start(ap, fmt);
21a8c466 3048 ret = trace_vprintk(ip, task_curr_ret_stack(current), fmt, ap);
801fe400
PP
3049 va_end(ap);
3050 return ret;
3051}
dd0e545f
SR
3052EXPORT_SYMBOL_GPL(__ftrace_printk);
3053
9011262a
ACM
3054int __ftrace_vprintk(unsigned long ip, const char *fmt, va_list ap)
3055{
3056 if (!(trace_flags & TRACE_ITER_PRINTK))
3057 return 0;
3058
3059 return trace_vprintk(ip, task_curr_ret_stack(current), fmt, ap);
3060}
3061EXPORT_SYMBOL_GPL(__ftrace_vprintk);
3062
3f5a54e3
SR
3063static int trace_panic_handler(struct notifier_block *this,
3064 unsigned long event, void *unused)
3065{
944ac425
SR
3066 if (ftrace_dump_on_oops)
3067 ftrace_dump();
3f5a54e3
SR
3068 return NOTIFY_OK;
3069}
3070
3071static struct notifier_block trace_panic_notifier = {
3072 .notifier_call = trace_panic_handler,
3073 .next = NULL,
3074 .priority = 150 /* priority: INT_MAX >= x >= 0 */
3075};
3076
3077static int trace_die_handler(struct notifier_block *self,
3078 unsigned long val,
3079 void *data)
3080{
3081 switch (val) {
3082 case DIE_OOPS:
944ac425
SR
3083 if (ftrace_dump_on_oops)
3084 ftrace_dump();
3f5a54e3
SR
3085 break;
3086 default:
3087 break;
3088 }
3089 return NOTIFY_OK;
3090}
3091
3092static struct notifier_block trace_die_notifier = {
3093 .notifier_call = trace_die_handler,
3094 .priority = 200
3095};
3096
3097/*
3098 * printk is set to max of 1024, we really don't need it that big.
3099 * Nothing should be printing 1000 characters anyway.
3100 */
3101#define TRACE_MAX_PRINT 1000
3102
3103/*
3104 * Define here KERN_TRACE so that we have one place to modify
3105 * it if we decide to change what log level the ftrace dump
3106 * should be at.
3107 */
428aee14 3108#define KERN_TRACE KERN_EMERG
3f5a54e3
SR
3109
3110static void
3111trace_printk_seq(struct trace_seq *s)
3112{
3113 /* Probably should print a warning here. */
3114 if (s->len >= 1000)
3115 s->len = 1000;
3116
3117 /* should be zero ended, but we are paranoid. */
3118 s->buffer[s->len] = 0;
3119
3120 printk(KERN_TRACE "%s", s->buffer);
3121
3122 trace_seq_reset(s);
3123}
3124
3f5a54e3
SR
3125void ftrace_dump(void)
3126{
3127 static DEFINE_SPINLOCK(ftrace_dump_lock);
3128 /* use static because iter can be a bit big for the stack */
3129 static struct trace_iterator iter;
3f5a54e3 3130 static int dump_ran;
d769041f
SR
3131 unsigned long flags;
3132 int cnt = 0, cpu;
3f5a54e3
SR
3133
3134 /* only one dump */
3135 spin_lock_irqsave(&ftrace_dump_lock, flags);
3136 if (dump_ran)
3137 goto out;
3138
3139 dump_ran = 1;
3140
3141 /* No turning back! */
0ee6b6cf 3142 tracing_off();
81adbdc0 3143 ftrace_kill();
3f5a54e3 3144
d769041f
SR
3145 for_each_tracing_cpu(cpu) {
3146 atomic_inc(&global_trace.data[cpu]->disabled);
3147 }
3148
b54d3de9
TE
3149 /* don't look at user memory in panic mode */
3150 trace_flags &= ~TRACE_ITER_SYM_USEROBJ;
3151
3f5a54e3
SR
3152 printk(KERN_TRACE "Dumping ftrace buffer:\n");
3153
3154 iter.tr = &global_trace;
3155 iter.trace = current_trace;
3156
3157 /*
3158 * We need to stop all tracing on all CPUS to read the
3159 * the next buffer. This is a bit expensive, but is
3160 * not done often. We fill all what we can read,
3161 * and then release the locks again.
