]>
Commit | Line | Data |
---|---|---|
7a8e76a3 SR |
1 | /* |
2 | * Generic ring buffer | |
3 | * | |
4 | * Copyright (C) 2008 Steven Rostedt <srostedt@redhat.com> | |
5 | */ | |
6 | #include <linux/ring_buffer.h> | |
7 | #include <linux/spinlock.h> | |
8 | #include <linux/debugfs.h> | |
9 | #include <linux/uaccess.h> | |
10 | #include <linux/module.h> | |
11 | #include <linux/percpu.h> | |
12 | #include <linux/mutex.h> | |
13 | #include <linux/sched.h> /* used for sched_clock() (for now) */ | |
14 | #include <linux/init.h> | |
15 | #include <linux/hash.h> | |
16 | #include <linux/list.h> | |
17 | #include <linux/fs.h> | |
18 | ||
182e9f5f SR |
19 | #include "trace.h" |
20 | ||
7a8e76a3 SR |
21 | /* Up this if you want to test the TIME_EXTENTS and normalization */ |
22 | #define DEBUG_SHIFT 0 | |
23 | ||
24 | /* FIXME!!! */ | |
25 | u64 ring_buffer_time_stamp(int cpu) | |
26 | { | |
27 | /* shift to debug/test normalization and TIME_EXTENTS */ | |
28 | return sched_clock() << DEBUG_SHIFT; | |
29 | } | |
30 | ||
31 | void ring_buffer_normalize_time_stamp(int cpu, u64 *ts) | |
32 | { | |
33 | /* Just stupid testing the normalize function and deltas */ | |
34 | *ts >>= DEBUG_SHIFT; | |
35 | } | |
36 | ||
37 | #define RB_EVNT_HDR_SIZE (sizeof(struct ring_buffer_event)) | |
38 | #define RB_ALIGNMENT_SHIFT 2 | |
39 | #define RB_ALIGNMENT (1 << RB_ALIGNMENT_SHIFT) | |
40 | #define RB_MAX_SMALL_DATA 28 | |
41 | ||
42 | enum { | |
43 | RB_LEN_TIME_EXTEND = 8, | |
44 | RB_LEN_TIME_STAMP = 16, | |
45 | }; | |
46 | ||
47 | /* inline for ring buffer fast paths */ | |
48 | static inline unsigned | |
49 | rb_event_length(struct ring_buffer_event *event) | |
50 | { | |
51 | unsigned length; | |
52 | ||
53 | switch (event->type) { | |
54 | case RINGBUF_TYPE_PADDING: | |
55 | /* undefined */ | |
56 | return -1; | |
57 | ||
58 | case RINGBUF_TYPE_TIME_EXTEND: | |
59 | return RB_LEN_TIME_EXTEND; | |
60 | ||
61 | case RINGBUF_TYPE_TIME_STAMP: | |
62 | return RB_LEN_TIME_STAMP; | |
63 | ||
64 | case RINGBUF_TYPE_DATA: | |
65 | if (event->len) | |
66 | length = event->len << RB_ALIGNMENT_SHIFT; | |
67 | else | |
68 | length = event->array[0]; | |
69 | return length + RB_EVNT_HDR_SIZE; | |
70 | default: | |
71 | BUG(); | |
72 | } | |
73 | /* not hit */ | |
74 | return 0; | |
75 | } | |
76 | ||
77 | /** | |
78 | * ring_buffer_event_length - return the length of the event | |
79 | * @event: the event to get the length of | |
80 | */ | |
81 | unsigned ring_buffer_event_length(struct ring_buffer_event *event) | |
82 | { | |
83 | return rb_event_length(event); | |
84 | } | |
85 | ||
86 | /* inline for ring buffer fast paths */ | |
87 | static inline void * | |
88 | rb_event_data(struct ring_buffer_event *event) | |
89 | { | |
90 | BUG_ON(event->type != RINGBUF_TYPE_DATA); | |
91 | /* If length is in len field, then array[0] has the data */ | |
92 | if (event->len) | |
93 | return (void *)&event->array[0]; | |
94 | /* Otherwise length is in array[0] and array[1] has the data */ | |
95 | return (void *)&event->array[1]; | |
96 | } | |
97 | ||
98 | /** | |
99 | * ring_buffer_event_data - return the data of the event | |
100 | * @event: the event to get the data from | |
101 | */ | |
102 | void *ring_buffer_event_data(struct ring_buffer_event *event) | |
103 | { | |
104 | return rb_event_data(event); | |
105 | } | |
106 | ||
107 | #define for_each_buffer_cpu(buffer, cpu) \ | |
108 | for_each_cpu_mask(cpu, buffer->cpumask) | |
109 | ||
110 | #define TS_SHIFT 27 | |
111 | #define TS_MASK ((1ULL << TS_SHIFT) - 1) | |
112 | #define TS_DELTA_TEST (~TS_MASK) | |
113 | ||
114 | /* | |
115 | * This hack stolen from mm/slob.c. | |
116 | * We can store per page timing information in the page frame of the page. | |
117 | * Thanks to Peter Zijlstra for suggesting this idea. | |
118 | */ | |
119 | struct buffer_page { | |
e4c2ce82 | 120 | u64 time_stamp; /* page time stamp */ |
bf41a158 SR |
121 | local_t write; /* index for next write */ |
122 | local_t commit; /* write commited index */ | |
6f807acd | 123 | unsigned read; /* index for next read */ |
e4c2ce82 SR |
124 | struct list_head list; /* list of free pages */ |
125 | void *page; /* Actual data page */ | |
7a8e76a3 SR |
126 | }; |
127 | ||
ed56829c SR |
128 | /* |
