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> | |
17 | #include <linux/debugfs.h> | |
4c11d7ae | 18 | #include <linux/pagemap.h> |
bc0c38d1 SR |
19 | #include <linux/hardirq.h> |
20 | #include <linux/linkage.h> | |
21 | #include <linux/uaccess.h> | |
22 | #include <linux/ftrace.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/percpu.h> | |
25 | #include <linux/ctype.h> | |
26 | #include <linux/init.h> | |
2a2cc8f7 | 27 | #include <linux/poll.h> |
bc0c38d1 SR |
28 | #include <linux/gfp.h> |
29 | #include <linux/fs.h> | |
3eefae99 | 30 | #include <linux/writeback.h> |
bc0c38d1 | 31 | |
86387f7e IM |
32 | #include <linux/stacktrace.h> |
33 | ||
bc0c38d1 SR |
34 | #include "trace.h" |
35 | ||
36 | unsigned long __read_mostly tracing_max_latency = (cycle_t)ULONG_MAX; | |
37 | unsigned long __read_mostly tracing_thresh; | |
38 | ||
ab46428c SR |
39 | static unsigned long __read_mostly tracing_nr_buffers; |
40 | static cpumask_t __read_mostly tracing_buffer_mask; | |
41 | ||
42 | #define for_each_tracing_cpu(cpu) \ | |
43 | for_each_cpu_mask(cpu, tracing_buffer_mask) | |
44 | ||
a98a3c3f | 45 | /* dummy trace to disable tracing */ |
cffae437 | 46 | static struct tracer no_tracer __read_mostly = { |
a98a3c3f SR |
47 | .name = "none", |
48 | }; | |
49 | ||
50 | static int trace_alloc_page(void); | |
51 | static int trace_free_page(void); | |
52 | ||
60a11774 SR |
53 | static int tracing_disabled = 1; |
54 | ||
3eefae99 SR |
55 | static unsigned long tracing_pages_allocated; |
56 | ||
72829bc3 | 57 | long |
bc0c38d1 SR |
58 | ns2usecs(cycle_t nsec) |
59 | { | |
60 | nsec += 500; | |
61 | do_div(nsec, 1000); | |
62 | return nsec; | |
63 | } | |
64 | ||
e309b41d | 65 | cycle_t ftrace_now(int cpu) |
750ed1a4 | 66 | { |
0fd9e0da | 67 | return cpu_clock(cpu); |
750ed1a4 IM |
68 | } |
69 | ||
4fcdae83 SR |
70 | /* |
71 | * The global_trace is the descriptor that holds the tracing | |
72 | * buffers for the live tracing. For each CPU, it contains | |
73 | * a link list of pages that will store trace entries. The | |
74 | * page descriptor of the pages in the memory is used to hold | |
75 | * the link list by linking the lru item in the page descriptor | |
76 | * to each of the pages in the buffer per CPU. | |
77 | * | |
78 | * For each active CPU there is a data field that holds the | |
79 | * pages for the buffer for that CPU. Each CPU has the same number | |
80 | * of pages allocated for its buffer. | |
81 | */ | |
bc0c38d1 SR |
82 | static struct trace_array global_trace; |
83 | ||
84 | static DEFINE_PER_CPU(struct trace_array_cpu, global_trace_cpu); | |
85 | ||
4fcdae83 SR |
86 | /* |
87 | * The max_tr is used to snapshot the global_trace when a maximum | |
88 | * latency is reached. Some tracers will use this to store a maximum | |
89 | * trace while it continues examining live traces. | |
90 | * | |
91 | * The buffers for the max_tr are set up the same as the global_trace. | |
92 | * When a snapshot is taken, the link list of the max_tr is swapped | |
93 | * with the link list of the global_trace and the buffers are reset for | |
94 | * the global_trace so the tracing can continue. | |
95 | */ | |
bc0c38d1 SR |
96 | static struct trace_array max_tr; |
97 | ||
98 | static DEFINE_PER_CPU(struct trace_array_cpu, max_data); | |
99 | ||
4fcdae83 | 100 | /* tracer_enabled is used to toggle activation of a tracer */ |
26994ead | 101 | static int tracer_enabled = 1; |
4fcdae83 SR |
102 | |
103 | /* | |
104 | * trace_nr_entries is the number of entries that is allocated | |
105 | * for a buffer. Note, the number of entries is always rounded | |
106 | * to ENTRIES_PER_PAGE. | |
107 | */ | |
57422797 | 108 | static unsigned long trace_nr_entries = 65536UL; |
bc0c38d1 | 109 | |
4fcdae83 | 110 | /* trace_types holds a link list of available tracers. */ |
bc0c38d1 | 111 | static struct tracer *trace_types __read_mostly; |
4fcdae83 SR |
112 | |
113 | /* current_trace points to the tracer that is currently active */ | |
bc0c38d1 | 114 | static struct tracer *current_trace __read_mostly; |
4fcdae83 SR |
115 | |
116 | /* | |
117 | * max_tracer_type_len is used to simplify the allocating of | |
118 | * buffers to read userspace tracer names. We keep track of | |
119 | * the longest tracer name registered. | |
120 | */ | |
bc0c38d1 SR |
121 | static int max_tracer_type_len; |
122 | ||
4fcdae83 SR |
123 | /* |
124 | * trace_types_lock is used to protect the trace_types list. | |
125 | * This lock is also used to keep user access serialized. | |
126 | * Accesses from userspace will grab this lock while userspace | |
127 | * activities happen inside the kernel. | |
128 | */ | |
bc0c38d1 | 129 | static DEFINE_MUTEX(trace_types_lock); |
4fcdae83 SR |
130 | |
131 | /* trace_wait is a waitqueue for tasks blocked on trace_poll */ | |
4e655519 IM |
132 | static DECLARE_WAIT_QUEUE_HEAD(trace_wait); |
133 | ||
4fcdae83 | 134 | /* trace_flags holds iter_ctrl options */ |
4e655519 IM |
135 | unsigned long trace_flags = TRACE_ITER_PRINT_PARENT; |
136 | ||
4fcdae83 SR |
137 | /** |
138 | * trace_wake_up - wake up tasks waiting for trace input | |
139 | * | |
140 | * Simply wakes up any task that is blocked on the trace_wait | |
141 | * queue. These is used with trace_poll for tasks polling the trace. | |
142 | */ | |
4e655519 IM |
143 | void trace_wake_up(void) |
144 | { | |
017730c1 IM |
145 | /* |
146 | * The runqueue_is_locked() can fail, but this is the best we | |
147 | * have for now: | |
148 | */ | |
149 | if (!(trace_flags & TRACE_ITER_BLOCK) && !runqueue_is_locked()) | |
4e655519 IM |
150 | wake_up(&trace_wait); |
151 | } | |
bc0c38d1 | 152 | |
4c11d7ae SR |
153 | #define ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(struct trace_entry)) |
154 | ||
bc0c38d1 SR |
155 | static int __init set_nr_entries(char *str) |
156 | { | |
c6caeeb1 SR |
157 | unsigned long nr_entries; |
158 | int ret; | |
159 | ||
bc0c38d1 SR |
160 | if (!str) |
161 | return 0; | |
c6caeeb1 SR |
162 | ret = strict_strtoul(str, 0, &nr_entries); |
163 | /* nr_entries can not be zero */ | |
164 | if (ret < 0 || nr_entries == 0) | |
165 | return 0; | |
166 | trace_nr_entries = nr_entries; | |
bc0c38d1 SR |
167 | return 1; |
168 | } | |
169 | __setup("trace_entries=", set_nr_entries); | |
170 | ||
57f50be1 SR |
171 | unsigned long nsecs_to_usecs(unsigned long nsecs) |
172 | { | |
173 | return nsecs / 1000; | |
174 | } | |
175 | ||
4fcdae83 SR |
176 | /* |
177 | * trace_flag_type is an enumeration that holds different | |
178 | * states when a trace occurs. These are: | |
179 | * IRQS_OFF - interrupts were disabled | |
180 | * NEED_RESCED - reschedule is requested | |
181 | * HARDIRQ - inside an interrupt handler | |
182 | * SOFTIRQ - inside a softirq handler | |
183 | */ | |
bc0c38d1 SR |
184 | enum trace_flag_type { |
185 | TRACE_FLAG_IRQS_OFF = 0x01, | |
186 | TRACE_FLAG_NEED_RESCHED = 0x02, | |
187 | TRACE_FLAG_HARDIRQ = 0x04, | |
188 | TRACE_FLAG_SOFTIRQ = 0x08, | |
189 | }; | |
190 | ||
4fcdae83 SR |
191 | /* |
192 | * TRACE_ITER_SYM_MASK masks the options in trace_flags that | |
193 | * control the output of kernel symbols. | |
194 | */ | |
bc0c38d1 SR |
195 | #define TRACE_ITER_SYM_MASK \ |
196 | (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR) | |
197 | ||
4fcdae83 | 198 | /* These must match the bit postions in trace_iterator_flags */ |
bc0c38d1 SR |
199 | static const char *trace_options[] = { |
200 | "print-parent", | |
201 | "sym-offset", | |
202 | "sym-addr", | |
203 | "verbose", | |
f9896bf3 | 204 | "raw", |
5e3ca0ec | 205 | "hex", |
cb0f12aa | 206 | "bin", |
2a2cc8f7 | 207 | "block", |
86387f7e | 208 | "stacktrace", |
4ac3ba41 | 209 | "sched-tree", |
bc0c38d1 SR |
210 | NULL |
211 | }; | |
212 | ||
4fcdae83 SR |
213 | /* |
214 | * ftrace_max_lock is used to protect the swapping of buffers | |
215 | * when taking a max snapshot. The buffers themselves are | |
216 | * protected by per_cpu spinlocks. But the action of the swap | |
217 | * needs its own lock. | |
218 | * | |
219 | * This is defined as a raw_spinlock_t in order to help | |
220 | * with performance when lockdep debugging is enabled. | |
221 | */ | |
92205c23 SR |
222 | static raw_spinlock_t ftrace_max_lock = |
223 | (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; | |
bc0c38d1 SR |
224 | |
225 | /* | |
226 | * Copy the new maximum trace into the separate maximum-trace | |
227 | * structure. (this way the maximum trace is permanently saved, | |
228 | * for later retrieval via /debugfs/tracing/latency_trace) | |
229 | */ | |
e309b41d | 230 | static void |
bc0c38d1 SR |
231 | __update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu) |
232 | { | |
233 | struct trace_array_cpu *data = tr->data[cpu]; | |
234 | ||
235 | max_tr.cpu = cpu; | |
236 | max_tr.time_start = data->preempt_timestamp; | |
237 | ||
238 | data = max_tr.data[cpu]; | |
239 | data->saved_latency = tracing_max_latency; | |
240 | ||
241 | memcpy(data->comm, tsk->comm, TASK_COMM_LEN); | |
242 | data->pid = tsk->pid; | |
243 | data->uid = tsk->uid; | |
244 | data->nice = tsk->static_prio - 20 - MAX_RT_PRIO; | |
245 | data->policy = tsk->policy; | |
246 | data->rt_priority = tsk->rt_priority; | |
247 | ||
248 | /* record this tasks comm */ | |
249 | tracing_record_cmdline(current); | |
250 | } | |
251 | ||
19384c03 SR |
252 | #define CHECK_COND(cond) \ |
253 | if (unlikely(cond)) { \ | |
254 | tracing_disabled = 1; \ | |
255 | WARN_ON(1); \ | |
256 | return -1; \ | |
257 | } | |
258 | ||
4fcdae83 SR |
259 | /** |
260 | * check_pages - integrity check of trace buffers | |
261 | * | |
262 | * As a safty measure we check to make sure the data pages have not | |
19384c03 | 263 | * been corrupted. |
4fcdae83 | 264 | */ |
19384c03 | 265 | int check_pages(struct trace_array_cpu *data) |
c7aafc54 IM |
266 | { |
267 | struct page *page, *tmp; | |
268 | ||
19384c03 SR |
269 | CHECK_COND(data->trace_pages.next->prev != &data->trace_pages); |
270 | CHECK_COND(data->trace_pages.prev->next != &data->trace_pages); | |
c7aafc54 IM |
271 | |
272 | list_for_each_entry_safe(page, tmp, &data->trace_pages, lru) { | |
19384c03 SR |
273 | CHECK_COND(page->lru.next->prev != &page->lru); |
274 | CHECK_COND(page->lru.prev->next != &page->lru); | |
c7aafc54 | 275 | } |
19384c03 SR |
276 | |
277 | return 0; | |
c7aafc54 IM |
278 | } |
279 | ||
4fcdae83 SR |
280 | /** |
281 | * head_page - page address of the first page in per_cpu buffer. | |
282 | * | |
283 | * head_page returns the page address of the first page in | |
284 | * a per_cpu buffer. This also preforms various consistency | |
285 | * checks to make sure the buffer has not been corrupted. | |
286 | */ | |
c7aafc54 IM |
287 | void *head_page(struct trace_array_cpu *data) |
288 | { | |
289 | struct page *page; | |
290 | ||
c7aafc54 IM |
291 | if (list_empty(&data->trace_pages)) |
292 | return NULL; | |
293 | ||
294 | page = list_entry(data->trace_pages.next, struct page, lru); | |
295 | BUG_ON(&page->lru == &data->trace_pages); | |
296 | ||
297 | return page_address(page); | |
298 | } | |
299 | ||
4fcdae83 SR |
300 | /** |
301 | * trace_seq_printf - sequence printing of trace information | |
302 | * @s: trace sequence descriptor | |
303 | * @fmt: printf format string | |
304 | * | |
305 | * The tracer may use either sequence operations or its own | |
306 | * copy to user routines. To simplify formating of a trace | |
307 | * trace_seq_printf is used to store strings into a special | |
308 | * buffer (@s). Then the output may be either used by | |
309 | * the sequencer or pulled into another buffer. | |
310 | */ | |
72829bc3 | 311 | int |
214023c3 SR |
312 | trace_seq_printf(struct trace_seq *s, const char *fmt, ...) |
313 | { | |
314 | int len = (PAGE_SIZE - 1) - s->len; | |
315 | va_list ap; | |
b3806b43 | 316 | int ret; |
214023c3 SR |
317 | |
318 | if (!len) | |
319 | return 0; | |
320 | ||
321 | va_start(ap, fmt); | |
b3806b43 | 322 | ret = vsnprintf(s->buffer + s->len, len, fmt, ap); |
214023c3 SR |
323 | va_end(ap); |
324 | ||
b3806b43 | 325 | /* If we can't write it all, don't bother writing anything */ |
72829bc3 | 326 | if (ret >= len) |
b3806b43 SR |
327 | return 0; |
328 | ||
329 | s->len += ret; | |
214023c3 SR |
330 | |
331 | return len; | |
332 | } | |
333 | ||
4fcdae83 SR |
334 | /** |
335 | * trace_seq_puts - trace sequence printing of simple string | |
336 | * @s: trace sequence descriptor | |
337 | * @str: simple string to record | |
338 | * | |
339 | * The tracer may use either the sequence operations or its own | |
340 | * copy to user routines. This function records a simple string | |
341 | * into a special buffer (@s) for later retrieval by a sequencer | |
342 | * or other mechanism. | |
343 | */ | |
e309b41d | 344 | static int |
214023c3 SR |
345 | trace_seq_puts(struct trace_seq *s, const char *str) |
346 | { | |
347 | int len = strlen(str); | |
348 | ||
349 | if (len > ((PAGE_SIZE - 1) - s->len)) | |
b3806b43 | 350 | return 0; |
214023c3 SR |
351 | |
352 | memcpy(s->buffer + s->len, str, len); | |
353 | s->len += len; | |
354 | ||
355 | return len; | |
356 | } | |
357 | ||
e309b41d | 358 | static int |
214023c3 SR |
359 | trace_seq_putc(struct trace_seq *s, unsigned char c) |
360 | { | |
361 | if (s->len >= (PAGE_SIZE - 1)) | |
362 | return 0; | |
363 | ||
364 | s->buffer[s->len++] = c; | |
365 | ||
366 | return 1; | |
