]>
Commit | Line | Data |
---|---|---|
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> |
bc0c38d1 | 34 | |
86387f7e IM |
35 | #include <linux/stacktrace.h> |
36 | ||
bc0c38d1 SR |
37 | #include "trace.h" |
38 | ||
39 | unsigned long __read_mostly tracing_max_latency = (cycle_t)ULONG_MAX; | |
40 | unsigned long __read_mostly tracing_thresh; | |
41 | ||
ab46428c SR |
42 | static unsigned long __read_mostly tracing_nr_buffers; |
43 | static cpumask_t __read_mostly tracing_buffer_mask; | |
44 | ||
45 | #define for_each_tracing_cpu(cpu) \ | |
46 | for_each_cpu_mask(cpu, tracing_buffer_mask) | |
47 | ||
a98a3c3f SR |
48 | static int trace_alloc_page(void); |
49 | static int trace_free_page(void); | |
50 | ||
60a11774 SR |
51 | static int tracing_disabled = 1; |
52 | ||
3eefae99 SR |
53 | static unsigned long tracing_pages_allocated; |
54 | ||
72829bc3 | 55 | long |
bc0c38d1 SR |
56 | ns2usecs(cycle_t nsec) |
57 | { | |
58 | nsec += 500; | |
59 | do_div(nsec, 1000); | |
60 | return nsec; | |
61 | } | |
62 | ||
e309b41d | 63 | cycle_t ftrace_now(int cpu) |
750ed1a4 | 64 | { |
0fd9e0da | 65 | return cpu_clock(cpu); |
750ed1a4 IM |
66 | } |
67 | ||
4fcdae83 SR |
68 | /* |
69 | * The global_trace is the descriptor that holds the tracing | |
70 | * buffers for the live tracing. For each CPU, it contains | |
71 | * a link list of pages that will store trace entries. The | |
72 | * page descriptor of the pages in the memory is used to hold | |
73 | * the link list by linking the lru item in the page descriptor | |
74 | * to each of the pages in the buffer per CPU. | |
75 | * | |
76 | * For each active CPU there is a data field that holds the | |
77 | * pages for the buffer for that CPU. Each CPU has the same number | |
78 | * of pages allocated for its buffer. | |
79 | */ | |
bc0c38d1 SR |
80 | static struct trace_array global_trace; |
81 | ||
82 | static DEFINE_PER_CPU(struct trace_array_cpu, global_trace_cpu); | |
83 | ||
4fcdae83 SR |
84 | /* |
85 | * The max_tr is used to snapshot the global_trace when a maximum | |
86 | * latency is reached. Some tracers will use this to store a maximum | |
87 | * trace while it continues examining live traces. | |
88 | * | |
89 | * The buffers for the max_tr are set up the same as the global_trace. | |
90 | * When a snapshot is taken, the link list of the max_tr is swapped | |
91 | * with the link list of the global_trace and the buffers are reset for | |
92 | * the global_trace so the tracing can continue. | |
93 | */ | |
bc0c38d1 SR |
94 | static struct trace_array max_tr; |
95 | ||
96 | static DEFINE_PER_CPU(struct trace_array_cpu, max_data); | |
97 | ||
4fcdae83 | 98 | /* tracer_enabled is used to toggle activation of a tracer */ |
26994ead | 99 | static int tracer_enabled = 1; |
4fcdae83 | 100 | |
60bc0800 SR |
101 | /* function tracing enabled */ |
102 | int ftrace_function_enabled; | |
103 | ||
4fcdae83 SR |
104 | /* |
105 | * trace_nr_entries is the number of entries that is allocated | |
106 | * for a buffer. Note, the number of entries is always rounded | |
107 | * to ENTRIES_PER_PAGE. | |
3f5a54e3 SR |
108 | * |
109 | * This number is purposely set to a low number of 16384. | |
110 | * If the dump on oops happens, it will be much appreciated | |
111 | * to not have to wait for all that output. Anyway this can be | |
112 | * boot time and run time configurable. | |
4fcdae83 | 113 | */ |
3f5a54e3 SR |
114 | #define TRACE_ENTRIES_DEFAULT 16384UL |
115 | ||
116 | static unsigned long trace_nr_entries = TRACE_ENTRIES_DEFAULT; | |
bc0c38d1 | 117 | |
4fcdae83 | 118 | /* trace_types holds a link list of available tracers. */ |
bc0c38d1 | 119 | static struct tracer *trace_types __read_mostly; |
4fcdae83 SR |
120 | |
121 | /* current_trace points to the tracer that is currently active */ | |
bc0c38d1 | 122 | static struct tracer *current_trace __read_mostly; |
4fcdae83 SR |
123 | |
124 | /* | |
125 | * max_tracer_type_len is used to simplify the allocating of | |
126 | * buffers to read userspace tracer names. We keep track of | |
127 | * the longest tracer name registered. | |
128 | */ | |
bc0c38d1 SR |
129 | static int max_tracer_type_len; |
130 | ||
4fcdae83 SR |
131 | /* |
132 | * trace_types_lock is used to protect the trace_types list. | |
133 | * This lock is also used to keep user access serialized. | |
134 | * Accesses from userspace will grab this lock while userspace | |
135 | * activities happen inside the kernel. | |
136 | */ | |
bc0c38d1 | 137 | static DEFINE_MUTEX(trace_types_lock); |
4fcdae83 SR |
138 | |
139 | /* trace_wait is a waitqueue for tasks blocked on trace_poll */ | |
4e655519 IM |
140 | static DECLARE_WAIT_QUEUE_HEAD(trace_wait); |
141 | ||
4fcdae83 | 142 | /* trace_flags holds iter_ctrl options */ |
4e655519 IM |
143 | unsigned long trace_flags = TRACE_ITER_PRINT_PARENT; |
144 | ||
2b1bce17 AG |
145 | static notrace void no_trace_init(struct trace_array *tr) |
146 | { | |
147 | int cpu; | |
148 | ||
60bc0800 | 149 | ftrace_function_enabled = 0; |
2b1bce17 AG |
150 | if(tr->ctrl) |
151 | for_each_online_cpu(cpu) | |
152 | tracing_reset(tr->data[cpu]); | |
153 | tracer_enabled = 0; | |
154 | } | |
155 | ||
156 | /* dummy trace to disable tracing */ | |
157 | static struct tracer no_tracer __read_mostly = { | |
158 | .name = "none", | |
159 | .init = no_trace_init | |
160 | }; | |
161 | ||
162 | ||
4fcdae83 SR |
163 | /** |
164 | * trace_wake_up - wake up tasks waiting for trace input | |
165 | * | |
166 | * Simply wakes up any task that is blocked on the trace_wait | |
167 | * queue. These is used with trace_poll for tasks polling the trace. | |
168 | */ | |
4e655519 IM |
169 | void trace_wake_up(void) |
170 | { | |
017730c1 IM |
171 | /* |
172 | * The runqueue_is_locked() can fail, but this is the best we | |
173 | * have for now: | |
174 | */ | |
175 | if (!(trace_flags & TRACE_ITER_BLOCK) && !runqueue_is_locked()) | |
4e655519 IM |
176 | wake_up(&trace_wait); |
177 | } | |
bc0c38d1 | 178 | |
4c11d7ae SR |
179 | #define ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(struct trace_entry)) |
180 | ||
bc0c38d1 SR |
181 | static int __init set_nr_entries(char *str) |
182 | { | |
c6caeeb1 SR |
183 | unsigned long nr_entries; |
184 | int ret; | |
185 | ||
bc0c38d1 SR |
186 | if (!str) |
187 | return 0; | |
c6caeeb1 SR |
188 | ret = strict_strtoul(str, 0, &nr_entries); |
189 | /* nr_entries can not be zero */ | |
190 | if (ret < 0 || nr_entries == 0) | |
191 | return 0; | |
192 | trace_nr_entries = nr_entries; | |
bc0c38d1 SR |
193 | return 1; |
194 | } | |
195 | __setup("trace_entries=", set_nr_entries); | |
196 | ||
57f50be1 SR |
197 | unsigned long nsecs_to_usecs(unsigned long nsecs) |
198 | { | |
199 | return nsecs / 1000; | |
200 | } | |
201 | ||
4fcdae83 SR |
202 | /* |
203 | * trace_flag_type is an enumeration that holds different | |
204 | * states when a trace occurs. These are: | |
205 | * IRQS_OFF - interrupts were disabled | |
206 | * NEED_RESCED - reschedule is requested | |
207 | * HARDIRQ - inside an interrupt handler | |
208 | * SOFTIRQ - inside a softirq handler | |
dd0e545f | 209 | * CONT - multiple entries hold the trace item |
4fcdae83 | 210 | */ |
bc0c38d1 SR |
211 | enum trace_flag_type { |
212 | TRACE_FLAG_IRQS_OFF = 0x01, | |
213 | TRACE_FLAG_NEED_RESCHED = 0x02, | |
214 | TRACE_FLAG_HARDIRQ = 0x04, | |
215 | TRACE_FLAG_SOFTIRQ = 0x08, | |
dd0e545f | 216 | TRACE_FLAG_CONT = 0x10, |
bc0c38d1 SR |
217 | }; |
218 | ||
4fcdae83 SR |
219 | /* |
220 | * TRACE_ITER_SYM_MASK masks the options in trace_flags that | |
221 | * control the output of kernel symbols. | |
222 | */ | |
bc0c38d1 SR |
223 | #define TRACE_ITER_SYM_MASK \ |
224 | (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR) | |
225 | ||
4fcdae83 | 226 | /* These must match the bit postions in trace_iterator_flags */ |
bc0c38d1 SR |
227 | static const char *trace_options[] = { |
228 | "print-parent", | |
229 | "sym-offset", | |
230 | "sym-addr", | |
231 | "verbose", | |
f9896bf3 | 232 | "raw", |
5e3ca0ec | 233 | "hex", |
cb0f12aa | 234 | "bin", |
2a2cc8f7 | 235 | "block", |
86387f7e | 236 | "stacktrace", |
4ac3ba41 | 237 | "sched-tree", |
f09ce573 | 238 | "ftrace_printk", |
bc0c38d1 SR |
239 | NULL |
240 | }; | |
241 | ||
4fcdae83 SR |
242 | /* |
243 | * ftrace_max_lock is used to protect the swapping of buffers | |
244 | * when taking a max snapshot. The buffers themselves are | |
245 | * protected by per_cpu spinlocks. But the action of the swap | |
246 | * needs its own lock. | |
247 | * | |
248 | * This is defined as a raw_spinlock_t in order to help | |
249 | * with performance when lockdep debugging is enabled. | |
250 | */ | |
92205c23 SR |
251 | static raw_spinlock_t ftrace_max_lock = |
252 | (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; | |
bc0c38d1 SR |
253 | |
254 | /* | |
255 | * Copy the new maximum trace into the separate maximum-trace | |
256 | * structure. (this way the maximum trace is permanently saved, | |
257 | * for later retrieval via /debugfs/tracing/latency_trace) | |
258 | */ | |
e309b41d | 259 | static void |
bc0c38d1 SR |
260 | __update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu) |
261 | { | |
262 | struct trace_array_cpu *data = tr->data[cpu]; | |
263 | ||
264 | max_tr.cpu = cpu; | |
265 | max_tr.time_start = data->preempt_timestamp; | |
266 | ||
267 | data = max_tr.data[cpu]; | |
268 | data->saved_latency = tracing_max_latency; | |
269 | ||
270 | memcpy(data->comm, tsk->comm, TASK_COMM_LEN); | |
271 | data->pid = tsk->pid; | |
272 | data->uid = tsk->uid; | |
273 | data->nice = tsk->static_prio - 20 - MAX_RT_PRIO; | |
274 | data->policy = tsk->policy; | |
275 | data->rt_priority = tsk->rt_priority; | |
276 | ||
277 | /* record this tasks comm */ | |
278 | tracing_record_cmdline(current); | |
279 | } | |
280 | ||
19384c03 SR |
281 | #define CHECK_COND(cond) \ |
282 | if (unlikely(cond)) { \ | |
283 | tracing_disabled = 1; \ | |
284 | WARN_ON(1); \ | |
285 | return -1; \ | |
286 | } | |
287 | ||
4fcdae83 SR |
288 | /** |
289 | * check_pages - integrity check of trace buffers | |
290 | * | |
291 | * As a safty measure we check to make sure the data pages have not | |
19384c03 | 292 | * been corrupted. |
4fcdae83 | 293 | */ |
19384c03 | 294 | int check_pages(struct trace_array_cpu *data) |
c7aafc54 IM |
295 | { |
296 | struct page *page, *tmp; | |
297 | ||
19384c03 SR |
298 | CHECK_COND(data->trace_pages.next->prev != &data->trace_pages); |
299 | CHECK_COND(data->trace_pages.prev->next != &data->trace_pages); | |
c7aafc54 IM |
300 | |
301 | list_for_each_entry_safe(page, tmp, &data->trace_pages, lru) { | |
19384c03 SR |
302 | CHECK_COND(page->lru.next->prev != &page->lru); |
303 | CHECK_COND(page->lru.prev->next != &page->lru); | |
c7aafc54 | 304 | } |
19384c03 SR |
305 | |
306 | return 0; | |
c7aafc54 IM |
307 | } |
308 | ||
4fcdae83 SR |
309 | /** |
310 | * head_page - page address of the first page in per_cpu buffer. | |
311 | * | |
312 | * head_page returns the page address of the first page in | |
313 | * a per_cpu buffer. This also preforms various consistency | |
314 | * checks to make sure the buffer has not been corrupted. | |
315 | */ | |
c7aafc54 IM |
316 | void *head_page(struct trace_array_cpu *data) |
317 | { | |
318 | struct page *page; | |
319 | ||
c7aafc54 IM |
320 | if (list_empty(&data->trace_pages)) |
321 | return NULL; | |
322 | ||
323 | page = list_entry(data->trace_pages.next, struct page, lru); | |
324 | BUG_ON(&page->lru == &data->trace_pages); | |
325 | ||
326 | return page_address(page); | |
327 | } | |
328 | ||
4fcdae83 SR |
329 | /** |
330 | * trace_seq_printf - sequence printing of trace information | |
331 | * @s: trace sequence descriptor | |
332 | * @fmt: printf format string | |
333 | * | |
334 | * The tracer may use either sequence operations or its own | |
335 | * copy to user routines. To simplify formating of a trace | |
336 | * trace_seq_printf is used to store strings into a special | |
337 | * buffer (@s). Then the output may be either used by | |
338 | * the sequencer or pulled into another buffer. | |
339 | */ | |
72829bc3 | 340 | int |
214023c3 SR |
341 | trace_seq_printf(struct trace_seq *s, const char *fmt, ...) |
342 | { | |
343 | int len = (PAGE_SIZE - 1) - s->len; | |
344 | va_list ap; | |
b3806b43 | 345 | int ret; |
214023c3 SR |
346 | |
347 | if (!len) | |
348 | return 0; | |
349 | ||
350 | va_start(ap, fmt); | |
b3806b43 | 351 | ret = vsnprintf(s->buffer + s->len, len, fmt, ap); |
214023c3 SR |
352 | va_end(ap); |
353 | ||
b3806b43 | 354 | /* If we can't write it all, don't bother writing anything */ |
72829bc3 | 355 | if (ret >= len) |
b3806b43 SR |
356 | return 0; |
357 | ||
358 | s->len += ret; | |
214023c3 SR |
359 | |
360 | return len; | |
361 | } | |
362 | ||
4fcdae83 SR |
363 | /** |
364 | * trace_seq_puts - trace sequence printing of simple string | |
365 | * @s: trace sequence descriptor | |
366 | * @str: simple string to record | |
367 | * | |
368 | * The tracer may use either the sequence operations or its own | |
369 | * copy to user routines. This function records a simple string | |
370 | * into a special buffer (@s) for later retrieval by a sequencer | |
371 | * or other mechanism. | |
372 | */ | |
e309b41d | 373 | static int |
214023c3 SR |
374 | trace_seq_puts(struct trace_seq *s, const char *str) |
375 | { | |
376 | int len = strlen(str); | |
377 | ||
378 | if (len > ((PAGE_SIZE - 1) - s->len)) | |
b3806b43 | 379 | return 0; |
214023c3 SR |
380 | |
381 | memcpy(s->buffer + s->len, str, len); | |
382 | s->len += len; | |
383 | ||
384 | return len; | |
385 | } | |
386 | ||
e309b41d | 387 | static int |
214023c3 SR |
388 | trace_seq_putc(struct trace_seq *s, unsigned char c) |
389 | { | |
390 | if (s->len >= (PAGE_SIZE - 1)) | |
391 | return 0; | |
392 | ||
393 | s->buffer[s->len++] = c; | |
394 | ||
395 | return 1; | |
396 | } | |
397 | ||
e309b41d | 398 | static int |
cb0f12aa IM |
399 | trace_seq_putmem(struct trace_seq *s, void *mem, size_t len) |
400 | { | |
401 | if (len > ((PAGE_SIZE - 1) - s->len)) | |
402 | return 0; | |
403 | ||
404 | memcpy(s->buffer + s->len, mem, len); | |
405 | s->len += len; | |
406 | ||
407 | return len; | |
408 | } | |
409 | ||
5e3ca0ec | 410 | #define HEX_CHARS 17 |
93dcc6ea | 411 | static const char hex2asc[] = "0123456789abcdef"; |
5e3ca0ec | 412 | |
e309b41d | 413 | static int |
5e3ca0ec IM |
414 | trace_seq_putmem_hex(struct trace_seq *s, void *mem, size_t len) |
415 | { | |
416 | unsigned char hex[HEX_CHARS]; | |
93dcc6ea | 417 | unsigned char *data = mem; |
5e3ca0ec IM |
418 | unsigned char byte; |
419 | int i, j; | |
420 | ||
421 | BUG_ON(len >= HEX_CHARS); | |
422 | ||
5e3ca0ec IM |
423 | #ifdef __BIG_ENDIAN |
424 | for (i = 0, j = 0; i < len; i++) { | |
425 | #else | |
426 | for (i = len-1, j = 0; i >= 0; i--) { | |
427 | #endif | |
428 | byte = data[i]; | |
429 | ||
93dcc6ea TG |
430 | hex[j++] = hex2asc[byte & 0x0f]; |
431 | hex[j++] = hex2asc[byte >> 4]; | |
5e3ca0ec | 432 | } |
93dcc6ea | 433 | hex[j++] = ' '; |
5e3ca0ec IM |
434 | |
435 | return trace_seq_putmem(s, hex, j); | |
436 | } | |
437 | ||
e309b41d | 438 | static void |
214023c3 SR |
439 | trace_seq_reset(struct trace_seq *s) |
440 | { | |
441 | s->len = 0; | |
6c6c2796 PP |
442 | s->readpos = 0; |
443 | } | |
444 | ||
445 | ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, size_t cnt) | |
446 | { | |
447 | int len; | |