129 | * Also stolen from mm/slob.c. Thanks to Mathieu Desnoyers for pointing | |
130 | * this issue out. | |
131 | */ | |
132 | static inline void free_buffer_page(struct buffer_page *bpage) | |
133 | { | |
e4c2ce82 | 134 | if (bpage->page) |
6ae2a076 | 135 | free_page((unsigned long)bpage->page); |
e4c2ce82 | 136 | kfree(bpage); |
ed56829c SR |
137 | } |
138 | ||
7a8e76a3 SR |
139 | /* |
140 | * We need to fit the time_stamp delta into 27 bits. | |
141 | */ | |
142 | static inline int test_time_stamp(u64 delta) | |
143 | { | |
144 | if (delta & TS_DELTA_TEST) | |
145 | return 1; | |
146 | return 0; | |
147 | } | |
148 | ||
149 | #define BUF_PAGE_SIZE PAGE_SIZE | |
150 | ||
151 | /* | |
152 | * head_page == tail_page && head == tail then buffer is empty. | |
153 | */ | |
154 | struct ring_buffer_per_cpu { | |
155 | int cpu; | |
156 | struct ring_buffer *buffer; | |
157 | spinlock_t lock; | |
158 | struct lock_class_key lock_key; | |
159 | struct list_head pages; | |
6f807acd SR |
160 | struct buffer_page *head_page; /* read from head */ |
161 | struct buffer_page *tail_page; /* write to tail */ | |
bf41a158 | 162 | struct buffer_page *commit_page; /* commited pages */ |
d769041f | 163 | struct buffer_page *reader_page; |
7a8e76a3 SR |
164 | unsigned long overrun; |
165 | unsigned long entries; | |
166 | u64 write_stamp; | |
167 | u64 read_stamp; | |
168 | atomic_t record_disabled; | |
169 | }; | |
170 | ||
171 | struct ring_buffer { | |
172 | unsigned long size; | |
173 | unsigned pages; | |
174 | unsigned flags; | |
175 | int cpus; | |
176 | cpumask_t cpumask; | |
177 | atomic_t record_disabled; | |
178 | ||
179 | struct mutex mutex; | |
180 | ||
181 | struct ring_buffer_per_cpu **buffers; | |
182 | }; | |
183 | ||
184 | struct ring_buffer_iter { | |
185 | struct ring_buffer_per_cpu *cpu_buffer; | |
186 | unsigned long head; | |
187 | struct buffer_page *head_page; | |
188 | u64 read_stamp; | |
189 | }; | |
190 | ||
bf41a158 SR |
191 | #define RB_WARN_ON(buffer, cond) \ |
192 | do { \ | |
193 | if (unlikely(cond)) { \ | |
194 | atomic_inc(&buffer->record_disabled); \ | |
195 | WARN_ON(1); \ | |
196 | } \ | |
197 | } while (0) | |
198 | ||
199 | #define RB_WARN_ON_RET(buffer, cond) \ | |
200 | do { \ | |
201 | if (unlikely(cond)) { \ | |
202 | atomic_inc(&buffer->record_disabled); \ | |
203 | WARN_ON(1); \ | |
204 | return -1; \ | |
205 | } \ | |
206 | } while (0) | |
207 | ||
208 | #define RB_WARN_ON_ONCE(buffer, cond) \ | |
209 | do { \ | |
210 | static int once; \ | |
211 | if (unlikely(cond) && !once) { \ | |
212 | once++; \ | |
213 | atomic_inc(&buffer->record_disabled); \ | |
214 | WARN_ON(1); \ | |
215 | } \ | |
216 | } while (0) | |
7a8e76a3 SR |
217 | |
218 | /** | |
219 | * check_pages - integrity check of buffer pages | |
220 | * @cpu_buffer: CPU buffer with pages to test | |
221 | * | |
222 | * As a safty measure we check to make sure the data pages have not | |
223 | * been corrupted. | |
224 | */ | |
225 | static int rb_check_pages(struct ring_buffer_per_cpu *cpu_buffer) | |
226 | { | |
227 | struct list_head *head = &cpu_buffer->pages; | |
228 | struct buffer_page *page, *tmp; | |
229 | ||
bf41a158 SR |
230 | RB_WARN_ON_RET(cpu_buffer, head->next->prev != head); |
231 | RB_WARN_ON_RET(cpu_buffer, head->prev->next != head); | |
7a8e76a3 SR |
232 | |
233 | list_for_each_entry_safe(page, tmp, head, list) { | |
bf41a158 SR |
234 | RB_WARN_ON_RET(cpu_buffer, |
235 | page->list.next->prev != &page->list); | |
236 | RB_WARN_ON_RET(cpu_buffer, | |
237 | page->list.prev->next != &page->list); | |
7a8e76a3 SR |
238 | } |
239 | ||
240 | return 0; | |
241 | } | |
242 | ||
7a8e76a3 SR |
243 | static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, |
244 | unsigned nr_pages) | |
245 | { | |
246 | struct list_head *head = &cpu_buffer->pages; | |
247 | struct buffer_page *page, *tmp; | |
248 | unsigned long addr; | |
249 | LIST_HEAD(pages); | |
250 | unsigned i; | |
251 | ||
252 | for (i = 0; i < nr_pages; i++) { | |
e4c2ce82 | 253 | page = kzalloc_node(ALIGN(sizeof(*page), cache_line_size()), |
aa1e0e3b | 254 | GFP_KERNEL, cpu_to_node(cpu_buffer->cpu)); |
e4c2ce82 SR |
255 | if (!page) |
256 | goto free_pages; | |
257 | list_add(&page->list, &pages); | |
258 | ||
7a8e76a3 SR |
259 | addr = __get_free_page(GFP_KERNEL); |