367 | } | |
368 | ||
e309b41d | 369 | static int |
cb0f12aa IM |
370 | trace_seq_putmem(struct trace_seq *s, void *mem, size_t len) |
371 | { | |
372 | if (len > ((PAGE_SIZE - 1) - s->len)) | |
373 | return 0; | |
374 | ||
375 | memcpy(s->buffer + s->len, mem, len); | |
376 | s->len += len; | |
377 | ||
378 | return len; | |
379 | } | |
380 | ||
5e3ca0ec | 381 | #define HEX_CHARS 17 |
93dcc6ea | 382 | static const char hex2asc[] = "0123456789abcdef"; |
5e3ca0ec | 383 | |
e309b41d | 384 | static int |
5e3ca0ec IM |
385 | trace_seq_putmem_hex(struct trace_seq *s, void *mem, size_t len) |
386 | { | |
387 | unsigned char hex[HEX_CHARS]; | |
93dcc6ea | 388 | unsigned char *data = mem; |
5e3ca0ec IM |
389 | unsigned char byte; |
390 | int i, j; | |
391 | ||
392 | BUG_ON(len >= HEX_CHARS); | |
393 | ||
5e3ca0ec IM |
394 | #ifdef __BIG_ENDIAN |
395 | for (i = 0, j = 0; i < len; i++) { | |
396 | #else | |
397 | for (i = len-1, j = 0; i >= 0; i--) { | |
398 | #endif | |
399 | byte = data[i]; | |
400 | ||
93dcc6ea TG |
401 | hex[j++] = hex2asc[byte & 0x0f]; |
402 | hex[j++] = hex2asc[byte >> 4]; | |
5e3ca0ec | 403 | } |
93dcc6ea | 404 | hex[j++] = ' '; |
5e3ca0ec IM |
405 | |
406 | return trace_seq_putmem(s, hex, j); | |
407 | } | |
408 | ||
e309b41d | 409 | static void |
214023c3 SR |
410 | trace_seq_reset(struct trace_seq *s) |
411 | { | |
412 | s->len = 0; | |
6c6c2796 PP |
413 | s->readpos = 0; |
414 | } | |
415 | ||
416 | ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, size_t cnt) | |
417 | { | |
418 | int len; | |
419 | int ret; | |
420 | ||
421 | if (s->len <= s->readpos) | |
422 | return -EBUSY; | |
423 | ||
424 | len = s->len - s->readpos; | |
425 | if (cnt > len) | |
426 | cnt = len; | |
427 | ret = copy_to_user(ubuf, s->buffer + s->readpos, cnt); | |
428 | if (ret) | |
429 | return -EFAULT; | |
430 | ||
431 | s->readpos += len; | |
432 | return cnt; | |
214023c3 SR |
433 | } |
434 | ||
e309b41d | 435 | static void |
214023c3 SR |
436 | trace_print_seq(struct seq_file *m, struct trace_seq *s) |
437 | { | |
438 | int len = s->len >= PAGE_SIZE ? PAGE_SIZE - 1 : s->len; | |
439 | ||
440 | s->buffer[len] = 0; | |
441 | seq_puts(m, s->buffer); | |
442 | ||
443 | trace_seq_reset(s); | |
444 | } | |
445 | ||
4fcdae83 SR |
446 | /* |
447 | * flip the trace buffers between two trace descriptors. | |
448 | * This usually is the buffers between the global_trace and | |
449 | * the max_tr to record a snapshot of a current trace. | |
450 | * | |
451 | * The ftrace_max_lock must be held. | |
452 | */ | |
e309b41d | 453 | static void |
c7aafc54 IM |
454 | flip_trace(struct trace_array_cpu *tr1, struct trace_array_cpu *tr2) |
455 | { | |
456 | struct list_head flip_pages; | |
457 | ||
458 | INIT_LIST_HEAD(&flip_pages); | |
459 | ||
93a588f4 | 460 | memcpy(&tr1->trace_head_idx, &tr2->trace_head_idx, |
c7aafc54 | 461 | sizeof(struct trace_array_cpu) - |
93a588f4 | 462 | offsetof(struct trace_array_cpu, trace_head_idx)); |
c7aafc54 IM |
463 | |
464 | check_pages(tr1); | |
465 | check_pages(tr2); | |
466 | list_splice_init(&tr1->trace_pages, &flip_pages); | |
467 | list_splice_init(&tr2->trace_pages, &tr1->trace_pages); | |
468 | list_splice_init(&flip_pages, &tr2->trace_pages); | |
469 | BUG_ON(!list_empty(&flip_pages)); | |
470 | check_pages(tr1); | |
471 | check_pages(tr2); | |
472 | } | |
473 | ||
4fcdae83 SR |
474 | /** |
475 | * update_max_tr - snapshot all trace buffers from global_trace to max_tr | |
476 | * @tr: tracer | |
477 | * @tsk: the task with the latency | |
478 | * @cpu: The cpu that initiated the trace. | |
479 | * | |
480 | * Flip the buffers between the @tr and the max_tr and record information | |
481 | * about which task was the cause of this latency. | |
482 | */ | |
e309b41d | 483 | void |
bc0c38d1 SR |
484 | update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu) |
485 | { | |
486 | struct trace_array_cpu *data; | |
bc0c38d1 SR |
487 | int i; |
488 | ||
4c11d7ae | 489 | WARN_ON_ONCE(!irqs_disabled()); |
92205c23 | 490 | __raw_spin_lock(&ftrace_max_lock); |
bc0c38d1 | 491 | /* clear out all the previous traces */ |
ab46428c | 492 | for_each_tracing_cpu(i) { |
bc0c38d1 | 493 | data = tr->data[i]; |
c7aafc54 | 494 | flip_trace(max_tr.data[i], data); |
89b2f978 | 495 | tracing_reset(data); |
bc0c38d1 SR |
496 | } |
497 | ||
498 | __update_max_tr(tr, tsk, cpu); | |
92205c23 | 499 | __raw_spin_unlock(&ftrace_max_lock); |
bc0c38d1 SR |
500 | } |
501 | ||
502 | /** | |
503 | * update_max_tr_single - only copy one trace over, and reset the rest | |
504 | * @tr - tracer | |
505 | * @tsk - task with the latency | |
506 | * @cpu - the cpu of the buffer to copy. | |
4fcdae83 SR |
507 | * |
508 | * Flip the trace of a single CPU buffer between the @tr and the max_tr. | |
bc0c38d1 | 509 | */ |
e309b41d | 510 | void |
bc0c38d1 SR |
511 | update_max_tr_single(struct trace_array *tr, struct task_struct *tsk, int cpu) |
512 | { | |
513 | struct trace_array_cpu *data = tr->data[cpu]; | |
bc0c38d1 SR |
514 | int i; |
515 | ||
4c11d7ae | 516 | WARN_ON_ONCE(!irqs_disabled()); |
92205c23 | 517 | __raw_spin_lock(&ftrace_max_lock); |
ab46428c | 518 | for_each_tracing_cpu(i) |
bc0c38d1 SR |
519 | tracing_reset(max_tr.data[i]); |
520 | ||
c7aafc54 | 521 | flip_trace(max_tr.data[cpu], data); |
89b2f978 | 522 | tracing_reset(data); |
bc0c38d1 SR |
523 | |
524 | __update_max_tr(tr, tsk, cpu); | |
92205c23 | 525 | __raw_spin_unlock(&ftrace_max_lock); |
bc0c38d1 SR |
526 | } |
527 | ||
4fcdae83 SR |
528 | /** |
529 | * register_tracer - register a tracer with the ftrace system. | |
530 | * @type - the plugin for the tracer | |
531 | * | |
532 | * Register a new plugin tracer. | |
533 | */ | |
bc0c38d1 SR |
534 | int register_tracer(struct tracer *type) |
535 | { | |
536 | struct tracer *t; | |
537 | int len; | |
538 | int ret = 0; | |
539 | ||
540 | if (!type->name) { | |
541 | pr_info("Tracer must have a name\n"); | |
542 | return -1; | |
543 | } | |
544 | ||
545 | mutex_lock(&trace_types_lock); | |
546 | for (t = trace_types; t; t = t->next) { | |
547 | if (strcmp(type->name, t->name) == 0) { | |
548 | /* already found */ | |
549 | pr_info("Trace %s already registered\n", | |
550 | type->name); | |
551 | ret = -1; | |
552 | goto out; | |
553 | } | |
554 | } | |
555 | ||
60a11774 SR |
556 | #ifdef CONFIG_FTRACE_STARTUP_TEST |
557 | if (type->selftest) { | |
558 | struct tracer *saved_tracer = current_trace; | |
559 | struct trace_array_cpu *data; | |
560 | struct trace_array *tr = &global_trace; | |
561 | int saved_ctrl = tr->ctrl; | |
562 | int i; | |
563 | /* | |
564 | * Run a selftest on this tracer. | |
565 | * Here we reset the trace buffer, and set the current | |
566 | * tracer to be this tracer. The tracer can then run some | |
567 | * internal tracing to verify that everything is in order. | |
568 | * If we fail, we do not register this tracer. | |
569 | */ | |
ab46428c | 570 | for_each_tracing_cpu(i) { |
60a11774 | 571 | data = tr->data[i]; |
c7aafc54 IM |
572 | if (!head_page(data)) |
573 | continue; | |
60a11774 SR |
574 | tracing_reset(data); |
575 | } | |
576 | current_trace = type; | |
577 | tr->ctrl = 0; | |
578 | /* the test is responsible for initializing and enabling */ | |
579 | pr_info("Testing tracer %s: ", type->name); | |
580 | ret = type->selftest(type, tr); | |
581 | /* the test is responsible for resetting too */ | |
582 | current_trace = saved_tracer; | |
583 | tr->ctrl = saved_ctrl; | |
584 | if (ret) { | |
585 | printk(KERN_CONT "FAILED!\n"); | |
586 | goto out; | |
587 | } | |
1d4db00a | 588 | /* Only reset on passing, to avoid touching corrupted buffers */ |
ab46428c | 589 | for_each_tracing_cpu(i) { |
1d4db00a SR |
590 | data = tr->data[i]; |
591 | if (!head_page(data)) | |
592 | continue; | |
593 | tracing_reset(data); | |
594 | } | |
60a11774 SR |
595 | printk(KERN_CONT "PASSED\n"); |
596 | } | |
597 | #endif | |
598 | ||
bc0c38d1 SR |
599 | type->next = trace_types; |
600 | trace_types = type; | |
601 | len = strlen(type->name); | |
602 | if (len > max_tracer_type_len) | |
603 | max_tracer_type_len = len; | |
60a11774 | 604 | |
bc0c38d1 SR |
605 | out: |
606 | mutex_unlock(&trace_types_lock); | |
607 | ||
608 | return ret; | |
609 | } | |
610 | ||
611 | void unregister_tracer(struct tracer *type) | |
612 | { | |
613 | struct tracer **t; | |
614 | int len; | |
615 | ||
616 | mutex_lock(&trace_types_lock); | |
617 | for (t = &trace_types; *t; t = &(*t)->next) { | |
618 | if (*t == type) | |
619 | goto found; | |
620 | } | |
621 | pr_info("Trace %s not registered\n", type->name); | |
622 | goto out; | |
623 | ||
624 | found: | |
625 | *t = (*t)->next; | |
626 | if (strlen(type->name) != max_tracer_type_len) | |
627 | goto out; | |
628 | ||
629 | max_tracer_type_len = 0; | |
630 | for (t = &trace_types; *t; t = &(*t)->next) { | |
631 | len = strlen((*t)->name); | |
632 | if (len > max_tracer_type_len) | |
633 | max_tracer_type_len = len; | |
634 | } | |
635 | out: | |
636 | mutex_unlock(&trace_types_lock); | |
637 | } | |
638 | ||
e309b41d | 639 | void tracing_reset(struct trace_array_cpu *data) |
bc0c38d1 SR |
640 | { |
641 | data->trace_idx = 0; | |
53d0aa77 | 642 | data->overrun = 0; |
93a588f4 SR |
643 | data->trace_head = data->trace_tail = head_page(data); |
644 | data->trace_head_idx = 0; | |
645 | data->trace_tail_idx = 0; | |
bc0c38d1 SR |
646 | } |
647 | ||
bc0c38d1 SR |
648 | #define SAVED_CMDLINES 128 |
649 | static unsigned map_pid_to_cmdline[PID_MAX_DEFAULT+1]; | |
650 | static unsigned map_cmdline_to_pid[SAVED_CMDLINES]; | |
651 | static char saved_cmdlines[SAVED_CMDLINES][TASK_COMM_LEN]; | |
652 | static int cmdline_idx; | |
653 | static DEFINE_SPINLOCK(trace_cmdline_lock); | |
25b0b44a | 654 | |
25b0b44a SR |
655 | /* temporary disable recording */ |
656 | atomic_t trace_record_cmdline_disabled __read_mostly; | |
bc0c38d1 SR |
657 | |
658 | static void trace_init_cmdlines(void) | |
659 | { | |
660 | memset(&map_pid_to_cmdline, -1, sizeof(map_pid_to_cmdline)); | |
661 | memset(&map_cmdline_to_pid, -1, sizeof(map_cmdline_to_pid)); | |
662 | cmdline_idx = 0; | |
663 | } | |
664 | ||
e309b41d | 665 | void trace_stop_cmdline_recording(void); |
bc0c38d1 | 666 | |
e309b41d | 667 | static void trace_save_cmdline(struct task_struct *tsk) |
bc0c38d1 SR |
668 | { |
669 | unsigned map; | |
670 | unsigned idx; | |
671 | ||
672 | if (!tsk->pid || unlikely(tsk->pid > PID_MAX_DEFAULT)) | |
673 | return; | |
674 | ||
675 | /* | |
676 | * It's not the end of the world if we don't get | |
677 | * the lock, but we also don't want to spin | |
678 | * nor do we want to disable interrupts, | |
679 | * so if we miss here, then better luck next time. | |
680 | */ | |
681 | if (!spin_trylock(&trace_cmdline_lock)) | |
682 | return; | |
683 | ||
684 | idx = map_pid_to_cmdline[tsk->pid]; | |
685 | if (idx >= SAVED_CMDLINES) { | |
686 | idx = (cmdline_idx + 1) % SAVED_CMDLINES; | |
687 | ||
688 | map = map_cmdline_to_pid[idx]; | |
689 | if (map <= PID_MAX_DEFAULT) | |
690 | map_pid_to_cmdline[map] = (unsigned)-1; | |
691 | ||
692 | map_pid_to_cmdline[tsk->pid] = idx; | |
693 | ||
694 | cmdline_idx = idx; | |
695 | } | |
696 | ||
697 | memcpy(&saved_cmdlines[idx], tsk->comm, TASK_COMM_LEN); | |
698 | ||
699 | spin_unlock(&trace_cmdline_lock); | |
700 | } | |
701 | ||
e309b41d | 702 | static char *trace_find_cmdline(int pid) |
bc0c38d1 SR |
703 | { |
704 | char *cmdline = "<...>"; | |
705 | unsigned map; | |
706 | ||
707 | if (!pid) | |
708 | return "<idle>"; | |
709 | ||
710 | if (pid > PID_MAX_DEFAULT) | |
711 | goto out; | |
712 | ||
713 | map = map_pid_to_cmdline[pid]; | |
714 | if (map >= SAVED_CMDLINES) | |
715 | goto out; | |
716 | ||
717 | cmdline = saved_cmdlines[map]; | |
718 | ||
719 | out: | |
720 | return cmdline; | |
721 | } | |
722 | ||
e309b41d | 723 | void tracing_record_cmdline(struct task_struct *tsk) |
bc0c38d1 SR |
724 | { |
725 | if (atomic_read(&trace_record_cmdline_disabled)) | |
726 | return; | |
727 | ||
728 | trace_save_cmdline(tsk); | |
729 | } | |
730 | ||
e309b41d | 731 | static inline struct list_head * |
93a588f4 SR |
732 | trace_next_list(struct trace_array_cpu *data, struct list_head *next) |
733 | { | |
734 | /* | |
735 | * Roundrobin - but skip the head (which is not a real page): | |
736 | */ | |
737 | next = next->next; | |
738 | if (unlikely(next == &data->trace_pages)) | |
739 | next = next->next; | |
740 | BUG_ON(next == &data->trace_pages); | |
741 | ||
742 | return next; | |
743 | } | |
744 | ||
e309b41d | 745 | static inline void * |
93a588f4 SR |
746 | trace_next_page(struct trace_array_cpu *data, void *addr) |
747 | { | |
748 | struct list_head *next; | |
749 | struct page *page; | |
750 | ||
751 | page = virt_to_page(addr); | |
752 | ||
753 | next = trace_next_list(data, &page->lru); | |
754 | page = list_entry(next, struct page, lru); | |
755 | ||
756 | return page_address(page); | |
757 | } | |
758 | ||
e309b41d | 759 | static inline struct trace_entry * |
c7aafc54 | 760 | tracing_get_trace_entry(struct trace_array *tr, struct trace_array_cpu *data) |
bc0c38d1 SR |
761 | { |
762 | unsigned long idx, idx_next; | |
763 | struct trace_entry *entry; | |
764 | ||
4c11d7ae | 765 | data->trace_idx++; |
93a588f4 | 766 | idx = data->trace_head_idx; |
bc0c38d1 SR |
767 | idx_next = idx + 1; |
768 | ||
c7aafc54 IM |