448 | int ret; | |
449 | ||
450 | if (s->len <= s->readpos) | |
451 | return -EBUSY; | |
452 | ||
453 | len = s->len - s->readpos; | |
454 | if (cnt > len) | |
455 | cnt = len; | |
456 | ret = copy_to_user(ubuf, s->buffer + s->readpos, cnt); | |
457 | if (ret) | |
458 | return -EFAULT; | |
459 | ||
460 | s->readpos += len; | |
461 | return cnt; | |
214023c3 SR |
462 | } |
463 | ||
e309b41d | 464 | static void |
214023c3 SR |
465 | trace_print_seq(struct seq_file *m, struct trace_seq *s) |
466 | { | |
467 | int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len; | |
468 | ||
469 | s->buffer[len] = 0; | |
470 | seq_puts(m, s->buffer); | |
471 | ||
472 | trace_seq_reset(s); | |
473 | } | |
474 | ||
4fcdae83 SR |
475 | /* |
476 | * flip the trace buffers between two trace descriptors. | |
477 | * This usually is the buffers between the global_trace and | |
478 | * the max_tr to record a snapshot of a current trace. | |
479 | * | |
480 | * The ftrace_max_lock must be held. | |
481 | */ | |
e309b41d | 482 | static void |
c7aafc54 IM |
483 | flip_trace(struct trace_array_cpu *tr1, struct trace_array_cpu *tr2) |
484 | { | |
485 | struct list_head flip_pages; | |
486 | ||
487 | INIT_LIST_HEAD(&flip_pages); | |
488 | ||
93a588f4 | 489 | memcpy(&tr1->trace_head_idx, &tr2->trace_head_idx, |
c7aafc54 | 490 | sizeof(struct trace_array_cpu) - |
93a588f4 | 491 | offsetof(struct trace_array_cpu, trace_head_idx)); |
c7aafc54 IM |
492 | |
493 | check_pages(tr1); | |
494 | check_pages(tr2); | |
495 | list_splice_init(&tr1->trace_pages, &flip_pages); | |
496 | list_splice_init(&tr2->trace_pages, &tr1->trace_pages); | |
497 | list_splice_init(&flip_pages, &tr2->trace_pages); | |
498 | BUG_ON(!list_empty(&flip_pages)); | |
499 | check_pages(tr1); | |
500 | check_pages(tr2); | |
501 | } | |
502 | ||
4fcdae83 SR |
503 | /** |
504 | * update_max_tr - snapshot all trace buffers from global_trace to max_tr | |
505 | * @tr: tracer | |
506 | * @tsk: the task with the latency | |
507 | * @cpu: The cpu that initiated the trace. | |
508 | * | |
509 | * Flip the buffers between the @tr and the max_tr and record information | |
510 | * about which task was the cause of this latency. | |
511 | */ | |
e309b41d | 512 | void |
bc0c38d1 SR |
513 | update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu) |
514 | { | |
515 | struct trace_array_cpu *data; | |
bc0c38d1 SR |
516 | int i; |
517 | ||
4c11d7ae | 518 | WARN_ON_ONCE(!irqs_disabled()); |
92205c23 | 519 | __raw_spin_lock(&ftrace_max_lock); |
bc0c38d1 | 520 | /* clear out all the previous traces */ |
ab46428c | 521 | for_each_tracing_cpu(i) { |
bc0c38d1 | 522 | data = tr->data[i]; |
c7aafc54 | 523 | flip_trace(max_tr.data[i], data); |
89b2f978 | 524 | tracing_reset(data); |
bc0c38d1 SR |
525 | } |
526 | ||
527 | __update_max_tr(tr, tsk, cpu); | |
92205c23 | 528 | __raw_spin_unlock(&ftrace_max_lock); |
bc0c38d1 SR |
529 | } |
530 | ||
531 | /** | |
532 | * update_max_tr_single - only copy one trace over, and reset the rest | |
533 | * @tr - tracer | |
534 | * @tsk - task with the latency | |
535 | * @cpu - the cpu of the buffer to copy. | |
4fcdae83 SR |
536 | * |
537 | * Flip the trace of a single CPU buffer between the @tr and the max_tr. | |
bc0c38d1 | 538 | */ |
e309b41d | 539 | void |
bc0c38d1 SR |
540 | update_max_tr_single(struct trace_array *tr, struct task_struct *tsk, int cpu) |
541 | { | |
542 | struct trace_array_cpu *data = tr->data[cpu]; | |
bc0c38d1 SR |
543 | int i; |
544 | ||
4c11d7ae | 545 | WARN_ON_ONCE(!irqs_disabled()); |
92205c23 | 546 | __raw_spin_lock(&ftrace_max_lock); |
ab46428c | 547 | for_each_tracing_cpu(i) |
bc0c38d1 SR |
548 | tracing_reset(max_tr.data[i]); |
549 | ||
c7aafc54 | 550 | flip_trace(max_tr.data[cpu], data); |
89b2f978 | 551 | tracing_reset(data); |
bc0c38d1 SR |
552 | |
553 | __update_max_tr(tr, tsk, cpu); | |
92205c23 | 554 | __raw_spin_unlock(&ftrace_max_lock); |
bc0c38d1 SR |
555 | } |
556 | ||
4fcdae83 SR |
557 | /** |
558 | * register_tracer - register a tracer with the ftrace system. | |
559 | * @type - the plugin for the tracer | |
560 | * | |
561 | * Register a new plugin tracer. | |
562 | */ | |
bc0c38d1 SR |
563 | int register_tracer(struct tracer *type) |
564 | { | |
565 | struct tracer *t; | |
566 | int len; | |
567 | int ret = 0; | |
568 | ||
569 | if (!type->name) { | |
570 | pr_info("Tracer must have a name\n"); | |
571 | return -1; | |
572 | } | |
573 | ||
574 | mutex_lock(&trace_types_lock); | |
575 | for (t = trace_types; t; t = t->next) { | |
576 | if (strcmp(type->name, t->name) == 0) { | |
577 | /* already found */ | |
578 | pr_info("Trace %s already registered\n", | |
579 | type->name); | |
580 | ret = -1; | |
581 | goto out; | |
582 | } | |
583 | } | |
584 | ||
60a11774 SR |
585 | #ifdef CONFIG_FTRACE_STARTUP_TEST |
586 | if (type->selftest) { | |
587 | struct tracer *saved_tracer = current_trace; | |
588 | struct trace_array_cpu *data; | |
589 | struct trace_array *tr = &global_trace; | |
590 | int saved_ctrl = tr->ctrl; | |
591 | int i; | |
592 | /* | |
593 | * Run a selftest on this tracer. | |
594 | * Here we reset the trace buffer, and set the current | |
595 | * tracer to be this tracer. The tracer can then run some | |
596 | * internal tracing to verify that everything is in order. | |
597 | * If we fail, we do not register this tracer. | |
598 | */ | |
ab46428c | 599 | for_each_tracing_cpu(i) { |
60a11774 | 600 | data = tr->data[i]; |
c7aafc54 IM |
601 | if (!head_page(data)) |
602 | continue; | |
60a11774 SR |
603 | tracing_reset(data); |
604 | } | |
605 | current_trace = type; | |
606 | tr->ctrl = 0; | |
607 | /* the test is responsible for initializing and enabling */ | |
608 | pr_info("Testing tracer %s: ", type->name); | |
609 | ret = type->selftest(type, tr); | |
610 | /* the test is responsible for resetting too */ | |
611 | current_trace = saved_tracer; | |
612 | tr->ctrl = saved_ctrl; | |
613 | if (ret) { | |
614 | printk(KERN_CONT "FAILED!\n"); | |
615 | goto out; | |
616 | } | |
1d4db00a | 617 | /* Only reset on passing, to avoid touching corrupted buffers */ |
ab46428c | 618 | for_each_tracing_cpu(i) { |
1d4db00a SR |
619 | data = tr->data[i]; |
620 | if (!head_page(data)) | |
621 | continue; | |
622 | tracing_reset(data); | |
623 | } | |
60a11774 SR |
624 | printk(KERN_CONT "PASSED\n"); |
625 | } | |
626 | #endif | |
627 | ||
bc0c38d1 SR |
628 | type->next = trace_types; |
629 | trace_types = type; | |
630 | len = strlen(type->name); | |
631 | if (len > max_tracer_type_len) | |
632 | max_tracer_type_len = len; | |
60a11774 | 633 | |
bc0c38d1 SR |
634 | out: |
635 | mutex_unlock(&trace_types_lock); | |
636 | ||
637 | return ret; | |
638 | } | |
639 | ||
640 | void unregister_tracer(struct tracer *type) | |
641 | { | |
642 | struct tracer **t; | |
643 | int len; | |
644 | ||
645 | mutex_lock(&trace_types_lock); | |
646 | for (t = &trace_types; *t; t = &(*t)->next) { | |
647 | if (*t == type) | |
648 | goto found; | |
649 | } | |
650 | pr_info("Trace %s not registered\n", type->name); | |
651 | goto out; | |
652 | ||
653 | found: | |
654 | *t = (*t)->next; | |
655 | if (strlen(type->name) != max_tracer_type_len) | |
656 | goto out; | |
657 | ||
658 | max_tracer_type_len = 0; | |
659 | for (t = &trace_types; *t; t = &(*t)->next) { | |
660 | len = strlen((*t)->name); | |
661 | if (len > max_tracer_type_len) | |
662 | max_tracer_type_len = len; | |
663 | } | |
664 | out: | |
665 | mutex_unlock(&trace_types_lock); | |
666 | } | |
667 | ||
e309b41d | 668 | void tracing_reset(struct trace_array_cpu *data) |
bc0c38d1 SR |
669 | { |
670 | data->trace_idx = 0; | |
53d0aa77 | 671 | data->overrun = 0; |
93a588f4 SR |
672 | data->trace_head = data->trace_tail = head_page(data); |
673 | data->trace_head_idx = 0; | |
674 | data->trace_tail_idx = 0; | |
bc0c38d1 SR |
675 | } |
676 | ||
bc0c38d1 SR |
677 | #define SAVED_CMDLINES 128 |
678 | static unsigned map_pid_to_cmdline[PID_MAX_DEFAULT+1]; | |
679 | static unsigned map_cmdline_to_pid[SAVED_CMDLINES]; | |
680 | static char saved_cmdlines[SAVED_CMDLINES][TASK_COMM_LEN]; | |
681 | static int cmdline_idx; | |
682 | static DEFINE_SPINLOCK(trace_cmdline_lock); | |
25b0b44a | 683 | |
25b0b44a SR |
684 | /* temporary disable recording */ |
685 | atomic_t trace_record_cmdline_disabled __read_mostly; | |
bc0c38d1 SR |
686 | |
687 | static void trace_init_cmdlines(void) | |
688 | { | |
689 | memset(&map_pid_to_cmdline, -1, sizeof(map_pid_to_cmdline)); | |
690 | memset(&map_cmdline_to_pid, -1, sizeof(map_cmdline_to_pid)); | |
691 | cmdline_idx = 0; | |
692 | } | |
693 | ||
e309b41d | 694 | void trace_stop_cmdline_recording(void); |
bc0c38d1 | 695 | |
e309b41d | 696 | static void trace_save_cmdline(struct task_struct *tsk) |
bc0c38d1 SR |
697 | { |
698 | unsigned map; | |
699 | unsigned idx; | |
700 | ||
701 | if (!tsk->pid || unlikely(tsk->pid > PID_MAX_DEFAULT)) | |
702 | return; | |
703 | ||
704 | /* | |
705 | * It's not the end of the world if we don't get | |
706 | * the lock, but we also don't want to spin | |
707 | * nor do we want to disable interrupts, | |
708 | * so if we miss here, then better luck next time. | |
709 | */ | |
710 | if (!spin_trylock(&trace_cmdline_lock)) | |
711 | return; | |
712 | ||
713 | idx = map_pid_to_cmdline[tsk->pid]; | |
714 | if (idx >= SAVED_CMDLINES) { | |
715 | idx = (cmdline_idx + 1) % SAVED_CMDLINES; | |
716 | ||
717 | map = map_cmdline_to_pid[idx]; | |
718 | if (map <= PID_MAX_DEFAULT) | |
719 | map_pid_to_cmdline[map] = (unsigned)-1; | |
720 | ||
721 | map_pid_to_cmdline[tsk->pid] = idx; | |
722 | ||
723 | cmdline_idx = idx; | |
724 | } | |
725 | ||
726 | memcpy(&saved_cmdlines[idx], tsk->comm, TASK_COMM_LEN); | |
727 | ||
728 | spin_unlock(&trace_cmdline_lock); | |
729 | } | |
730 | ||
e309b41d | 731 | static char *trace_find_cmdline(int pid) |
bc0c38d1 SR |
732 | { |
733 | char *cmdline = "<...>"; | |
734 | unsigned map; | |
735 | ||
736 | if (!pid) | |
737 | return "<idle>"; | |
738 | ||
739 | if (pid > PID_MAX_DEFAULT) | |
740 | goto out; | |
741 | ||
742 | map = map_pid_to_cmdline[pid]; | |
743 | if (map >= SAVED_CMDLINES) | |
744 | goto out; | |
745 | ||
746 | cmdline = saved_cmdlines[map]; | |
747 | ||
748 | out: | |
749 | return cmdline; | |
750 | } | |
751 | ||
e309b41d | 752 | void tracing_record_cmdline(struct task_struct *tsk) |
bc0c38d1 SR |
753 | { |
754 | if (atomic_read(&trace_record_cmdline_disabled)) | |
755 | return; | |
756 | ||
757 | trace_save_cmdline(tsk); | |
758 | } | |
759 | ||
e309b41d | 760 | static inline struct list_head * |
93a588f4 SR |
761 | trace_next_list(struct trace_array_cpu *data, struct list_head *next) |
762 | { | |
763 | /* | |
764 | * Roundrobin - but skip the head (which is not a real page): | |
765 | */ | |
766 | next = next->next; | |
767 | if (unlikely(next == &data->trace_pages)) | |
768 | next = next->next; | |
769 | BUG_ON(next == &data->trace_pages); | |
770 | ||
771 | return next; | |
772 | } | |
773 | ||
e309b41d | 774 | static inline void * |
93a588f4 SR |
775 | trace_next_page(struct trace_array_cpu *data, void *addr) |
776 | { | |
777 | struct list_head *next; | |
778 | struct page *page; | |
779 | ||
780 | page = virt_to_page(addr); | |
781 | ||
782 | next = trace_next_list(data, &page->lru); | |
783 | page = list_entry(next, struct page, lru); | |
784 | ||
785 | return page_address(page); | |
786 | } | |
787 | ||
45dcd8b8 | 788 | struct trace_entry * |
c7aafc54 | 789 | tracing_get_trace_entry(struct trace_array *tr, struct trace_array_cpu *data) |
bc0c38d1 SR |
790 | { |
791 | unsigned long idx, idx_next; | |
792 | struct trace_entry *entry; | |
793 | ||
4c11d7ae | 794 | data->trace_idx++; |
93a588f4 | 795 | idx = data->trace_head_idx; |
bc0c38d1 SR |
796 | idx_next = idx + 1; |
797 | ||
c7aafc54 IM |
798 | BUG_ON(idx * TRACE_ENTRY_SIZE >= PAGE_SIZE); |
799 | ||
93a588f4 | 800 | entry = data->trace_head + idx * TRACE_ENTRY_SIZE; |
4c11d7ae SR |
801 | |
802 | if (unlikely(idx_next >= ENTRIES_PER_PAGE)) { | |
93a588f4 | 803 | data->trace_head = trace_next_page(data, data->trace_head); |
bc0c38d1 SR |
804 | idx_next = 0; |
805 | } | |
806 | ||
93a588f4 SR |
807 | if (data->trace_head == data->trace_tail && |
808 | idx_next == data->trace_tail_idx) { | |
809 | /* overrun */ | |
53d0aa77 | 810 | data->overrun++; |
93a588f4 SR |
811 | data->trace_tail_idx++; |
812 | if (data->trace_tail_idx >= ENTRIES_PER_PAGE) { | |
813 | data->trace_tail = | |
814 | trace_next_page(data, data->trace_tail); | |
815 | data->trace_tail_idx = 0; | |
816 | } | |
817 | } | |
818 | ||
819 | data->trace_head_idx = idx_next; | |
bc0c38d1 SR |
820 | |
821 | return entry; | |
822 | } | |
823 | ||
45dcd8b8 | 824 | void |
c7aafc54 | 825 | tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags) |
bc0c38d1 SR |
826 | { |
827 | struct task_struct *tsk = current; | |
828 | unsigned long pc; | |
829 | ||
830 | pc = preempt_count(); | |
831 | ||
2e2ca155 SR |
832 | entry->field.preempt_count = pc & 0xff; |
833 | entry->field.pid = (tsk) ? tsk->pid : 0; | |
834 | entry->field.t = ftrace_now(raw_smp_processor_id()); | |
835 | entry->field.flags = | |
836 | (irqs_disabled_flags(flags) ? TRACE_FLAG_IRQS_OFF : 0) | | |
bc0c38d1 SR |
837 | ((pc & HARDIRQ_MASK) ? TRACE_FLAG_HARDIRQ : 0) | |
838 | ((pc & SOFTIRQ_MASK) ? TRACE_FLAG_SOFTIRQ : 0) | | |
839 | (need_resched() ? TRACE_FLAG_NEED_RESCHED : 0); | |
840 | } | |
841 | ||
e309b41d | 842 | void |
6fb44b71 SR |
843 | trace_function(struct trace_array *tr, struct trace_array_cpu *data, |
844 | unsigned long ip, unsigned long parent_ip, unsigned long flags) | |
bc0c38d1 SR |
845 | { |
846 | struct trace_entry *entry; | |
dcb6308f | 847 | unsigned long irq_flags; |
bc0c38d1 | 848 | |
92205c23 SR |
849 | raw_local_irq_save(irq_flags); |
850 | __raw_spin_lock(&data->lock); | |
2e2ca155 | 851 | entry = tracing_get_trace_entry(tr, data); |
bc0c38d1 | 852 | tracing_generic_entry_update(entry, flags); |
2e2ca155 SR |
853 | entry->type = TRACE_FN; |
854 | entry->field.fn.ip = ip; | |
855 | entry->field.fn.parent_ip = parent_ip; | |
92205c23 SR |
856 | __raw_spin_unlock(&data->lock); |
857 | raw_local_irq_restore(irq_flags); | |
bc0c38d1 SR |
858 | } |
859 | ||
e309b41d | 860 | void |
2e0f5761 IM |
861 | ftrace(struct trace_array *tr, struct trace_array_cpu *data, |
862 | unsigned long ip, unsigned long parent_ip, unsigned long flags) | |
863 | { | |
864 | if (likely(!atomic_read(&data->disabled))) | |
6fb44b71 | 865 | trace_function(tr, data, ip, parent_ip, flags); |
2e0f5761 IM |
866 | } |
867 | ||
86387f7e IM |
868 | void __trace_stack(struct trace_array *tr, |
869 | struct trace_array_cpu *data, | |
870 | unsigned long flags, | |
871 | int skip) | |
872 | { | |
873 | struct trace_entry *entry; | |
874 | struct stack_trace trace; | |
875 | ||
876 | if (!(trace_flags & TRACE_ITER_STACKTRACE)) | |
877 | return; | |
878 | ||
879 | entry = tracing_get_trace_entry(tr, data); | |
880 | tracing_generic_entry_update(entry, flags); | |
881 | entry->type = TRACE_STACK; | |
882 | ||
2e2ca155 | 883 | memset(&entry->field.stack, 0, sizeof(entry->field.stack)); |
86387f7e IM |
884 | |
885 | trace.nr_entries = 0; | |
886 | trace.max_entries = FTRACE_STACK_ENTRIES; | |
887 | trace.skip = skip; | |
2e2ca155 | 888 | trace.entries = entry->field.stack.caller; |
86387f7e IM |
889 | |
890 | save_stack_trace(&trace); | |
f0a920d5 IM |
891 | } |
892 | ||
a4feb834 IM |
893 | void |