260 | if (!addr) | |
261 | goto free_pages; | |
e4c2ce82 | 262 | page->page = (void *)addr; |
7a8e76a3 SR |
263 | } |
264 | ||
265 | list_splice(&pages, head); | |
266 | ||
267 | rb_check_pages(cpu_buffer); | |
268 | ||
269 | return 0; | |
270 | ||
271 | free_pages: | |
272 | list_for_each_entry_safe(page, tmp, &pages, list) { | |
273 | list_del_init(&page->list); | |
ed56829c | 274 | free_buffer_page(page); |
7a8e76a3 SR |
275 | } |
276 | return -ENOMEM; | |
277 | } | |
278 | ||
279 | static struct ring_buffer_per_cpu * | |
280 | rb_allocate_cpu_buffer(struct ring_buffer *buffer, int cpu) | |
281 | { | |
282 | struct ring_buffer_per_cpu *cpu_buffer; | |
e4c2ce82 | 283 | struct buffer_page *page; |
d769041f | 284 | unsigned long addr; |
7a8e76a3 SR |
285 | int ret; |
286 | ||
287 | cpu_buffer = kzalloc_node(ALIGN(sizeof(*cpu_buffer), cache_line_size()), | |
288 | GFP_KERNEL, cpu_to_node(cpu)); | |
289 | if (!cpu_buffer) | |
290 | return NULL; | |
291 | ||
292 | cpu_buffer->cpu = cpu; | |
293 | cpu_buffer->buffer = buffer; | |
294 | spin_lock_init(&cpu_buffer->lock); | |
295 | INIT_LIST_HEAD(&cpu_buffer->pages); | |
296 | ||
e4c2ce82 SR |
297 | page = kzalloc_node(ALIGN(sizeof(*page), cache_line_size()), |
298 | GFP_KERNEL, cpu_to_node(cpu)); | |
299 | if (!page) | |
300 | goto fail_free_buffer; | |
301 | ||
302 | cpu_buffer->reader_page = page; | |
d769041f SR |
303 | addr = __get_free_page(GFP_KERNEL); |
304 | if (!addr) | |
e4c2ce82 SR |
305 | goto fail_free_reader; |
306 | page->page = (void *)addr; | |
307 | ||
d769041f | 308 | INIT_LIST_HEAD(&cpu_buffer->reader_page->list); |
d769041f | 309 | |
7a8e76a3 SR |
310 | ret = rb_allocate_pages(cpu_buffer, buffer->pages); |
311 | if (ret < 0) | |
d769041f | 312 | goto fail_free_reader; |
7a8e76a3 SR |
313 | |
314 | cpu_buffer->head_page | |
315 | = list_entry(cpu_buffer->pages.next, struct buffer_page, list); | |
bf41a158 | 316 | cpu_buffer->tail_page = cpu_buffer->commit_page = cpu_buffer->head_page; |
7a8e76a3 SR |
317 | |
318 | return cpu_buffer; | |
319 | ||
d769041f SR |
320 | fail_free_reader: |
321 | free_buffer_page(cpu_buffer->reader_page); | |
322 | ||
7a8e76a3 SR |
323 | fail_free_buffer: |
324 | kfree(cpu_buffer); | |
325 | return NULL; | |
326 | } | |
327 | ||
328 | static void rb_free_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer) | |
329 | { | |
330 | struct list_head *head = &cpu_buffer->pages; | |
331 | struct buffer_page *page, *tmp; | |
332 | ||
d769041f SR |
333 | list_del_init(&cpu_buffer->reader_page->list); |
334 | free_buffer_page(cpu_buffer->reader_page); | |
335 | ||
7a8e76a3 SR |
336 | list_for_each_entry_safe(page, tmp, head, list) { |
337 | list_del_init(&page->list); | |
ed56829c | 338 | free_buffer_page(page); |
7a8e76a3 SR |
339 | } |
340 | kfree(cpu_buffer); | |
341 | } | |
342 | ||
a7b13743 SR |
343 | /* |
344 | * Causes compile errors if the struct buffer_page gets bigger | |
345 | * than the struct page. | |
346 | */ | |
347 | extern int ring_buffer_page_too_big(void); | |
348 | ||
7a8e76a3 SR |
349 | /** |
350 | * ring_buffer_alloc - allocate a new ring_buffer | |
351 | * @size: the size in bytes that is needed. | |
352 | * @flags: attributes to set for the ring buffer. | |
353 | * | |
354 | * Currently the only flag that is available is the RB_FL_OVERWRITE | |
355 | * flag. This flag means that the buffer will overwrite old data | |
356 | * when the buffer wraps. If this flag is not set, the buffer will | |
357 | * drop data when the tail hits the head. | |
358 | */ | |
359 | struct ring_buffer *ring_buffer_alloc(unsigned long size, unsigned flags) | |
360 | { | |
361 | struct ring_buffer *buffer; | |
362 | int bsize; | |
363 | int cpu; | |
364 | ||
a7b13743 SR |
365 | /* Paranoid! Optimizes out when all is well */ |
366 | if (sizeof(struct buffer_page) > sizeof(struct page)) | |
367 | ring_buffer_page_too_big(); | |
368 | ||
369 | ||
7a8e76a3 SR |
370 | /* keep it in its own cache line */ |
371 | buffer = kzalloc(ALIGN(sizeof(*buffer), cache_line_size()), | |
372 | GFP_KERNEL); | |
373 | if (!buffer) | |
374 | return NULL; | |
375 | ||
376 | buffer->pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE); | |
377 | buffer->flags = flags; | |
378 | ||
379 | /* need at least two pages */ | |
380 | if (buffer->pages == 1) | |