769 | BUG_ON(idx * TRACE_ENTRY_SIZE >= PAGE_SIZE); |
770 | ||
93a588f4 | 771 | entry = data->trace_head + idx * TRACE_ENTRY_SIZE; |
4c11d7ae SR |
772 | |
773 | if (unlikely(idx_next >= ENTRIES_PER_PAGE)) { | |
93a588f4 | 774 | data->trace_head = trace_next_page(data, data->trace_head); |
bc0c38d1 SR |
775 | idx_next = 0; |
776 | } | |
777 | ||
93a588f4 SR |
778 | if (data->trace_head == data->trace_tail && |
779 | idx_next == data->trace_tail_idx) { | |
780 | /* overrun */ | |
53d0aa77 | 781 | data->overrun++; |
93a588f4 SR |
782 | data->trace_tail_idx++; |
783 | if (data->trace_tail_idx >= ENTRIES_PER_PAGE) { | |
784 | data->trace_tail = | |
785 | trace_next_page(data, data->trace_tail); | |
786 | data->trace_tail_idx = 0; | |
787 | } | |
788 | } | |
789 | ||
790 | data->trace_head_idx = idx_next; | |
bc0c38d1 SR |
791 | |
792 | return entry; | |
793 | } | |
794 | ||
e309b41d | 795 | static inline void |
c7aafc54 | 796 | tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags) |
bc0c38d1 SR |
797 | { |
798 | struct task_struct *tsk = current; | |
799 | unsigned long pc; | |
800 | ||
801 | pc = preempt_count(); | |
802 | ||
c7aafc54 | 803 | entry->preempt_count = pc & 0xff; |
72829bc3 | 804 | entry->pid = (tsk) ? tsk->pid : 0; |
750ed1a4 | 805 | entry->t = ftrace_now(raw_smp_processor_id()); |
bc0c38d1 SR |
806 | entry->flags = (irqs_disabled_flags(flags) ? TRACE_FLAG_IRQS_OFF : 0) | |
807 | ((pc & HARDIRQ_MASK) ? TRACE_FLAG_HARDIRQ : 0) | | |
808 | ((pc & SOFTIRQ_MASK) ? TRACE_FLAG_SOFTIRQ : 0) | | |
809 | (need_resched() ? TRACE_FLAG_NEED_RESCHED : 0); | |
810 | } | |
811 | ||
e309b41d | 812 | void |
6fb44b71 SR |
813 | trace_function(struct trace_array *tr, struct trace_array_cpu *data, |
814 | unsigned long ip, unsigned long parent_ip, unsigned long flags) | |
bc0c38d1 SR |
815 | { |
816 | struct trace_entry *entry; | |
dcb6308f | 817 | unsigned long irq_flags; |
bc0c38d1 | 818 | |
92205c23 SR |
819 | raw_local_irq_save(irq_flags); |
820 | __raw_spin_lock(&data->lock); | |
c7aafc54 | 821 | entry = tracing_get_trace_entry(tr, data); |
bc0c38d1 | 822 | tracing_generic_entry_update(entry, flags); |
c7aafc54 IM |
823 | entry->type = TRACE_FN; |
824 | entry->fn.ip = ip; | |
825 | entry->fn.parent_ip = parent_ip; | |
92205c23 SR |
826 | __raw_spin_unlock(&data->lock); |
827 | raw_local_irq_restore(irq_flags); | |
bc0c38d1 SR |
828 | } |
829 | ||
e309b41d | 830 | void |
2e0f5761 IM |
831 | ftrace(struct trace_array *tr, struct trace_array_cpu *data, |
832 | unsigned long ip, unsigned long parent_ip, unsigned long flags) | |
833 | { | |
834 | if (likely(!atomic_read(&data->disabled))) | |
6fb44b71 | 835 | trace_function(tr, data, ip, parent_ip, flags); |
2e0f5761 IM |
836 | } |
837 | ||
86387f7e IM |
838 | void __trace_stack(struct trace_array *tr, |
839 | struct trace_array_cpu *data, | |
840 | unsigned long flags, | |
841 | int skip) | |
842 | { | |
843 | struct trace_entry *entry; | |
844 | struct stack_trace trace; | |
845 | ||
846 | if (!(trace_flags & TRACE_ITER_STACKTRACE)) | |
847 | return; | |
848 | ||
849 | entry = tracing_get_trace_entry(tr, data); | |
850 | tracing_generic_entry_update(entry, flags); | |
851 | entry->type = TRACE_STACK; | |
852 | ||
853 | memset(&entry->stack, 0, sizeof(entry->stack)); | |
854 | ||
855 | trace.nr_entries = 0; | |
856 | trace.max_entries = FTRACE_STACK_ENTRIES; | |
857 | trace.skip = skip; | |
858 | trace.entries = entry->stack.caller; | |
859 | ||
860 | save_stack_trace(&trace); | |
f0a920d5 IM |
861 | } |
862 | ||
a4feb834 IM |
863 | void |
864 | __trace_special(void *__tr, void *__data, | |
865 | unsigned long arg1, unsigned long arg2, unsigned long arg3) | |
866 | { | |
867 | struct trace_array_cpu *data = __data; | |
868 | struct trace_array *tr = __tr; | |
869 | struct trace_entry *entry; | |
870 | unsigned long irq_flags; | |
871 | ||
872 | raw_local_irq_save(irq_flags); | |
873 | __raw_spin_lock(&data->lock); | |
874 | entry = tracing_get_trace_entry(tr, data); | |
875 | tracing_generic_entry_update(entry, 0); | |
876 | entry->type = TRACE_SPECIAL; | |
877 | entry->special.arg1 = arg1; | |
878 | entry->special.arg2 = arg2; | |
879 | entry->special.arg3 = arg3; | |
880 | __trace_stack(tr, data, irq_flags, 4); | |
881 | __raw_spin_unlock(&data->lock); | |
882 | raw_local_irq_restore(irq_flags); | |
883 | ||
884 | trace_wake_up(); | |
885 | } | |
886 | ||
e309b41d | 887 | void |
bc0c38d1 SR |
888 | tracing_sched_switch_trace(struct trace_array *tr, |
889 | struct trace_array_cpu *data, | |
86387f7e IM |
890 | struct task_struct *prev, |
891 | struct task_struct *next, | |
bc0c38d1 SR |
892 | unsigned long flags) |
893 | { | |
894 | struct trace_entry *entry; | |
dcb6308f | 895 | unsigned long irq_flags; |
bc0c38d1 | 896 | |
92205c23 SR |
897 | raw_local_irq_save(irq_flags); |
898 | __raw_spin_lock(&data->lock); | |
c7aafc54 | 899 | entry = tracing_get_trace_entry(tr, data); |
bc0c38d1 SR |
900 | tracing_generic_entry_update(entry, flags); |
901 | entry->type = TRACE_CTX; | |
902 | entry->ctx.prev_pid = prev->pid; | |
903 | entry->ctx.prev_prio = prev->prio; | |
904 | entry->ctx.prev_state = prev->state; | |
905 | entry->ctx.next_pid = next->pid; | |
906 | entry->ctx.next_prio = next->prio; | |
bac524d3 | 907 | entry->ctx.next_state = next->state; |
74f4e369 | 908 | __trace_stack(tr, data, flags, 5); |
92205c23 SR |
909 | __raw_spin_unlock(&data->lock); |
910 | raw_local_irq_restore(irq_flags); | |
bc0c38d1 SR |
911 | } |
912 | ||
57422797 IM |
913 | void |
914 | tracing_sched_wakeup_trace(struct trace_array *tr, | |
915 | struct trace_array_cpu *data, | |
86387f7e IM |
916 | struct task_struct *wakee, |
917 | struct task_struct *curr, | |
57422797 IM |
918 | unsigned long flags) |
919 | { | |
920 | struct trace_entry *entry; | |
921 | unsigned long irq_flags; | |
922 | ||
92205c23 SR |
923 | raw_local_irq_save(irq_flags); |
924 | __raw_spin_lock(&data->lock); | |
57422797 IM |
925 | entry = tracing_get_trace_entry(tr, data); |
926 | tracing_generic_entry_update(entry, flags); | |
927 | entry->type = TRACE_WAKE; | |
928 | entry->ctx.prev_pid = curr->pid; | |
929 | entry->ctx.prev_prio = curr->prio; | |
930 | entry->ctx.prev_state = curr->state; | |
931 | entry->ctx.next_pid = wakee->pid; | |
932 | entry->ctx.next_prio = wakee->prio; | |
bac524d3 | 933 | entry->ctx.next_state = wakee->state; |
74f4e369 | 934 | __trace_stack(tr, data, flags, 6); |
92205c23 SR |
935 | __raw_spin_unlock(&data->lock); |
936 | raw_local_irq_restore(irq_flags); | |
017730c1 IM |
937 | |
938 | trace_wake_up(); | |
57422797 IM |
939 | } |
940 | ||
4902f884 SR |
941 | void |
942 | ftrace_special(unsigned long arg1, unsigned long arg2, unsigned long arg3) | |
943 | { | |
944 | struct trace_array *tr = &global_trace; | |
945 | struct trace_array_cpu *data; | |
946 | unsigned long flags; | |
947 | long disabled; | |
948 | int cpu; | |
949 | ||
950 | if (tracing_disabled || current_trace == &no_tracer || !tr->ctrl) | |
951 | return; | |
952 | ||
953 | local_irq_save(flags); | |
954 | cpu = raw_smp_processor_id(); | |
955 | data = tr->data[cpu]; | |
956 | disabled = atomic_inc_return(&data->disabled); | |
957 | ||
958 | if (likely(disabled == 1)) | |
959 | __trace_special(tr, data, arg1, arg2, arg3); | |
960 | ||
961 | atomic_dec(&data->disabled); | |
962 | local_irq_restore(flags); | |
963 | } | |
964 | ||
2e0f5761 | 965 | #ifdef CONFIG_FTRACE |
e309b41d | 966 | static void |
2e0f5761 IM |
967 | function_trace_call(unsigned long ip, unsigned long parent_ip) |
968 | { | |
969 | struct trace_array *tr = &global_trace; | |
970 | struct trace_array_cpu *data; | |
971 | unsigned long flags; | |
972 | long disabled; | |
973 | int cpu; | |
974 | ||
975 | if (unlikely(!tracer_enabled)) | |
976 | return; | |
977 | ||
978 | local_irq_save(flags); | |
979 | cpu = raw_smp_processor_id(); | |
980 | data = tr->data[cpu]; | |
981 | disabled = atomic_inc_return(&data->disabled); | |
982 | ||
983 | if (likely(disabled == 1)) | |
6fb44b71 | 984 | trace_function(tr, data, ip, parent_ip, flags); |
2e0f5761 IM |
985 | |
986 | atomic_dec(&data->disabled); | |
987 | local_irq_restore(flags); | |
988 | } | |
989 | ||
990 | static struct ftrace_ops trace_ops __read_mostly = | |
991 | { | |
992 | .func = function_trace_call, | |
993 | }; | |
994 | ||
e309b41d | 995 | void tracing_start_function_trace(void) |
2e0f5761 IM |
996 | { |
997 | register_ftrace_function(&trace_ops); | |
998 | } | |
999 | ||
e309b41d | 1000 | void tracing_stop_function_trace(void) |
2e0f5761 IM |
1001 | { |
1002 | unregister_ftrace_function(&trace_ops); | |
1003 | } | |
1004 | #endif | |
1005 | ||
bc0c38d1 SR |
1006 | enum trace_file_type { |
1007 | TRACE_FILE_LAT_FMT = 1, | |
1008 | }; | |
1009 | ||
1010 | static struct trace_entry * | |
4c11d7ae SR |
1011 | trace_entry_idx(struct trace_array *tr, struct trace_array_cpu *data, |
1012 | struct trace_iterator *iter, int cpu) | |
bc0c38d1 | 1013 | { |
4c11d7ae SR |
1014 | struct page *page; |
1015 | struct trace_entry *array; | |
bc0c38d1 | 1016 | |
4c11d7ae | 1017 | if (iter->next_idx[cpu] >= tr->entries || |
b3806b43 SR |
1018 | iter->next_idx[cpu] >= data->trace_idx || |
1019 | (data->trace_head == data->trace_tail && | |
1020 | data->trace_head_idx == data->trace_tail_idx)) | |
bc0c38d1 SR |
1021 | return NULL; |
1022 | ||
4c11d7ae | 1023 | if (!iter->next_page[cpu]) { |
93a588f4 SR |
1024 | /* Initialize the iterator for this cpu trace buffer */ |
1025 | WARN_ON(!data->trace_tail); | |
1026 | page = virt_to_page(data->trace_tail); | |
1027 | iter->next_page[cpu] = &page->lru; | |
1028 | iter->next_page_idx[cpu] = data->trace_tail_idx; | |
4c11d7ae | 1029 | } |
bc0c38d1 | 1030 | |
4c11d7ae | 1031 | page = list_entry(iter->next_page[cpu], struct page, lru); |
c7aafc54 IM |
1032 | BUG_ON(&data->trace_pages == &page->lru); |
1033 | ||
4c11d7ae SR |
1034 | array = page_address(page); |
1035 | ||
93a588f4 | 1036 | WARN_ON(iter->next_page_idx[cpu] >= ENTRIES_PER_PAGE); |
4c11d7ae | 1037 | return &array[iter->next_page_idx[cpu]]; |
bc0c38d1 SR |
1038 | } |
1039 | ||
e309b41d | 1040 | static struct trace_entry * |
bc0c38d1 SR |
1041 | find_next_entry(struct trace_iterator *iter, int *ent_cpu) |
1042 | { | |
1043 | struct trace_array *tr = iter->tr; | |
1044 | struct trace_entry *ent, *next = NULL; | |
1045 | int next_cpu = -1; | |
1046 | int cpu; | |
1047 | ||
ab46428c | 1048 | for_each_tracing_cpu(cpu) { |
c7aafc54 | 1049 | if (!head_page(tr->data[cpu])) |
bc0c38d1 | 1050 | continue; |
4c11d7ae | 1051 | ent = trace_entry_idx(tr, tr->data[cpu], iter, cpu); |
cdd31cd2 IM |
1052 | /* |
1053 | * Pick the entry with the smallest timestamp: | |
1054 | */ | |
1055 | if (ent && (!next || ent->t < next->t)) { | |
bc0c38d1 SR |
1056 | next = ent; |
1057 | next_cpu = cpu; | |
1058 | } | |
1059 | } | |
1060 | ||
1061 | if (ent_cpu) | |
1062 | *ent_cpu = next_cpu; | |
1063 | ||
1064 | return next; | |
1065 | } | |
1066 | ||
e309b41d | 1067 | static void trace_iterator_increment(struct trace_iterator *iter) |
bc0c38d1 | 1068 | { |
b3806b43 SR |
1069 | iter->idx++; |
1070 | iter->next_idx[iter->cpu]++; | |
1071 | iter->next_page_idx[iter->cpu]++; | |
8c523a9d | 1072 | |
b3806b43 SR |
1073 | if (iter->next_page_idx[iter->cpu] >= ENTRIES_PER_PAGE) { |
1074 | struct trace_array_cpu *data = iter->tr->data[iter->cpu]; | |
bc0c38d1 | 1075 | |
b3806b43 SR |
1076 | iter->next_page_idx[iter->cpu] = 0; |
1077 | iter->next_page[iter->cpu] = | |
1078 | trace_next_list(data, iter->next_page[iter->cpu]); | |
1079 | } | |
1080 | } | |
bc0c38d1 | 1081 | |
e309b41d | 1082 | static void trace_consume(struct trace_iterator *iter) |
b3806b43 SR |
1083 | { |
1084 | struct trace_array_cpu *data = iter->tr->data[iter->cpu]; | |
1085 | ||
1086 | data->trace_tail_idx++; | |
1087 | if (data->trace_tail_idx >= ENTRIES_PER_PAGE) { | |
1088 | data->trace_tail = trace_next_page(data, data->trace_tail); | |
1089 | data->trace_tail_idx = 0; | |
1090 | } | |
4e3c3333 | 1091 | |
b3806b43 SR |
1092 | /* Check if we empty it, then reset the index */ |
1093 | if (data->trace_head == data->trace_tail && | |
1094 | data->trace_head_idx == data->trace_tail_idx) | |
1095 | data->trace_idx = 0; | |
b3806b43 SR |
1096 | } |
1097 | ||
e309b41d | 1098 | static void *find_next_entry_inc(struct trace_iterator *iter) |
b3806b43 SR |
1099 | { |
1100 | struct trace_entry *next; | |
1101 | int next_cpu = -1; | |
1102 | ||
1103 | next = find_next_entry(iter, &next_cpu); | |
93a588f4 | 1104 | |
4e3c3333 IM |
1105 | iter->prev_ent = iter->ent; |
1106 | iter->prev_cpu = iter->cpu; | |
1107 | ||
bc0c38d1 SR |
1108 | iter->ent = next; |
1109 | iter->cpu = next_cpu; | |
1110 | ||
b3806b43 SR |
1111 | if (next) |
1112 | trace_iterator_increment(iter); | |
1113 | ||
bc0c38d1 SR |
1114 | return next ? iter : NULL; |
1115 | } | |
1116 | ||
e309b41d | 1117 | static void *s_next(struct seq_file *m, void *v, loff_t *pos) |
bc0c38d1 SR |
1118 | { |
1119 | struct trace_iterator *iter = m->private; | |
bc0c38d1 SR |
1120 | void *last_ent = iter->ent; |
1121 | int i = (int)*pos; | |
4e3c3333 | 1122 | void *ent; |
bc0c38d1 SR |
1123 | |
1124 | (*pos)++; | |
1125 | ||
1126 | /* can't go backwards */ | |
1127 | if (iter->idx > i) | |
1128 | return NULL; | |
1129 | ||
1130 | if (iter->idx < 0) | |
1131 | ent = find_next_entry_inc(iter); | |
1132 | else | |
1133 | ent = iter; | |