894 | __trace_special(void *__tr, void *__data, | |
895 | unsigned long arg1, unsigned long arg2, unsigned long arg3) | |
896 | { | |
897 | struct trace_array_cpu *data = __data; | |
898 | struct trace_array *tr = __tr; | |
899 | struct trace_entry *entry; | |
900 | unsigned long irq_flags; | |
901 | ||
902 | raw_local_irq_save(irq_flags); | |
903 | __raw_spin_lock(&data->lock); | |
2e2ca155 | 904 | entry = tracing_get_trace_entry(tr, data); |
a4feb834 | 905 | tracing_generic_entry_update(entry, 0); |
2e2ca155 SR |
906 | entry->type = TRACE_SPECIAL; |
907 | entry->field.special.arg1 = arg1; | |
908 | entry->field.special.arg2 = arg2; | |
909 | entry->field.special.arg3 = arg3; | |
a4feb834 IM |
910 | __trace_stack(tr, data, irq_flags, 4); |
911 | __raw_spin_unlock(&data->lock); | |
912 | raw_local_irq_restore(irq_flags); | |
913 | ||
914 | trace_wake_up(); | |
915 | } | |
916 | ||
e309b41d | 917 | void |
bc0c38d1 SR |
918 | tracing_sched_switch_trace(struct trace_array *tr, |
919 | struct trace_array_cpu *data, | |
86387f7e IM |
920 | struct task_struct *prev, |
921 | struct task_struct *next, | |
bc0c38d1 SR |
922 | unsigned long flags) |
923 | { | |
924 | struct trace_entry *entry; | |
dcb6308f | 925 | unsigned long irq_flags; |
bc0c38d1 | 926 | |
92205c23 SR |
927 | raw_local_irq_save(irq_flags); |
928 | __raw_spin_lock(&data->lock); | |
2e2ca155 | 929 | entry = tracing_get_trace_entry(tr, data); |
bc0c38d1 | 930 | tracing_generic_entry_update(entry, flags); |
2e2ca155 SR |
931 | entry->type = TRACE_CTX; |
932 | entry->field.ctx.prev_pid = prev->pid; | |
933 | entry->field.ctx.prev_prio = prev->prio; | |
934 | entry->field.ctx.prev_state = prev->state; | |
935 | entry->field.ctx.next_pid = next->pid; | |
936 | entry->field.ctx.next_prio = next->prio; | |
937 | entry->field.ctx.next_state = next->state; | |
80b5e940 | 938 | entry->field.ctx.next_cpu = task_cpu(next); |
74f4e369 | 939 | __trace_stack(tr, data, flags, 5); |
92205c23 SR |
940 | __raw_spin_unlock(&data->lock); |
941 | raw_local_irq_restore(irq_flags); | |
bc0c38d1 SR |
942 | } |
943 | ||
57422797 IM |
944 | void |
945 | tracing_sched_wakeup_trace(struct trace_array *tr, | |
946 | struct trace_array_cpu *data, | |
86387f7e IM |
947 | struct task_struct *wakee, |
948 | struct task_struct *curr, | |
57422797 IM |
949 | unsigned long flags) |
950 | { | |
951 | struct trace_entry *entry; | |
952 | unsigned long irq_flags; | |
953 | ||
92205c23 SR |
954 | raw_local_irq_save(irq_flags); |
955 | __raw_spin_lock(&data->lock); | |
57422797 IM |
956 | entry = tracing_get_trace_entry(tr, data); |
957 | tracing_generic_entry_update(entry, flags); | |
958 | entry->type = TRACE_WAKE; | |
2e2ca155 SR |
959 | entry->field.ctx.prev_pid = curr->pid; |
960 | entry->field.ctx.prev_prio = curr->prio; | |
961 | entry->field.ctx.prev_state = curr->state; | |
962 | entry->field.ctx.next_pid = wakee->pid; | |
963 | entry->field.ctx.next_prio = wakee->prio; | |
964 | entry->field.ctx.next_state = wakee->state; | |
80b5e940 | 965 | entry->field.ctx.next_cpu = task_cpu(wakee); |
74f4e369 | 966 | __trace_stack(tr, data, flags, 6); |
92205c23 SR |
967 | __raw_spin_unlock(&data->lock); |
968 | raw_local_irq_restore(irq_flags); | |
017730c1 IM |
969 | |
970 | trace_wake_up(); | |
57422797 IM |
971 | } |
972 | ||
4902f884 SR |
973 | void |
974 | ftrace_special(unsigned long arg1, unsigned long arg2, unsigned long arg3) | |
975 | { | |
976 | struct trace_array *tr = &global_trace; | |
977 | struct trace_array_cpu *data; | |
978 | unsigned long flags; | |
979 | long disabled; | |
980 | int cpu; | |
981 | ||
982 | if (tracing_disabled || current_trace == &no_tracer || !tr->ctrl) | |
983 | return; | |
984 | ||
985 | local_irq_save(flags); | |
986 | cpu = raw_smp_processor_id(); | |
987 | data = tr->data[cpu]; | |
988 | disabled = atomic_inc_return(&data->disabled); | |
989 | ||
990 | if (likely(disabled == 1)) | |
991 | __trace_special(tr, data, arg1, arg2, arg3); | |
992 | ||
993 | atomic_dec(&data->disabled); | |
994 | local_irq_restore(flags); | |
995 | } | |
996 | ||
2e0f5761 | 997 | #ifdef CONFIG_FTRACE |
e309b41d | 998 | static void |
2e0f5761 IM |
999 | function_trace_call(unsigned long ip, unsigned long parent_ip) |
1000 | { | |
1001 | struct trace_array *tr = &global_trace; | |
1002 | struct trace_array_cpu *data; | |
1003 | unsigned long flags; | |
1004 | long disabled; | |
1005 | int cpu; | |
1006 | ||
60bc0800 | 1007 | if (unlikely(!ftrace_function_enabled)) |
2e0f5761 IM |
1008 | return; |
1009 | ||
ecea656d AS |
1010 | if (skip_trace(ip)) |
1011 | return; | |
1012 | ||
2e0f5761 IM |
1013 | local_irq_save(flags); |
1014 | cpu = raw_smp_processor_id(); | |
1015 | data = tr->data[cpu]; | |
1016 | disabled = atomic_inc_return(&data->disabled); | |
1017 | ||
1018 | if (likely(disabled == 1)) | |
6fb44b71 | 1019 | trace_function(tr, data, ip, parent_ip, flags); |
2e0f5761 IM |
1020 | |
1021 | atomic_dec(&data->disabled); | |
1022 | local_irq_restore(flags); | |
1023 | } | |
1024 | ||
1025 | static struct ftrace_ops trace_ops __read_mostly = | |
1026 | { | |
1027 | .func = function_trace_call, | |
1028 | }; | |
1029 | ||
e309b41d | 1030 | void tracing_start_function_trace(void) |
2e0f5761 | 1031 | { |
60bc0800 | 1032 | ftrace_function_enabled = 0; |
2e0f5761 | 1033 | register_ftrace_function(&trace_ops); |
60bc0800 SR |
1034 | if (tracer_enabled) |
1035 | ftrace_function_enabled = 1; | |
2e0f5761 IM |
1036 | } |
1037 | ||
e309b41d | 1038 | void tracing_stop_function_trace(void) |
2e0f5761 | 1039 | { |
60bc0800 | 1040 | ftrace_function_enabled = 0; |
2e0f5761 IM |
1041 | unregister_ftrace_function(&trace_ops); |
1042 | } | |
1043 | #endif | |
1044 | ||
bc0c38d1 SR |
1045 | enum trace_file_type { |
1046 | TRACE_FILE_LAT_FMT = 1, | |
1047 | }; | |
1048 | ||
dd0e545f | 1049 | /* Return the current entry. */ |
bc0c38d1 | 1050 | static struct trace_entry * |
4c11d7ae SR |
1051 | trace_entry_idx(struct trace_array *tr, struct trace_array_cpu *data, |
1052 | struct trace_iterator *iter, int cpu) | |
bc0c38d1 | 1053 | { |
4c11d7ae SR |
1054 | struct page *page; |
1055 | struct trace_entry *array; | |
bc0c38d1 | 1056 | |
4c11d7ae | 1057 | if (iter->next_idx[cpu] >= tr->entries || |
b3806b43 SR |
1058 | iter->next_idx[cpu] >= data->trace_idx || |
1059 | (data->trace_head == data->trace_tail && | |
1060 | data->trace_head_idx == data->trace_tail_idx)) | |
bc0c38d1 SR |
1061 | return NULL; |
1062 | ||
4c11d7ae | 1063 | if (!iter->next_page[cpu]) { |
93a588f4 SR |
1064 | /* Initialize the iterator for this cpu trace buffer */ |
1065 | WARN_ON(!data->trace_tail); | |
1066 | page = virt_to_page(data->trace_tail); | |
1067 | iter->next_page[cpu] = &page->lru; | |
1068 | iter->next_page_idx[cpu] = data->trace_tail_idx; | |
4c11d7ae | 1069 | } |
bc0c38d1 | 1070 | |
4c11d7ae | 1071 | page = list_entry(iter->next_page[cpu], struct page, lru); |
c7aafc54 IM |
1072 | BUG_ON(&data->trace_pages == &page->lru); |
1073 | ||
4c11d7ae SR |
1074 | array = page_address(page); |
1075 | ||
93a588f4 | 1076 | WARN_ON(iter->next_page_idx[cpu] >= ENTRIES_PER_PAGE); |
4c11d7ae | 1077 | return &array[iter->next_page_idx[cpu]]; |
bc0c38d1 SR |
1078 | } |
1079 | ||
dd0e545f | 1080 | /* Increment the index counter of an iterator by one */ |
5a90f577 | 1081 | static void __trace_iterator_increment(struct trace_iterator *iter, int cpu) |
dd0e545f | 1082 | { |
dd0e545f SR |
1083 | iter->next_idx[cpu]++; |
1084 | iter->next_page_idx[cpu]++; | |
1085 | ||
1086 | if (iter->next_page_idx[cpu] >= ENTRIES_PER_PAGE) { | |
1087 | struct trace_array_cpu *data = iter->tr->data[cpu]; | |
1088 | ||
1089 | iter->next_page_idx[cpu] = 0; | |
1090 | iter->next_page[cpu] = | |
1091 | trace_next_list(data, iter->next_page[cpu]); | |
1092 | } | |
1093 | } | |
1094 | ||
5a90f577 SR |
1095 | static void trace_iterator_increment(struct trace_iterator *iter, int cpu) |
1096 | { | |
1097 | iter->idx++; | |
1098 | __trace_iterator_increment(iter, cpu); | |
1099 | } | |
1100 | ||
e309b41d | 1101 | static struct trace_entry * |
dd0e545f SR |
1102 | trace_entry_next(struct trace_array *tr, struct trace_array_cpu *data, |
1103 | struct trace_iterator *iter, int cpu) | |
1104 | { | |
1105 | struct list_head *next_page; | |
1106 | struct trace_entry *ent; | |
1107 | int idx, next_idx, next_page_idx; | |
1108 | ||
1109 | ent = trace_entry_idx(tr, tr->data[cpu], iter, cpu); | |
1110 | ||
1111 | if (likely(!ent || ent->type != TRACE_CONT)) | |
1112 | return ent; | |
1113 | ||
1114 | /* save the iterator details */ | |
1115 | idx = iter->idx; | |
1116 | next_idx = iter->next_idx[cpu]; | |
1117 | next_page_idx = iter->next_page_idx[cpu]; | |
1118 | next_page = iter->next_page[cpu]; | |
1119 | ||
1120 | /* find a real entry */ | |
1121 | do { | |
5a90f577 | 1122 | __trace_iterator_increment(iter, cpu); |
dd0e545f SR |
1123 | ent = trace_entry_idx(tr, tr->data[cpu], iter, cpu); |
1124 | } while (ent && ent->type != TRACE_CONT); | |
1125 | ||
1126 | /* reset the iterator */ | |
1127 | iter->idx = idx; | |
1128 | iter->next_idx[cpu] = next_idx; | |
1129 | iter->next_page_idx[cpu] = next_page_idx; | |
1130 | iter->next_page[cpu] = next_page; | |
1131 | ||
1132 | return ent; | |
1133 | } | |
1134 | ||
1135 | static struct trace_entry * | |
1136 | __find_next_entry(struct trace_iterator *iter, int *ent_cpu, int inc) | |
bc0c38d1 SR |
1137 | { |
1138 | struct trace_array *tr = iter->tr; | |
1139 | struct trace_entry *ent, *next = NULL; | |
1140 | int next_cpu = -1; | |
1141 | int cpu; | |
1142 | ||
ab46428c | 1143 | for_each_tracing_cpu(cpu) { |
c7aafc54 | 1144 | if (!head_page(tr->data[cpu])) |
bc0c38d1 | 1145 | continue; |
dd0e545f | 1146 | |
4c11d7ae | 1147 | ent = trace_entry_idx(tr, tr->data[cpu], iter, cpu); |
dd0e545f SR |
1148 | |
1149 | if (ent && ent->type == TRACE_CONT) { | |
1150 | struct trace_array_cpu *data = tr->data[cpu]; | |
1151 | ||
1152 | if (!inc) | |
1153 | ent = trace_entry_next(tr, data, iter, cpu); | |
1154 | else { | |
1155 | while (ent && ent->type == TRACE_CONT) { | |
5a90f577 | 1156 | __trace_iterator_increment(iter, cpu); |
dd0e545f SR |
1157 | ent = trace_entry_idx(tr, tr->data[cpu], |
1158 | iter, cpu); | |
1159 | } | |
1160 | } | |
1161 | } | |
1162 | ||
cdd31cd2 IM |
1163 | /* |
1164 | * Pick the entry with the smallest timestamp: | |
1165 | */ | |
2e2ca155 | 1166 | if (ent && (!next || ent->field.t < next->field.t)) { |
bc0c38d1 SR |
1167 | next = ent; |
1168 | next_cpu = cpu; | |
1169 | } | |
1170 | } | |
1171 | ||
1172 | if (ent_cpu) | |
1173 | *ent_cpu = next_cpu; | |
1174 | ||
1175 | return next; | |
1176 | } | |
1177 | ||
dd0e545f SR |
1178 | /* Find the next real entry, without updating the iterator itself */ |
1179 | static struct trace_entry * | |
1180 | find_next_entry(struct trace_iterator *iter, int *ent_cpu) | |
bc0c38d1 | 1181 | { |
dd0e545f SR |
1182 | return __find_next_entry(iter, ent_cpu, 0); |
1183 | } | |
1184 | ||
1185 | /* Find the next real entry, and increment the iterator to the next entry */ | |
1186 | static void *find_next_entry_inc(struct trace_iterator *iter) | |
1187 | { | |
1188 | struct trace_entry *next; | |
1189 | int next_cpu = -1; | |
8c523a9d | 1190 | |
dd0e545f | 1191 | next = __find_next_entry(iter, &next_cpu, 1); |
bc0c38d1 | 1192 | |
dd0e545f SR |
1193 | iter->prev_ent = iter->ent; |
1194 | iter->prev_cpu = iter->cpu; | |
1195 | ||
1196 | iter->ent = next; | |
1197 | iter->cpu = next_cpu; | |
1198 | ||
1199 | if (next) | |
1200 | trace_iterator_increment(iter, iter->cpu); | |
1201 | ||
1202 | return next ? iter : NULL; | |
b3806b43 | 1203 | } |
bc0c38d1 | 1204 | |
e309b41d | 1205 | static void trace_consume(struct trace_iterator *iter) |
b3806b43 SR |
1206 | { |
1207 | struct trace_array_cpu *data = iter->tr->data[iter->cpu]; | |
dd0e545f | 1208 | struct trace_entry *ent; |
b3806b43 | 1209 | |
dd0e545f | 1210 | again: |
b3806b43 SR |
1211 | data->trace_tail_idx++; |
1212 | if (data->trace_tail_idx >= ENTRIES_PER_PAGE) { | |
1213 | data->trace_tail = trace_next_page(data, data->trace_tail); | |
1214 | data->trace_tail_idx = 0; | |
1215 | } | |
4e3c3333 | 1216 | |
b3806b43 SR |
1217 | /* Check if we empty it, then reset the index */ |
1218 | if (data->trace_head == data->trace_tail && | |
1219 | data->trace_head_idx == data->trace_tail_idx) | |
1220 | data->trace_idx = 0; | |
b3806b43 | 1221 | |
dd0e545f SR |
1222 | ent = trace_entry_idx(iter->tr, iter->tr->data[iter->cpu], |
1223 | iter, iter->cpu); | |
1224 | if (ent && ent->type == TRACE_CONT) | |
1225 | goto again; | |
bc0c38d1 SR |
1226 | } |
1227 | ||
e309b41d | 1228 | static void *s_next(struct seq_file *m, void *v, loff_t *pos) |
bc0c38d1 SR |
1229 | { |
1230 | struct trace_iterator *iter = m->private; | |
bc0c38d1 | 1231 | int i = (int)*pos; |
4e3c3333 | 1232 | void *ent; |
bc0c38d1 SR |
1233 | |
1234 | (*pos)++; | |
1235 | ||
1236 | /* can't go backwards */ | |
1237 | if (iter->idx > i) | |
1238 | return NULL; | |
1239 | ||
1240 | if (iter->idx < 0) | |
1241 | ent = find_next_entry_inc(iter); | |
1242 | else | |
1243 | ent = iter; | |
1244 | ||
1245 | while (ent && iter->idx < i) | |
1246 | ent = find_next_entry_inc(iter); | |
1247 | ||
1248 | iter->pos = *pos; | |
1249 | ||
bc0c38d1 SR |
1250 | return ent; |
1251 | } | |
1252 | ||
1253 | static void *s_start(struct seq_file *m, loff_t *pos) | |
1254 | { | |
1255 | struct trace_iterator *iter = m->private; | |
1256 | void *p = NULL; | |
1257 | loff_t l = 0; | |
1258 | int i; | |
1259 | ||
1260 | mutex_lock(&trace_types_lock); | |
1261 | ||
d15f57f2 SR |
1262 | if (!current_trace || current_trace != iter->trace) { |
1263 | mutex_unlock(&trace_types_lock); | |
bc0c38d1 | 1264 | return NULL; |
d15f57f2 | 1265 | } |
bc0c38d1 SR |
1266 | |
1267 | atomic_inc(&trace_record_cmdline_disabled); | |
1268 | ||
1269 | /* let the tracer grab locks here if needed */ | |
1270 | if (current_trace->start) | |
1271 | current_trace->start(iter); | |
1272 | ||
1273 | if (*pos != iter->pos) { | |
1274 | iter->ent = NULL; | |
1275 | iter->cpu = 0; | |
1276 | iter->idx = -1; | |
4e3c3333 IM |
1277 | iter->prev_ent = NULL; |
1278 | iter->prev_cpu = -1; | |
bc0c38d1 | 1279 | |
ab46428c | 1280 | for_each_tracing_cpu(i) { |
bc0c38d1 | 1281 | iter->next_idx[i] = 0; |
4c11d7ae SR |
1282 | iter->next_page[i] = NULL; |
1283 | } | |
bc0c38d1 SR |
1284 | |
1285 | for (p = iter; p && l < *pos; p = s_next(m, p, &l)) | |
1286 | ; | |
1287 | ||
1288 | } else { | |
4c11d7ae | 1289 | l = *pos - 1; |
bc0c38d1 SR |
1290 | p = s_next(m, p, &l); |
1291 | } | |
1292 | ||
1293 | return p; | |
1294 | } | |
1295 | ||
1296 | static void s_stop(struct seq_file *m, void *p) | |
1297 | { | |
1298 | struct trace_iterator *iter = m->private; | |
1299 | ||
1300 | atomic_dec(&trace_record_cmdline_disabled); | |
1301 | ||
1302 | /* let the tracer release locks here if needed */ | |
1303 | if (current_trace && current_trace == iter->trace && iter->trace->stop) | |
1304 | iter->trace->stop(iter); | |
1305 | ||
1306 | mutex_unlock(&trace_types_lock); | |
1307 | } | |
1308 | ||
76094a2c AS |
1309 | #define KRETPROBE_MSG "[unknown/kretprobe'd]" |
1310 | ||
1311 | #ifdef CONFIG_KRETPROBES | |
1312 | static inline int kretprobed(unsigned long addr) | |
1313 | { | |
1314 | return addr == (unsigned long)kretprobe_trampoline; | |
1315 | } | |
1316 | #else | |
1317 | static inline int kretprobed(unsigned long addr) | |