381 | buffer->pages++; | |
382 | ||
383 | buffer->cpumask = cpu_possible_map; | |
384 | buffer->cpus = nr_cpu_ids; | |
385 | ||
386 | bsize = sizeof(void *) * nr_cpu_ids; | |
387 | buffer->buffers = kzalloc(ALIGN(bsize, cache_line_size()), | |
388 | GFP_KERNEL); | |
389 | if (!buffer->buffers) | |
390 | goto fail_free_buffer; | |
391 | ||
392 | for_each_buffer_cpu(buffer, cpu) { | |
393 | buffer->buffers[cpu] = | |
394 | rb_allocate_cpu_buffer(buffer, cpu); | |
395 | if (!buffer->buffers[cpu]) | |
396 | goto fail_free_buffers; | |
397 | } | |
398 | ||
399 | mutex_init(&buffer->mutex); | |
400 | ||
401 | return buffer; | |
402 | ||
403 | fail_free_buffers: | |
404 | for_each_buffer_cpu(buffer, cpu) { | |
405 | if (buffer->buffers[cpu]) | |
406 | rb_free_cpu_buffer(buffer->buffers[cpu]); | |
407 | } | |
408 | kfree(buffer->buffers); | |
409 | ||
410 | fail_free_buffer: | |
411 | kfree(buffer); | |
412 | return NULL; | |
413 | } | |
414 | ||
415 | /** | |
416 | * ring_buffer_free - free a ring buffer. | |
417 | * @buffer: the buffer to free. | |
418 | */ | |
419 | void | |
420 | ring_buffer_free(struct ring_buffer *buffer) | |
421 | { | |
422 | int cpu; | |
423 | ||
424 | for_each_buffer_cpu(buffer, cpu) | |
425 | rb_free_cpu_buffer(buffer->buffers[cpu]); | |
426 | ||
427 | kfree(buffer); | |
428 | } | |
429 | ||
430 | static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer); | |
431 | ||
432 | static void | |
433 | rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned nr_pages) | |
434 | { | |
435 | struct buffer_page *page; | |
436 | struct list_head *p; | |
437 | unsigned i; | |
438 | ||
439 | atomic_inc(&cpu_buffer->record_disabled); | |
440 | synchronize_sched(); | |
441 | ||
442 | for (i = 0; i < nr_pages; i++) { | |
443 | BUG_ON(list_empty(&cpu_buffer->pages)); | |
444 | p = cpu_buffer->pages.next; | |
445 | page = list_entry(p, struct buffer_page, list); | |
446 | list_del_init(&page->list); | |
ed56829c | 447 | free_buffer_page(page); |
7a8e76a3 SR |
448 | } |
449 | BUG_ON(list_empty(&cpu_buffer->pages)); | |
450 | ||
451 | rb_reset_cpu(cpu_buffer); | |
452 | ||
453 | rb_check_pages(cpu_buffer); | |
454 | ||
455 | atomic_dec(&cpu_buffer->record_disabled); | |
456 | ||
457 | } | |
458 | ||
459 | static void | |
460 | rb_insert_pages(struct ring_buffer_per_cpu *cpu_buffer, | |
461 | struct list_head *pages, unsigned nr_pages) | |
462 | { | |
463 | struct buffer_page *page; | |
464 | struct list_head *p; | |
465 | unsigned i; | |
466 | ||
467 | atomic_inc(&cpu_buffer->record_disabled); | |
468 | synchronize_sched(); | |
469 | ||
470 | for (i = 0; i < nr_pages; i++) { | |
471 | BUG_ON(list_empty(pages)); | |
472 | p = pages->next; | |
473 | page = list_entry(p, struct buffer_page, list); | |
474 | list_del_init(&page->list); | |
475 | list_add_tail(&page->list, &cpu_buffer->pages); | |
476 | } | |
477 | rb_reset_cpu(cpu_buffer); | |
478 | ||
479 | rb_check_pages(cpu_buffer); | |
480 | ||
481 | atomic_dec(&cpu_buffer->record_disabled); | |
482 | } | |
483 | ||
484 | /** | |
485 | * ring_buffer_resize - resize the ring buffer | |
486 | * @buffer: the buffer to resize. | |
487 | * @size: the new size. | |
488 | * | |
489 | * The tracer is responsible for making sure that the buffer is | |
490 | * not being used while changing the size. | |
491 | * Note: We may be able to change the above requirement by using | |
492 | * RCU synchronizations. | |
493 | * | |
494 | * Minimum size is 2 * BUF_PAGE_SIZE. | |
495 | * | |
496 | * Returns -1 on failure. | |
497 | */ | |
498 | int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size) | |
499 | { | |
500 | struct ring_buffer_per_cpu *cpu_buffer; | |
501 | unsigned nr_pages, rm_pages, new_pages; | |
502 | struct buffer_page *page, *tmp; | |
503 | unsigned long buffer_size; | |
504 | unsigned long addr; | |
505 | LIST_HEAD(pages); | |
506 | int i, cpu; | |
507 | ||
508 | size = DIV_ROUND_UP(size, BUF_PAGE_SIZE); | |
509 | size *= BUF_PAGE_SIZE; | |
510 | buffer_size = buffer->pages * BUF_PAGE_SIZE; | |
511 | ||
512 | /* we need a minimum of two pages */ | |
513 | if (size < BUF_PAGE_SIZE * 2) | |
514 | size = BUF_PAGE_SIZE * 2; | |
515 | ||
516 | if (size == buffer_size) | |
517 | return size; | |
518 | ||
519 | mutex_lock(&buffer->mutex); | |