1134 | ||
1135 | while (ent && iter->idx < i) | |
1136 | ent = find_next_entry_inc(iter); | |
1137 | ||
1138 | iter->pos = *pos; | |
1139 | ||
1140 | if (last_ent && !ent) | |
1141 | seq_puts(m, "\n\nvim:ft=help\n"); | |
1142 | ||
1143 | return ent; | |
1144 | } | |
1145 | ||
1146 | static void *s_start(struct seq_file *m, loff_t *pos) | |
1147 | { | |
1148 | struct trace_iterator *iter = m->private; | |
1149 | void *p = NULL; | |
1150 | loff_t l = 0; | |
1151 | int i; | |
1152 | ||
1153 | mutex_lock(&trace_types_lock); | |
1154 | ||
d15f57f2 SR |
1155 | if (!current_trace || current_trace != iter->trace) { |
1156 | mutex_unlock(&trace_types_lock); | |
bc0c38d1 | 1157 | return NULL; |
d15f57f2 | 1158 | } |
bc0c38d1 SR |
1159 | |
1160 | atomic_inc(&trace_record_cmdline_disabled); | |
1161 | ||
1162 | /* let the tracer grab locks here if needed */ | |
1163 | if (current_trace->start) | |
1164 | current_trace->start(iter); | |
1165 | ||
1166 | if (*pos != iter->pos) { | |
1167 | iter->ent = NULL; | |
1168 | iter->cpu = 0; | |
1169 | iter->idx = -1; | |
4e3c3333 IM |
1170 | iter->prev_ent = NULL; |
1171 | iter->prev_cpu = -1; | |
bc0c38d1 | 1172 | |
ab46428c | 1173 | for_each_tracing_cpu(i) { |
bc0c38d1 | 1174 | iter->next_idx[i] = 0; |
4c11d7ae SR |
1175 | iter->next_page[i] = NULL; |
1176 | } | |
bc0c38d1 SR |
1177 | |
1178 | for (p = iter; p && l < *pos; p = s_next(m, p, &l)) | |
1179 | ; | |
1180 | ||
1181 | } else { | |
4c11d7ae | 1182 | l = *pos - 1; |
bc0c38d1 SR |
1183 | p = s_next(m, p, &l); |
1184 | } | |
1185 | ||
1186 | return p; | |
1187 | } | |
1188 | ||
1189 | static void s_stop(struct seq_file *m, void *p) | |
1190 | { | |
1191 | struct trace_iterator *iter = m->private; | |
1192 | ||
1193 | atomic_dec(&trace_record_cmdline_disabled); | |
1194 | ||
1195 | /* let the tracer release locks here if needed */ | |
1196 | if (current_trace && current_trace == iter->trace && iter->trace->stop) | |
1197 | iter->trace->stop(iter); | |
1198 | ||
1199 | mutex_unlock(&trace_types_lock); | |
1200 | } | |
1201 | ||
b3806b43 | 1202 | static int |
214023c3 | 1203 | seq_print_sym_short(struct trace_seq *s, const char *fmt, unsigned long address) |
bc0c38d1 SR |
1204 | { |
1205 | #ifdef CONFIG_KALLSYMS | |
1206 | char str[KSYM_SYMBOL_LEN]; | |
1207 | ||
1208 | kallsyms_lookup(address, NULL, NULL, NULL, str); | |
1209 | ||
b3806b43 | 1210 | return trace_seq_printf(s, fmt, str); |
bc0c38d1 | 1211 | #endif |
b3806b43 | 1212 | return 1; |
bc0c38d1 SR |
1213 | } |
1214 | ||
b3806b43 | 1215 | static int |
214023c3 SR |
1216 | seq_print_sym_offset(struct trace_seq *s, const char *fmt, |
1217 | unsigned long address) | |
bc0c38d1 SR |
1218 | { |
1219 | #ifdef CONFIG_KALLSYMS | |
1220 | char str[KSYM_SYMBOL_LEN]; | |
1221 | ||
1222 | sprint_symbol(str, address); | |
b3806b43 | 1223 | return trace_seq_printf(s, fmt, str); |
bc0c38d1 | 1224 | #endif |
b3806b43 | 1225 | return 1; |
bc0c38d1 SR |
1226 | } |
1227 | ||
1228 | #ifndef CONFIG_64BIT | |
1229 | # define IP_FMT "%08lx" | |
1230 | #else | |
1231 | # define IP_FMT "%016lx" | |
1232 | #endif | |
1233 | ||
e309b41d | 1234 | static int |
214023c3 | 1235 | seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags) |
bc0c38d1 | 1236 | { |
b3806b43 SR |
1237 | int ret; |
1238 | ||
1239 | if (!ip) | |
1240 | return trace_seq_printf(s, "0"); | |
bc0c38d1 SR |
1241 | |
1242 | if (sym_flags & TRACE_ITER_SYM_OFFSET) | |
b3806b43 | 1243 | ret = seq_print_sym_offset(s, "%s", ip); |
bc0c38d1 | 1244 | else |
b3806b43 SR |
1245 | ret = seq_print_sym_short(s, "%s", ip); |
1246 | ||
1247 | if (!ret) | |
1248 | return 0; | |
bc0c38d1 SR |
1249 | |
1250 | if (sym_flags & TRACE_ITER_SYM_ADDR) | |
b3806b43 SR |
1251 | ret = trace_seq_printf(s, " <" IP_FMT ">", ip); |
1252 | return ret; | |
bc0c38d1 SR |
1253 | } |
1254 | ||
e309b41d | 1255 | static void print_lat_help_header(struct seq_file *m) |
bc0c38d1 SR |
1256 | { |
1257 | seq_puts(m, "# _------=> CPU# \n"); | |
1258 | seq_puts(m, "# / _-----=> irqs-off \n"); | |
1259 | seq_puts(m, "# | / _----=> need-resched \n"); | |
1260 | seq_puts(m, "# || / _---=> hardirq/softirq \n"); | |
1261 | seq_puts(m, "# ||| / _--=> preempt-depth \n"); | |
1262 | seq_puts(m, "# |||| / \n"); | |
1263 | seq_puts(m, "# ||||| delay \n"); | |
1264 | seq_puts(m, "# cmd pid ||||| time | caller \n"); | |
1265 | seq_puts(m, "# \\ / ||||| \\ | / \n"); | |
1266 | } | |
1267 | ||
e309b41d | 1268 | static void print_func_help_header(struct seq_file *m) |
bc0c38d1 SR |
1269 | { |
1270 | seq_puts(m, "# TASK-PID CPU# TIMESTAMP FUNCTION\n"); | |
1271 | seq_puts(m, "# | | | | |\n"); | |
1272 | } | |
1273 | ||
1274 | ||
e309b41d | 1275 | static void |
bc0c38d1 SR |
1276 | print_trace_header(struct seq_file *m, struct trace_iterator *iter) |
1277 | { | |
1278 | unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK); | |
1279 | struct trace_array *tr = iter->tr; | |
1280 | struct trace_array_cpu *data = tr->data[tr->cpu]; | |
1281 | struct tracer *type = current_trace; | |
4c11d7ae SR |
1282 | unsigned long total = 0; |
1283 | unsigned long entries = 0; | |
bc0c38d1 SR |
1284 | int cpu; |
1285 | const char *name = "preemption"; | |
1286 | ||
1287 | if (type) | |
1288 | name = type->name; | |
1289 | ||
ab46428c | 1290 | for_each_tracing_cpu(cpu) { |
c7aafc54 | 1291 | if (head_page(tr->data[cpu])) { |
4c11d7ae SR |
1292 | total += tr->data[cpu]->trace_idx; |
1293 | if (tr->data[cpu]->trace_idx > tr->entries) | |
bc0c38d1 | 1294 | entries += tr->entries; |
4c11d7ae | 1295 | else |
bc0c38d1 SR |
1296 | entries += tr->data[cpu]->trace_idx; |
1297 | } | |
1298 | } | |
1299 | ||
1300 | seq_printf(m, "%s latency trace v1.1.5 on %s\n", | |
1301 | name, UTS_RELEASE); | |
1302 | seq_puts(m, "-----------------------------------" | |
1303 | "---------------------------------\n"); | |
1304 | seq_printf(m, " latency: %lu us, #%lu/%lu, CPU#%d |" | |
1305 | " (M:%s VP:%d, KP:%d, SP:%d HP:%d", | |
57f50be1 | 1306 | nsecs_to_usecs(data->saved_latency), |
bc0c38d1 | 1307 | entries, |
4c11d7ae | 1308 | total, |
bc0c38d1 SR |
1309 | tr->cpu, |
1310 | #if defined(CONFIG_PREEMPT_NONE) | |
1311 | "server", | |
1312 | #elif defined(CONFIG_PREEMPT_VOLUNTARY) | |
1313 | "desktop", | |
1314 | #elif defined(CONFIG_PREEMPT_DESKTOP) | |
1315 | "preempt", | |
1316 | #else | |
1317 | "unknown", | |
1318 | #endif | |
1319 | /* These are reserved for later use */ | |
1320 | 0, 0, 0, 0); | |
1321 | #ifdef CONFIG_SMP | |
1322 | seq_printf(m, " #P:%d)\n", num_online_cpus()); | |
1323 | #else | |
1324 | seq_puts(m, ")\n"); | |
1325 | #endif | |
1326 | seq_puts(m, " -----------------\n"); | |
1327 | seq_printf(m, " | task: %.16s-%d " | |
1328 | "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n", | |
1329 | data->comm, data->pid, data->uid, data->nice, | |
1330 | data->policy, data->rt_priority); | |
1331 | seq_puts(m, " -----------------\n"); | |
1332 | ||
1333 | if (data->critical_start) { | |
1334 | seq_puts(m, " => started at: "); | |
214023c3 SR |
1335 | seq_print_ip_sym(&iter->seq, data->critical_start, sym_flags); |
1336 | trace_print_seq(m, &iter->seq); | |
bc0c38d1 | 1337 | seq_puts(m, "\n => ended at: "); |
214023c3 SR |
1338 | seq_print_ip_sym(&iter->seq, data->critical_end, sym_flags); |
1339 | trace_print_seq(m, &iter->seq); | |
bc0c38d1 SR |
1340 | seq_puts(m, "\n"); |
1341 | } | |
1342 | ||
1343 | seq_puts(m, "\n"); | |
1344 | } | |
1345 | ||
e309b41d | 1346 | static void |
214023c3 | 1347 | lat_print_generic(struct trace_seq *s, struct trace_entry *entry, int cpu) |
bc0c38d1 SR |
1348 | { |
1349 | int hardirq, softirq; | |
1350 | char *comm; | |
1351 | ||
1352 | comm = trace_find_cmdline(entry->pid); | |
1353 | ||
214023c3 SR |
1354 | trace_seq_printf(s, "%8.8s-%-5d ", comm, entry->pid); |
1355 | trace_seq_printf(s, "%d", cpu); | |
1356 | trace_seq_printf(s, "%c%c", | |
1357 | (entry->flags & TRACE_FLAG_IRQS_OFF) ? 'd' : '.', | |
1358 | ((entry->flags & TRACE_FLAG_NEED_RESCHED) ? 'N' : '.')); | |
bc0c38d1 SR |
1359 | |
1360 | hardirq = entry->flags & TRACE_FLAG_HARDIRQ; | |
1361 | softirq = entry->flags & TRACE_FLAG_SOFTIRQ; | |
afc2abc0 | 1362 | if (hardirq && softirq) { |
214023c3 | 1363 | trace_seq_putc(s, 'H'); |
afc2abc0 IM |
1364 | } else { |
1365 | if (hardirq) { | |
214023c3 | 1366 | trace_seq_putc(s, 'h'); |
afc2abc0 | 1367 | } else { |
bc0c38d1 | 1368 | if (softirq) |
214023c3 | 1369 | trace_seq_putc(s, 's'); |
bc0c38d1 | 1370 | else |
214023c3 | 1371 | trace_seq_putc(s, '.'); |
bc0c38d1 SR |
1372 | } |
1373 | } | |
1374 | ||
1375 | if (entry->preempt_count) | |
214023c3 | 1376 | trace_seq_printf(s, "%x", entry->preempt_count); |
bc0c38d1 | 1377 | else |
214023c3 | 1378 | trace_seq_puts(s, "."); |
bc0c38d1 SR |
1379 | } |
1380 | ||
1381 | unsigned long preempt_mark_thresh = 100; | |
1382 | ||
e309b41d | 1383 | static void |
214023c3 | 1384 | lat_print_timestamp(struct trace_seq *s, unsigned long long abs_usecs, |
bc0c38d1 SR |
1385 | unsigned long rel_usecs) |
1386 | { | |
214023c3 | 1387 | trace_seq_printf(s, " %4lldus", abs_usecs); |
bc0c38d1 | 1388 | if (rel_usecs > preempt_mark_thresh) |
214023c3 | 1389 | trace_seq_puts(s, "!: "); |
bc0c38d1 | 1390 | else if (rel_usecs > 1) |
214023c3 | 1391 | trace_seq_puts(s, "+: "); |
bc0c38d1 | 1392 | else |
214023c3 | 1393 | trace_seq_puts(s, " : "); |
bc0c38d1 SR |
1394 | } |
1395 | ||
1396 | static const char state_to_char[] = TASK_STATE_TO_CHAR_STR; | |
1397 | ||
e309b41d | 1398 | static int |
214023c3 | 1399 | print_lat_fmt(struct trace_iterator *iter, unsigned int trace_idx, int cpu) |
bc0c38d1 | 1400 | { |
214023c3 | 1401 | struct trace_seq *s = &iter->seq; |
bc0c38d1 SR |
1402 | unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK); |
1403 | struct trace_entry *next_entry = find_next_entry(iter, NULL); | |
1404 | unsigned long verbose = (trace_flags & TRACE_ITER_VERBOSE); | |
1405 | struct trace_entry *entry = iter->ent; | |
1406 | unsigned long abs_usecs; | |
1407 | unsigned long rel_usecs; | |
1408 | char *comm; | |
bac524d3 | 1409 | int S, T; |
86387f7e | 1410 | int i; |
d17d9691 | 1411 | unsigned state; |
bc0c38d1 SR |
1412 | |
1413 | if (!next_entry) | |
1414 | next_entry = entry; | |
1415 | rel_usecs = ns2usecs(next_entry->t - entry->t); | |
1416 | abs_usecs = ns2usecs(entry->t - iter->tr->time_start); | |
1417 | ||
1418 | if (verbose) { | |
1419 | comm = trace_find_cmdline(entry->pid); | |
214023c3 SR |
1420 | trace_seq_printf(s, "%16s %5d %d %d %08x %08x [%08lx]" |
1421 | " %ld.%03ldms (+%ld.%03ldms): ", | |
1422 | comm, | |
1423 | entry->pid, cpu, entry->flags, | |
1424 | entry->preempt_count, trace_idx, | |
1425 | ns2usecs(entry->t), | |
1426 | abs_usecs/1000, | |
1427 | abs_usecs % 1000, rel_usecs/1000, | |
1428 | rel_usecs % 1000); | |
bc0c38d1 | 1429 | } else { |
f29c73fe IM |
1430 | lat_print_generic(s, entry, cpu); |
1431 | lat_print_timestamp(s, abs_usecs, rel_usecs); | |
bc0c38d1 SR |
1432 | } |
1433 | switch (entry->type) { | |
1434 | case TRACE_FN: | |
214023c3 SR |
1435 | seq_print_ip_sym(s, entry->fn.ip, sym_flags); |
1436 | trace_seq_puts(s, " ("); | |
1437 | seq_print_ip_sym(s, entry->fn.parent_ip, sym_flags); | |
1438 | trace_seq_puts(s, ")\n"); | |
bc0c38d1 SR |
1439 | break; |
1440 | case TRACE_CTX: | |
57422797 | 1441 | case TRACE_WAKE: |
bac524d3 PZ |
1442 | T = entry->ctx.next_state < sizeof(state_to_char) ? |
1443 | state_to_char[entry->ctx.next_state] : 'X'; | |
1444 | ||
d17d9691 AG |
1445 | state = entry->ctx.prev_state ? __ffs(entry->ctx.prev_state) + 1 : 0; |
1446 | S = state < sizeof(state_to_char) - 1 ? state_to_char[state] : 'X'; | |
bc0c38d1 | 1447 | comm = trace_find_cmdline(entry->ctx.next_pid); |
bac524d3 | 1448 | trace_seq_printf(s, " %5d:%3d:%c %s %5d:%3d:%c %s\n", |
214023c3 SR |
1449 | entry->ctx.prev_pid, |
1450 | entry->ctx.prev_prio, | |
57422797 | 1451 | S, entry->type == TRACE_CTX ? "==>" : " +", |
214023c3 SR |
1452 | entry->ctx.next_pid, |
1453 | entry->ctx.next_prio, | |
bac524d3 | 1454 | T, comm); |
bc0c38d1 | 1455 | break; |
f0a920d5 | 1456 | case TRACE_SPECIAL: |
88a4216c | 1457 | trace_seq_printf(s, "# %ld %ld %ld\n", |
f0a920d5 IM |
1458 | entry->special.arg1, |
1459 | entry->special.arg2, | |
1460 | entry->special.arg3); | |
1461 | break; | |
86387f7e IM |
1462 | case TRACE_STACK: |
1463 | for (i = 0; i < FTRACE_STACK_ENTRIES; i++) { | |
1464 | if (i) | |
1465 | trace_seq_puts(s, " <= "); | |
1466 | seq_print_ip_sym(s, entry->stack.caller[i], sym_flags); | |
1467 | } | |
1468 | trace_seq_puts(s, "\n"); | |
1469 | break; | |
89b2f978 | 1470 | default: |
214023c3 | 1471 | trace_seq_printf(s, "Unknown type %d\n", entry->type); |
bc0c38d1 | 1472 | } |
f9896bf3 | 1473 | return 1; |
bc0c38d1 SR |
1474 | } |
1475 | ||
e309b41d | 1476 | static int print_trace_fmt(struct trace_iterator *iter) |
bc0c38d1 | 1477 | { |
214023c3 | 1478 | struct trace_seq *s = &iter->seq; |
bc0c38d1 | 1479 | unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK); |
4e3c3333 | 1480 | struct trace_entry *entry; |
bc0c38d1 SR |
1481 | unsigned long usec_rem; |
1482 | unsigned long long t; | |
1483 | unsigned long secs; | |
1484 | char *comm; | |
b3806b43 | 1485 | int ret; |
bac524d3 | 1486 | int S, T; |
86387f7e | 1487 | int i; |
bc0c38d1 | 1488 | |
4e3c3333 IM |
1489 | entry = iter->ent; |
1490 | ||
bc0c38d1 SR |
1491 | comm = trace_find_cmdline(iter->ent->pid); |