1318 | { | |
1319 | return 0; | |
1320 | } | |
1321 | #endif /* CONFIG_KRETPROBES */ | |
1322 | ||
b3806b43 | 1323 | static int |
214023c3 | 1324 | seq_print_sym_short(struct trace_seq *s, const char *fmt, unsigned long address) |
bc0c38d1 SR |
1325 | { |
1326 | #ifdef CONFIG_KALLSYMS | |
1327 | char str[KSYM_SYMBOL_LEN]; | |
1328 | ||
1329 | kallsyms_lookup(address, NULL, NULL, NULL, str); | |
1330 | ||
b3806b43 | 1331 | return trace_seq_printf(s, fmt, str); |
bc0c38d1 | 1332 | #endif |
b3806b43 | 1333 | return 1; |
bc0c38d1 SR |
1334 | } |
1335 | ||
b3806b43 | 1336 | static int |
214023c3 SR |
1337 | seq_print_sym_offset(struct trace_seq *s, const char *fmt, |
1338 | unsigned long address) | |
bc0c38d1 SR |
1339 | { |
1340 | #ifdef CONFIG_KALLSYMS | |
1341 | char str[KSYM_SYMBOL_LEN]; | |
1342 | ||
1343 | sprint_symbol(str, address); | |
b3806b43 | 1344 | return trace_seq_printf(s, fmt, str); |
bc0c38d1 | 1345 | #endif |
b3806b43 | 1346 | return 1; |
bc0c38d1 SR |
1347 | } |
1348 | ||
1349 | #ifndef CONFIG_64BIT | |
1350 | # define IP_FMT "%08lx" | |
1351 | #else | |
1352 | # define IP_FMT "%016lx" | |
1353 | #endif | |
1354 | ||
e309b41d | 1355 | static int |
214023c3 | 1356 | seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags) |
bc0c38d1 | 1357 | { |
b3806b43 SR |
1358 | int ret; |
1359 | ||
1360 | if (!ip) | |
1361 | return trace_seq_printf(s, "0"); | |
bc0c38d1 SR |
1362 | |
1363 | if (sym_flags & TRACE_ITER_SYM_OFFSET) | |
b3806b43 | 1364 | ret = seq_print_sym_offset(s, "%s", ip); |
bc0c38d1 | 1365 | else |
b3806b43 SR |
1366 | ret = seq_print_sym_short(s, "%s", ip); |
1367 | ||
1368 | if (!ret) | |
1369 | return 0; | |
bc0c38d1 SR |
1370 | |
1371 | if (sym_flags & TRACE_ITER_SYM_ADDR) | |
b3806b43 SR |
1372 | ret = trace_seq_printf(s, " <" IP_FMT ">", ip); |
1373 | return ret; | |
bc0c38d1 SR |
1374 | } |
1375 | ||
e309b41d | 1376 | static void print_lat_help_header(struct seq_file *m) |
bc0c38d1 | 1377 | { |
a6168353 ME |
1378 | seq_puts(m, "# _------=> CPU# \n"); |
1379 | seq_puts(m, "# / _-----=> irqs-off \n"); | |
1380 | seq_puts(m, "# | / _----=> need-resched \n"); | |
1381 | seq_puts(m, "# || / _---=> hardirq/softirq \n"); | |
1382 | seq_puts(m, "# ||| / _--=> preempt-depth \n"); | |
1383 | seq_puts(m, "# |||| / \n"); | |
1384 | seq_puts(m, "# ||||| delay \n"); | |
1385 | seq_puts(m, "# cmd pid ||||| time | caller \n"); | |
1386 | seq_puts(m, "# \\ / ||||| \\ | / \n"); | |
bc0c38d1 SR |
1387 | } |
1388 | ||
e309b41d | 1389 | static void print_func_help_header(struct seq_file *m) |
bc0c38d1 | 1390 | { |
a6168353 ME |
1391 | seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n"); |
1392 | seq_puts(m, "# | | | | |\n"); | |
bc0c38d1 SR |
1393 | } |
1394 | ||
1395 | ||
e309b41d | 1396 | static void |
bc0c38d1 SR |
1397 | print_trace_header(struct seq_file *m, struct trace_iterator *iter) |
1398 | { | |
1399 | unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK); | |
1400 | struct trace_array *tr = iter->tr; | |
1401 | struct trace_array_cpu *data = tr->data[tr->cpu]; | |
1402 | struct tracer *type = current_trace; | |
4c11d7ae SR |
1403 | unsigned long total = 0; |
1404 | unsigned long entries = 0; | |
bc0c38d1 SR |
1405 | int cpu; |
1406 | const char *name = "preemption"; | |
1407 | ||
1408 | if (type) | |
1409 | name = type->name; | |
1410 | ||
ab46428c | 1411 | for_each_tracing_cpu(cpu) { |
c7aafc54 | 1412 | if (head_page(tr->data[cpu])) { |
4c11d7ae SR |
1413 | total += tr->data[cpu]->trace_idx; |
1414 | if (tr->data[cpu]->trace_idx > tr->entries) | |
bc0c38d1 | 1415 | entries += tr->entries; |
4c11d7ae | 1416 | else |
bc0c38d1 SR |
1417 | entries += tr->data[cpu]->trace_idx; |
1418 | } | |
1419 | } | |
1420 | ||
1421 | seq_printf(m, "%s latency trace v1.1.5 on %s\n", | |
1422 | name, UTS_RELEASE); | |
1423 | seq_puts(m, "-----------------------------------" | |
1424 | "---------------------------------\n"); | |
1425 | seq_printf(m, " latency: %lu us, #%lu/%lu, CPU#%d |" | |
1426 | " (M:%s VP:%d, KP:%d, SP:%d HP:%d", | |
57f50be1 | 1427 | nsecs_to_usecs(data->saved_latency), |
bc0c38d1 | 1428 | entries, |
4c11d7ae | 1429 | total, |
bc0c38d1 SR |
1430 | tr->cpu, |
1431 | #if defined(CONFIG_PREEMPT_NONE) | |
1432 | "server", | |
1433 | #elif defined(CONFIG_PREEMPT_VOLUNTARY) | |
1434 | "desktop", | |
b5c21b45 | 1435 | #elif defined(CONFIG_PREEMPT) |
bc0c38d1 SR |
1436 | "preempt", |
1437 | #else | |
1438 | "unknown", | |
1439 | #endif | |
1440 | /* These are reserved for later use */ | |
1441 | 0, 0, 0, 0); | |
1442 | #ifdef CONFIG_SMP | |
1443 | seq_printf(m, " #P:%d)\n", num_online_cpus()); | |
1444 | #else | |
1445 | seq_puts(m, ")\n"); | |
1446 | #endif | |
1447 | seq_puts(m, " -----------------\n"); | |
1448 | seq_printf(m, " | task: %.16s-%d " | |
1449 | "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n", | |
1450 | data->comm, data->pid, data->uid, data->nice, | |
1451 | data->policy, data->rt_priority); | |
1452 | seq_puts(m, " -----------------\n"); | |
1453 | ||
1454 | if (data->critical_start) { | |
1455 | seq_puts(m, " => started at: "); | |
214023c3 SR |
1456 | seq_print_ip_sym(&iter->seq, data->critical_start, sym_flags); |
1457 | trace_print_seq(m, &iter->seq); | |
bc0c38d1 | 1458 | seq_puts(m, "\n => ended at: "); |
214023c3 SR |
1459 | seq_print_ip_sym(&iter->seq, data->critical_end, sym_flags); |
1460 | trace_print_seq(m, &iter->seq); | |
bc0c38d1 SR |
1461 | seq_puts(m, "\n"); |
1462 | } | |
1463 | ||
1464 | seq_puts(m, "\n"); | |
1465 | } | |
1466 | ||
e309b41d | 1467 | static void |
214023c3 | 1468 | lat_print_generic(struct trace_seq *s, struct trace_entry *entry, int cpu) |
bc0c38d1 | 1469 | { |
2e2ca155 | 1470 | struct trace_field *field = &entry->field; |
bc0c38d1 SR |
1471 | int hardirq, softirq; |
1472 | char *comm; | |
1473 | ||
2e2ca155 | 1474 | comm = trace_find_cmdline(field->pid); |
bc0c38d1 | 1475 | |
2e2ca155 | 1476 | trace_seq_printf(s, "%8.8s-%-5d ", comm, field->pid); |
a6168353 | 1477 | trace_seq_printf(s, "%3d", cpu); |
214023c3 | 1478 | trace_seq_printf(s, "%c%c", |
2e2ca155 SR |
1479 | (field->flags & TRACE_FLAG_IRQS_OFF) ? 'd' : '.', |
1480 | ((field->flags & TRACE_FLAG_NEED_RESCHED) ? 'N' : '.')); | |
bc0c38d1 | 1481 | |
2e2ca155 SR |
1482 | hardirq = field->flags & TRACE_FLAG_HARDIRQ; |
1483 | softirq = field->flags & TRACE_FLAG_SOFTIRQ; | |
afc2abc0 | 1484 | if (hardirq && softirq) { |
214023c3 | 1485 | trace_seq_putc(s, 'H'); |
afc2abc0 IM |
1486 | } else { |
1487 | if (hardirq) { | |
214023c3 | 1488 | trace_seq_putc(s, 'h'); |
afc2abc0 | 1489 | } else { |
bc0c38d1 | 1490 | if (softirq) |
214023c3 | 1491 | trace_seq_putc(s, 's'); |
bc0c38d1 | 1492 | else |
214023c3 | 1493 | trace_seq_putc(s, '.'); |
bc0c38d1 SR |
1494 | } |
1495 | } | |
1496 | ||
2e2ca155 SR |
1497 | if (field->preempt_count) |
1498 | trace_seq_printf(s, "%x", field->preempt_count); | |
bc0c38d1 | 1499 | else |
214023c3 | 1500 | trace_seq_puts(s, "."); |
bc0c38d1 SR |
1501 | } |
1502 | ||
1503 | unsigned long preempt_mark_thresh = 100; | |
1504 | ||
e309b41d | 1505 | static void |
214023c3 | 1506 | lat_print_timestamp(struct trace_seq *s, unsigned long long abs_usecs, |
bc0c38d1 SR |
1507 | unsigned long rel_usecs) |
1508 | { | |
214023c3 | 1509 | trace_seq_printf(s, " %4lldus", abs_usecs); |
bc0c38d1 | 1510 | if (rel_usecs > preempt_mark_thresh) |
214023c3 | 1511 | trace_seq_puts(s, "!: "); |
bc0c38d1 | 1512 | else if (rel_usecs > 1) |
214023c3 | 1513 | trace_seq_puts(s, "+: "); |
bc0c38d1 | 1514 | else |
214023c3 | 1515 | trace_seq_puts(s, " : "); |
bc0c38d1 SR |
1516 | } |
1517 | ||
1518 | static const char state_to_char[] = TASK_STATE_TO_CHAR_STR; | |
1519 | ||
dd0e545f SR |
1520 | static void |
1521 | trace_seq_print_cont(struct trace_seq *s, struct trace_iterator *iter) | |
1522 | { | |
1523 | struct trace_array *tr = iter->tr; | |
1524 | struct trace_array_cpu *data = tr->data[iter->cpu]; | |
1525 | struct trace_entry *ent; | |
1526 | ||
1527 | ent = trace_entry_idx(tr, data, iter, iter->cpu); | |
1528 | if (!ent || ent->type != TRACE_CONT) { | |
1529 | trace_seq_putc(s, '\n'); | |
1530 | return; | |
1531 | } | |
1532 | ||
1533 | do { | |
1534 | trace_seq_printf(s, "%s", ent->cont.buf); | |
5a90f577 | 1535 | __trace_iterator_increment(iter, iter->cpu); |
dd0e545f SR |
1536 | ent = trace_entry_idx(tr, data, iter, iter->cpu); |
1537 | } while (ent && ent->type == TRACE_CONT); | |
1538 | } | |
1539 | ||
e309b41d | 1540 | static int |
214023c3 | 1541 | print_lat_fmt(struct trace_iterator *iter, unsigned int trace_idx, int cpu) |
bc0c38d1 | 1542 | { |
214023c3 | 1543 | struct trace_seq *s = &iter->seq; |
bc0c38d1 SR |
1544 | unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK); |
1545 | struct trace_entry *next_entry = find_next_entry(iter, NULL); | |
1546 | unsigned long verbose = (trace_flags & TRACE_ITER_VERBOSE); | |
1547 | struct trace_entry *entry = iter->ent; | |
2e2ca155 | 1548 | struct trace_field *field = &entry->field; |
bc0c38d1 SR |
1549 | unsigned long abs_usecs; |
1550 | unsigned long rel_usecs; | |
1551 | char *comm; | |
bac524d3 | 1552 | int S, T; |
86387f7e | 1553 | int i; |
d17d9691 | 1554 | unsigned state; |
bc0c38d1 SR |
1555 | |
1556 | if (!next_entry) | |
1557 | next_entry = entry; | |
dd0e545f SR |
1558 | |
1559 | if (entry->type == TRACE_CONT) | |
1560 | return 1; | |
1561 | ||
2e2ca155 SR |
1562 | rel_usecs = ns2usecs(next_entry->field.t - entry->field.t); |
1563 | abs_usecs = ns2usecs(entry->field.t - iter->tr->time_start); | |
bc0c38d1 SR |
1564 | |
1565 | if (verbose) { | |
2e2ca155 | 1566 | comm = trace_find_cmdline(field->pid); |
a6168353 | 1567 | trace_seq_printf(s, "%16s %5d %3d %d %08x %08x [%08lx]" |
214023c3 SR |
1568 | " %ld.%03ldms (+%ld.%03ldms): ", |
1569 | comm, | |
2e2ca155 SR |
1570 | field->pid, cpu, field->flags, |
1571 | field->preempt_count, trace_idx, | |
1572 | ns2usecs(field->t), | |
214023c3 SR |
1573 | abs_usecs/1000, |
1574 | abs_usecs % 1000, rel_usecs/1000, | |
1575 | rel_usecs % 1000); | |
bc0c38d1 | 1576 | } else { |
f29c73fe IM |
1577 | lat_print_generic(s, entry, cpu); |
1578 | lat_print_timestamp(s, abs_usecs, rel_usecs); | |
bc0c38d1 SR |
1579 | } |
1580 | switch (entry->type) { | |
1581 | case TRACE_FN: | |
2e2ca155 | 1582 | seq_print_ip_sym(s, field->fn.ip, sym_flags); |
214023c3 | 1583 | trace_seq_puts(s, " ("); |
2e2ca155 | 1584 | if (kretprobed(field->fn.parent_ip)) |
76094a2c AS |
1585 | trace_seq_puts(s, KRETPROBE_MSG); |
1586 | else | |
2e2ca155 | 1587 | seq_print_ip_sym(s, field->fn.parent_ip, sym_flags); |
214023c3 | 1588 | trace_seq_puts(s, ")\n"); |
bc0c38d1 SR |
1589 | break; |
1590 | case TRACE_CTX: | |
57422797 | 1591 | case TRACE_WAKE: |
2e2ca155 SR |
1592 | T = field->ctx.next_state < sizeof(state_to_char) ? |
1593 | state_to_char[field->ctx.next_state] : 'X'; | |
bac524d3 | 1594 | |
2e2ca155 SR |
1595 | state = field->ctx.prev_state ? |
1596 | __ffs(field->ctx.prev_state) + 1 : 0; | |
d17d9691 | 1597 | S = state < sizeof(state_to_char) - 1 ? state_to_char[state] : 'X'; |
2e2ca155 | 1598 | comm = trace_find_cmdline(field->ctx.next_pid); |
80b5e940 | 1599 | trace_seq_printf(s, " %5d:%3d:%c %s [%03d] %5d:%3d:%c %s\n", |
2e2ca155 SR |
1600 | field->ctx.prev_pid, |
1601 | field->ctx.prev_prio, | |
57422797 | 1602 | S, entry->type == TRACE_CTX ? "==>" : " +", |
80b5e940 | 1603 | field->ctx.next_cpu, |
2e2ca155 SR |
1604 | field->ctx.next_pid, |
1605 | field->ctx.next_prio, | |
bac524d3 | 1606 | T, comm); |
bc0c38d1 | 1607 | break; |
f0a920d5 | 1608 | case TRACE_SPECIAL: |
88a4216c | 1609 | trace_seq_printf(s, "# %ld %ld %ld\n", |
2e2ca155 SR |
1610 | field->special.arg1, |
1611 | field->special.arg2, | |
1612 | field->special.arg3); | |
f0a920d5 | 1613 | break; |
86387f7e IM |
1614 | case TRACE_STACK: |
1615 | for (i = 0; i < FTRACE_STACK_ENTRIES; i++) { | |
1616 | if (i) | |
1617 | trace_seq_puts(s, " <= "); | |
2e2ca155 | 1618 | seq_print_ip_sym(s, field->stack.caller[i], sym_flags); |
86387f7e IM |
1619 | } |
1620 | trace_seq_puts(s, "\n"); | |
1621 | break; | |
dd0e545f SR |
1622 | case TRACE_PRINT: |
1623 | seq_print_ip_sym(s, field->print.ip, sym_flags); | |
1624 | trace_seq_printf(s, ": %s", field->print.buf); | |
652567aa | 1625 | if (field->flags & TRACE_FLAG_CONT) |
dd0e545f SR |
1626 | trace_seq_print_cont(s, iter); |
1627 | break; | |
89b2f978 | 1628 | default: |
214023c3 | 1629 | trace_seq_printf(s, "Unknown type %d\n", entry->type); |
bc0c38d1 | 1630 | } |
f9896bf3 | 1631 | return 1; |
bc0c38d1 SR |
1632 | } |
1633 | ||
e309b41d | 1634 | static int print_trace_fmt(struct trace_iterator *iter) |
bc0c38d1 | 1635 | { |
214023c3 | 1636 | struct trace_seq *s = &iter->seq; |
bc0c38d1 | 1637 | unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK); |
4e3c3333 | 1638 | struct trace_entry *entry; |
2e2ca155 | 1639 | struct trace_field *field; |
bc0c38d1 SR |
1640 | unsigned long usec_rem; |
1641 | unsigned long long t; | |
1642 | unsigned long secs; | |
1643 | char *comm; | |
b3806b43 | 1644 | int ret; |
bac524d3 | 1645 | int S, T; |
86387f7e | 1646 | int i; |
bc0c38d1 | 1647 | |
4e3c3333 | 1648 | entry = iter->ent; |
dd0e545f SR |
1649 | |
1650 | if (entry->type == TRACE_CONT) | |
1651 | return 1; | |
1652 | ||
2e2ca155 | 1653 | field = &entry->field; |
4e3c3333 | 1654 | |
2e2ca155 | 1655 | comm = trace_find_cmdline(iter->ent->field.pid); |
bc0c38d1 | 1656 | |
2e2ca155 | 1657 | t = ns2usecs(field->t); |
bc0c38d1 SR |
1658 | usec_rem = do_div(t, 1000000ULL); |
1659 | secs = (unsigned long)t; | |
1660 | ||
2e2ca155 | 1661 | ret = trace_seq_printf(s, "%16s-%-5d ", comm, field->pid); |
f29c73fe IM |
1662 | if (!ret) |
1663 | return 0; | |
a6168353 | 1664 | ret = trace_seq_printf(s, "[%03d] ", iter->cpu); |
f29c73fe IM |
1665 | if (!ret) |
1666 | return 0; | |
1667 | ret = trace_seq_printf(s, "%5lu.%06lu: ", secs, usec_rem); | |
1668 | if (!ret) | |
1669 | return 0; | |
bc0c38d1 SR |
1670 | |
1671 | switch (entry->type) { | |
1672 | case TRACE_FN: | |
2e2ca155 | 1673 | ret = seq_print_ip_sym(s, field->fn.ip, sym_flags); |
b3806b43 SR |
1674 | if (!ret) |
1675 | return 0; | |
bc0c38d1 | 1676 | if ((sym_flags & TRACE_ITER_PRINT_PARENT) && |
2e2ca155 | 1677 | field->fn.parent_ip) { |
b3806b43 SR |
1678 | ret = trace_seq_printf(s, " <-"); |
1679 | if (!ret) | |
1680 | return 0; | |
2e2ca155 | 1681 | if (kretprobed(field->fn.parent_ip)) |
76094a2c AS |
1682 | ret = trace_seq_puts(s, KRETPROBE_MSG); |
1683 | else | |
2e2ca155 SR |
1684 | ret = seq_print_ip_sym(s, |
1685 | field->fn.parent_ip, | |
76094a2c | 1686 | sym_flags); |
b3806b43 SR |
1687 | if (!ret) |
1688 | return 0; | |
bc0c38d1 | 1689 | } |
b3806b43 SR |
1690 | ret = trace_seq_printf(s, "\n"); |
1691 | if (!ret) | |
1692 | return 0; | |
bc0c38d1 SR |
1693 | break; |
1694 | case TRACE_CTX: | |
57422797 | 1695 | case TRACE_WAKE: |
2e2ca155 SR |
1696 | S = field->ctx.prev_state < sizeof(state_to_char) ? |
1697 | state_to_char[field->ctx.prev_state] : 'X'; | |
1698 | T = field->ctx.next_state < sizeof(state_to_char) ? | |
1699 | state_to_char[field->ctx.next_state] : 'X'; | |
80b5e940 | 1700 | ret = trace_seq_printf(s, " %5d:%3d:%c %s [%03d] %5d:%3d:%c\n", |
2e2ca155 SR |
1701 | field->ctx.prev_pid, |
1702 | field->ctx.prev_prio, | |
b3806b43 | 1703 | S, |
57422797 | 1704 | entry->type == TRACE_CTX ? "==>" : " +", |
80b5e940 | 1705 | field->ctx.next_cpu, |
2e2ca155 SR |
1706 | field->ctx.next_pid, |