520 | ||
521 | nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE); | |
522 | ||
523 | if (size < buffer_size) { | |
524 | ||
525 | /* easy case, just free pages */ | |
526 | BUG_ON(nr_pages >= buffer->pages); | |
527 | ||
528 | rm_pages = buffer->pages - nr_pages; | |
529 | ||
530 | for_each_buffer_cpu(buffer, cpu) { | |
531 | cpu_buffer = buffer->buffers[cpu]; | |
532 | rb_remove_pages(cpu_buffer, rm_pages); | |
533 | } | |
534 | goto out; | |
535 | } | |
536 | ||
537 | /* | |
538 | * This is a bit more difficult. We only want to add pages | |
539 | * when we can allocate enough for all CPUs. We do this | |
540 | * by allocating all the pages and storing them on a local | |
541 | * link list. If we succeed in our allocation, then we | |
542 | * add these pages to the cpu_buffers. Otherwise we just free | |
543 | * them all and return -ENOMEM; | |
544 | */ | |
545 | BUG_ON(nr_pages <= buffer->pages); | |
546 | new_pages = nr_pages - buffer->pages; | |
547 | ||
548 | for_each_buffer_cpu(buffer, cpu) { | |
549 | for (i = 0; i < new_pages; i++) { | |
e4c2ce82 SR |
550 | page = kzalloc_node(ALIGN(sizeof(*page), |
551 | cache_line_size()), | |
552 | GFP_KERNEL, cpu_to_node(cpu)); | |
553 | if (!page) | |
554 | goto free_pages; | |
555 | list_add(&page->list, &pages); | |
7a8e76a3 SR |
556 | addr = __get_free_page(GFP_KERNEL); |
557 | if (!addr) | |
558 | goto free_pages; | |
e4c2ce82 | 559 | page->page = (void *)addr; |
7a8e76a3 SR |
560 | } |
561 | } | |
562 | ||
563 | for_each_buffer_cpu(buffer, cpu) { | |
564 | cpu_buffer = buffer->buffers[cpu]; | |
565 | rb_insert_pages(cpu_buffer, &pages, new_pages); | |
566 | } | |
567 | ||
568 | BUG_ON(!list_empty(&pages)); | |
569 | ||
570 | out: | |
571 | buffer->pages = nr_pages; | |
572 | mutex_unlock(&buffer->mutex); | |
573 | ||
574 | return size; | |
575 | ||
576 | free_pages: | |
577 | list_for_each_entry_safe(page, tmp, &pages, list) { | |
578 | list_del_init(&page->list); | |
ed56829c | 579 | free_buffer_page(page); |
7a8e76a3 SR |
580 | } |
581 | return -ENOMEM; | |
582 | } | |
583 | ||
7a8e76a3 SR |
584 | static inline int rb_null_event(struct ring_buffer_event *event) |
585 | { | |
586 | return event->type == RINGBUF_TYPE_PADDING; | |
587 | } | |
588 | ||
6f807acd | 589 | static inline void *__rb_page_index(struct buffer_page *page, unsigned index) |
7a8e76a3 | 590 | { |
e4c2ce82 | 591 | return page->page + index; |
7a8e76a3 SR |
592 | } |
593 | ||
594 | static inline struct ring_buffer_event * | |
d769041f | 595 | rb_reader_event(struct ring_buffer_per_cpu *cpu_buffer) |
7a8e76a3 | 596 | { |
6f807acd SR |
597 | return __rb_page_index(cpu_buffer->reader_page, |
598 | cpu_buffer->reader_page->read); | |
599 | } | |
600 | ||
601 | static inline struct ring_buffer_event * | |
602 | rb_head_event(struct ring_buffer_per_cpu *cpu_buffer) | |
603 | { | |
604 | return __rb_page_index(cpu_buffer->head_page, | |
605 | cpu_buffer->head_page->read); | |
7a8e76a3 SR |
606 | } |
607 | ||
608 | static inline struct ring_buffer_event * | |
609 | rb_iter_head_event(struct ring_buffer_iter *iter) | |
610 | { | |
6f807acd | 611 | return __rb_page_index(iter->head_page, iter->head); |
7a8e76a3 SR |
612 | } |
613 | ||
bf41a158 SR |
614 | static inline unsigned rb_page_write(struct buffer_page *bpage) |
615 | { | |
616 | return local_read(&bpage->write); | |
617 | } | |
618 | ||
619 | static inline unsigned rb_page_commit(struct buffer_page *bpage) | |
620 | { | |
621 | return local_read(&bpage->commit); | |
622 | } | |
623 | ||
624 | /* Size is determined by what has been commited */ | |
625 | static inline unsigned rb_page_size(struct buffer_page *bpage) | |
626 | { | |
627 | return rb_page_commit(bpage); | |
628 | } | |
629 | ||
630 | static inline unsigned | |
631 | rb_commit_index(struct ring_buffer_per_cpu *cpu_buffer) | |
632 | { | |
633 | return rb_page_commit(cpu_buffer->commit_page); | |
634 | } | |
635 | ||
636 | static inline unsigned rb_head_size(struct ring_buffer_per_cpu *cpu_buffer) | |
637 | { | |
638 | return rb_page_commit(cpu_buffer->head_page); | |
639 | } | |
640 | ||
7a8e76a3 SR |
641 | /* |
642 | * When the tail hits the head and the buffer is in overwrite mode, | |
643 | * the head jumps to the next page and all content on the previous | |