1492 | ||
cdd31cd2 | 1493 | t = ns2usecs(entry->t); |
bc0c38d1 SR |
1494 | usec_rem = do_div(t, 1000000ULL); |
1495 | secs = (unsigned long)t; | |
1496 | ||
f29c73fe IM |
1497 | ret = trace_seq_printf(s, "%16s-%-5d ", comm, entry->pid); |
1498 | if (!ret) | |
1499 | return 0; | |
1500 | ret = trace_seq_printf(s, "[%02d] ", iter->cpu); | |
1501 | if (!ret) | |
1502 | return 0; | |
1503 | ret = trace_seq_printf(s, "%5lu.%06lu: ", secs, usec_rem); | |
1504 | if (!ret) | |
1505 | return 0; | |
bc0c38d1 SR |
1506 | |
1507 | switch (entry->type) { | |
1508 | case TRACE_FN: | |
b3806b43 SR |
1509 | ret = seq_print_ip_sym(s, entry->fn.ip, sym_flags); |
1510 | if (!ret) | |
1511 | return 0; | |
bc0c38d1 SR |
1512 | if ((sym_flags & TRACE_ITER_PRINT_PARENT) && |
1513 | entry->fn.parent_ip) { | |
b3806b43 SR |
1514 | ret = trace_seq_printf(s, " <-"); |
1515 | if (!ret) | |
1516 | return 0; | |
1517 | ret = seq_print_ip_sym(s, entry->fn.parent_ip, | |
1518 | sym_flags); | |
1519 | if (!ret) | |
1520 | return 0; | |
bc0c38d1 | 1521 | } |
b3806b43 SR |
1522 | ret = trace_seq_printf(s, "\n"); |
1523 | if (!ret) | |
1524 | return 0; | |
bc0c38d1 SR |
1525 | break; |
1526 | case TRACE_CTX: | |
57422797 | 1527 | case TRACE_WAKE: |
bc0c38d1 SR |
1528 | S = entry->ctx.prev_state < sizeof(state_to_char) ? |
1529 | state_to_char[entry->ctx.prev_state] : 'X'; | |
bac524d3 PZ |
1530 | T = entry->ctx.next_state < sizeof(state_to_char) ? |
1531 | state_to_char[entry->ctx.next_state] : 'X'; | |
1532 | ret = trace_seq_printf(s, " %5d:%3d:%c %s %5d:%3d:%c\n", | |
b3806b43 SR |
1533 | entry->ctx.prev_pid, |
1534 | entry->ctx.prev_prio, | |
1535 | S, | |
57422797 | 1536 | entry->type == TRACE_CTX ? "==>" : " +", |
b3806b43 | 1537 | entry->ctx.next_pid, |
bac524d3 PZ |
1538 | entry->ctx.next_prio, |
1539 | T); | |
b3806b43 SR |
1540 | if (!ret) |
1541 | return 0; | |
bc0c38d1 | 1542 | break; |
f0a920d5 | 1543 | case TRACE_SPECIAL: |
88a4216c | 1544 | ret = trace_seq_printf(s, "# %ld %ld %ld\n", |
f0a920d5 IM |
1545 | entry->special.arg1, |
1546 | entry->special.arg2, | |
1547 | entry->special.arg3); | |
1548 | if (!ret) | |
1549 | return 0; | |
1550 | break; | |
86387f7e IM |
1551 | case TRACE_STACK: |
1552 | for (i = 0; i < FTRACE_STACK_ENTRIES; i++) { | |
1553 | if (i) { | |
1554 | ret = trace_seq_puts(s, " <= "); | |
1555 | if (!ret) | |
1556 | return 0; | |
1557 | } | |
1558 | ret = seq_print_ip_sym(s, entry->stack.caller[i], | |
1559 | sym_flags); | |
1560 | if (!ret) | |
1561 | return 0; | |
1562 | } | |
1563 | ret = trace_seq_puts(s, "\n"); | |
1564 | if (!ret) | |
1565 | return 0; | |
1566 | break; | |
bc0c38d1 | 1567 | } |
b3806b43 | 1568 | return 1; |
bc0c38d1 SR |
1569 | } |
1570 | ||
e309b41d | 1571 | static int print_raw_fmt(struct trace_iterator *iter) |
f9896bf3 IM |
1572 | { |
1573 | struct trace_seq *s = &iter->seq; | |
1574 | struct trace_entry *entry; | |
1575 | int ret; | |
bac524d3 | 1576 | int S, T; |
f9896bf3 IM |
1577 | |
1578 | entry = iter->ent; | |
1579 | ||
1580 | ret = trace_seq_printf(s, "%d %d %llu ", | |
1581 | entry->pid, iter->cpu, entry->t); | |
1582 | if (!ret) | |
1583 | return 0; | |
1584 | ||
1585 | switch (entry->type) { | |
1586 | case TRACE_FN: | |
1587 | ret = trace_seq_printf(s, "%x %x\n", | |
1588 | entry->fn.ip, entry->fn.parent_ip); | |
1589 | if (!ret) | |
1590 | return 0; | |
1591 | break; | |
1592 | case TRACE_CTX: | |
57422797 | 1593 | case TRACE_WAKE: |
f9896bf3 IM |
1594 | S = entry->ctx.prev_state < sizeof(state_to_char) ? |
1595 | state_to_char[entry->ctx.prev_state] : 'X'; | |
bac524d3 PZ |
1596 | T = entry->ctx.next_state < sizeof(state_to_char) ? |
1597 | state_to_char[entry->ctx.next_state] : 'X'; | |
57422797 IM |
1598 | if (entry->type == TRACE_WAKE) |
1599 | S = '+'; | |
bac524d3 | 1600 | ret = trace_seq_printf(s, "%d %d %c %d %d %c\n", |
f9896bf3 IM |
1601 | entry->ctx.prev_pid, |
1602 | entry->ctx.prev_prio, | |
1603 | S, | |
1604 | entry->ctx.next_pid, | |
bac524d3 PZ |
1605 | entry->ctx.next_prio, |
1606 | T); | |
f9896bf3 IM |
1607 | if (!ret) |
1608 | return 0; | |
1609 | break; | |
f0a920d5 | 1610 | case TRACE_SPECIAL: |
86387f7e | 1611 | case TRACE_STACK: |
88a4216c | 1612 | ret = trace_seq_printf(s, "# %ld %ld %ld\n", |
f0a920d5 IM |
1613 | entry->special.arg1, |
1614 | entry->special.arg2, | |
1615 | entry->special.arg3); | |
1616 | if (!ret) | |
1617 | return 0; | |
1618 | break; | |
f9896bf3 IM |
1619 | } |
1620 | return 1; | |
1621 | } | |
1622 | ||
cb0f12aa IM |
1623 | #define SEQ_PUT_FIELD_RET(s, x) \ |
1624 | do { \ | |
1625 | if (!trace_seq_putmem(s, &(x), sizeof(x))) \ | |
1626 | return 0; \ | |
1627 | } while (0) | |
1628 | ||
5e3ca0ec IM |
1629 | #define SEQ_PUT_HEX_FIELD_RET(s, x) \ |
1630 | do { \ | |
1631 | if (!trace_seq_putmem_hex(s, &(x), sizeof(x))) \ | |
1632 | return 0; \ | |
1633 | } while (0) | |
1634 | ||
e309b41d | 1635 | static int print_hex_fmt(struct trace_iterator *iter) |
5e3ca0ec IM |
1636 | { |
1637 | struct trace_seq *s = &iter->seq; | |
1638 | unsigned char newline = '\n'; | |
1639 | struct trace_entry *entry; | |
bac524d3 | 1640 | int S, T; |
5e3ca0ec IM |
1641 | |
1642 | entry = iter->ent; | |
1643 | ||
1644 | SEQ_PUT_HEX_FIELD_RET(s, entry->pid); | |
1645 | SEQ_PUT_HEX_FIELD_RET(s, iter->cpu); | |
1646 | SEQ_PUT_HEX_FIELD_RET(s, entry->t); | |
1647 | ||
1648 | switch (entry->type) { | |
1649 | case TRACE_FN: | |
1650 | SEQ_PUT_HEX_FIELD_RET(s, entry->fn.ip); | |
1651 | SEQ_PUT_HEX_FIELD_RET(s, entry->fn.parent_ip); | |
1652 | break; | |
1653 | case TRACE_CTX: | |
57422797 | 1654 | case TRACE_WAKE: |
5e3ca0ec IM |
1655 | S = entry->ctx.prev_state < sizeof(state_to_char) ? |
1656 | state_to_char[entry->ctx.prev_state] : 'X'; | |
bac524d3 PZ |
1657 | T = entry->ctx.next_state < sizeof(state_to_char) ? |
1658 | state_to_char[entry->ctx.next_state] : 'X'; | |
57422797 IM |
1659 | if (entry->type == TRACE_WAKE) |
1660 | S = '+'; | |
5e3ca0ec IM |
1661 | SEQ_PUT_HEX_FIELD_RET(s, entry->ctx.prev_pid); |
1662 | SEQ_PUT_HEX_FIELD_RET(s, entry->ctx.prev_prio); | |
1663 | SEQ_PUT_HEX_FIELD_RET(s, S); | |
1664 | SEQ_PUT_HEX_FIELD_RET(s, entry->ctx.next_pid); | |
1665 | SEQ_PUT_HEX_FIELD_RET(s, entry->ctx.next_prio); | |
1666 | SEQ_PUT_HEX_FIELD_RET(s, entry->fn.parent_ip); | |
bac524d3 | 1667 | SEQ_PUT_HEX_FIELD_RET(s, T); |
5e3ca0ec IM |
1668 | break; |
1669 | case TRACE_SPECIAL: | |
86387f7e | 1670 | case TRACE_STACK: |
5e3ca0ec IM |
1671 | SEQ_PUT_HEX_FIELD_RET(s, entry->special.arg1); |
1672 | SEQ_PUT_HEX_FIELD_RET(s, entry->special.arg2); | |
1673 | SEQ_PUT_HEX_FIELD_RET(s, entry->special.arg3); | |
1674 | break; | |
1675 | } | |
1676 | SEQ_PUT_FIELD_RET(s, newline); | |
1677 | ||
1678 | return 1; | |
1679 | } | |
1680 | ||
e309b41d | 1681 | static int print_bin_fmt(struct trace_iterator *iter) |
cb0f12aa IM |
1682 | { |
1683 | struct trace_seq *s = &iter->seq; | |
1684 | struct trace_entry *entry; | |
1685 | ||
1686 | entry = iter->ent; | |
1687 | ||
1688 | SEQ_PUT_FIELD_RET(s, entry->pid); | |
1689 | SEQ_PUT_FIELD_RET(s, entry->cpu); | |
1690 | SEQ_PUT_FIELD_RET(s, entry->t); | |
1691 | ||
1692 | switch (entry->type) { | |
1693 | case TRACE_FN: | |
1694 | SEQ_PUT_FIELD_RET(s, entry->fn.ip); | |
1695 | SEQ_PUT_FIELD_RET(s, entry->fn.parent_ip); | |
1696 | break; | |
1697 | case TRACE_CTX: | |
1698 | SEQ_PUT_FIELD_RET(s, entry->ctx.prev_pid); | |
1699 | SEQ_PUT_FIELD_RET(s, entry->ctx.prev_prio); | |
1700 | SEQ_PUT_FIELD_RET(s, entry->ctx.prev_state); | |
1701 | SEQ_PUT_FIELD_RET(s, entry->ctx.next_pid); | |
1702 | SEQ_PUT_FIELD_RET(s, entry->ctx.next_prio); | |
bac524d3 | 1703 | SEQ_PUT_FIELD_RET(s, entry->ctx.next_state); |
cb0f12aa | 1704 | break; |
f0a920d5 | 1705 | case TRACE_SPECIAL: |
86387f7e | 1706 | case TRACE_STACK: |
f0a920d5 IM |
1707 | SEQ_PUT_FIELD_RET(s, entry->special.arg1); |
1708 | SEQ_PUT_FIELD_RET(s, entry->special.arg2); | |
1709 | SEQ_PUT_FIELD_RET(s, entry->special.arg3); | |
1710 | break; | |
cb0f12aa IM |
1711 | } |
1712 | return 1; | |
1713 | } | |
1714 | ||
bc0c38d1 SR |
1715 | static int trace_empty(struct trace_iterator *iter) |
1716 | { | |
1717 | struct trace_array_cpu *data; | |
1718 | int cpu; | |
1719 | ||
ab46428c | 1720 | for_each_tracing_cpu(cpu) { |
bc0c38d1 SR |
1721 | data = iter->tr->data[cpu]; |
1722 | ||
b3806b43 SR |
1723 | if (head_page(data) && data->trace_idx && |
1724 | (data->trace_tail != data->trace_head || | |
1725 | data->trace_tail_idx != data->trace_head_idx)) | |
bc0c38d1 SR |
1726 | return 0; |
1727 | } | |
1728 | return 1; | |
1729 | } | |
1730 | ||
f9896bf3 IM |
1731 | static int print_trace_line(struct trace_iterator *iter) |
1732 | { | |
72829bc3 TG |
1733 | if (iter->trace && iter->trace->print_line) |
1734 | return iter->trace->print_line(iter); | |
1735 | ||
cb0f12aa IM |
1736 | if (trace_flags & TRACE_ITER_BIN) |
1737 | return print_bin_fmt(iter); | |
1738 | ||
5e3ca0ec IM |
1739 | if (trace_flags & TRACE_ITER_HEX) |
1740 | return print_hex_fmt(iter); | |
1741 | ||
f9896bf3 IM |
1742 | if (trace_flags & TRACE_ITER_RAW) |
1743 | return print_raw_fmt(iter); | |
1744 | ||
1745 | if (iter->iter_flags & TRACE_FILE_LAT_FMT) | |
1746 | return print_lat_fmt(iter, iter->idx, iter->cpu); | |
1747 | ||
1748 | return print_trace_fmt(iter); | |
1749 | } | |
1750 | ||
bc0c38d1 SR |
1751 | static int s_show(struct seq_file *m, void *v) |
1752 | { | |
1753 | struct trace_iterator *iter = v; | |
1754 | ||
1755 | if (iter->ent == NULL) { | |
1756 | if (iter->tr) { | |
1757 | seq_printf(m, "# tracer: %s\n", iter->trace->name); | |
1758 | seq_puts(m, "#\n"); | |
1759 | } | |
1760 | if (iter->iter_flags & TRACE_FILE_LAT_FMT) { | |
1761 | /* print nothing if the buffers are empty */ | |
1762 | if (trace_empty(iter)) | |
1763 | return 0; | |
1764 | print_trace_header(m, iter); | |
1765 | if (!(trace_flags & TRACE_ITER_VERBOSE)) | |
1766 | print_lat_help_header(m); | |
1767 | } else { | |
1768 | if (!(trace_flags & TRACE_ITER_VERBOSE)) | |
1769 | print_func_help_header(m); | |
1770 | } | |
1771 | } else { | |
f9896bf3 | 1772 | print_trace_line(iter); |
214023c3 | 1773 | trace_print_seq(m, &iter->seq); |
bc0c38d1 SR |
1774 | } |
1775 | ||
1776 | return 0; | |
1777 | } | |
1778 | ||
1779 | static struct seq_operations tracer_seq_ops = { | |
4bf39a94 IM |
1780 | .start = s_start, |
1781 | .next = s_next, | |
1782 | .stop = s_stop, | |
1783 | .show = s_show, | |
bc0c38d1 SR |
1784 | }; |
1785 | ||
e309b41d | 1786 | static struct trace_iterator * |
bc0c38d1 SR |
1787 | __tracing_open(struct inode *inode, struct file *file, int *ret) |
1788 | { | |
1789 | struct trace_iterator *iter; | |
1790 | ||
60a11774 SR |
1791 | if (tracing_disabled) { |
1792 | *ret = -ENODEV; | |
1793 | return NULL; | |
1794 | } | |
1795 | ||
bc0c38d1 SR |
1796 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); |
1797 | if (!iter) { | |
1798 | *ret = -ENOMEM; | |
1799 | goto out; | |
1800 | } | |
1801 | ||
1802 | mutex_lock(&trace_types_lock); | |
1803 | if (current_trace && current_trace->print_max) | |
1804 | iter->tr = &max_tr; | |
1805 | else | |
1806 | iter->tr = inode->i_private; | |
1807 | iter->trace = current_trace; | |
1808 | iter->pos = -1; | |
1809 | ||
1810 | /* TODO stop tracer */ | |
1811 | *ret = seq_open(file, &tracer_seq_ops); | |
1812 | if (!*ret) { | |
1813 | struct seq_file *m = file->private_data; | |
1814 | m->private = iter; | |
1815 | ||
1816 | /* stop the trace while dumping */ | |
1817 | if (iter->tr->ctrl) | |
1818 | tracer_enabled = 0; | |
1819 | ||
1820 | if (iter->trace && iter->trace->open) | |
1821 | iter->trace->open(iter); | |
1822 | } else { | |
1823 | kfree(iter); | |
1824 | iter = NULL; | |
1825 | } | |
1826 | mutex_unlock(&trace_types_lock); | |
1827 | ||
1828 | out: | |
1829 | return iter; | |
1830 | } | |
1831 | ||
1832 | int tracing_open_generic(struct inode *inode, struct file *filp) | |
1833 | { | |
60a11774 SR |
1834 | if (tracing_disabled) |
1835 | return -ENODEV; | |
1836 | ||
bc0c38d1 SR |
1837 | filp->private_data = inode->i_private; |
1838 | return 0; | |
1839 | } | |
1840 | ||
1841 | int tracing_release(struct inode *inode, struct file *file) | |
1842 | { | |
1843 | struct seq_file *m = (struct seq_file *)file->private_data; | |
1844 | struct trace_iterator *iter = m->private; | |
1845 | ||
1846 | mutex_lock(&trace_types_lock); | |
1847 | if (iter->trace && iter->trace->close) | |
1848 | iter->trace->close(iter); | |
1849 | ||
1850 | /* reenable tracing if it was previously enabled */ | |
1851 | if (iter->tr->ctrl) | |
1852 | tracer_enabled = 1; | |
1853 | mutex_unlock(&trace_types_lock); | |
1854 | ||
1855 | seq_release(inode, file); | |
1856 | kfree(iter); | |
1857 | return 0; | |
1858 | } | |
1859 | ||
1860 | static int tracing_open(struct inode *inode, struct file *file) | |
1861 | { | |
1862 | int ret; | |
1863 | ||
1864 | __tracing_open(inode, file, &ret); | |
1865 | ||
1866 | return ret; | |
1867 | } | |
1868 | ||
1869 | static int tracing_lt_open(struct inode *inode, struct file *file) | |
1870 | { | |
1871 | struct trace_iterator *iter; | |
1872 | int ret; | |
1873 | ||
1874 | iter = __tracing_open(inode, file, &ret); | |
1875 | ||
1876 | if (!ret) | |
1877 | iter->iter_flags |= TRACE_FILE_LAT_FMT; | |
1878 | ||
1879 | return ret; | |
1880 | } | |
1881 | ||
1882 | ||
e309b41d | 1883 | static void * |
bc0c38d1 SR |