1707 | field->ctx.next_prio, | |
bac524d3 | 1708 | T); |
b3806b43 SR |
1709 | if (!ret) |
1710 | return 0; | |
bc0c38d1 | 1711 | break; |
f0a920d5 | 1712 | case TRACE_SPECIAL: |
88a4216c | 1713 | ret = trace_seq_printf(s, "# %ld %ld %ld\n", |
2e2ca155 SR |
1714 | field->special.arg1, |
1715 | field->special.arg2, | |
1716 | field->special.arg3); | |
f0a920d5 IM |
1717 | if (!ret) |
1718 | return 0; | |
1719 | break; | |
86387f7e IM |
1720 | case TRACE_STACK: |
1721 | for (i = 0; i < FTRACE_STACK_ENTRIES; i++) { | |
1722 | if (i) { | |
1723 | ret = trace_seq_puts(s, " <= "); | |
1724 | if (!ret) | |
1725 | return 0; | |
1726 | } | |
2e2ca155 | 1727 | ret = seq_print_ip_sym(s, field->stack.caller[i], |
86387f7e IM |
1728 | sym_flags); |
1729 | if (!ret) | |
1730 | return 0; | |
1731 | } | |
1732 | ret = trace_seq_puts(s, "\n"); | |
1733 | if (!ret) | |
1734 | return 0; | |
1735 | break; | |
dd0e545f SR |
1736 | case TRACE_PRINT: |
1737 | seq_print_ip_sym(s, field->print.ip, sym_flags); | |
1738 | trace_seq_printf(s, ": %s", field->print.buf); | |
652567aa | 1739 | if (field->flags & TRACE_FLAG_CONT) |
dd0e545f SR |
1740 | trace_seq_print_cont(s, iter); |
1741 | break; | |
bc0c38d1 | 1742 | } |
b3806b43 | 1743 | return 1; |
bc0c38d1 SR |
1744 | } |
1745 | ||
e309b41d | 1746 | static int print_raw_fmt(struct trace_iterator *iter) |
f9896bf3 IM |
1747 | { |
1748 | struct trace_seq *s = &iter->seq; | |
1749 | struct trace_entry *entry; | |
2e2ca155 | 1750 | struct trace_field *field; |
f9896bf3 | 1751 | int ret; |
bac524d3 | 1752 | int S, T; |
f9896bf3 IM |
1753 | |
1754 | entry = iter->ent; | |
dd0e545f SR |
1755 | |
1756 | if (entry->type == TRACE_CONT) | |
1757 | return 1; | |
1758 | ||
2e2ca155 | 1759 | field = &entry->field; |
f9896bf3 IM |
1760 | |
1761 | ret = trace_seq_printf(s, "%d %d %llu ", | |
2e2ca155 | 1762 | field->pid, iter->cpu, field->t); |
f9896bf3 IM |
1763 | if (!ret) |
1764 | return 0; | |
1765 | ||
1766 | switch (entry->type) { | |
1767 | case TRACE_FN: | |
1768 | ret = trace_seq_printf(s, "%x %x\n", | |
2e2ca155 SR |
1769 | field->fn.ip, |
1770 | field->fn.parent_ip); | |
f9896bf3 IM |
1771 | if (!ret) |
1772 | return 0; | |
1773 | break; | |
1774 | case TRACE_CTX: | |
57422797 | 1775 | case TRACE_WAKE: |
2e2ca155 SR |
1776 | S = field->ctx.prev_state < sizeof(state_to_char) ? |
1777 | state_to_char[field->ctx.prev_state] : 'X'; | |
1778 | T = field->ctx.next_state < sizeof(state_to_char) ? | |
1779 | state_to_char[field->ctx.next_state] : 'X'; | |
57422797 IM |
1780 | if (entry->type == TRACE_WAKE) |
1781 | S = '+'; | |
80b5e940 | 1782 | ret = trace_seq_printf(s, "%d %d %c %d %d %d %c\n", |
2e2ca155 SR |
1783 | field->ctx.prev_pid, |
1784 | field->ctx.prev_prio, | |
f9896bf3 | 1785 | S, |
80b5e940 | 1786 | field->ctx.next_cpu, |
2e2ca155 SR |
1787 | field->ctx.next_pid, |
1788 | field->ctx.next_prio, | |
bac524d3 | 1789 | T); |
f9896bf3 IM |
1790 | if (!ret) |
1791 | return 0; | |
1792 | break; | |
f0a920d5 | 1793 | case TRACE_SPECIAL: |
86387f7e | 1794 | case TRACE_STACK: |
88a4216c | 1795 | ret = trace_seq_printf(s, "# %ld %ld %ld\n", |
2e2ca155 SR |
1796 | field->special.arg1, |
1797 | field->special.arg2, | |
1798 | field->special.arg3); | |
f0a920d5 IM |
1799 | if (!ret) |
1800 | return 0; | |
1801 | break; | |
dd0e545f SR |
1802 | case TRACE_PRINT: |
1803 | trace_seq_printf(s, "# %lx %s", | |
1804 | field->print.ip, field->print.buf); | |
652567aa | 1805 | if (field->flags & TRACE_FLAG_CONT) |
dd0e545f SR |
1806 | trace_seq_print_cont(s, iter); |
1807 | break; | |
f9896bf3 IM |
1808 | } |
1809 | return 1; | |
1810 | } | |
1811 | ||
cb0f12aa IM |
1812 | #define SEQ_PUT_FIELD_RET(s, x) \ |
1813 | do { \ | |
1814 | if (!trace_seq_putmem(s, &(x), sizeof(x))) \ | |
1815 | return 0; \ | |
1816 | } while (0) | |
1817 | ||
5e3ca0ec IM |
1818 | #define SEQ_PUT_HEX_FIELD_RET(s, x) \ |
1819 | do { \ | |
1820 | if (!trace_seq_putmem_hex(s, &(x), sizeof(x))) \ | |
1821 | return 0; \ | |
1822 | } while (0) | |
1823 | ||
e309b41d | 1824 | static int print_hex_fmt(struct trace_iterator *iter) |
5e3ca0ec IM |
1825 | { |
1826 | struct trace_seq *s = &iter->seq; | |
1827 | unsigned char newline = '\n'; | |
1828 | struct trace_entry *entry; | |
2e2ca155 | 1829 | struct trace_field *field; |
bac524d3 | 1830 | int S, T; |
5e3ca0ec IM |
1831 | |
1832 | entry = iter->ent; | |
dd0e545f SR |
1833 | |
1834 | if (entry->type == TRACE_CONT) | |
1835 | return 1; | |
1836 | ||
2e2ca155 | 1837 | field = &entry->field; |
5e3ca0ec | 1838 | |
2e2ca155 | 1839 | SEQ_PUT_HEX_FIELD_RET(s, field->pid); |
5e3ca0ec | 1840 | SEQ_PUT_HEX_FIELD_RET(s, iter->cpu); |
2e2ca155 | 1841 | SEQ_PUT_HEX_FIELD_RET(s, field->t); |
5e3ca0ec IM |
1842 | |
1843 | switch (entry->type) { | |
1844 | case TRACE_FN: | |
2e2ca155 SR |
1845 | SEQ_PUT_HEX_FIELD_RET(s, field->fn.ip); |
1846 | SEQ_PUT_HEX_FIELD_RET(s, field->fn.parent_ip); | |
5e3ca0ec IM |
1847 | break; |
1848 | case TRACE_CTX: | |
57422797 | 1849 | case TRACE_WAKE: |
2e2ca155 SR |
1850 | S = field->ctx.prev_state < sizeof(state_to_char) ? |
1851 | state_to_char[field->ctx.prev_state] : 'X'; | |
1852 | T = field->ctx.next_state < sizeof(state_to_char) ? | |
1853 | state_to_char[field->ctx.next_state] : 'X'; | |
57422797 IM |
1854 | if (entry->type == TRACE_WAKE) |
1855 | S = '+'; | |
2e2ca155 SR |
1856 | SEQ_PUT_HEX_FIELD_RET(s, field->ctx.prev_pid); |
1857 | SEQ_PUT_HEX_FIELD_RET(s, field->ctx.prev_prio); | |
5e3ca0ec | 1858 | SEQ_PUT_HEX_FIELD_RET(s, S); |
80b5e940 | 1859 | SEQ_PUT_HEX_FIELD_RET(s, field->ctx.next_cpu); |
2e2ca155 SR |
1860 | SEQ_PUT_HEX_FIELD_RET(s, field->ctx.next_pid); |
1861 | SEQ_PUT_HEX_FIELD_RET(s, field->ctx.next_prio); | |
bac524d3 | 1862 | SEQ_PUT_HEX_FIELD_RET(s, T); |
5e3ca0ec IM |
1863 | break; |
1864 | case TRACE_SPECIAL: | |
86387f7e | 1865 | case TRACE_STACK: |
2e2ca155 SR |
1866 | SEQ_PUT_HEX_FIELD_RET(s, field->special.arg1); |
1867 | SEQ_PUT_HEX_FIELD_RET(s, field->special.arg2); | |
1868 | SEQ_PUT_HEX_FIELD_RET(s, field->special.arg3); | |
5e3ca0ec IM |
1869 | break; |
1870 | } | |
1871 | SEQ_PUT_FIELD_RET(s, newline); | |
1872 | ||
1873 | return 1; | |
1874 | } | |
1875 | ||
e309b41d | 1876 | static int print_bin_fmt(struct trace_iterator *iter) |
cb0f12aa IM |
1877 | { |
1878 | struct trace_seq *s = &iter->seq; | |
1879 | struct trace_entry *entry; | |
2e2ca155 | 1880 | struct trace_field *field; |
cb0f12aa IM |
1881 | |
1882 | entry = iter->ent; | |
dd0e545f SR |
1883 | |
1884 | if (entry->type == TRACE_CONT) | |
1885 | return 1; | |
1886 | ||
2e2ca155 | 1887 | field = &entry->field; |
cb0f12aa | 1888 | |
2e2ca155 SR |
1889 | SEQ_PUT_FIELD_RET(s, field->pid); |
1890 | SEQ_PUT_FIELD_RET(s, field->cpu); | |
1891 | SEQ_PUT_FIELD_RET(s, field->t); | |
cb0f12aa IM |
1892 | |
1893 | switch (entry->type) { | |
1894 | case TRACE_FN: | |
2e2ca155 SR |
1895 | SEQ_PUT_FIELD_RET(s, field->fn.ip); |
1896 | SEQ_PUT_FIELD_RET(s, field->fn.parent_ip); | |
cb0f12aa IM |
1897 | break; |
1898 | case TRACE_CTX: | |
2e2ca155 SR |
1899 | SEQ_PUT_FIELD_RET(s, field->ctx.prev_pid); |
1900 | SEQ_PUT_FIELD_RET(s, field->ctx.prev_prio); | |
1901 | SEQ_PUT_FIELD_RET(s, field->ctx.prev_state); | |
1902 | SEQ_PUT_FIELD_RET(s, field->ctx.next_pid); | |
1903 | SEQ_PUT_FIELD_RET(s, field->ctx.next_prio); | |
1904 | SEQ_PUT_FIELD_RET(s, field->ctx.next_state); | |
cb0f12aa | 1905 | break; |
f0a920d5 | 1906 | case TRACE_SPECIAL: |
86387f7e | 1907 | case TRACE_STACK: |
2e2ca155 SR |
1908 | SEQ_PUT_FIELD_RET(s, field->special.arg1); |
1909 | SEQ_PUT_FIELD_RET(s, field->special.arg2); | |
1910 | SEQ_PUT_FIELD_RET(s, field->special.arg3); | |
f0a920d5 | 1911 | break; |
cb0f12aa IM |
1912 | } |
1913 | return 1; | |
1914 | } | |
1915 | ||
bc0c38d1 SR |
1916 | static int trace_empty(struct trace_iterator *iter) |
1917 | { | |
1918 | struct trace_array_cpu *data; | |
1919 | int cpu; | |
1920 | ||
ab46428c | 1921 | for_each_tracing_cpu(cpu) { |
bc0c38d1 SR |
1922 | data = iter->tr->data[cpu]; |
1923 | ||
b3806b43 SR |
1924 | if (head_page(data) && data->trace_idx && |
1925 | (data->trace_tail != data->trace_head || | |
1926 | data->trace_tail_idx != data->trace_head_idx)) | |
bc0c38d1 SR |
1927 | return 0; |
1928 | } | |
1929 | return 1; | |
1930 | } | |
1931 | ||
f9896bf3 IM |
1932 | static int print_trace_line(struct trace_iterator *iter) |
1933 | { | |
72829bc3 TG |
1934 | if (iter->trace && iter->trace->print_line) |
1935 | return iter->trace->print_line(iter); | |
1936 | ||
cb0f12aa IM |
1937 | if (trace_flags & TRACE_ITER_BIN) |
1938 | return print_bin_fmt(iter); | |
1939 | ||
5e3ca0ec IM |
1940 | if (trace_flags & TRACE_ITER_HEX) |
1941 | return print_hex_fmt(iter); | |
1942 | ||
f9896bf3 IM |
1943 | if (trace_flags & TRACE_ITER_RAW) |
1944 | return print_raw_fmt(iter); | |
1945 | ||
1946 | if (iter->iter_flags & TRACE_FILE_LAT_FMT) | |
1947 | return print_lat_fmt(iter, iter->idx, iter->cpu); | |
1948 | ||
1949 | return print_trace_fmt(iter); | |
1950 | } | |
1951 | ||
bc0c38d1 SR |
1952 | static int s_show(struct seq_file *m, void *v) |
1953 | { | |
1954 | struct trace_iterator *iter = v; | |
1955 | ||
1956 | if (iter->ent == NULL) { | |
1957 | if (iter->tr) { | |
1958 | seq_printf(m, "# tracer: %s\n", iter->trace->name); | |
1959 | seq_puts(m, "#\n"); | |
1960 | } | |
1961 | if (iter->iter_flags & TRACE_FILE_LAT_FMT) { | |
1962 | /* print nothing if the buffers are empty */ | |
1963 | if (trace_empty(iter)) | |
1964 | return 0; | |
1965 | print_trace_header(m, iter); | |
1966 | if (!(trace_flags & TRACE_ITER_VERBOSE)) | |
1967 | print_lat_help_header(m); | |
1968 | } else { | |
1969 | if (!(trace_flags & TRACE_ITER_VERBOSE)) | |
1970 | print_func_help_header(m); | |
1971 | } | |
1972 | } else { | |
f9896bf3 | 1973 | print_trace_line(iter); |
214023c3 | 1974 | trace_print_seq(m, &iter->seq); |
bc0c38d1 SR |
1975 | } |
1976 | ||
1977 | return 0; | |
1978 | } | |
1979 | ||
1980 | static struct seq_operations tracer_seq_ops = { | |
4bf39a94 IM |
1981 | .start = s_start, |
1982 | .next = s_next, | |
1983 | .stop = s_stop, | |
1984 | .show = s_show, | |
bc0c38d1 SR |
1985 | }; |
1986 | ||
e309b41d | 1987 | static struct trace_iterator * |
bc0c38d1 SR |
1988 | __tracing_open(struct inode *inode, struct file *file, int *ret) |
1989 | { | |
1990 | struct trace_iterator *iter; | |
1991 | ||
60a11774 SR |
1992 | if (tracing_disabled) { |
1993 | *ret = -ENODEV; | |
1994 | return NULL; | |
1995 | } | |
1996 | ||
bc0c38d1 SR |
1997 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); |
1998 | if (!iter) { | |
1999 | *ret = -ENOMEM; | |
2000 | goto out; | |
2001 | } | |
2002 | ||
2003 | mutex_lock(&trace_types_lock); | |
2004 | if (current_trace && current_trace->print_max) | |
2005 | iter->tr = &max_tr; | |
2006 | else | |
2007 | iter->tr = inode->i_private; | |
2008 | iter->trace = current_trace; | |
2009 | iter->pos = -1; | |
2010 | ||
2011 | /* TODO stop tracer */ | |
2012 | *ret = seq_open(file, &tracer_seq_ops); | |
2013 | if (!*ret) { | |
2014 | struct seq_file *m = file->private_data; | |
2015 | m->private = iter; | |
2016 | ||
2017 | /* stop the trace while dumping */ | |
60bc0800 | 2018 | if (iter->tr->ctrl) { |
bc0c38d1 | 2019 | tracer_enabled = 0; |
60bc0800 SR |
2020 | ftrace_function_enabled = 0; |
2021 | } | |
bc0c38d1 SR |
2022 | |
2023 | if (iter->trace && iter->trace->open) | |
2024 | iter->trace->open(iter); | |
2025 | } else { | |
2026 | kfree(iter); | |
2027 | iter = NULL; | |
2028 | } | |
2029 | mutex_unlock(&trace_types_lock); | |
2030 | ||
2031 | out: | |
2032 | return iter; | |
2033 | } | |
2034 | ||
2035 | int tracing_open_generic(struct inode *inode, struct file *filp) | |
2036 | { | |
60a11774 SR |
2037 | if (tracing_disabled) |
2038 | return -ENODEV; | |
2039 | ||
bc0c38d1 SR |
2040 | filp->private_data = inode->i_private; |
2041 | return 0; | |
2042 | } | |
2043 | ||
2044 | int tracing_release(struct inode *inode, struct file *file) | |
2045 | { | |
2046 | struct seq_file *m = (struct seq_file *)file->private_data; | |
2047 | struct trace_iterator *iter = m->private; | |
2048 | ||
2049 | mutex_lock(&trace_types_lock); | |
2050 | if (iter->trace && iter->trace->close) | |
2051 | iter->trace->close(iter); | |
2052 | ||
2053 | /* reenable tracing if it was previously enabled */ | |
60bc0800 | 2054 | if (iter->tr->ctrl) { |
bc0c38d1 | 2055 | tracer_enabled = 1; |
60bc0800 SR |
2056 | /* |
2057 | * It is safe to enable function tracing even if it | |
2058 | * isn't used | |
2059 | */ | |
2060 | ftrace_function_enabled = 1; | |
2061 | } | |
bc0c38d1 SR |
2062 | mutex_unlock(&trace_types_lock); |
2063 | ||
2064 | seq_release(inode, file); | |
2065 | kfree(iter); | |
2066 | return 0; | |
2067 | } | |
2068 | ||
2069 | static int tracing_open(struct inode *inode, struct file *file) | |
2070 | { | |
2071 | int ret; | |
2072 | ||
2073 | __tracing_open(inode, file, &ret); | |
2074 | ||
2075 | return ret; | |
2076 | } | |
2077 | ||
2078 | static int tracing_lt_open(struct inode *inode, struct file *file) | |
2079 | { | |
2080 | struct trace_iterator *iter; | |
2081 | int ret; | |
2082 | ||
2083 | iter = __tracing_open(inode, file, &ret); | |
2084 | ||
2085 | if (!ret) | |
2086 | iter->iter_flags |= TRACE_FILE_LAT_FMT; | |
2087 | ||
2088 | return ret; | |
2089 | } | |
2090 | ||
2091 | ||
e309b41d | 2092 | static void * |
bc0c38d1 SR |
2093 | t_next(struct seq_file *m, void *v, loff_t *pos) |
2094 | { | |
2095 | struct tracer *t = m->private; | |
2096 | ||
2097 | (*pos)++; | |
2098 | ||
2099 | if (t) | |
2100 | t = t->next; | |
2101 | ||
2102 | m->private = t; | |
2103 | ||
2104 | return t; | |
2105 | } | |
2106 | ||
2107 | static void *t_start(struct seq_file *m, loff_t *pos) | |
2108 | { | |
2109 | struct tracer *t = m->private; | |
2110 | loff_t l = 0; | |
2111 | ||
2112 | mutex_lock(&trace_types_lock); | |
2113 | for (; t && l < *pos; t = t_next(m, t, &l)) | |
2114 | ; | |
2115 | ||
2116 | return t; | |
2117 | } | |
2118 | ||
2119 | static void t_stop(struct seq_file *m, void *p) | |
2120 | { | |
2121 | mutex_unlock(&trace_types_lock); | |
2122 | } | |
2123 | ||
2124 | static int t_show(struct seq_file *m, void *v) | |
2125 | { | |
2126 | struct tracer *t = v; | |
2127 | ||
2128 | if (!t) | |
2129 | return 0; | |
2130 | ||
2131 | seq_printf(m, "%s", t->name); | |
2132 | if (t->next) | |
2133 | seq_putc(m, ' '); | |
2134 | else | |
2135 | seq_putc(m, '\n'); | |
2136 | ||
2137 | return 0; | |
2138 | } | |
2139 | ||
2140 | static struct seq_operations show_traces_seq_ops = { | |
4bf39a94 IM |
2141 | .start = t_start, |
2142 | .next = t_next, | |
2143 | .stop = t_stop, | |
2144 | .show = t_show, | |
bc0c38d1 SR |
2145 | }; |
2146 | ||
2147 | static int show_traces_open(struct inode *inode, struct file *file) | |
2148 | { | |
2149 | int ret; | |
2150 | ||
60a11774 SR |
2151 | if (tracing_disabled) |
2152 | return -ENODEV; | |
2153 | ||
bc0c38d1 SR |
2154 | ret = seq_open(file, &show_traces_seq_ops); |
2155 | if (!ret) { | |
2156 | struct seq_file *m = file->private_data; | |
2157 | m->private = trace_types; | |
2158 | } | |
2159 | ||
2160 | return ret; | |
2161 | } | |
2162 | ||
2163 | static struct file_operations tracing_fops = { | |
4bf39a94 IM |
2164 | .open = tracing_open, |
2165 | .read = seq_read, | |
2166 | .llseek = seq_lseek, | |
2167 | .release = tracing_release, | |
bc0c38d1 SR |
2168 | }; |
2169 | ||
2170 | static struct file_operations tracing_lt_fops = { | |