644 | * page is discarded. But before doing so, we update the overrun | |
645 | * variable of the buffer. | |
646 | */ | |
647 | static void rb_update_overflow(struct ring_buffer_per_cpu *cpu_buffer) | |
648 | { | |
649 | struct ring_buffer_event *event; | |
650 | unsigned long head; | |
651 | ||
652 | for (head = 0; head < rb_head_size(cpu_buffer); | |
653 | head += rb_event_length(event)) { | |
654 | ||
6f807acd | 655 | event = __rb_page_index(cpu_buffer->head_page, head); |
7a8e76a3 SR |
656 | BUG_ON(rb_null_event(event)); |
657 | /* Only count data entries */ | |
658 | if (event->type != RINGBUF_TYPE_DATA) | |
659 | continue; | |
660 | cpu_buffer->overrun++; | |
661 | cpu_buffer->entries--; | |
662 | } | |
663 | } | |
664 | ||
665 | static inline void rb_inc_page(struct ring_buffer_per_cpu *cpu_buffer, | |
666 | struct buffer_page **page) | |
667 | { | |
668 | struct list_head *p = (*page)->list.next; | |
669 | ||
670 | if (p == &cpu_buffer->pages) | |
671 | p = p->next; | |
672 | ||
673 | *page = list_entry(p, struct buffer_page, list); | |
674 | } | |
675 | ||
bf41a158 SR |
676 | static inline unsigned |
677 | rb_event_index(struct ring_buffer_event *event) | |
678 | { | |
679 | unsigned long addr = (unsigned long)event; | |
680 | ||
681 | return (addr & ~PAGE_MASK) - (PAGE_SIZE - BUF_PAGE_SIZE); | |
682 | } | |
683 | ||
684 | static inline int | |
685 | rb_is_commit(struct ring_buffer_per_cpu *cpu_buffer, | |
686 | struct ring_buffer_event *event) | |
687 | { | |
688 | unsigned long addr = (unsigned long)event; | |
689 | unsigned long index; | |
690 | ||
691 | index = rb_event_index(event); | |
692 | addr &= PAGE_MASK; | |
693 | ||
694 | return cpu_buffer->commit_page->page == (void *)addr && | |
695 | rb_commit_index(cpu_buffer) == index; | |
696 | } | |
697 | ||
7a8e76a3 | 698 | static inline void |
bf41a158 SR |
699 | rb_set_commit_event(struct ring_buffer_per_cpu *cpu_buffer, |
700 | struct ring_buffer_event *event) | |
7a8e76a3 | 701 | { |
bf41a158 SR |
702 | unsigned long addr = (unsigned long)event; |
703 | unsigned long index; | |
704 | ||
705 | index = rb_event_index(event); | |
706 | addr &= PAGE_MASK; | |
707 | ||
708 | while (cpu_buffer->commit_page->page != (void *)addr) { | |
709 | RB_WARN_ON(cpu_buffer, | |
710 | cpu_buffer->commit_page == cpu_buffer->tail_page); | |
711 | cpu_buffer->commit_page->commit = | |
712 | cpu_buffer->commit_page->write; | |
713 | rb_inc_page(cpu_buffer, &cpu_buffer->commit_page); | |
714 | cpu_buffer->write_stamp = cpu_buffer->commit_page->time_stamp; | |
715 | } | |
716 | ||
717 | /* Now set the commit to the event's index */ | |
718 | local_set(&cpu_buffer->commit_page->commit, index); | |
7a8e76a3 SR |
719 | } |
720 | ||
bf41a158 SR |
721 | static inline void |
722 | rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer) | |
7a8e76a3 | 723 | { |
bf41a158 SR |
724 | /* |
725 | * We only race with interrupts and NMIs on this CPU. | |
726 | * If we own the commit event, then we can commit | |
727 | * all others that interrupted us, since the interruptions | |
728 | * are in stack format (they finish before they come | |
729 | * back to us). This allows us to do a simple loop to | |
730 | * assign the commit to the tail. | |
731 | */ | |
732 | while (cpu_buffer->commit_page != cpu_buffer->tail_page) { | |
733 | cpu_buffer->commit_page->commit = | |
734 | cpu_buffer->commit_page->write; | |
735 | rb_inc_page(cpu_buffer, &cpu_buffer->commit_page); | |
736 | cpu_buffer->write_stamp = cpu_buffer->commit_page->time_stamp; | |
737 | /* add barrier to keep gcc from optimizing too much */ | |
738 | barrier(); | |
739 | } | |
740 | while (rb_commit_index(cpu_buffer) != | |
741 | rb_page_write(cpu_buffer->commit_page)) { | |
742 | cpu_buffer->commit_page->commit = | |
743 | cpu_buffer->commit_page->write; | |
744 | barrier(); | |
745 | } | |
7a8e76a3 SR |
746 | } |
747 | ||
d769041f | 748 | static void rb_reset_reader_page(struct ring_buffer_per_cpu *cpu_buffer) |
7a8e76a3 | 749 | { |
d769041f | 750 | cpu_buffer->read_stamp = cpu_buffer->reader_page->time_stamp; |
6f807acd | 751 | cpu_buffer->reader_page->read = 0; |
d769041f SR |
752 | } |
753 | ||
754 | static inline void rb_inc_iter(struct ring_buffer_iter *iter) | |
755 | { | |
756 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