1884 | t_next(struct seq_file *m, void *v, loff_t *pos) |
1885 | { | |
1886 | struct tracer *t = m->private; | |
1887 | ||
1888 | (*pos)++; | |
1889 | ||
1890 | if (t) | |
1891 | t = t->next; | |
1892 | ||
1893 | m->private = t; | |
1894 | ||
1895 | return t; | |
1896 | } | |
1897 | ||
1898 | static void *t_start(struct seq_file *m, loff_t *pos) | |
1899 | { | |
1900 | struct tracer *t = m->private; | |
1901 | loff_t l = 0; | |
1902 | ||
1903 | mutex_lock(&trace_types_lock); | |
1904 | for (; t && l < *pos; t = t_next(m, t, &l)) | |
1905 | ; | |
1906 | ||
1907 | return t; | |
1908 | } | |
1909 | ||
1910 | static void t_stop(struct seq_file *m, void *p) | |
1911 | { | |
1912 | mutex_unlock(&trace_types_lock); | |
1913 | } | |
1914 | ||
1915 | static int t_show(struct seq_file *m, void *v) | |
1916 | { | |
1917 | struct tracer *t = v; | |
1918 | ||
1919 | if (!t) | |
1920 | return 0; | |
1921 | ||
1922 | seq_printf(m, "%s", t->name); | |
1923 | if (t->next) | |
1924 | seq_putc(m, ' '); | |
1925 | else | |
1926 | seq_putc(m, '\n'); | |
1927 | ||
1928 | return 0; | |
1929 | } | |
1930 | ||
1931 | static struct seq_operations show_traces_seq_ops = { | |
4bf39a94 IM |
1932 | .start = t_start, |
1933 | .next = t_next, | |
1934 | .stop = t_stop, | |
1935 | .show = t_show, | |
bc0c38d1 SR |
1936 | }; |
1937 | ||
1938 | static int show_traces_open(struct inode *inode, struct file *file) | |
1939 | { | |
1940 | int ret; | |
1941 | ||
60a11774 SR |
1942 | if (tracing_disabled) |
1943 | return -ENODEV; | |
1944 | ||
bc0c38d1 SR |
1945 | ret = seq_open(file, &show_traces_seq_ops); |
1946 | if (!ret) { | |
1947 | struct seq_file *m = file->private_data; | |
1948 | m->private = trace_types; | |
1949 | } | |
1950 | ||
1951 | return ret; | |
1952 | } | |
1953 | ||
1954 | static struct file_operations tracing_fops = { | |
4bf39a94 IM |
1955 | .open = tracing_open, |
1956 | .read = seq_read, | |
1957 | .llseek = seq_lseek, | |
1958 | .release = tracing_release, | |
bc0c38d1 SR |
1959 | }; |
1960 | ||
1961 | static struct file_operations tracing_lt_fops = { | |
4bf39a94 IM |
1962 | .open = tracing_lt_open, |
1963 | .read = seq_read, | |
1964 | .llseek = seq_lseek, | |
1965 | .release = tracing_release, | |
bc0c38d1 SR |
1966 | }; |
1967 | ||
1968 | static struct file_operations show_traces_fops = { | |
c7078de1 IM |
1969 | .open = show_traces_open, |
1970 | .read = seq_read, | |
1971 | .release = seq_release, | |
1972 | }; | |
1973 | ||
36dfe925 IM |
1974 | /* |
1975 | * Only trace on a CPU if the bitmask is set: | |
1976 | */ | |
1977 | static cpumask_t tracing_cpumask = CPU_MASK_ALL; | |
1978 | ||
1979 | /* | |
1980 | * When tracing/tracing_cpu_mask is modified then this holds | |
1981 | * the new bitmask we are about to install: | |
1982 | */ | |
1983 | static cpumask_t tracing_cpumask_new; | |
1984 | ||
1985 | /* | |
1986 | * The tracer itself will not take this lock, but still we want | |
1987 | * to provide a consistent cpumask to user-space: | |
1988 | */ | |
1989 | static DEFINE_MUTEX(tracing_cpumask_update_lock); | |
1990 | ||
1991 | /* | |
1992 | * Temporary storage for the character representation of the | |
1993 | * CPU bitmask (and one more byte for the newline): | |
1994 | */ | |
1995 | static char mask_str[NR_CPUS + 1]; | |
1996 | ||
c7078de1 IM |
1997 | static ssize_t |
1998 | tracing_cpumask_read(struct file *filp, char __user *ubuf, | |
1999 | size_t count, loff_t *ppos) | |
2000 | { | |
36dfe925 | 2001 | int len; |
c7078de1 IM |
2002 | |
2003 | mutex_lock(&tracing_cpumask_update_lock); | |
36dfe925 IM |
2004 | |
2005 | len = cpumask_scnprintf(mask_str, count, tracing_cpumask); | |
2006 | if (count - len < 2) { | |
2007 | count = -EINVAL; | |
2008 | goto out_err; | |
2009 | } | |
2010 | len += sprintf(mask_str + len, "\n"); | |
2011 | count = simple_read_from_buffer(ubuf, count, ppos, mask_str, NR_CPUS+1); | |
2012 | ||
2013 | out_err: | |
c7078de1 IM |
2014 | mutex_unlock(&tracing_cpumask_update_lock); |
2015 | ||
2016 | return count; | |
2017 | } | |
2018 | ||
2019 | static ssize_t | |
2020 | tracing_cpumask_write(struct file *filp, const char __user *ubuf, | |
2021 | size_t count, loff_t *ppos) | |
2022 | { | |
36dfe925 | 2023 | int err, cpu; |
c7078de1 IM |
2024 | |
2025 | mutex_lock(&tracing_cpumask_update_lock); | |
36dfe925 | 2026 | err = cpumask_parse_user(ubuf, count, tracing_cpumask_new); |
c7078de1 | 2027 | if (err) |
36dfe925 IM |
2028 | goto err_unlock; |
2029 | ||
92205c23 SR |
2030 | raw_local_irq_disable(); |
2031 | __raw_spin_lock(&ftrace_max_lock); | |
ab46428c | 2032 | for_each_tracing_cpu(cpu) { |
36dfe925 IM |
2033 | /* |
2034 | * Increase/decrease the disabled counter if we are | |
2035 | * about to flip a bit in the cpumask: | |
2036 | */ | |
2037 | if (cpu_isset(cpu, tracing_cpumask) && | |
2038 | !cpu_isset(cpu, tracing_cpumask_new)) { | |
2039 | atomic_inc(&global_trace.data[cpu]->disabled); | |
2040 | } | |
2041 | if (!cpu_isset(cpu, tracing_cpumask) && | |
2042 | cpu_isset(cpu, tracing_cpumask_new)) { | |
2043 | atomic_dec(&global_trace.data[cpu]->disabled); | |
2044 | } | |
2045 | } | |
92205c23 SR |
2046 | __raw_spin_unlock(&ftrace_max_lock); |
2047 | raw_local_irq_enable(); | |
36dfe925 IM |
2048 | |
2049 | tracing_cpumask = tracing_cpumask_new; | |
2050 | ||
2051 | mutex_unlock(&tracing_cpumask_update_lock); | |
c7078de1 IM |
2052 | |
2053 | return count; | |
36dfe925 IM |
2054 | |
2055 | err_unlock: | |
2056 | mutex_unlock(&tracing_cpumask_update_lock); | |
2057 | ||
2058 | return err; | |
c7078de1 IM |
2059 | } |
2060 | ||
2061 | static struct file_operations tracing_cpumask_fops = { | |
2062 | .open = tracing_open_generic, | |
2063 | .read = tracing_cpumask_read, | |
2064 | .write = tracing_cpumask_write, | |
bc0c38d1 SR |
2065 | }; |
2066 | ||
2067 | static ssize_t | |
2068 | tracing_iter_ctrl_read(struct file *filp, char __user *ubuf, | |
2069 | size_t cnt, loff_t *ppos) | |
2070 | { | |
2071 | char *buf; | |
2072 | int r = 0; | |
2073 | int len = 0; | |
2074 | int i; | |
2075 | ||
2076 | /* calulate max size */ | |
2077 | for (i = 0; trace_options[i]; i++) { | |
2078 | len += strlen(trace_options[i]); | |
2079 | len += 3; /* "no" and space */ | |
2080 | } | |
2081 | ||
2082 | /* +2 for \n and \0 */ | |
2083 | buf = kmalloc(len + 2, GFP_KERNEL); | |
2084 | if (!buf) | |
2085 | return -ENOMEM; | |
2086 | ||
2087 | for (i = 0; trace_options[i]; i++) { | |
2088 | if (trace_flags & (1 << i)) | |
2089 | r += sprintf(buf + r, "%s ", trace_options[i]); | |
2090 | else | |
2091 | r += sprintf(buf + r, "no%s ", trace_options[i]); | |
2092 | } | |
2093 | ||
2094 | r += sprintf(buf + r, "\n"); | |
2095 | WARN_ON(r >= len + 2); | |
2096 | ||
36dfe925 | 2097 | r = simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2098 | |
2099 | kfree(buf); | |
2100 | ||
2101 | return r; | |
2102 | } | |
2103 | ||
2104 | static ssize_t | |
2105 | tracing_iter_ctrl_write(struct file *filp, const char __user *ubuf, | |
2106 | size_t cnt, loff_t *ppos) | |
2107 | { | |
2108 | char buf[64]; | |
2109 | char *cmp = buf; | |
2110 | int neg = 0; | |
2111 | int i; | |
2112 | ||
cffae437 SR |
2113 | if (cnt >= sizeof(buf)) |
2114 | return -EINVAL; | |
bc0c38d1 SR |
2115 | |
2116 | if (copy_from_user(&buf, ubuf, cnt)) | |
2117 | return -EFAULT; | |
2118 | ||
2119 | buf[cnt] = 0; | |
2120 | ||
2121 | if (strncmp(buf, "no", 2) == 0) { | |
2122 | neg = 1; | |
2123 | cmp += 2; | |
2124 | } | |
2125 | ||
2126 | for (i = 0; trace_options[i]; i++) { | |
2127 | int len = strlen(trace_options[i]); | |
2128 | ||
2129 | if (strncmp(cmp, trace_options[i], len) == 0) { | |
2130 | if (neg) | |
2131 | trace_flags &= ~(1 << i); | |
2132 | else | |
2133 | trace_flags |= (1 << i); | |
2134 | break; | |
2135 | } | |
2136 | } | |
442e544c IM |
2137 | /* |
2138 | * If no option could be set, return an error: | |
2139 | */ | |
2140 | if (!trace_options[i]) | |
2141 | return -EINVAL; | |
bc0c38d1 SR |
2142 | |
2143 | filp->f_pos += cnt; | |
2144 | ||
2145 | return cnt; | |
2146 | } | |
2147 | ||
2148 | static struct file_operations tracing_iter_fops = { | |
c7078de1 IM |
2149 | .open = tracing_open_generic, |
2150 | .read = tracing_iter_ctrl_read, | |
2151 | .write = tracing_iter_ctrl_write, | |
bc0c38d1 SR |
2152 | }; |
2153 | ||
7bd2f24c IM |
2154 | static const char readme_msg[] = |
2155 | "tracing mini-HOWTO:\n\n" | |
2156 | "# mkdir /debug\n" | |
2157 | "# mount -t debugfs nodev /debug\n\n" | |
2158 | "# cat /debug/tracing/available_tracers\n" | |
2159 | "wakeup preemptirqsoff preemptoff irqsoff ftrace sched_switch none\n\n" | |
2160 | "# cat /debug/tracing/current_tracer\n" | |
2161 | "none\n" | |
2162 | "# echo sched_switch > /debug/tracing/current_tracer\n" | |
2163 | "# cat /debug/tracing/current_tracer\n" | |
2164 | "sched_switch\n" | |
2165 | "# cat /debug/tracing/iter_ctrl\n" | |
2166 | "noprint-parent nosym-offset nosym-addr noverbose\n" | |
2167 | "# echo print-parent > /debug/tracing/iter_ctrl\n" | |
2168 | "# echo 1 > /debug/tracing/tracing_enabled\n" | |
2169 | "# cat /debug/tracing/trace > /tmp/trace.txt\n" | |
2170 | "echo 0 > /debug/tracing/tracing_enabled\n" | |
2171 | ; | |
2172 | ||
2173 | static ssize_t | |
2174 | tracing_readme_read(struct file *filp, char __user *ubuf, | |
2175 | size_t cnt, loff_t *ppos) | |
2176 | { | |
2177 | return simple_read_from_buffer(ubuf, cnt, ppos, | |
2178 | readme_msg, strlen(readme_msg)); | |
2179 | } | |
2180 | ||
2181 | static struct file_operations tracing_readme_fops = { | |
c7078de1 IM |
2182 | .open = tracing_open_generic, |
2183 | .read = tracing_readme_read, | |
7bd2f24c IM |
2184 | }; |
2185 | ||
bc0c38d1 SR |
2186 | static ssize_t |
2187 | tracing_ctrl_read(struct file *filp, char __user *ubuf, | |
2188 | size_t cnt, loff_t *ppos) | |
2189 | { | |
2190 | struct trace_array *tr = filp->private_data; | |
2191 | char buf[64]; | |
2192 | int r; | |
2193 | ||
2194 | r = sprintf(buf, "%ld\n", tr->ctrl); | |
4e3c3333 | 2195 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2196 | } |
2197 | ||
2198 | static ssize_t | |
2199 | tracing_ctrl_write(struct file *filp, const char __user *ubuf, | |
2200 | size_t cnt, loff_t *ppos) | |
2201 | { | |
2202 | struct trace_array *tr = filp->private_data; | |
bc0c38d1 | 2203 | char buf[64]; |
c6caeeb1 SR |
2204 | long val; |
2205 | int ret; | |
bc0c38d1 | 2206 | |
cffae437 SR |
2207 | if (cnt >= sizeof(buf)) |
2208 | return -EINVAL; | |
bc0c38d1 SR |
2209 | |
2210 | if (copy_from_user(&buf, ubuf, cnt)) | |
2211 | return -EFAULT; | |
2212 | ||
2213 | buf[cnt] = 0; | |
2214 | ||
c6caeeb1 SR |
2215 | ret = strict_strtoul(buf, 10, &val); |
2216 | if (ret < 0) | |
2217 | return ret; | |
bc0c38d1 SR |
2218 | |
2219 | val = !!val; | |
2220 | ||
2221 | mutex_lock(&trace_types_lock); | |
2222 | if (tr->ctrl ^ val) { | |
2223 | if (val) | |
2224 | tracer_enabled = 1; | |
2225 | else | |
2226 | tracer_enabled = 0; | |
2227 | ||
2228 | tr->ctrl = val; | |
2229 | ||
2230 | if (current_trace && current_trace->ctrl_update) | |
2231 | current_trace->ctrl_update(tr); | |
2232 | } | |
2233 | mutex_unlock(&trace_types_lock); | |
2234 | ||
2235 | filp->f_pos += cnt; | |
2236 | ||
2237 | return cnt; | |
2238 | } | |
2239 | ||
2240 | static ssize_t | |
2241 | tracing_set_trace_read(struct file *filp, char __user *ubuf, | |
2242 | size_t cnt, loff_t *ppos) | |
2243 | { | |
2244 | char buf[max_tracer_type_len+2]; | |
2245 | int r; | |
2246 | ||
2247 | mutex_lock(&trace_types_lock); | |
2248 | if (current_trace) | |
2249 | r = sprintf(buf, "%s\n", current_trace->name); | |
2250 | else | |
2251 | r = sprintf(buf, "\n"); | |
2252 | mutex_unlock(&trace_types_lock); | |
2253 | ||
4bf39a94 | 2254 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2255 | } |
2256 | ||
2257 | static ssize_t | |
2258 | tracing_set_trace_write(struct file *filp, const char __user *ubuf, | |
2259 | size_t cnt, loff_t *ppos) | |
2260 | { | |
2261 | struct trace_array *tr = &global_trace; | |
2262 | struct tracer *t; | |
2263 | char buf[max_tracer_type_len+1]; | |
2264 | int i; | |
2265 | ||
2266 | if (cnt > max_tracer_type_len) | |
2267 | cnt = max_tracer_type_len; | |
2268 | ||
2269 | if (copy_from_user(&buf, ubuf, cnt)) | |
2270 | return -EFAULT; | |
2271 | ||
2272 | buf[cnt] = 0; | |
2273 | ||
2274 | /* strip ending whitespace. */ | |
2275 | for (i = cnt - 1; i > 0 && isspace(buf[i]); i--) | |
2276 | buf[i] = 0; | |
2277 | ||
2278 | mutex_lock(&trace_types_lock); | |
2279 | for (t = trace_types; t; t = t->next) { | |
2280 | if (strcmp(t->name, buf) == 0) | |
2281 | break; | |
2282 | } | |
2283 | if (!t || t == current_trace) | |
2284 | goto out; | |
2285 | ||
2286 | if (current_trace && current_trace->reset) | |
2287 | current_trace->reset(tr); | |
2288 | ||
2289 | current_trace = t; | |
2290 | if (t->init) | |
2291 | t->init(tr); | |
2292 | ||
2293 | out: | |
2294 | mutex_unlock(&trace_types_lock); | |
2295 | ||
2296 | filp->f_pos += cnt; | |
2297 | ||
2298 | return cnt; | |
2299 | } | |
2300 | ||
2301 | static ssize_t | |
2302 | tracing_max_lat_read(struct file *filp, char __user *ubuf, | |
2303 | size_t cnt, loff_t *ppos) | |
2304 | { | |
2305 | unsigned long *ptr = filp->private_data; | |
2306 | char buf[64]; | |
2307 | int r; | |
2308 | ||
cffae437 | 2309 | r = snprintf(buf, sizeof(buf), "%ld\n", |
bc0c38d1 | 2310 | *ptr == (unsigned long)-1 ? -1 : nsecs_to_usecs(*ptr)); |