4bf39a94 IM |
2171 | .open = tracing_lt_open, |
2172 | .read = seq_read, | |
2173 | .llseek = seq_lseek, | |
2174 | .release = tracing_release, | |
bc0c38d1 SR |
2175 | }; |
2176 | ||
2177 | static struct file_operations show_traces_fops = { | |
c7078de1 IM |
2178 | .open = show_traces_open, |
2179 | .read = seq_read, | |
2180 | .release = seq_release, | |
2181 | }; | |
2182 | ||
36dfe925 IM |
2183 | /* |
2184 | * Only trace on a CPU if the bitmask is set: | |
2185 | */ | |
2186 | static cpumask_t tracing_cpumask = CPU_MASK_ALL; | |
2187 | ||
2188 | /* | |
2189 | * When tracing/tracing_cpu_mask is modified then this holds | |
2190 | * the new bitmask we are about to install: | |
2191 | */ | |
2192 | static cpumask_t tracing_cpumask_new; | |
2193 | ||
2194 | /* | |
2195 | * The tracer itself will not take this lock, but still we want | |
2196 | * to provide a consistent cpumask to user-space: | |
2197 | */ | |
2198 | static DEFINE_MUTEX(tracing_cpumask_update_lock); | |
2199 | ||
2200 | /* | |
2201 | * Temporary storage for the character representation of the | |
2202 | * CPU bitmask (and one more byte for the newline): | |
2203 | */ | |
2204 | static char mask_str[NR_CPUS + 1]; | |
2205 | ||
c7078de1 IM |
2206 | static ssize_t |
2207 | tracing_cpumask_read(struct file *filp, char __user *ubuf, | |
2208 | size_t count, loff_t *ppos) | |
2209 | { | |
36dfe925 | 2210 | int len; |
c7078de1 IM |
2211 | |
2212 | mutex_lock(&tracing_cpumask_update_lock); | |
36dfe925 IM |
2213 | |
2214 | len = cpumask_scnprintf(mask_str, count, tracing_cpumask); | |
2215 | if (count - len < 2) { | |
2216 | count = -EINVAL; | |
2217 | goto out_err; | |
2218 | } | |
2219 | len += sprintf(mask_str + len, "\n"); | |
2220 | count = simple_read_from_buffer(ubuf, count, ppos, mask_str, NR_CPUS+1); | |
2221 | ||
2222 | out_err: | |
c7078de1 IM |
2223 | mutex_unlock(&tracing_cpumask_update_lock); |
2224 | ||
2225 | return count; | |
2226 | } | |
2227 | ||
2228 | static ssize_t | |
2229 | tracing_cpumask_write(struct file *filp, const char __user *ubuf, | |
2230 | size_t count, loff_t *ppos) | |
2231 | { | |
36dfe925 | 2232 | int err, cpu; |
c7078de1 IM |
2233 | |
2234 | mutex_lock(&tracing_cpumask_update_lock); | |
36dfe925 | 2235 | err = cpumask_parse_user(ubuf, count, tracing_cpumask_new); |
c7078de1 | 2236 | if (err) |
36dfe925 IM |
2237 | goto err_unlock; |
2238 | ||
92205c23 SR |
2239 | raw_local_irq_disable(); |
2240 | __raw_spin_lock(&ftrace_max_lock); | |
ab46428c | 2241 | for_each_tracing_cpu(cpu) { |
36dfe925 IM |
2242 | /* |
2243 | * Increase/decrease the disabled counter if we are | |
2244 | * about to flip a bit in the cpumask: | |
2245 | */ | |
2246 | if (cpu_isset(cpu, tracing_cpumask) && | |
2247 | !cpu_isset(cpu, tracing_cpumask_new)) { | |
2248 | atomic_inc(&global_trace.data[cpu]->disabled); | |
2249 | } | |
2250 | if (!cpu_isset(cpu, tracing_cpumask) && | |
2251 | cpu_isset(cpu, tracing_cpumask_new)) { | |
2252 | atomic_dec(&global_trace.data[cpu]->disabled); | |
2253 | } | |
2254 | } | |
92205c23 SR |
2255 | __raw_spin_unlock(&ftrace_max_lock); |
2256 | raw_local_irq_enable(); | |
36dfe925 IM |
2257 | |
2258 | tracing_cpumask = tracing_cpumask_new; | |
2259 | ||
2260 | mutex_unlock(&tracing_cpumask_update_lock); | |
c7078de1 IM |
2261 | |
2262 | return count; | |
36dfe925 IM |
2263 | |
2264 | err_unlock: | |
2265 | mutex_unlock(&tracing_cpumask_update_lock); | |
2266 | ||
2267 | return err; | |
c7078de1 IM |
2268 | } |
2269 | ||
2270 | static struct file_operations tracing_cpumask_fops = { | |
2271 | .open = tracing_open_generic, | |
2272 | .read = tracing_cpumask_read, | |
2273 | .write = tracing_cpumask_write, | |
bc0c38d1 SR |
2274 | }; |
2275 | ||
2276 | static ssize_t | |
2277 | tracing_iter_ctrl_read(struct file *filp, char __user *ubuf, | |
2278 | size_t cnt, loff_t *ppos) | |
2279 | { | |
2280 | char *buf; | |
2281 | int r = 0; | |
2282 | int len = 0; | |
2283 | int i; | |
2284 | ||
2285 | /* calulate max size */ | |
2286 | for (i = 0; trace_options[i]; i++) { | |
2287 | len += strlen(trace_options[i]); | |
2288 | len += 3; /* "no" and space */ | |
2289 | } | |
2290 | ||
2291 | /* +2 for \n and \0 */ | |
2292 | buf = kmalloc(len + 2, GFP_KERNEL); | |
2293 | if (!buf) | |
2294 | return -ENOMEM; | |
2295 | ||
2296 | for (i = 0; trace_options[i]; i++) { | |
2297 | if (trace_flags & (1 << i)) | |
2298 | r += sprintf(buf + r, "%s ", trace_options[i]); | |
2299 | else | |
2300 | r += sprintf(buf + r, "no%s ", trace_options[i]); | |
2301 | } | |
2302 | ||
2303 | r += sprintf(buf + r, "\n"); | |
2304 | WARN_ON(r >= len + 2); | |
2305 | ||
36dfe925 | 2306 | r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2307 | |
2308 | kfree(buf); | |
2309 | ||
2310 | return r; | |
2311 | } | |
2312 | ||
2313 | static ssize_t | |
2314 | tracing_iter_ctrl_write(struct file *filp, const char __user *ubuf, | |
2315 | size_t cnt, loff_t *ppos) | |
2316 | { | |
2317 | char buf[64]; | |
2318 | char *cmp = buf; | |
2319 | int neg = 0; | |
2320 | int i; | |
2321 | ||
cffae437 SR |
2322 | if (cnt >= sizeof(buf)) |
2323 | return -EINVAL; | |
bc0c38d1 SR |
2324 | |
2325 | if (copy_from_user(&buf, ubuf, cnt)) | |
2326 | return -EFAULT; | |
2327 | ||
2328 | buf[cnt] = 0; | |
2329 | ||
2330 | if (strncmp(buf, "no", 2) == 0) { | |
2331 | neg = 1; | |
2332 | cmp += 2; | |
2333 | } | |
2334 | ||
2335 | for (i = 0; trace_options[i]; i++) { | |
2336 | int len = strlen(trace_options[i]); | |
2337 | ||
2338 | if (strncmp(cmp, trace_options[i], len) == 0) { | |
2339 | if (neg) | |
2340 | trace_flags &= ~(1 << i); | |
2341 | else | |
2342 | trace_flags |= (1 << i); | |
2343 | break; | |
2344 | } | |
2345 | } | |
442e544c IM |
2346 | /* |
2347 | * If no option could be set, return an error: | |
2348 | */ | |
2349 | if (!trace_options[i]) | |
2350 | return -EINVAL; | |
bc0c38d1 SR |
2351 | |
2352 | filp->f_pos += cnt; | |
2353 | ||
2354 | return cnt; | |
2355 | } | |
2356 | ||
2357 | static struct file_operations tracing_iter_fops = { | |
c7078de1 IM |
2358 | .open = tracing_open_generic, |
2359 | .read = tracing_iter_ctrl_read, | |
2360 | .write = tracing_iter_ctrl_write, | |
bc0c38d1 SR |
2361 | }; |
2362 | ||
7bd2f24c IM |
2363 | static const char readme_msg[] = |
2364 | "tracing mini-HOWTO:\n\n" | |
2365 | "# mkdir /debug\n" | |
2366 | "# mount -t debugfs nodev /debug\n\n" | |
2367 | "# cat /debug/tracing/available_tracers\n" | |
2368 | "wakeup preemptirqsoff preemptoff irqsoff ftrace sched_switch none\n\n" | |
2369 | "# cat /debug/tracing/current_tracer\n" | |
2370 | "none\n" | |
2371 | "# echo sched_switch > /debug/tracing/current_tracer\n" | |
2372 | "# cat /debug/tracing/current_tracer\n" | |
2373 | "sched_switch\n" | |
2374 | "# cat /debug/tracing/iter_ctrl\n" | |
2375 | "noprint-parent nosym-offset nosym-addr noverbose\n" | |
2376 | "# echo print-parent > /debug/tracing/iter_ctrl\n" | |
2377 | "# echo 1 > /debug/tracing/tracing_enabled\n" | |
2378 | "# cat /debug/tracing/trace > /tmp/trace.txt\n" | |
2379 | "echo 0 > /debug/tracing/tracing_enabled\n" | |
2380 | ; | |
2381 | ||
2382 | static ssize_t | |
2383 | tracing_readme_read(struct file *filp, char __user *ubuf, | |
2384 | size_t cnt, loff_t *ppos) | |
2385 | { | |
2386 | return simple_read_from_buffer(ubuf, cnt, ppos, | |
2387 | readme_msg, strlen(readme_msg)); | |
2388 | } | |
2389 | ||
2390 | static struct file_operations tracing_readme_fops = { | |
c7078de1 IM |
2391 | .open = tracing_open_generic, |
2392 | .read = tracing_readme_read, | |
7bd2f24c IM |
2393 | }; |
2394 | ||
bc0c38d1 SR |
2395 | static ssize_t |
2396 | tracing_ctrl_read(struct file *filp, char __user *ubuf, | |
2397 | size_t cnt, loff_t *ppos) | |
2398 | { | |
2399 | struct trace_array *tr = filp->private_data; | |
2400 | char buf[64]; | |
2401 | int r; | |
2402 | ||
2403 | r = sprintf(buf, "%ld\n", tr->ctrl); | |
4e3c3333 | 2404 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2405 | } |
2406 | ||
2407 | static ssize_t | |
2408 | tracing_ctrl_write(struct file *filp, const char __user *ubuf, | |
2409 | size_t cnt, loff_t *ppos) | |
2410 | { | |
2411 | struct trace_array *tr = filp->private_data; | |
bc0c38d1 | 2412 | char buf[64]; |
c6caeeb1 SR |
2413 | long val; |
2414 | int ret; | |
bc0c38d1 | 2415 | |
cffae437 SR |
2416 | if (cnt >= sizeof(buf)) |
2417 | return -EINVAL; | |
bc0c38d1 SR |
2418 | |
2419 | if (copy_from_user(&buf, ubuf, cnt)) | |
2420 | return -EFAULT; | |
2421 | ||
2422 | buf[cnt] = 0; | |
2423 | ||
c6caeeb1 SR |
2424 | ret = strict_strtoul(buf, 10, &val); |
2425 | if (ret < 0) | |
2426 | return ret; | |
bc0c38d1 SR |
2427 | |
2428 | val = !!val; | |
2429 | ||
2430 | mutex_lock(&trace_types_lock); | |
2431 | if (tr->ctrl ^ val) { | |
2432 | if (val) | |
2433 | tracer_enabled = 1; | |
2434 | else | |
2435 | tracer_enabled = 0; | |
2436 | ||
2437 | tr->ctrl = val; | |
2438 | ||
2439 | if (current_trace && current_trace->ctrl_update) | |
2440 | current_trace->ctrl_update(tr); | |
2441 | } | |
2442 | mutex_unlock(&trace_types_lock); | |
2443 | ||
2444 | filp->f_pos += cnt; | |
2445 | ||
2446 | return cnt; | |
2447 | } | |
2448 | ||
2449 | static ssize_t | |
2450 | tracing_set_trace_read(struct file *filp, char __user *ubuf, | |
2451 | size_t cnt, loff_t *ppos) | |
2452 | { | |
2453 | char buf[max_tracer_type_len+2]; | |
2454 | int r; | |
2455 | ||
2456 | mutex_lock(&trace_types_lock); | |
2457 | if (current_trace) | |
2458 | r = sprintf(buf, "%s\n", current_trace->name); | |
2459 | else | |
2460 | r = sprintf(buf, "\n"); | |
2461 | mutex_unlock(&trace_types_lock); | |
2462 | ||
4bf39a94 | 2463 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2464 | } |
2465 | ||
2466 | static ssize_t | |
2467 | tracing_set_trace_write(struct file *filp, const char __user *ubuf, | |
2468 | size_t cnt, loff_t *ppos) | |
2469 | { | |
2470 | struct trace_array *tr = &global_trace; | |
2471 | struct tracer *t; | |
2472 | char buf[max_tracer_type_len+1]; | |
2473 | int i; | |
2474 | ||
2475 | if (cnt > max_tracer_type_len) | |
2476 | cnt = max_tracer_type_len; | |
2477 | ||
2478 | if (copy_from_user(&buf, ubuf, cnt)) | |
2479 | return -EFAULT; | |
2480 | ||
2481 | buf[cnt] = 0; | |
2482 | ||
2483 | /* strip ending whitespace. */ | |
2484 | for (i = cnt - 1; i > 0 && isspace(buf[i]); i--) | |
2485 | buf[i] = 0; | |
2486 | ||
2487 | mutex_lock(&trace_types_lock); | |
2488 | for (t = trace_types; t; t = t->next) { | |
2489 | if (strcmp(t->name, buf) == 0) | |
2490 | break; | |
2491 | } | |
2492 | if (!t || t == current_trace) | |
2493 | goto out; | |
2494 | ||
2495 | if (current_trace && current_trace->reset) | |
2496 | current_trace->reset(tr); | |
2497 | ||
2498 | current_trace = t; | |
2499 | if (t->init) | |
2500 | t->init(tr); | |
2501 | ||
2502 | out: | |
2503 | mutex_unlock(&trace_types_lock); | |
2504 | ||
2505 | filp->f_pos += cnt; | |
2506 | ||
2507 | return cnt; | |
2508 | } | |
2509 | ||
2510 | static ssize_t | |
2511 | tracing_max_lat_read(struct file *filp, char __user *ubuf, | |
2512 | size_t cnt, loff_t *ppos) | |
2513 | { | |
2514 | unsigned long *ptr = filp->private_data; | |
2515 | char buf[64]; | |
2516 | int r; | |
2517 | ||
cffae437 | 2518 | r = snprintf(buf, sizeof(buf), "%ld\n", |
bc0c38d1 | 2519 | *ptr == (unsigned long)-1 ? -1 : nsecs_to_usecs(*ptr)); |
cffae437 SR |
2520 | if (r > sizeof(buf)) |
2521 | r = sizeof(buf); | |
4bf39a94 | 2522 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2523 | } |
2524 | ||
2525 | static ssize_t | |
2526 | tracing_max_lat_write(struct file *filp, const char __user *ubuf, | |
2527 | size_t cnt, loff_t *ppos) | |
2528 | { | |
2529 | long *ptr = filp->private_data; | |
bc0c38d1 | 2530 | char buf[64]; |
c6caeeb1 SR |
2531 | long val; |
2532 | int ret; | |
bc0c38d1 | 2533 | |
cffae437 SR |
2534 | if (cnt >= sizeof(buf)) |
2535 | return -EINVAL; | |
bc0c38d1 SR |
2536 | |
2537 | if (copy_from_user(&buf, ubuf, cnt)) | |
2538 | return -EFAULT; | |
2539 | ||
2540 | buf[cnt] = 0; | |
2541 | ||
c6caeeb1 SR |
2542 | ret = strict_strtoul(buf, 10, &val); |
2543 | if (ret < 0) | |
2544 | return ret; | |
bc0c38d1 SR |
2545 | |
2546 | *ptr = val * 1000; | |
2547 | ||
2548 | return cnt; | |
2549 | } | |
2550 | ||
b3806b43 SR |
2551 | static atomic_t tracing_reader; |
2552 | ||
2553 | static int tracing_open_pipe(struct inode *inode, struct file *filp) | |
2554 | { | |
2555 | struct trace_iterator *iter; | |
2556 | ||
2557 | if (tracing_disabled) | |
2558 | return -ENODEV; | |
2559 | ||
2560 | /* We only allow for reader of the pipe */ | |
2561 | if (atomic_inc_return(&tracing_reader) != 1) { | |
2562 | atomic_dec(&tracing_reader); | |
2563 | return -EBUSY; | |
2564 | } | |
2565 | ||
2566 | /* create a buffer to store the information to pass to userspace */ | |
2567 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); | |
2568 | if (!iter) | |
2569 | return -ENOMEM; | |
2570 | ||
107bad8b | 2571 | mutex_lock(&trace_types_lock); |
b3806b43 | 2572 | iter->tr = &global_trace; |
72829bc3 | 2573 | iter->trace = current_trace; |
b3806b43 SR |
2574 | filp->private_data = iter; |
2575 | ||
107bad8b SR |
2576 | if (iter->trace->pipe_open) |
2577 | iter->trace->pipe_open(iter); | |
2578 | mutex_unlock(&trace_types_lock); | |
2579 | ||
b3806b43 SR |
2580 | return 0; |
2581 | } | |
2582 | ||
2583 | static int tracing_release_pipe(struct inode *inode, struct file *file) | |
2584 | { | |
2585 | struct trace_iterator *iter = file->private_data; | |
2586 | ||
2587 | kfree(iter); | |
2588 | atomic_dec(&tracing_reader); | |
2589 | ||
2590 | return 0; | |
2591 | } | |
2592 | ||
2a2cc8f7 SSP |
2593 | static unsigned int |
2594 | tracing_poll_pipe(struct file *filp, poll_table *poll_table) | |
2595 | { | |
2596 | struct trace_iterator *iter = filp->private_data; | |
2597 | ||
2598 | if (trace_flags & TRACE_ITER_BLOCK) { | |
2599 | /* | |
2600 | * Always select as readable when in blocking mode | |
2601 | */ | |
2602 | return POLLIN | POLLRDNORM; | |
afc2abc0 | 2603 | } else { |
2a2cc8f7 SSP |
2604 | if (!trace_empty(iter)) |
2605 | return POLLIN | POLLRDNORM; | |
2606 | poll_wait(filp, &trace_wait, poll_table); | |
2607 | if (!trace_empty(iter)) | |
2608 | return POLLIN | POLLRDNORM; | |
2609 | ||
2610 | return 0; | |
2611 | } | |
2612 | } | |
2613 | ||
b3806b43 SR |
2614 | /* |
2615 | * Consumer reader. | |
2616 | */ | |
2617 | static ssize_t | |
2618 | tracing_read_pipe(struct file *filp, char __user *ubuf, | |
2619 | size_t cnt, loff_t *ppos) | |
2620 | { | |
2621 | struct trace_iterator *iter = filp->private_data; | |
2622 | struct trace_array_cpu *data; | |
2623 | static cpumask_t mask; | |
b3806b43 | 2624 | unsigned long flags; |
25770467 | 2625 | #ifdef CONFIG_FTRACE |
2e0f5761 | 2626 | int ftrace_save; |
25770467 | 2627 | #endif |
b3806b43 | 2628 | int cpu; |
6c6c2796 | 2629 | ssize_t sret; |
b3806b43 SR |
2630 | |
2631 | /* return any leftover data */ | |
6c6c2796 PP |
2632 | sret = trace_seq_to_user(&iter->seq, ubuf, cnt); |
2633 | if (sret != -EBUSY) | |
2634 | return sret; | |
2635 | sret = 0; | |
b3806b43 | 2636 | |
6c6c2796 | 2637 | trace_seq_reset(&iter->seq); |
b3806b43 | 2638 | |
107bad8b SR |
2639 | mutex_lock(&trace_types_lock); |
2640 | if (iter->trace->read) { | |
6c6c2796 PP |
2641 | sret = iter->trace->read(iter, filp, ubuf, cnt, ppos); |
2642 | if (sret) | |
107bad8b | 2643 | goto out; |
107bad8b SR |
2644 | } |
2645 | ||