757 | ||
758 | /* | |
759 | * The iterator could be on the reader page (it starts there). | |
760 | * But the head could have moved, since the reader was | |
761 | * found. Check for this case and assign the iterator | |
762 | * to the head page instead of next. | |
763 | */ | |
764 | if (iter->head_page == cpu_buffer->reader_page) | |
765 | iter->head_page = cpu_buffer->head_page; | |
766 | else | |
767 | rb_inc_page(cpu_buffer, &iter->head_page); | |
768 | ||
7a8e76a3 SR |
769 | iter->read_stamp = iter->head_page->time_stamp; |
770 | iter->head = 0; | |
771 | } | |
772 | ||
773 | /** | |
774 | * ring_buffer_update_event - update event type and data | |
775 | * @event: the even to update | |
776 | * @type: the type of event | |
777 | * @length: the size of the event field in the ring buffer | |
778 | * | |
779 | * Update the type and data fields of the event. The length | |
780 | * is the actual size that is written to the ring buffer, | |
781 | * and with this, we can determine what to place into the | |
782 | * data field. | |
783 | */ | |
784 | static inline void | |
785 | rb_update_event(struct ring_buffer_event *event, | |
786 | unsigned type, unsigned length) | |
787 | { | |
788 | event->type = type; | |
789 | ||
790 | switch (type) { | |
791 | ||
792 | case RINGBUF_TYPE_PADDING: | |
793 | break; | |
794 | ||
795 | case RINGBUF_TYPE_TIME_EXTEND: | |
796 | event->len = | |
797 | (RB_LEN_TIME_EXTEND + (RB_ALIGNMENT-1)) | |
798 | >> RB_ALIGNMENT_SHIFT; | |
799 | break; | |
800 | ||
801 | case RINGBUF_TYPE_TIME_STAMP: | |
802 | event->len = | |
803 | (RB_LEN_TIME_STAMP + (RB_ALIGNMENT-1)) | |
804 | >> RB_ALIGNMENT_SHIFT; | |
805 | break; | |
806 | ||
807 | case RINGBUF_TYPE_DATA: | |
808 | length -= RB_EVNT_HDR_SIZE; | |
809 | if (length > RB_MAX_SMALL_DATA) { | |
810 | event->len = 0; | |
811 | event->array[0] = length; | |
812 | } else | |
813 | event->len = | |
814 | (length + (RB_ALIGNMENT-1)) | |
815 | >> RB_ALIGNMENT_SHIFT; | |
816 | break; | |
817 | default: | |
818 | BUG(); | |
819 | } | |
820 | } | |
821 | ||
822 | static inline unsigned rb_calculate_event_length(unsigned length) | |
823 | { | |
824 | struct ring_buffer_event event; /* Used only for sizeof array */ | |
825 | ||
826 | /* zero length can cause confusions */ | |
827 | if (!length) | |
828 | length = 1; | |
829 | ||
830 | if (length > RB_MAX_SMALL_DATA) | |
831 | length += sizeof(event.array[0]); | |
832 | ||
833 | length += RB_EVNT_HDR_SIZE; | |
834 | length = ALIGN(length, RB_ALIGNMENT); | |
835 | ||
836 | return length; | |
837 | } | |
838 | ||
839 | static struct ring_buffer_event * | |
840 | __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer, | |
841 | unsigned type, unsigned long length, u64 *ts) | |
842 | { | |
d769041f | 843 | struct buffer_page *tail_page, *head_page, *reader_page; |
bf41a158 | 844 | unsigned long tail, write; |
7a8e76a3 SR |
845 | struct ring_buffer *buffer = cpu_buffer->buffer; |
846 | struct ring_buffer_event *event; | |
bf41a158 | 847 | unsigned long flags; |
7a8e76a3 SR |
848 | |
849 | tail_page = cpu_buffer->tail_page; | |
bf41a158 SR |
850 | write = local_add_return(length, &tail_page->write); |
851 | tail = write - length; | |
7a8e76a3 | 852 | |
bf41a158 SR |
853 | /* See if we shot pass the end of this buffer page */ |
854 | if (write > BUF_PAGE_SIZE) { | |
7a8e76a3 SR |
855 | struct buffer_page *next_page = tail_page; |
856 | ||
bf41a158 SR |
857 | spin_lock_irqsave(&cpu_buffer->lock, flags); |
858 | ||
7a8e76a3 SR |
859 | rb_inc_page(cpu_buffer, &next_page); |
860 | ||
d769041f SR |
861 | head_page = cpu_buffer->head_page; |
862 | reader_page = cpu_buffer->reader_page; | |
863 | ||
864 | /* we grabbed the lock before incrementing */ | |
bf41a158 SR |
865 | RB_WARN_ON(cpu_buffer, next_page == reader_page); |
866 | ||
867 | /* | |
868 | * If for some reason, we had an interrupt storm that made | |
869 | * it all the way around the buffer, bail, and warn | |
870 | * about it. | |
871 | */ | |
872 | if (unlikely(next_page == cpu_buffer->commit_page)) { | |
873 | WARN_ON_ONCE(1); | |
874 | goto out_unlock; | |
875 | } | |
d769041f | 876 | |
7a8e76a3 | 877 | if (next_page == head_page) { |
d769041f | 878 | if (!(buffer->flags & RB_FL_OVERWRITE)) { |
bf41a158 SR |