cffae437 SR |
2311 | if (r > sizeof(buf)) |
2312 | r = sizeof(buf); | |
4bf39a94 | 2313 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); |
bc0c38d1 SR |
2314 | } |
2315 | ||
2316 | static ssize_t | |
2317 | tracing_max_lat_write(struct file *filp, const char __user *ubuf, | |
2318 | size_t cnt, loff_t *ppos) | |
2319 | { | |
2320 | long *ptr = filp->private_data; | |
bc0c38d1 | 2321 | char buf[64]; |
c6caeeb1 SR |
2322 | long val; |
2323 | int ret; | |
bc0c38d1 | 2324 | |
cffae437 SR |
2325 | if (cnt >= sizeof(buf)) |
2326 | return -EINVAL; | |
bc0c38d1 SR |
2327 | |
2328 | if (copy_from_user(&buf, ubuf, cnt)) | |
2329 | return -EFAULT; | |
2330 | ||
2331 | buf[cnt] = 0; | |
2332 | ||
c6caeeb1 SR |
2333 | ret = strict_strtoul(buf, 10, &val); |
2334 | if (ret < 0) | |
2335 | return ret; | |
bc0c38d1 SR |
2336 | |
2337 | *ptr = val * 1000; | |
2338 | ||
2339 | return cnt; | |
2340 | } | |
2341 | ||
b3806b43 SR |
2342 | static atomic_t tracing_reader; |
2343 | ||
2344 | static int tracing_open_pipe(struct inode *inode, struct file *filp) | |
2345 | { | |
2346 | struct trace_iterator *iter; | |
2347 | ||
2348 | if (tracing_disabled) | |
2349 | return -ENODEV; | |
2350 | ||
2351 | /* We only allow for reader of the pipe */ | |
2352 | if (atomic_inc_return(&tracing_reader) != 1) { | |
2353 | atomic_dec(&tracing_reader); | |
2354 | return -EBUSY; | |
2355 | } | |
2356 | ||
2357 | /* create a buffer to store the information to pass to userspace */ | |
2358 | iter = kzalloc(sizeof(*iter), GFP_KERNEL); | |
2359 | if (!iter) | |
2360 | return -ENOMEM; | |
2361 | ||
107bad8b | 2362 | mutex_lock(&trace_types_lock); |
b3806b43 | 2363 | iter->tr = &global_trace; |
72829bc3 | 2364 | iter->trace = current_trace; |
b3806b43 SR |
2365 | filp->private_data = iter; |
2366 | ||
107bad8b SR |
2367 | if (iter->trace->pipe_open) |
2368 | iter->trace->pipe_open(iter); | |
2369 | mutex_unlock(&trace_types_lock); | |
2370 | ||
b3806b43 SR |
2371 | return 0; |
2372 | } | |
2373 | ||
2374 | static int tracing_release_pipe(struct inode *inode, struct file *file) | |
2375 | { | |
2376 | struct trace_iterator *iter = file->private_data; | |
2377 | ||
2378 | kfree(iter); | |
2379 | atomic_dec(&tracing_reader); | |
2380 | ||
2381 | return 0; | |
2382 | } | |
2383 | ||
2a2cc8f7 SSP |
2384 | static unsigned int |
2385 | tracing_poll_pipe(struct file *filp, poll_table *poll_table) | |
2386 | { | |
2387 | struct trace_iterator *iter = filp->private_data; | |
2388 | ||
2389 | if (trace_flags & TRACE_ITER_BLOCK) { | |
2390 | /* | |
2391 | * Always select as readable when in blocking mode | |
2392 | */ | |
2393 | return POLLIN | POLLRDNORM; | |
afc2abc0 | 2394 | } else { |
2a2cc8f7 SSP |
2395 | if (!trace_empty(iter)) |
2396 | return POLLIN | POLLRDNORM; | |
2397 | poll_wait(filp, &trace_wait, poll_table); | |
2398 | if (!trace_empty(iter)) | |
2399 | return POLLIN | POLLRDNORM; | |
2400 | ||
2401 | return 0; | |
2402 | } | |
2403 | } | |
2404 | ||
b3806b43 SR |
2405 | /* |
2406 | * Consumer reader. | |
2407 | */ | |
2408 | static ssize_t | |
2409 | tracing_read_pipe(struct file *filp, char __user *ubuf, | |
2410 | size_t cnt, loff_t *ppos) | |
2411 | { | |
2412 | struct trace_iterator *iter = filp->private_data; | |
2413 | struct trace_array_cpu *data; | |
2414 | static cpumask_t mask; | |
b3806b43 | 2415 | unsigned long flags; |
25770467 | 2416 | #ifdef CONFIG_FTRACE |
2e0f5761 | 2417 | int ftrace_save; |
25770467 | 2418 | #endif |
b3806b43 | 2419 | int cpu; |
6c6c2796 | 2420 | ssize_t sret; |
b3806b43 SR |
2421 | |
2422 | /* return any leftover data */ | |
6c6c2796 PP |
2423 | sret = trace_seq_to_user(&iter->seq, ubuf, cnt); |
2424 | if (sret != -EBUSY) | |
2425 | return sret; | |
2426 | sret = 0; | |
b3806b43 | 2427 | |
6c6c2796 | 2428 | trace_seq_reset(&iter->seq); |
b3806b43 | 2429 | |
107bad8b SR |
2430 | mutex_lock(&trace_types_lock); |
2431 | if (iter->trace->read) { | |
6c6c2796 PP |
2432 | sret = iter->trace->read(iter, filp, ubuf, cnt, ppos); |
2433 | if (sret) | |
107bad8b | 2434 | goto out; |
107bad8b SR |
2435 | } |
2436 | ||
b3806b43 | 2437 | while (trace_empty(iter)) { |
2dc8f095 | 2438 | |
107bad8b | 2439 | if ((filp->f_flags & O_NONBLOCK)) { |
6c6c2796 | 2440 | sret = -EAGAIN; |
107bad8b SR |
2441 | goto out; |
2442 | } | |
2dc8f095 | 2443 | |
b3806b43 SR |
2444 | /* |
2445 | * This is a make-shift waitqueue. The reason we don't use | |
2446 | * an actual wait queue is because: | |
2447 | * 1) we only ever have one waiter | |
2448 | * 2) the tracing, traces all functions, we don't want | |
2449 | * the overhead of calling wake_up and friends | |
2450 | * (and tracing them too) | |
2451 | * Anyway, this is really very primitive wakeup. | |
2452 | */ | |
2453 | set_current_state(TASK_INTERRUPTIBLE); | |
2454 | iter->tr->waiter = current; | |
2455 | ||
107bad8b SR |
2456 | mutex_unlock(&trace_types_lock); |
2457 | ||
9fe068e9 IM |
2458 | /* sleep for 100 msecs, and try again. */ |
2459 | schedule_timeout(HZ/10); | |
b3806b43 | 2460 | |
107bad8b SR |
2461 | mutex_lock(&trace_types_lock); |
2462 | ||
b3806b43 SR |
2463 | iter->tr->waiter = NULL; |
2464 | ||
107bad8b | 2465 | if (signal_pending(current)) { |
6c6c2796 | 2466 | sret = -EINTR; |
107bad8b SR |
2467 | goto out; |
2468 | } | |
b3806b43 | 2469 | |
84527997 | 2470 | if (iter->trace != current_trace) |
107bad8b | 2471 | goto out; |
84527997 | 2472 | |
b3806b43 SR |
2473 | /* |
2474 | * We block until we read something and tracing is disabled. | |
2475 | * We still block if tracing is disabled, but we have never | |
2476 | * read anything. This allows a user to cat this file, and | |
2477 | * then enable tracing. But after we have read something, | |
2478 | * we give an EOF when tracing is again disabled. | |
2479 | * | |
2480 | * iter->pos will be 0 if we haven't read anything. | |
2481 | */ | |
2482 | if (!tracer_enabled && iter->pos) | |
2483 | break; | |
2484 | ||
2485 | continue; | |
2486 | } | |
2487 | ||
2488 | /* stop when tracing is finished */ | |
2489 | if (trace_empty(iter)) | |
107bad8b | 2490 | goto out; |
b3806b43 SR |
2491 | |
2492 | if (cnt >= PAGE_SIZE) | |
2493 | cnt = PAGE_SIZE - 1; | |
2494 | ||
53d0aa77 | 2495 | /* reset all but tr, trace, and overruns */ |
53d0aa77 SR |
2496 | memset(&iter->seq, 0, |
2497 | sizeof(struct trace_iterator) - | |
2498 | offsetof(struct trace_iterator, seq)); | |
4823ed7e | 2499 | iter->pos = -1; |
b3806b43 SR |
2500 | |
2501 | /* | |
2502 | * We need to stop all tracing on all CPUS to read the | |
2503 | * the next buffer. This is a bit expensive, but is | |
2504 | * not done often. We fill all what we can read, | |
2505 | * and then release the locks again. | |
2506 | */ | |
2507 | ||
2508 | cpus_clear(mask); | |
2509 | local_irq_save(flags); | |
25770467 | 2510 | #ifdef CONFIG_FTRACE |
2e0f5761 IM |
2511 | ftrace_save = ftrace_enabled; |
2512 | ftrace_enabled = 0; | |
25770467 | 2513 | #endif |
2e0f5761 | 2514 | smp_wmb(); |
ab46428c | 2515 | for_each_tracing_cpu(cpu) { |
b3806b43 SR |
2516 | data = iter->tr->data[cpu]; |
2517 | ||
2518 | if (!head_page(data) || !data->trace_idx) | |
2519 | continue; | |
2520 | ||
2521 | atomic_inc(&data->disabled); | |
b3806b43 SR |
2522 | cpu_set(cpu, mask); |
2523 | } | |
2524 | ||
2e0f5761 IM |
2525 | for_each_cpu_mask(cpu, mask) { |
2526 | data = iter->tr->data[cpu]; | |
92205c23 | 2527 | __raw_spin_lock(&data->lock); |
53d0aa77 SR |
2528 | |
2529 | if (data->overrun > iter->last_overrun[cpu]) | |
2530 | iter->overrun[cpu] += | |
2531 | data->overrun - iter->last_overrun[cpu]; | |
2532 | iter->last_overrun[cpu] = data->overrun; | |
2e0f5761 IM |
2533 | } |
2534 | ||
088b1e42 | 2535 | while (find_next_entry_inc(iter) != NULL) { |
6c6c2796 | 2536 | int ret; |
088b1e42 SR |
2537 | int len = iter->seq.len; |
2538 | ||
f9896bf3 | 2539 | ret = print_trace_line(iter); |
088b1e42 SR |
2540 | if (!ret) { |
2541 | /* don't print partial lines */ | |
2542 | iter->seq.len = len; | |
b3806b43 | 2543 | break; |
088b1e42 | 2544 | } |
b3806b43 SR |
2545 | |
2546 | trace_consume(iter); | |
2547 | ||
2548 | if (iter->seq.len >= cnt) | |
2549 | break; | |
b3806b43 SR |
2550 | } |
2551 | ||
d4c5a2f5 | 2552 | for_each_cpu_mask(cpu, mask) { |
b3806b43 | 2553 | data = iter->tr->data[cpu]; |
92205c23 | 2554 | __raw_spin_unlock(&data->lock); |
2e0f5761 IM |
2555 | } |
2556 | ||
2557 | for_each_cpu_mask(cpu, mask) { | |
2558 | data = iter->tr->data[cpu]; | |
b3806b43 SR |
2559 | atomic_dec(&data->disabled); |
2560 | } | |
25770467 | 2561 | #ifdef CONFIG_FTRACE |
2e0f5761 | 2562 | ftrace_enabled = ftrace_save; |
25770467 | 2563 | #endif |
b3806b43 SR |
2564 | local_irq_restore(flags); |
2565 | ||
2566 | /* Now copy what we have to the user */ | |
6c6c2796 PP |
2567 | sret = trace_seq_to_user(&iter->seq, ubuf, cnt); |
2568 | if (iter->seq.readpos >= iter->seq.len) | |
b3806b43 | 2569 | trace_seq_reset(&iter->seq); |
6c6c2796 PP |
2570 | if (sret == -EBUSY) |
2571 | sret = 0; | |
b3806b43 | 2572 | |
107bad8b SR |
2573 | out: |
2574 | mutex_unlock(&trace_types_lock); | |
2575 | ||
6c6c2796 | 2576 | return sret; |
b3806b43 SR |
2577 | } |
2578 | ||
a98a3c3f SR |
2579 | static ssize_t |
2580 | tracing_entries_read(struct file *filp, char __user *ubuf, | |
2581 | size_t cnt, loff_t *ppos) | |
2582 | { | |
2583 | struct trace_array *tr = filp->private_data; | |
2584 | char buf[64]; | |
2585 | int r; | |
2586 | ||
2587 | r = sprintf(buf, "%lu\n", tr->entries); | |
2588 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); | |
2589 | } | |
2590 | ||
2591 | static ssize_t | |
2592 | tracing_entries_write(struct file *filp, const char __user *ubuf, | |
2593 | size_t cnt, loff_t *ppos) | |
2594 | { | |
2595 | unsigned long val; | |
2596 | char buf[64]; | |
19384c03 | 2597 | int i, ret; |
a98a3c3f | 2598 | |
cffae437 SR |
2599 | if (cnt >= sizeof(buf)) |
2600 | return -EINVAL; | |
a98a3c3f SR |
2601 | |
2602 | if (copy_from_user(&buf, ubuf, cnt)) | |
2603 | return -EFAULT; | |
2604 | ||
2605 | buf[cnt] = 0; | |
2606 | ||
c6caeeb1 SR |
2607 | ret = strict_strtoul(buf, 10, &val); |
2608 | if (ret < 0) | |
2609 | return ret; | |
a98a3c3f SR |
2610 | |
2611 | /* must have at least 1 entry */ | |
2612 | if (!val) | |
2613 | return -EINVAL; | |
2614 | ||
2615 | mutex_lock(&trace_types_lock); | |
2616 | ||
2617 | if (current_trace != &no_tracer) { | |
2618 | cnt = -EBUSY; | |
2619 | pr_info("ftrace: set current_tracer to none" | |
2620 | " before modifying buffer size\n"); | |
2621 | goto out; | |
2622 | } | |
2623 | ||
2624 | if (val > global_trace.entries) { | |
3eefae99 SR |
2625 | long pages_requested; |
2626 | unsigned long freeable_pages; | |
2627 | ||
2628 | /* make sure we have enough memory before mapping */ | |
2629 | pages_requested = | |
2630 | (val + (ENTRIES_PER_PAGE-1)) / ENTRIES_PER_PAGE; | |
2631 | ||
2632 | /* account for each buffer (and max_tr) */ | |
2633 | pages_requested *= tracing_nr_buffers * 2; | |
2634 | ||
2635 | /* Check for overflow */ | |
2636 | if (pages_requested < 0) { | |
2637 | cnt = -ENOMEM; | |
2638 | goto out; | |
2639 | } | |
2640 | ||
2641 | freeable_pages = determine_dirtyable_memory(); | |
2642 | ||
2643 | /* we only allow to request 1/4 of useable memory */ | |
2644 | if (pages_requested > | |
2645 | ((freeable_pages + tracing_pages_allocated) / 4)) { | |
2646 | cnt = -ENOMEM; | |
2647 | goto out; | |
2648 | } | |
2649 | ||
a98a3c3f SR |
2650 | while (global_trace.entries < val) { |
2651 | if (trace_alloc_page()) { | |
2652 | cnt = -ENOMEM; | |
2653 | goto out; | |
2654 | } | |
3eefae99 SR |
2655 | /* double check that we don't go over the known pages */ |
2656 | if (tracing_pages_allocated > pages_requested) | |
2657 | break; | |
a98a3c3f | 2658 | } |
3eefae99 | 2659 | |
a98a3c3f SR |
2660 | } else { |
2661 | /* include the number of entries in val (inc of page entries) */ | |
2662 | while (global_trace.entries > val + (ENTRIES_PER_PAGE - 1)) | |
2663 | trace_free_page(); | |
2664 | } | |
2665 | ||
19384c03 SR |
2666 | /* check integrity */ |
2667 | for_each_tracing_cpu(i) | |
2668 | check_pages(global_trace.data[i]); | |
2669 | ||
a98a3c3f SR |
2670 | filp->f_pos += cnt; |
2671 | ||
19384c03 SR |
2672 | /* If check pages failed, return ENOMEM */ |
2673 | if (tracing_disabled) | |
2674 | cnt = -ENOMEM; | |
a98a3c3f SR |
2675 | out: |
2676 | max_tr.entries = global_trace.entries; | |
2677 | mutex_unlock(&trace_types_lock); | |
2678 | ||
2679 | return cnt; | |
2680 | } | |
2681 | ||
bc0c38d1 | 2682 | static struct file_operations tracing_max_lat_fops = { |
4bf39a94 IM |
2683 | .open = tracing_open_generic, |
2684 | .read = tracing_max_lat_read, | |
2685 | .write = tracing_max_lat_write, | |
bc0c38d1 SR |
2686 | }; |
2687 | ||
2688 | static struct file_operations tracing_ctrl_fops = { | |
4bf39a94 IM |
2689 | .open = tracing_open_generic, |
2690 | .read = tracing_ctrl_read, | |
2691 | .write = tracing_ctrl_write, | |
bc0c38d1 SR |
2692 | }; |
2693 | ||
2694 | static struct file_operations set_tracer_fops = { | |
4bf39a94 IM |
2695 | .open = tracing_open_generic, |
2696 | .read = tracing_set_trace_read, | |
2697 | .write = tracing_set_trace_write, | |
bc0c38d1 SR |
2698 | }; |
2699 | ||
b3806b43 | 2700 | static struct file_operations tracing_pipe_fops = { |
4bf39a94 | 2701 | .open = tracing_open_pipe, |
2a2cc8f7 | 2702 | .poll = tracing_poll_pipe, |