b3806b43 | 2646 | while (trace_empty(iter)) { |
2dc8f095 | 2647 | |
107bad8b | 2648 | if ((filp->f_flags & O_NONBLOCK)) { |
6c6c2796 | 2649 | sret = -EAGAIN; |
107bad8b SR |
2650 | goto out; |
2651 | } | |
2dc8f095 | 2652 | |
b3806b43 SR |
2653 | /* |
2654 | * This is a make-shift waitqueue. The reason we don't use | |
2655 | * an actual wait queue is because: | |
2656 | * 1) we only ever have one waiter | |
2657 | * 2) the tracing, traces all functions, we don't want | |
2658 | * the overhead of calling wake_up and friends | |
2659 | * (and tracing them too) | |
2660 | * Anyway, this is really very primitive wakeup. | |
2661 | */ | |
2662 | set_current_state(TASK_INTERRUPTIBLE); | |
2663 | iter->tr->waiter = current; | |
2664 | ||
107bad8b SR |
2665 | mutex_unlock(&trace_types_lock); |
2666 | ||
9fe068e9 IM |
2667 | /* sleep for 100 msecs, and try again. */ |
2668 | schedule_timeout(HZ/10); | |
b3806b43 | 2669 | |
107bad8b SR |
2670 | mutex_lock(&trace_types_lock); |
2671 | ||
b3806b43 SR |
2672 | iter->tr->waiter = NULL; |
2673 | ||
107bad8b | 2674 | if (signal_pending(current)) { |
6c6c2796 | 2675 | sret = -EINTR; |
107bad8b SR |
2676 | goto out; |
2677 | } | |
b3806b43 | 2678 | |
84527997 | 2679 | if (iter->trace != current_trace) |
107bad8b | 2680 | goto out; |
84527997 | 2681 | |
b3806b43 SR |
2682 | /* |
2683 | * We block until we read something and tracing is disabled. | |
2684 | * We still block if tracing is disabled, but we have never | |
2685 | * read anything. This allows a user to cat this file, and | |
2686 | * then enable tracing. But after we have read something, | |
2687 | * we give an EOF when tracing is again disabled. | |
2688 | * | |
2689 | * iter->pos will be 0 if we haven't read anything. | |
2690 | */ | |
2691 | if (!tracer_enabled && iter->pos) | |
2692 | break; | |
2693 | ||
2694 | continue; | |
2695 | } | |
2696 | ||
2697 | /* stop when tracing is finished */ | |
2698 | if (trace_empty(iter)) | |
107bad8b | 2699 | goto out; |
b3806b43 SR |
2700 | |
2701 | if (cnt >= PAGE_SIZE) | |
2702 | cnt = PAGE_SIZE - 1; | |
2703 | ||
53d0aa77 | 2704 | /* reset all but tr, trace, and overruns */ |
53d0aa77 SR |
2705 | memset(&iter->seq, 0, |
2706 | sizeof(struct trace_iterator) - | |
2707 | offsetof(struct trace_iterator, seq)); | |
4823ed7e | 2708 | iter->pos = -1; |
b3806b43 SR |
2709 | |
2710 | /* | |
2711 | * We need to stop all tracing on all CPUS to read the | |
2712 | * the next buffer. This is a bit expensive, but is | |
2713 | * not done often. We fill all what we can read, | |
2714 | * and then release the locks again. | |
2715 | */ | |
2716 | ||
2717 | cpus_clear(mask); | |
2718 | local_irq_save(flags); | |
25770467 | 2719 | #ifdef CONFIG_FTRACE |
2e0f5761 IM |
2720 | ftrace_save = ftrace_enabled; |
2721 | ftrace_enabled = 0; | |
25770467 | 2722 | #endif |
2e0f5761 | 2723 | smp_wmb(); |
ab46428c | 2724 | for_each_tracing_cpu(cpu) { |
b3806b43 SR |
2725 | data = iter->tr->data[cpu]; |
2726 | ||
2727 | if (!head_page(data) || !data->trace_idx) | |
2728 | continue; | |
2729 | ||
2730 | atomic_inc(&data->disabled); | |
b3806b43 SR |
2731 | cpu_set(cpu, mask); |
2732 | } | |
2733 | ||
2e0f5761 IM |
2734 | for_each_cpu_mask(cpu, mask) { |
2735 | data = iter->tr->data[cpu]; | |
92205c23 | 2736 | __raw_spin_lock(&data->lock); |
53d0aa77 SR |
2737 | |
2738 | if (data->overrun > iter->last_overrun[cpu]) | |
2739 | iter->overrun[cpu] += | |
2740 | data->overrun - iter->last_overrun[cpu]; | |
2741 | iter->last_overrun[cpu] = data->overrun; | |
2e0f5761 IM |
2742 | } |
2743 | ||
088b1e42 | 2744 | while (find_next_entry_inc(iter) != NULL) { |
6c6c2796 | 2745 | int ret; |
088b1e42 SR |
2746 | int len = iter->seq.len; |
2747 | ||
f9896bf3 | 2748 | ret = print_trace_line(iter); |
088b1e42 SR |
2749 | if (!ret) { |
2750 | /* don't print partial lines */ | |
2751 | iter->seq.len = len; | |
b3806b43 | 2752 | break; |
088b1e42 | 2753 | } |
b3806b43 SR |
2754 | |
2755 | trace_consume(iter); | |
2756 | ||
2757 | if (iter->seq.len >= cnt) | |
2758 | break; | |
b3806b43 SR |
2759 | } |
2760 | ||
d4c5a2f5 | 2761 | for_each_cpu_mask(cpu, mask) { |
b3806b43 | 2762 | data = iter->tr->data[cpu]; |
92205c23 | 2763 | __raw_spin_unlock(&data->lock); |
2e0f5761 IM |
2764 | } |
2765 | ||
2766 | for_each_cpu_mask(cpu, mask) { | |
2767 | data = iter->tr->data[cpu]; | |
b3806b43 SR |
2768 | atomic_dec(&data->disabled); |
2769 | } | |
25770467 | 2770 | #ifdef CONFIG_FTRACE |
2e0f5761 | 2771 | ftrace_enabled = ftrace_save; |
25770467 | 2772 | #endif |
b3806b43 SR |
2773 | local_irq_restore(flags); |
2774 | ||
2775 | /* Now copy what we have to the user */ | |
6c6c2796 PP |
2776 | sret = trace_seq_to_user(&iter->seq, ubuf, cnt); |
2777 | if (iter->seq.readpos >= iter->seq.len) | |
b3806b43 | 2778 | trace_seq_reset(&iter->seq); |
6c6c2796 PP |
2779 | if (sret == -EBUSY) |
2780 | sret = 0; | |
b3806b43 | 2781 | |
107bad8b SR |
2782 | out: |
2783 | mutex_unlock(&trace_types_lock); | |
2784 | ||
6c6c2796 | 2785 | return sret; |
b3806b43 SR |
2786 | } |
2787 | ||
a98a3c3f SR |
2788 | static ssize_t |
2789 | tracing_entries_read(struct file *filp, char __user *ubuf, | |
2790 | size_t cnt, loff_t *ppos) | |
2791 | { | |
2792 | struct trace_array *tr = filp->private_data; | |
2793 | char buf[64]; | |
2794 | int r; | |
2795 | ||
2796 | r = sprintf(buf, "%lu\n", tr->entries); | |
2797 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); | |
2798 | } | |
2799 | ||
2800 | static ssize_t | |
2801 | tracing_entries_write(struct file *filp, const char __user *ubuf, | |
2802 | size_t cnt, loff_t *ppos) | |
2803 | { | |
2804 | unsigned long val; | |
2805 | char buf[64]; | |
19384c03 | 2806 | int i, ret; |
a98a3c3f | 2807 | |
cffae437 SR |
2808 | if (cnt >= sizeof(buf)) |
2809 | return -EINVAL; | |
a98a3c3f SR |
2810 | |
2811 | if (copy_from_user(&buf, ubuf, cnt)) | |
2812 | return -EFAULT; | |
2813 | ||
2814 | buf[cnt] = 0; | |
2815 | ||
c6caeeb1 SR |
2816 | ret = strict_strtoul(buf, 10, &val); |
2817 | if (ret < 0) | |
2818 | return ret; | |
a98a3c3f SR |
2819 | |
2820 | /* must have at least 1 entry */ | |
2821 | if (!val) | |
2822 | return -EINVAL; | |
2823 | ||
2824 | mutex_lock(&trace_types_lock); | |
2825 | ||
2826 | if (current_trace != &no_tracer) { | |
2827 | cnt = -EBUSY; | |
2828 | pr_info("ftrace: set current_tracer to none" | |
2829 | " before modifying buffer size\n"); | |
2830 | goto out; | |
2831 | } | |
2832 | ||
2833 | if (val > global_trace.entries) { | |
3eefae99 SR |
2834 | long pages_requested; |
2835 | unsigned long freeable_pages; | |
2836 | ||
2837 | /* make sure we have enough memory before mapping */ | |
2838 | pages_requested = | |
2839 | (val + (ENTRIES_PER_PAGE-1)) / ENTRIES_PER_PAGE; | |
2840 | ||
2841 | /* account for each buffer (and max_tr) */ | |
2842 | pages_requested *= tracing_nr_buffers * 2; | |
2843 | ||
2844 | /* Check for overflow */ | |
2845 | if (pages_requested < 0) { | |
2846 | cnt = -ENOMEM; | |
2847 | goto out; | |
2848 | } | |
2849 | ||
2850 | freeable_pages = determine_dirtyable_memory(); | |
2851 | ||
2852 | /* we only allow to request 1/4 of useable memory */ | |
2853 | if (pages_requested > | |
2854 | ((freeable_pages + tracing_pages_allocated) / 4)) { | |
2855 | cnt = -ENOMEM; | |
2856 | goto out; | |
2857 | } | |
2858 | ||
a98a3c3f SR |
2859 | while (global_trace.entries < val) { |
2860 | if (trace_alloc_page()) { | |
2861 | cnt = -ENOMEM; | |
2862 | goto out; | |
2863 | } | |
3eefae99 SR |
2864 | /* double check that we don't go over the known pages */ |
2865 | if (tracing_pages_allocated > pages_requested) | |
2866 | break; | |
a98a3c3f | 2867 | } |
3eefae99 | 2868 | |
a98a3c3f SR |
2869 | } else { |
2870 | /* include the number of entries in val (inc of page entries) */ | |
2871 | while (global_trace.entries > val + (ENTRIES_PER_PAGE - 1)) | |
2872 | trace_free_page(); | |
2873 | } | |
2874 | ||
19384c03 SR |
2875 | /* check integrity */ |
2876 | for_each_tracing_cpu(i) | |
2877 | check_pages(global_trace.data[i]); | |
2878 | ||
a98a3c3f SR |
2879 | filp->f_pos += cnt; |
2880 | ||
19384c03 SR |
2881 | /* If check pages failed, return ENOMEM */ |
2882 | if (tracing_disabled) | |
2883 | cnt = -ENOMEM; | |
a98a3c3f SR |
2884 | out: |
2885 | max_tr.entries = global_trace.entries; | |
2886 | mutex_unlock(&trace_types_lock); | |
2887 | ||
2888 | return cnt; | |
2889 | } | |
2890 | ||
bc0c38d1 | 2891 | static struct file_operations tracing_max_lat_fops = { |
4bf39a94 IM |
2892 | .open = tracing_open_generic, |
2893 | .read = tracing_max_lat_read, | |
2894 | .write = tracing_max_lat_write, | |
bc0c38d1 SR |
2895 | }; |
2896 | ||
2897 | static struct file_operations tracing_ctrl_fops = { | |
4bf39a94 IM |
2898 | .open = tracing_open_generic, |
2899 | .read = tracing_ctrl_read, | |
2900 | .write = tracing_ctrl_write, | |
bc0c38d1 SR |
2901 | }; |
2902 | ||
2903 | static struct file_operations set_tracer_fops = { | |
4bf39a94 IM |
2904 | .open = tracing_open_generic, |
2905 | .read = tracing_set_trace_read, | |
2906 | .write = tracing_set_trace_write, | |
bc0c38d1 SR |
2907 | }; |
2908 | ||
b3806b43 | 2909 | static struct file_operations tracing_pipe_fops = { |
4bf39a94 | 2910 | .open = tracing_open_pipe, |
2a2cc8f7 | 2911 | .poll = tracing_poll_pipe, |
4bf39a94 IM |
2912 | .read = tracing_read_pipe, |
2913 | .release = tracing_release_pipe, | |
b3806b43 SR |
2914 | }; |
2915 | ||
a98a3c3f SR |
2916 | static struct file_operations tracing_entries_fops = { |
2917 | .open = tracing_open_generic, | |
2918 | .read = tracing_entries_read, | |
2919 | .write = tracing_entries_write, | |
2920 | }; | |
2921 | ||
bc0c38d1 SR |
2922 | #ifdef CONFIG_DYNAMIC_FTRACE |
2923 | ||
2924 | static ssize_t | |
2925 | tracing_read_long(struct file *filp, char __user *ubuf, | |
2926 | size_t cnt, loff_t *ppos) | |
2927 | { | |
2928 | unsigned long *p = filp->private_data; | |
2929 | char buf[64]; | |
2930 | int r; | |
2931 | ||
2932 | r = sprintf(buf, "%ld\n", *p); | |
4bf39a94 IM |
2933 | |
2934 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); | |
bc0c38d1 SR |
2935 | } |
2936 | ||
2937 | static struct file_operations tracing_read_long_fops = { | |
4bf39a94 IM |
2938 | .open = tracing_open_generic, |
2939 | .read = tracing_read_long, | |
bc0c38d1 SR |
2940 | }; |
2941 | #endif | |
2942 | ||
2943 | static struct dentry *d_tracer; | |
2944 | ||
2945 | struct dentry *tracing_init_dentry(void) | |
2946 | { | |
2947 | static int once; | |
2948 | ||
2949 | if (d_tracer) | |
2950 | return d_tracer; | |
2951 | ||
2952 | d_tracer = debugfs_create_dir("tracing", NULL); | |
2953 | ||
2954 | if (!d_tracer && !once) { | |
2955 | once = 1; | |
2956 | pr_warning("Could not create debugfs directory 'tracing'\n"); | |
2957 | return NULL; | |
2958 | } | |
2959 | ||
2960 | return d_tracer; | |
2961 | } | |
2962 | ||
60a11774 SR |
2963 | #ifdef CONFIG_FTRACE_SELFTEST |
2964 | /* Let selftest have access to static functions in this file */ | |
2965 | #include "trace_selftest.c" | |
2966 | #endif | |
2967 | ||
bc0c38d1 SR |
2968 | static __init void tracer_init_debugfs(void) |
2969 | { | |
2970 | struct dentry *d_tracer; | |
2971 | struct dentry *entry; | |
2972 | ||
2973 | d_tracer = tracing_init_dentry(); | |
2974 | ||
2975 | entry = debugfs_create_file("tracing_enabled", 0644, d_tracer, | |
2976 | &global_trace, &tracing_ctrl_fops); | |
2977 | if (!entry) | |
2978 | pr_warning("Could not create debugfs 'tracing_enabled' entry\n"); | |
2979 | ||
2980 | entry = debugfs_create_file("iter_ctrl", 0644, d_tracer, | |
2981 | NULL, &tracing_iter_fops); | |
2982 | if (!entry) | |
2983 | pr_warning("Could not create debugfs 'iter_ctrl' entry\n"); | |
2984 | ||
c7078de1 IM |
2985 | entry = debugfs_create_file("tracing_cpumask", 0644, d_tracer, |
2986 | NULL, &tracing_cpumask_fops); | |
2987 | if (!entry) | |
2988 | pr_warning("Could not create debugfs 'tracing_cpumask' entry\n"); | |
2989 | ||
bc0c38d1 SR |
2990 | entry = debugfs_create_file("latency_trace", 0444, d_tracer, |
2991 | &global_trace, &tracing_lt_fops); | |
2992 | if (!entry) | |
2993 | pr_warning("Could not create debugfs 'latency_trace' entry\n"); | |
2994 | ||
2995 | entry = debugfs_create_file("trace", 0444, d_tracer, | |
2996 | &global_trace, &tracing_fops); | |
2997 | if (!entry) | |
2998 | pr_warning("Could not create debugfs 'trace' entry\n"); | |
2999 | ||
3000 | entry = debugfs_create_file("available_tracers", 0444, d_tracer, | |
3001 | &global_trace, &show_traces_fops); | |
3002 | if (!entry) | |
98a983aa | 3003 | pr_warning("Could not create debugfs 'available_tracers' entry\n"); |
bc0c38d1 SR |
3004 | |
3005 | entry = debugfs_create_file("current_tracer", 0444, d_tracer, | |
3006 | &global_trace, &set_tracer_fops); | |
3007 | if (!entry) | |
98a983aa | 3008 | pr_warning("Could not create debugfs 'current_tracer' entry\n"); |
bc0c38d1 SR |
3009 | |
3010 | entry = debugfs_create_file("tracing_max_latency", 0644, d_tracer, | |
3011 | &tracing_max_latency, | |
3012 | &tracing_max_lat_fops); | |
3013 | if (!entry) | |
3014 | pr_warning("Could not create debugfs " | |
3015 | "'tracing_max_latency' entry\n"); | |
3016 | ||
3017 | entry = debugfs_create_file("tracing_thresh", 0644, d_tracer, | |
3018 | &tracing_thresh, &tracing_max_lat_fops); | |
3019 | if (!entry) | |
3020 | pr_warning("Could not create debugfs " | |
98a983aa | 3021 | "'tracing_thresh' entry\n"); |
7bd2f24c IM |
3022 | entry = debugfs_create_file("README", 0644, d_tracer, |
3023 | NULL, &tracing_readme_fops); | |
3024 | if (!entry) | |
3025 | pr_warning("Could not create debugfs 'README' entry\n"); | |
3026 | ||
b3806b43 SR |
3027 | entry = debugfs_create_file("trace_pipe", 0644, d_tracer, |
3028 | NULL, &tracing_pipe_fops); | |
3029 | if (!entry) | |
3030 | pr_warning("Could not create debugfs " | |
98a983aa | 3031 | "'trace_pipe' entry\n"); |
bc0c38d1 | 3032 | |
a98a3c3f SR |
3033 | entry = debugfs_create_file("trace_entries", 0644, d_tracer, |
3034 | &global_trace, &tracing_entries_fops); | |
3035 | if (!entry) | |
3036 | pr_warning("Could not create debugfs " | |
98a983aa | 3037 | "'trace_entries' entry\n"); |
a98a3c3f | 3038 | |
bc0c38d1 SR |
3039 | #ifdef CONFIG_DYNAMIC_FTRACE |
3040 | entry = debugfs_create_file("dyn_ftrace_total_info", 0444, d_tracer, | |
3041 | &ftrace_update_tot_cnt, | |
3042 | &tracing_read_long_fops); | |
3043 | if (!entry) | |
3044 | pr_warning("Could not create debugfs " | |
3045 | "'dyn_ftrace_total_info' entry\n"); | |
3046 | #endif | |
d618b3e6 IM |
3047 | #ifdef CONFIG_SYSPROF_TRACER |
3048 | init_tracer_sysprof_debugfs(d_tracer); | |
3049 | #endif | |
bc0c38d1 SR |
3050 | } |
3051 | ||
dd0e545f SR |
3052 | #define TRACE_BUF_SIZE 1024 |
3053 | #define TRACE_PRINT_BUF_SIZE \ | |
3054 | (sizeof(struct trace_field) - offsetof(struct trace_field, print.buf)) | |
3055 | #define TRACE_CONT_BUF_SIZE sizeof(struct trace_field) | |
3056 | ||
801fe400 | 3057 | int trace_vprintk(unsigned long ip, const char *fmt, va_list args) |
dd0e545f | 3058 | { |
dd0e545f SR |
3059 | static DEFINE_SPINLOCK(trace_buf_lock); |
3060 | static char trace_buf[TRACE_BUF_SIZE]; | |
f09ce573 PZ |
3061 | |
3062 | struct trace_array *tr = &global_trace; | |
dd0e545f SR |
3063 | struct trace_array_cpu *data; |
3064 | struct trace_entry *entry; | |
3065 | unsigned long flags; | |
3066 | long disabled; | |
dd0e545f SR |
3067 | int cpu, len = 0, write, written = 0; |
3068 | ||