3162 */
3163
3f5a54e3
SR
3164 while (!trace_empty(&iter)) {
3165
3166 if (!cnt)
3167 printk(KERN_TRACE "---------------------------------\n");
3168
3169 cnt++;
3170
3171 /* reset all but tr, trace, and overruns */
3172 memset(&iter.seq, 0,
3173 sizeof(struct trace_iterator) -
3174 offsetof(struct trace_iterator, seq));
3175 iter.iter_flags |= TRACE_FILE_LAT_FMT;
3176 iter.pos = -1;
3177
3178 if (find_next_entry_inc(&iter) != NULL) {
3179 print_trace_line(&iter);
3180 trace_consume(&iter);
3181 }
3182
3183 trace_printk_seq(&iter.seq);
3184 }
3185
3186 if (!cnt)
3187 printk(KERN_TRACE " (ftrace buffer empty)\n");
3188 else
3189 printk(KERN_TRACE "---------------------------------\n");
3190
3f5a54e3
SR
3191 out:
3192 spin_unlock_irqrestore(&ftrace_dump_lock, flags);
3193}
3194
3928a8a2 3195__init static int tracer_alloc_buffers(void)
bc0c38d1 3196{
4c11d7ae 3197 struct trace_array_cpu *data;
4c11d7ae 3198 int i;
9e01c1b7 3199 int ret = -ENOMEM;
4c11d7ae 3200
9e01c1b7
RR
3201 if (!alloc_cpumask_var(&tracing_buffer_mask, GFP_KERNEL))
3202 goto out;
3203
3204 if (!alloc_cpumask_var(&tracing_cpumask, GFP_KERNEL))
3205 goto out_free_buffer_mask;
4c11d7ae 3206
9e01c1b7
RR
3207 cpumask_copy(tracing_buffer_mask, cpu_possible_mask);
3208 cpumask_copy(tracing_cpumask, cpu_all_mask);
3209
3210 /* TODO: make the number of buffers hot pluggable with CPUS */
3928a8a2
SR
3211 global_trace.buffer = ring_buffer_alloc(trace_buf_size,
3212 TRACE_BUFFER_FLAGS);
3213 if (!global_trace.buffer) {
3214 printk(KERN_ERR "tracer: failed to allocate ring buffer!\n");
3215 WARN_ON(1);
9e01c1b7 3216 goto out_free_cpumask;
4c11d7ae 3217 }
3928a8a2 3218 global_trace.entries = ring_buffer_size(global_trace.buffer);
4c11d7ae 3219
9e01c1b7 3220
4c11d7ae 3221#ifdef CONFIG_TRACER_MAX_TRACE
3928a8a2
SR
3222 max_tr.buffer = ring_buffer_alloc(trace_buf_size,
3223 TRACE_BUFFER_FLAGS);
3224 if (!max_tr.buffer) {
3225 printk(KERN_ERR "tracer: failed to allocate max ring buffer!\n");
3226 WARN_ON(1);
3227 ring_buffer_free(global_trace.buffer);
9e01c1b7 3228 goto out_free_cpumask;
4c11d7ae 3229 }
3928a8a2
SR
3230 max_tr.entries = ring_buffer_size(max_tr.buffer);
3231 WARN_ON(max_tr.entries != global_trace.entries);
a98a3c3f 3232#endif
ab46428c 3233
4c11d7ae 3234 /* Allocate the first page for all buffers */
ab46428c 3235 for_each_tracing_cpu(i) {
4c11d7ae 3236 data = global_trace.data[i] = &per_cpu(global_trace_cpu, i);
bc0c38d1 3237 max_tr.data[i] = &per_cpu(max_data, i);
4c11d7ae 3238 }
bc0c38d1 3239
bc0c38d1
SR
3240 trace_init_cmdlines();
3241
43a15386 3242 register_tracer(&nop_trace);
79fb0768 3243 current_trace = &nop_trace;
b5ad384e
FW
3244#ifdef CONFIG_BOOT_TRACER
3245 register_tracer(&boot_tracer);
b5ad384e 3246#endif
60a11774
SR
3247 /* All seems OK, enable tracing */
3248 tracing_disabled = 0;
3928a8a2 3249
3f5a54e3
SR
3250 atomic_notifier_chain_register(&panic_notifier_list,
3251 &trace_panic_notifier);
3252
3253 register_die_notifier(&trace_die_notifier);
9e01c1b7 3254 ret = 0;
3f5a54e3 3255
9e01c1b7
RR
3256out_free_cpumask:
3257 free_cpumask_var(tracing_cpumask);
3258out_free_buffer_mask:
3259 free_cpumask_var(tracing_buffer_mask);
3260out:
3261 return ret;
bc0c38d1 3262}
b2821ae6
SR
3263
3264__init static int clear_boot_tracer(void)
3265{
3266 /*
3267 * The default tracer at boot buffer is an init section.
3268 * This function is called in lateinit. If we did not
3269 * find the boot tracer, then clear it out, to prevent
3270 * later registration from accessing the buffer that is
3271 * about to be freed.
3272 */
3273 if (!default_bootup_tracer)
3274 return 0;
3275
3276 printk(KERN_INFO "ftrace bootup tracer '%s' not registered.\n",
3277 default_bootup_tracer);
3278 default_bootup_tracer = NULL;
3279
3280 return 0;
3281}
3282
b5ad384e
FW
3283early_initcall(tracer_alloc_buffers);
3284fs_initcall(tracer_init_debugfs);
b2821ae6 3285late_initcall(clear_boot_tracer);