4bf39a94 IM |
2703 | .read = tracing_read_pipe, |
2704 | .release = tracing_release_pipe, | |
b3806b43 SR |
2705 | }; |
2706 | ||
a98a3c3f SR |
2707 | static struct file_operations tracing_entries_fops = { |
2708 | .open = tracing_open_generic, | |
2709 | .read = tracing_entries_read, | |
2710 | .write = tracing_entries_write, | |
2711 | }; | |
2712 | ||
bc0c38d1 SR |
2713 | #ifdef CONFIG_DYNAMIC_FTRACE |
2714 | ||
2715 | static ssize_t | |
2716 | tracing_read_long(struct file *filp, char __user *ubuf, | |
2717 | size_t cnt, loff_t *ppos) | |
2718 | { | |
2719 | unsigned long *p = filp->private_data; | |
2720 | char buf[64]; | |
2721 | int r; | |
2722 | ||
2723 | r = sprintf(buf, "%ld\n", *p); | |
4bf39a94 IM |
2724 | |
2725 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); | |
bc0c38d1 SR |
2726 | } |
2727 | ||
2728 | static struct file_operations tracing_read_long_fops = { | |
4bf39a94 IM |
2729 | .open = tracing_open_generic, |
2730 | .read = tracing_read_long, | |
bc0c38d1 SR |
2731 | }; |
2732 | #endif | |
2733 | ||
2734 | static struct dentry *d_tracer; | |
2735 | ||
2736 | struct dentry *tracing_init_dentry(void) | |
2737 | { | |
2738 | static int once; | |
2739 | ||
2740 | if (d_tracer) | |
2741 | return d_tracer; | |
2742 | ||
2743 | d_tracer = debugfs_create_dir("tracing", NULL); | |
2744 | ||
2745 | if (!d_tracer && !once) { | |
2746 | once = 1; | |
2747 | pr_warning("Could not create debugfs directory 'tracing'\n"); | |
2748 | return NULL; | |
2749 | } | |
2750 | ||
2751 | return d_tracer; | |
2752 | } | |
2753 | ||
60a11774 SR |
2754 | #ifdef CONFIG_FTRACE_SELFTEST |
2755 | /* Let selftest have access to static functions in this file */ | |
2756 | #include "trace_selftest.c" | |
2757 | #endif | |
2758 | ||
bc0c38d1 SR |
2759 | static __init void tracer_init_debugfs(void) |
2760 | { | |
2761 | struct dentry *d_tracer; | |
2762 | struct dentry *entry; | |
2763 | ||
2764 | d_tracer = tracing_init_dentry(); | |
2765 | ||
2766 | entry = debugfs_create_file("tracing_enabled", 0644, d_tracer, | |
2767 | &global_trace, &tracing_ctrl_fops); | |
2768 | if (!entry) | |
2769 | pr_warning("Could not create debugfs 'tracing_enabled' entry\n"); | |
2770 | ||
2771 | entry = debugfs_create_file("iter_ctrl", 0644, d_tracer, | |
2772 | NULL, &tracing_iter_fops); | |
2773 | if (!entry) | |
2774 | pr_warning("Could not create debugfs 'iter_ctrl' entry\n"); | |
2775 | ||
c7078de1 IM |
2776 | entry = debugfs_create_file("tracing_cpumask", 0644, d_tracer, |
2777 | NULL, &tracing_cpumask_fops); | |
2778 | if (!entry) | |
2779 | pr_warning("Could not create debugfs 'tracing_cpumask' entry\n"); | |
2780 | ||
bc0c38d1 SR |
2781 | entry = debugfs_create_file("latency_trace", 0444, d_tracer, |
2782 | &global_trace, &tracing_lt_fops); | |
2783 | if (!entry) | |
2784 | pr_warning("Could not create debugfs 'latency_trace' entry\n"); | |
2785 | ||
2786 | entry = debugfs_create_file("trace", 0444, d_tracer, | |
2787 | &global_trace, &tracing_fops); | |
2788 | if (!entry) | |
2789 | pr_warning("Could not create debugfs 'trace' entry\n"); | |
2790 | ||
2791 | entry = debugfs_create_file("available_tracers", 0444, d_tracer, | |
2792 | &global_trace, &show_traces_fops); | |
2793 | if (!entry) | |
2794 | pr_warning("Could not create debugfs 'trace' entry\n"); | |
2795 | ||
2796 | entry = debugfs_create_file("current_tracer", 0444, d_tracer, | |
2797 | &global_trace, &set_tracer_fops); | |
2798 | if (!entry) | |
2799 | pr_warning("Could not create debugfs 'trace' entry\n"); | |
2800 | ||
2801 | entry = debugfs_create_file("tracing_max_latency", 0644, d_tracer, | |
2802 | &tracing_max_latency, | |
2803 | &tracing_max_lat_fops); | |
2804 | if (!entry) | |
2805 | pr_warning("Could not create debugfs " | |
2806 | "'tracing_max_latency' entry\n"); | |
2807 | ||
2808 | entry = debugfs_create_file("tracing_thresh", 0644, d_tracer, | |
2809 | &tracing_thresh, &tracing_max_lat_fops); | |
2810 | if (!entry) | |
2811 | pr_warning("Could not create debugfs " | |
2812 | "'tracing_threash' entry\n"); | |
7bd2f24c IM |
2813 | entry = debugfs_create_file("README", 0644, d_tracer, |
2814 | NULL, &tracing_readme_fops); | |
2815 | if (!entry) | |
2816 | pr_warning("Could not create debugfs 'README' entry\n"); | |
2817 | ||
b3806b43 SR |
2818 | entry = debugfs_create_file("trace_pipe", 0644, d_tracer, |
2819 | NULL, &tracing_pipe_fops); | |
2820 | if (!entry) | |
2821 | pr_warning("Could not create debugfs " | |
2822 | "'tracing_threash' entry\n"); | |
bc0c38d1 | 2823 | |
a98a3c3f SR |
2824 | entry = debugfs_create_file("trace_entries", 0644, d_tracer, |
2825 | &global_trace, &tracing_entries_fops); | |
2826 | if (!entry) | |
2827 | pr_warning("Could not create debugfs " | |
2828 | "'tracing_threash' entry\n"); | |
2829 | ||
bc0c38d1 SR |
2830 | #ifdef CONFIG_DYNAMIC_FTRACE |
2831 | entry = debugfs_create_file("dyn_ftrace_total_info", 0444, d_tracer, | |
2832 | &ftrace_update_tot_cnt, | |
2833 | &tracing_read_long_fops); | |
2834 | if (!entry) | |
2835 | pr_warning("Could not create debugfs " | |
2836 | "'dyn_ftrace_total_info' entry\n"); | |
2837 | #endif | |
2838 | } | |
2839 | ||
4c11d7ae | 2840 | static int trace_alloc_page(void) |
bc0c38d1 | 2841 | { |
4c11d7ae | 2842 | struct trace_array_cpu *data; |
4c11d7ae SR |
2843 | struct page *page, *tmp; |
2844 | LIST_HEAD(pages); | |
c7aafc54 | 2845 | void *array; |
3eefae99 | 2846 | unsigned pages_allocated = 0; |
4c11d7ae SR |
2847 | int i; |
2848 | ||
2849 | /* first allocate a page for each CPU */ | |
ab46428c | 2850 | for_each_tracing_cpu(i) { |
4c11d7ae SR |
2851 | array = (void *)__get_free_page(GFP_KERNEL); |
2852 | if (array == NULL) { | |
2853 | printk(KERN_ERR "tracer: failed to allocate page" | |
2854 | "for trace buffer!\n"); | |
2855 | goto free_pages; | |
2856 | } | |
2857 | ||
3eefae99 | 2858 | pages_allocated++; |
4c11d7ae SR |
2859 | page = virt_to_page(array); |
2860 | list_add(&page->lru, &pages); | |
2861 | ||
2862 | /* Only allocate if we are actually using the max trace */ | |
2863 | #ifdef CONFIG_TRACER_MAX_TRACE | |
2864 | array = (void *)__get_free_page(GFP_KERNEL); | |
2865 | if (array == NULL) { | |
2866 | printk(KERN_ERR "tracer: failed to allocate page" | |
2867 | "for trace buffer!\n"); | |
2868 | goto free_pages; | |
2869 | } | |
3eefae99 | 2870 | pages_allocated++; |
4c11d7ae SR |
2871 | page = virt_to_page(array); |
2872 | list_add(&page->lru, &pages); | |
2873 | #endif | |
2874 | } | |
2875 | ||
2876 | /* Now that we successfully allocate a page per CPU, add them */ | |
ab46428c | 2877 | for_each_tracing_cpu(i) { |
4c11d7ae SR |
2878 | data = global_trace.data[i]; |
2879 | page = list_entry(pages.next, struct page, lru); | |
c7aafc54 | 2880 | list_del_init(&page->lru); |
4c11d7ae SR |
2881 | list_add_tail(&page->lru, &data->trace_pages); |
2882 | ClearPageLRU(page); | |
2883 | ||
2884 | #ifdef CONFIG_TRACER_MAX_TRACE | |
2885 | data = max_tr.data[i]; | |
2886 | page = list_entry(pages.next, struct page, lru); | |
c7aafc54 | 2887 | list_del_init(&page->lru); |
4c11d7ae SR |
2888 | list_add_tail(&page->lru, &data->trace_pages); |
2889 | SetPageLRU(page); | |
2890 | #endif | |
2891 | } | |
3eefae99 | 2892 | tracing_pages_allocated += pages_allocated; |
4c11d7ae SR |
2893 | global_trace.entries += ENTRIES_PER_PAGE; |
2894 | ||
2895 | return 0; | |
2896 | ||
2897 | free_pages: | |
2898 | list_for_each_entry_safe(page, tmp, &pages, lru) { | |
c7aafc54 | 2899 | list_del_init(&page->lru); |
4c11d7ae SR |
2900 | __free_page(page); |
2901 | } | |
2902 | return -ENOMEM; | |
bc0c38d1 SR |
2903 | } |
2904 | ||
a98a3c3f SR |
2905 | static int trace_free_page(void) |
2906 | { | |
2907 | struct trace_array_cpu *data; | |
2908 | struct page *page; | |
2909 | struct list_head *p; | |
2910 | int i; | |
2911 | int ret = 0; | |
2912 | ||
2913 | /* free one page from each buffer */ | |
ab46428c | 2914 | for_each_tracing_cpu(i) { |
a98a3c3f SR |
2915 | data = global_trace.data[i]; |
2916 | p = data->trace_pages.next; | |
2917 | if (p == &data->trace_pages) { | |
2918 | /* should never happen */ | |
2919 | WARN_ON(1); | |
2920 | tracing_disabled = 1; | |
2921 | ret = -1; | |
2922 | break; | |
2923 | } | |
2924 | page = list_entry(p, struct page, lru); | |
2925 | ClearPageLRU(page); | |
2926 | list_del(&page->lru); | |
3eefae99 SR |
2927 | tracing_pages_allocated--; |
2928 | tracing_pages_allocated--; | |
a98a3c3f SR |
2929 | __free_page(page); |
2930 | ||
2931 | tracing_reset(data); | |
2932 | ||
2933 | #ifdef CONFIG_TRACER_MAX_TRACE | |
2934 | data = max_tr.data[i]; | |
2935 | p = data->trace_pages.next; | |
2936 | if (p == &data->trace_pages) { | |
2937 | /* should never happen */ | |
2938 | WARN_ON(1); | |
2939 | tracing_disabled = 1; | |
2940 | ret = -1; | |
2941 | break; | |
2942 | } | |
2943 | page = list_entry(p, struct page, lru); | |
2944 | ClearPageLRU(page); | |
2945 | list_del(&page->lru); | |
2946 | __free_page(page); | |
2947 | ||
2948 | tracing_reset(data); | |
2949 | #endif | |
2950 | } | |
2951 | global_trace.entries -= ENTRIES_PER_PAGE; | |
2952 | ||
2953 | return ret; | |
2954 | } | |
2955 | ||
bc0c38d1 SR |
2956 | __init static int tracer_alloc_buffers(void) |
2957 | { | |
4c11d7ae SR |
2958 | struct trace_array_cpu *data; |
2959 | void *array; | |
2960 | struct page *page; | |
2961 | int pages = 0; | |
60a11774 | 2962 | int ret = -ENOMEM; |
bc0c38d1 SR |
2963 | int i; |
2964 | ||
ab46428c SR |
2965 | /* TODO: make the number of buffers hot pluggable with CPUS */ |
2966 | tracing_nr_buffers = num_possible_cpus(); | |
2967 | tracing_buffer_mask = cpu_possible_map; | |
2968 | ||
4c11d7ae | 2969 | /* Allocate the first page for all buffers */ |
ab46428c | 2970 | for_each_tracing_cpu(i) { |
4c11d7ae | 2971 | data = global_trace.data[i] = &per_cpu(global_trace_cpu, i); |
bc0c38d1 SR |
2972 | max_tr.data[i] = &per_cpu(max_data, i); |
2973 | ||
4c11d7ae | 2974 | array = (void *)__get_free_page(GFP_KERNEL); |
bc0c38d1 | 2975 | if (array == NULL) { |
4c11d7ae SR |
2976 | printk(KERN_ERR "tracer: failed to allocate page" |
2977 | "for trace buffer!\n"); | |
bc0c38d1 SR |
2978 | goto free_buffers; |
2979 | } | |
4c11d7ae SR |
2980 | |
2981 | /* set the array to the list */ | |
2982 | INIT_LIST_HEAD(&data->trace_pages); | |
2983 | page = virt_to_page(array); | |
2984 | list_add(&page->lru, &data->trace_pages); | |
2985 | /* use the LRU flag to differentiate the two buffers */ | |
2986 | ClearPageLRU(page); | |
bc0c38d1 | 2987 | |
a98a3c3f SR |
2988 | data->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; |
2989 | max_tr.data[i]->lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED; | |
2990 | ||
bc0c38d1 SR |
2991 | /* Only allocate if we are actually using the max trace */ |
2992 | #ifdef CONFIG_TRACER_MAX_TRACE | |
4c11d7ae | 2993 | array = (void *)__get_free_page(GFP_KERNEL); |
bc0c38d1 | 2994 | if (array == NULL) { |
4c11d7ae SR |
2995 | printk(KERN_ERR "tracer: failed to allocate page" |
2996 | "for trace buffer!\n"); | |
bc0c38d1 SR |
2997 | goto free_buffers; |
2998 | } | |
4c11d7ae SR |
2999 | |
3000 | INIT_LIST_HEAD(&max_tr.data[i]->trace_pages); | |
3001 | page = virt_to_page(array); | |
3002 | list_add(&page->lru, &max_tr.data[i]->trace_pages); | |
3003 | SetPageLRU(page); | |
bc0c38d1 SR |
3004 | #endif |
3005 | } | |
3006 | ||
3007 | /* | |
3008 | * Since we allocate by orders of pages, we may be able to | |
3009 | * round up a bit. | |
3010 | */ | |
4c11d7ae | 3011 | global_trace.entries = ENTRIES_PER_PAGE; |
4c11d7ae SR |
3012 | pages++; |
3013 | ||
3014 | while (global_trace.entries < trace_nr_entries) { | |
3015 | if (trace_alloc_page()) | |
3016 | break; | |
3017 | pages++; | |
3018 | } | |
89b2f978 | 3019 | max_tr.entries = global_trace.entries; |
bc0c38d1 | 3020 | |
4c11d7ae SR |
3021 | pr_info("tracer: %d pages allocated for %ld", |
3022 | pages, trace_nr_entries); | |
bc0c38d1 SR |
3023 | pr_info(" entries of %ld bytes\n", (long)TRACE_ENTRY_SIZE); |
3024 | pr_info(" actual entries %ld\n", global_trace.entries); | |
3025 | ||
3026 | tracer_init_debugfs(); | |
3027 | ||
3028 | trace_init_cmdlines(); | |
3029 | ||
3030 | register_tracer(&no_tracer); | |
3031 | current_trace = &no_tracer; | |
3032 | ||
60a11774 | 3033 | /* All seems OK, enable tracing */ |
4902f884 | 3034 | global_trace.ctrl = tracer_enabled; |
60a11774 SR |
3035 | tracing_disabled = 0; |
3036 | ||
bc0c38d1 SR |
3037 | return 0; |
3038 | ||
3039 | free_buffers: | |
3040 | for (i-- ; i >= 0; i--) { | |
4c11d7ae | 3041 | struct page *page, *tmp; |
bc0c38d1 SR |
3042 | struct trace_array_cpu *data = global_trace.data[i]; |
3043 | ||
c7aafc54 | 3044 | if (data) { |
4c11d7ae SR |
3045 | list_for_each_entry_safe(page, tmp, |
3046 | &data->trace_pages, lru) { | |
c7aafc54 | 3047 | list_del_init(&page->lru); |
4c11d7ae SR |
3048 | __free_page(page); |
3049 | } | |
bc0c38d1 SR |
3050 | } |
3051 | ||
3052 | #ifdef CONFIG_TRACER_MAX_TRACE | |
3053 | data = max_tr.data[i]; | |
c7aafc54 | 3054 | if (data) { |
4c11d7ae SR |
3055 | list_for_each_entry_safe(page, tmp, |
3056 | &data->trace_pages, lru) { | |
c7aafc54 | 3057 | list_del_init(&page->lru); |
4c11d7ae SR |
3058 | __free_page(page); |
3059 | } | |
bc0c38d1 SR |
3060 | } |
3061 | #endif | |
3062 | } | |
60a11774 | 3063 | return ret; |
bc0c38d1 | 3064 | } |
60a11774 | 3065 | fs_initcall(tracer_alloc_buffers); |