801fe400 | 3069 | if (current_trace == &no_tracer || !tr->ctrl || tracing_disabled) |
dd0e545f SR |
3070 | return 0; |
3071 | ||
3072 | local_irq_save(flags); | |
3073 | cpu = raw_smp_processor_id(); | |
3074 | data = tr->data[cpu]; | |
3075 | disabled = atomic_inc_return(&data->disabled); | |
3076 | ||
f09ce573 | 3077 | if (unlikely(disabled != 1)) |
dd0e545f SR |
3078 | goto out; |
3079 | ||
3080 | spin_lock(&trace_buf_lock); | |
801fe400 | 3081 | len = vsnprintf(trace_buf, TRACE_BUF_SIZE, fmt, args); |
dd0e545f SR |
3082 | |
3083 | len = min(len, TRACE_BUF_SIZE-1); | |
3084 | trace_buf[len] = 0; | |
3085 | ||
3086 | __raw_spin_lock(&data->lock); | |
3087 | entry = tracing_get_trace_entry(tr, data); | |
3088 | tracing_generic_entry_update(entry, flags); | |
3089 | entry->type = TRACE_PRINT; | |
3090 | entry->field.print.ip = ip; | |
3091 | ||
3092 | write = min(len, (int)(TRACE_PRINT_BUF_SIZE-1)); | |
3093 | ||
3094 | memcpy(&entry->field.print.buf, trace_buf, write); | |
3095 | entry->field.print.buf[write] = 0; | |
3096 | written = write; | |
3097 | ||
3098 | if (written != len) | |
3099 | entry->field.flags |= TRACE_FLAG_CONT; | |
3100 | ||
3101 | while (written != len) { | |
3102 | entry = tracing_get_trace_entry(tr, data); | |
3103 | ||
3104 | entry->type = TRACE_CONT; | |
3105 | write = min(len - written, (int)(TRACE_CONT_BUF_SIZE-1)); | |
3106 | memcpy(&entry->cont.buf, trace_buf+written, write); | |
3107 | entry->cont.buf[write] = 0; | |
3108 | written += write; | |
3109 | } | |
3110 | __raw_spin_unlock(&data->lock); | |
3111 | ||
3112 | spin_unlock(&trace_buf_lock); | |
3113 | ||
3114 | out: | |
3115 | atomic_dec(&data->disabled); | |
3116 | local_irq_restore(flags); | |
3117 | ||
3118 | return len; | |
3119 | } | |
801fe400 PP |
3120 | EXPORT_SYMBOL_GPL(trace_vprintk); |
3121 | ||
3122 | int __ftrace_printk(unsigned long ip, const char *fmt, ...) | |
3123 | { | |
3124 | int ret; | |
3125 | va_list ap; | |
3126 | ||
3127 | if (!(trace_flags & TRACE_ITER_PRINTK)) | |
3128 | return 0; | |
3129 | ||
3130 | va_start(ap, fmt); | |
3131 | ret = trace_vprintk(ip, fmt, ap); | |
3132 | va_end(ap); | |
3133 | return ret; | |
3134 | } | |
dd0e545f SR |
3135 | EXPORT_SYMBOL_GPL(__ftrace_printk); |
3136 | ||
3f5a54e3 SR |
3137 | static int trace_panic_handler(struct notifier_block *this, |
3138 | unsigned long event, void *unused) | |
3139 | { | |
3140 | ftrace_dump(); | |
3141 | return NOTIFY_OK; | |
3142 | } | |
3143 | ||
3144 | static struct notifier_block trace_panic_notifier = { | |
3145 | .notifier_call = trace_panic_handler, | |
3146 | .next = NULL, | |
3147 | .priority = 150 /* priority: INT_MAX >= x >= 0 */ | |
3148 | }; | |
3149 | ||
3150 | static int trace_die_handler(struct notifier_block *self, | |
3151 | unsigned long val, | |
3152 | void *data) | |
3153 | { | |
3154 | switch (val) { | |
3155 | case DIE_OOPS: | |
3156 | ftrace_dump(); | |
3157 | break; | |
3158 | default: | |
3159 | break; | |
3160 | } | |
3161 | return NOTIFY_OK; | |
3162 | } | |
3163 | ||
3164 | static struct notifier_block trace_die_notifier = { | |
3165 | .notifier_call = trace_die_handler, | |
3166 | .priority = 200 | |
3167 | }; | |
3168 | ||
3169 | /* | |
3170 | * printk is set to max of 1024, we really don't need it that big. | |
3171 | * Nothing should be printing 1000 characters anyway. | |
3172 | */ | |
3173 | #define TRACE_MAX_PRINT 1000 | |
3174 | ||
3175 | /* | |
3176 | * Define here KERN_TRACE so that we have one place to modify | |
3177 | * it if we decide to change what log level the ftrace dump | |
3178 | * should be at. | |
3179 | */ | |
3180 | #define KERN_TRACE KERN_INFO | |
3181 | ||
3182 | static void | |
3183 | trace_printk_seq(struct trace_seq *s) | |
3184 | { | |
3185 | /* Probably should print a warning here. */ | |
3186 | if (s->len >= 1000) | |
3187 | s->len = 1000; | |
3188 | ||
3189 | /* should be zero ended, but we are paranoid. */ | |
3190 | s->buffer[s->len] = 0; | |
3191 | ||
3192 | printk(KERN_TRACE "%s", s->buffer); | |
3193 | ||
3194 | trace_seq_reset(s); | |
3195 | } | |
3196 | ||
3197 | ||
3198 | void ftrace_dump(void) | |
3199 | { | |
3200 | static DEFINE_SPINLOCK(ftrace_dump_lock); | |
3201 | /* use static because iter can be a bit big for the stack */ | |
3202 | static struct trace_iterator iter; | |
3203 | struct trace_array_cpu *data; | |
3204 | static cpumask_t mask; | |
3205 | static int dump_ran; | |
3206 | unsigned long flags; | |
3207 | int cnt = 0; | |
3208 | int cpu; | |
3209 | ||
3210 | /* only one dump */ | |
3211 | spin_lock_irqsave(&ftrace_dump_lock, flags); | |
3212 | if (dump_ran) | |
3213 | goto out; | |
3214 | ||
3215 | dump_ran = 1; | |
3216 | ||
3217 | /* No turning back! */ | |
3218 | ftrace_kill_atomic(); | |
3219 | ||
3220 | printk(KERN_TRACE "Dumping ftrace buffer:\n"); | |
3221 | ||
3222 | iter.tr = &global_trace; | |
3223 | iter.trace = current_trace; | |
3224 | ||
3225 | /* | |
3226 | * We need to stop all tracing on all CPUS to read the | |
3227 | * the next buffer. This is a bit expensive, but is | |
3228 | * not done often. We fill all what we can read, | |
3229 | * and then release the locks again. | |
3230 | */ | |
3231 | ||
3232 | cpus_clear(mask); | |
3233 | ||
3234 | for_each_tracing_cpu(cpu) { | |
3235 | data = iter.tr->data[cpu]; | |
3236 | ||
3237 | if (!head_page(data) || !data->trace_idx) | |
3238 | continue; | |
3239 | ||
3240 | atomic_inc(&data->disabled); | |
3241 | cpu_set(cpu, mask); | |
3242 | } | |
3243 | ||
3244 | for_each_cpu_mask(cpu, mask) { | |
3245 | data = iter.tr->data[cpu]; | |
3246 | __raw_spin_lock(&data->lock); | |
3247 | ||
3248 | if (data->overrun > iter.last_overrun[cpu]) | |
3249 | iter.overrun[cpu] += | |
3250 | data->overrun - iter.last_overrun[cpu]; | |
3251 | iter.last_overrun[cpu] = data->overrun; | |
3252 | } | |
3253 | ||
3254 | while (!trace_empty(&iter)) { | |
3255 | ||
3256 | if (!cnt) | |
3257 | printk(KERN_TRACE "---------------------------------\n"); | |
3258 | ||
3259 | cnt++; | |
3260 | ||
3261 | /* reset all but tr, trace, and overruns */ | |
3262 | memset(&iter.seq, 0, | |
3263 | sizeof(struct trace_iterator) - | |
3264 | offsetof(struct trace_iterator, seq)); | |
3265 | iter.iter_flags |= TRACE_FILE_LAT_FMT; | |
3266 | iter.pos = -1; | |
3267 | ||
3268 | if (find_next_entry_inc(&iter) != NULL) { | |
3269 | print_trace_line(&iter); | |
3270 | trace_consume(&iter); | |
3271 | } | |
3272 | ||
3273 | trace_printk_seq(&iter.seq); | |
3274 | } | |
3275 | ||
3276 | if (!cnt) | |
3277 | printk(KERN_TRACE " (ftrace buffer empty)\n"); | |
3278 | else | |
3279 | printk(KERN_TRACE "---------------------------------\n"); | |
3280 | ||
3281 | for_each_cpu_mask(cpu, mask) { | |
3282 | data = iter.tr->data[cpu]; | |
3283 | __raw_spin_unlock(&data->lock); | |
3284 | } | |
3285 | ||
3286 | for_each_cpu_mask(cpu, mask) { | |
3287 | data = iter.tr->data[cpu]; | |
3288 | atomic_dec(&data->disabled); | |
3289 | } | |
3290 | ||
3291 | ||
3292 | out: | |
3293 | spin_unlock_irqrestore(&ftrace_dump_lock, flags); | |
3294 | } | |
3295 | ||
4c11d7ae | 3296 | static int trace_alloc_page(void) |
bc0c38d1 | 3297 | { |
4c11d7ae | 3298 | struct trace_array_cpu *data; |
4c11d7ae SR |
3299 | struct page *page, *tmp; |
3300 | LIST_HEAD(pages); | |
c7aafc54 | 3301 | void *array; |
3eefae99 | 3302 | unsigned pages_allocated = 0; |
4c11d7ae SR |
3303 | int i; |
3304 | ||
3305 | /* first allocate a page for each CPU */ | |
ab46428c | 3306 | for_each_tracing_cpu(i) { |
4c11d7ae SR |
3307 | array = (void *)__get_free_page(GFP_KERNEL); |
3308 | if (array == NULL) { | |
3309 | printk(KERN_ERR "tracer: failed to allocate page" | |
3310 | "for trace buffer!\n"); | |
3311 | goto free_pages; | |
3312 | } | |
3313 | ||
3eefae99 | 3314 | pages_allocated++; |
4c11d7ae SR |
3315 | page = virt_to_page(array); |
3316 | list_add(&page->lru, &pages); | |
3317 | ||
3318 | /* Only allocate if we are actually using the max trace */ | |
3319 | #ifdef CONFIG_TRACER_MAX_TRACE | |
3320 | array = (void *)__get_free_page(GFP_KERNEL); | |
3321 | if (array == NULL) { | |
3322 | printk(KERN_ERR "tracer: failed to allocate page" | |
3323 | "for trace buffer!\n"); | |
3324 | goto free_pages; | |
3325 | } | |
3eefae99 | 3326 | pages_allocated++; |
4c11d7ae SR |
3327 | page = virt_to_page(array); |
3328 | list_add(&page->lru, &pages); | |
3329 | #endif | |
3330 | } | |
3331 | ||
3332 | /* Now that we successfully allocate a page per CPU, add them */ | |
ab46428c | 3333 | for_each_tracing_cpu(i) { |
4c11d7ae SR |
3334 | data = global_trace.data[i]; |
3335 | page = list_entry(pages.next, struct page, lru); | |
c7aafc54 | 3336 | list_del_init(&page->lru); |
4c11d7ae SR |
3337 | list_add_tail(&page->lru, &data->trace_pages); |
3338 | ClearPageLRU(page); | |
3339 | ||
3340 | #ifdef CONFIG_TRACER_MAX_TRACE | |
3341 | data = max_tr.data[i]; | |
3342 | page = list_entry(pages.next, struct page, lru); | |
c7aafc54 | 3343 | list_del_init(&page->lru); |
4c11d7ae SR |
3344 | list_add_tail(&page->lru, &data->trace_pages); |
3345 | SetPageLRU(page); | |
3346 | #endif | |
3347 | } | |
3eefae99 | 3348 | tracing_pages_allocated += pages_allocated; |
4c11d7ae SR |
3349 | global_trace.entries += ENTRIES_PER_PAGE; |
3350 | ||
3351 | return 0; | |
3352 | ||
3353 | free_pages: | |
3354 | list_for_each_entry_safe(page, tmp, &pages, lru) { | |
c7aafc54 | 3355 | list_del_init(&page->lru); |
4c11d7ae SR |
3356 | __free_page(page); |
3357 | } | |
3358 | return -ENOMEM; | |
bc0c38d1 SR |
3359 | } |
3360 | ||
a98a3c3f SR |
3361 | static int trace_free_page(void) |
3362 | { | |
3363 | struct trace_array_cpu *data; | |
3364 | struct page *page; | |
3365 | struct list_head *p; | |
3366 | int i; | |
3367 | int ret = 0; | |
3368 | ||
3369 | /* free one page from each buffer */ | |
ab46428c | 3370 | for_each_tracing_cpu(i) { |
a98a3c3f SR |
3371 | data = global_trace.data[i]; |
3372 | p = data->trace_pages.next; | |
3373 | if (p == &data->trace_pages) { | |
3374 | /* should never happen */ | |
3375 | WARN_ON(1); | |
3376 | tracing_disabled = 1; | |
3377 | ret = -1; | |
3378 | break; | |
3379 | } | |
3380 | page = list_entry(p, struct page, lru); | |
3381 | ClearPageLRU(page); | |
3382 | list_del(&page->lru); | |
3eefae99 SR |
3383 | tracing_pages_allocated--; |
3384 | tracing_pages_allocated--; | |
a98a3c3f SR |
3385 | __free_page(page); |
3386 | ||
3387 | tracing_reset(data); | |
3388 | ||
3389 | #ifdef CONFIG_TRACER_MAX_TRACE | |
3390 | data = max_tr.data[i]; | |
3391 | p = data->trace_pages.next; | |
3392 | if (p == &data->trace_pages) { | |
3393 | /* should never happen */ | |
3394 | WARN_ON(1); | |
3395 | tracing_disabled = 1; | |
3396 | ret = -1; | |
3397 | break; | |
3398 | } | |
3399 | page = list_entry(p, struct page, lru); | |
3400 | ClearPageLRU(page); | |
3401 | list_del(&page->lru); | |
3402 | __free_page(page); | |
3403 | ||
3404 | tracing_reset(data); | |
3405 | #endif | |
3406 | } | |
3407 | global_trace.entries -= ENTRIES_PER_PAGE; | |
3408 | ||
3409 | return ret; | |
3410 | } | |
3411 | ||
bc0c38d1 SR |
3412 | __init static int tracer_alloc_buffers(void) |
3413 | { | |
4c11d7ae SR |
3414 | struct trace_array_cpu *data; |
3415 | void *array; | |
3416 | struct page *page; | |
3417 | int pages = 0; | |
60a11774 | 3418 | int ret = -ENOMEM; |
bc0c38d1 SR |
3419 | int i; |
3420 | ||
ab46428c SR |
3421 | /* TODO: make the number of buffers hot pluggable with CPUS */ |
3422 | tracing_nr_buffers = num_possible_cpus(); | |
3423 | tracing_buffer_mask = cpu_possible_map; | |
3424 | ||
4c11d7ae | 3425 | /* Allocate the first page for all buffers */ |
ab46428c | 3426 | for_each_tracing_cpu(i) { |
4c11d7ae | 3427 | data = global_trace.data[i] = &per_cpu(global_trace_cpu, i); |
bc0c38d1 SR |
3428 | max_tr.data[i] = &per_cpu(max_data, i); |
3429 | ||
4c11d7ae | 3430 | array = (void *)__get_free_page(GFP_KERNEL); |
bc0c38d1 | 3431 | if (array == NULL) { |
4c11d7ae SR |
3432 | printk(KERN_ERR "tracer: failed to allocate page" |
3433 | "for trace buffer!\n"); | |
bc0c38d1 SR |
3434 | goto free_buffers; |
3435 | } | |
4c11d7ae SR |
3436 | |
3437 | /* set the array to the list */ | |
3438 | INIT_LIST_HEAD(&data->trace_pages); | |
3439 | page = virt_to_page(array); | |
3440 | list_add(&page->lru, &data->trace_pages); | |
3441 | /* use the LRU flag to differentiate the two buffers */ | |
3442 | ClearPageLRU(page); | |
bc0c38d1 | 3443 | |
a98a3c3f SR |
3444 | data->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; |
3445 | max_tr.data[i]->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; | |
3446 | ||
bc0c38d1 SR |
3447 | /* Only allocate if we are actually using the max trace */ |
3448 | #ifdef CONFIG_TRACER_MAX_TRACE | |
4c11d7ae | 3449 | array = (void *)__get_free_page(GFP_KERNEL); |
bc0c38d1 | 3450 | if (array == NULL) { |
4c11d7ae SR |
3451 | printk(KERN_ERR "tracer: failed to allocate page" |
3452 | "for trace buffer!\n"); | |
bc0c38d1 SR |
3453 | goto free_buffers; |
3454 | } | |
4c11d7ae SR |
3455 | |
3456 | INIT_LIST_HEAD(&max_tr.data[i]->trace_pages); | |
3457 | page = virt_to_page(array); | |
3458 | list_add(&page->lru, &max_tr.data[i]->trace_pages); | |
3459 | SetPageLRU(page); | |
bc0c38d1 SR |
3460 | #endif |
3461 | } | |
3462 | ||
3463 | /* | |
3464 | * Since we allocate by orders of pages, we may be able to | |
3465 | * round up a bit. | |
3466 | */ | |
4c11d7ae | 3467 | global_trace.entries = ENTRIES_PER_PAGE; |
4c11d7ae SR |
3468 | pages++; |
3469 | ||
3470 | while (global_trace.entries < trace_nr_entries) { | |
3471 | if (trace_alloc_page()) | |
3472 | break; | |
3473 | pages++; | |
3474 | } | |
89b2f978 | 3475 | max_tr.entries = global_trace.entries; |
bc0c38d1 | 3476 | |
20764ff1 JS |
3477 | pr_info("tracer: %d pages allocated for %ld entries of %ld bytes\n", |
3478 | pages, trace_nr_entries, (long)TRACE_ENTRY_SIZE); | |
bc0c38d1 SR |
3479 | pr_info(" actual entries %ld\n", global_trace.entries); |
3480 | ||
3481 | tracer_init_debugfs(); | |
3482 | ||
3483 | trace_init_cmdlines(); | |
3484 | ||
3485 | register_tracer(&no_tracer); | |
3486 | current_trace = &no_tracer; | |
3487 | ||
60a11774 | 3488 | /* All seems OK, enable tracing */ |
4902f884 | 3489 | global_trace.ctrl = tracer_enabled; |
60a11774 SR |
3490 | tracing_disabled = 0; |
3491 | ||
3f5a54e3 SR |
3492 | atomic_notifier_chain_register(&panic_notifier_list, |
3493 | &trace_panic_notifier); | |
3494 | ||
3495 | register_die_notifier(&trace_die_notifier); | |
3496 | ||
bc0c38d1 SR |
3497 | return 0; |
3498 | ||
3499 | free_buffers: | |
3500 | for (i-- ; i >= 0; i--) { | |
4c11d7ae | 3501 | struct page *page, *tmp; |
bc0c38d1 SR |
3502 | struct trace_array_cpu *data = global_trace.data[i]; |
3503 | ||
c7aafc54 | 3504 | if (data) { |
4c11d7ae SR |
3505 | list_for_each_entry_safe(page, tmp, |
3506 | &data->trace_pages, lru) { | |
c7aafc54 | 3507 | list_del_init(&page->lru); |
4c11d7ae SR |
3508 | __free_page(page); |
3509 | } | |
bc0c38d1 SR |
3510 | } |
3511 | ||
3512 | #ifdef CONFIG_TRACER_MAX_TRACE | |
3513 | data = max_tr.data[i]; | |
c7aafc54 | 3514 | if (data) { |
4c11d7ae SR |
3515 | list_for_each_entry_safe(page, tmp, |
3516 | &data->trace_pages, lru) { | |
c7aafc54 | 3517 | list_del_init(&page->lru); |
4c11d7ae SR |
3518 | __free_page(page); |
3519 | } | |
bc0c38d1 SR |
3520 | } |
3521 | #endif | |
3522 | } | |
60a11774 | 3523 | return ret; |
bc0c38d1 | 3524 | } |
60a11774 | 3525 | fs_initcall(tracer_alloc_buffers); |