879 | /* reset write */ |
880 | if (tail <= BUF_PAGE_SIZE) | |
881 | local_set(&tail_page->write, tail); | |
882 | goto out_unlock; | |
d769041f | 883 | } |
7a8e76a3 | 884 | |
bf41a158 SR |
885 | /* tail_page has not moved yet? */ |
886 | if (tail_page == cpu_buffer->tail_page) { | |
887 | /* count overflows */ | |
888 | rb_update_overflow(cpu_buffer); | |
889 | ||
890 | rb_inc_page(cpu_buffer, &head_page); | |
891 | cpu_buffer->head_page = head_page; | |
892 | cpu_buffer->head_page->read = 0; | |
893 | } | |
894 | } | |
7a8e76a3 | 895 | |
bf41a158 SR |
896 | /* |
897 | * If the tail page is still the same as what we think | |
898 | * it is, then it is up to us to update the tail | |
899 | * pointer. | |
900 | */ | |
901 | if (tail_page == cpu_buffer->tail_page) { | |
902 | local_set(&next_page->write, 0); | |
903 | local_set(&next_page->commit, 0); | |
904 | cpu_buffer->tail_page = next_page; | |
905 | ||
906 | /* reread the time stamp */ | |
907 | *ts = ring_buffer_time_stamp(cpu_buffer->cpu); | |
908 | cpu_buffer->tail_page->time_stamp = *ts; | |
7a8e76a3 SR |
909 | } |
910 | ||
bf41a158 SR |
911 | /* |
912 | * The actual tail page has moved forward. | |
913 | */ | |
914 | if (tail < BUF_PAGE_SIZE) { | |
915 | /* Mark the rest of the page with padding */ | |
6f807acd | 916 | event = __rb_page_index(tail_page, tail); |
7a8e76a3 SR |
917 | event->type = RINGBUF_TYPE_PADDING; |
918 | } | |
919 | ||
bf41a158 SR |
920 | if (tail <= BUF_PAGE_SIZE) |
921 | /* Set the write back to the previous setting */ | |
922 | local_set(&tail_page->write, tail); | |
923 | ||
924 | /* | |
925 | * If this was a commit entry that failed, | |
926 | * increment that too | |
927 | */ | |
928 | if (tail_page == cpu_buffer->commit_page && | |
929 | tail == rb_commit_index(cpu_buffer)) { | |
930 | rb_set_commit_to_write(cpu_buffer); | |
931 | } | |
932 | ||
933 | spin_unlock_irqrestore(&cpu_buffer->lock, flags); | |
934 | ||
935 | /* fail and let the caller try again */ | |
936 | return ERR_PTR(-EAGAIN); | |
7a8e76a3 SR |
937 | } |
938 | ||
bf41a158 SR |
939 | /* We reserved something on the buffer */ |
940 | ||
941 | BUG_ON(write > BUF_PAGE_SIZE); | |
7a8e76a3 | 942 | |
6f807acd | 943 | event = __rb_page_index(tail_page, tail); |
7a8e76a3 SR |
944 | rb_update_event(event, type, length); |
945 | ||
bf41a158 SR |
946 | /* |
947 | * If this is a commit and the tail is zero, then update | |
948 | * this page's time stamp. | |
949 | */ | |
950 | if (!tail && rb_is_commit(cpu_buffer, event)) | |
951 | cpu_buffer->commit_page->time_stamp = *ts; | |
952 | ||
7a8e76a3 | 953 | return event; |
bf41a158 SR |
954 | |
955 | out_unlock: | |
956 | spin_unlock_irqrestore(&cpu_buffer->lock, flags); | |
957 | return NULL; | |
7a8e76a3 SR |
958 | } |
959 | ||
960 | static int | |
961 | rb_add_time_stamp(struct ring_buffer_per_cpu *cpu_buffer, | |
962 | u64 *ts, u64 *delta) | |
963 | { | |
964 | struct ring_buffer_event *event; | |
965 | static int once; | |
bf41a158 | 966 | int ret; |
7a8e76a3 SR |
967 | |
968 | if (unlikely(*delta > (1ULL << 59) && !once++)) { | |
969 | printk(KERN_WARNING "Delta way too big! %llu" | |
970 | " ts=%llu write stamp = %llu\n", | |
e2862c94 SR |
971 | (unsigned long long)*delta, |
972 | (unsigned long long)*ts, | |
973 | (unsigned long long)cpu_buffer->write_stamp); | |
7a8e76a3 SR |
974 | WARN_ON(1); |
975 | } | |
976 | ||
977 | /* | |
978 | * The delta is too big, we to add a | |
979 | * new timestamp. | |
980 | */ | |
981 | event = __rb_reserve_next(cpu_buffer, | |
982 | RINGBUF_TYPE_TIME_EXTEND, | |
983 | RB_LEN_TIME_EXTEND, | |
984 | ts); | |
985 | if (!event) | |
bf41a158 | 986 | return -EBUSY; |
7a8e76a3 | 987 | |
bf41a158 SR |
988 | if (PTR_ERR(event) == -EAGAIN) |
989 | return -EAGAIN; | |
990 | ||
991 | /* Only a commited time event can update the write stamp */ | |
992 | if (rb_is_commit(cpu_buffer, event)) { | |
993 | /* | |
994 | * If this is the first on the page, then we need to | |
995 | * update the page itself, and just put in a zero. | |
996 | */ | |
997 | if (rb_event_index(event)) { | |
998 | event->time_delta = *delta & TS_MASK; | |
999 | event->array[0] = *delta >> TS_SHIFT; | |
1000 | } else { | |
1001 | cpu_buffer->commit_page->time_stamp = *ts; | |
1002 | event->time_delta = 0; | |
1003 | event->array[0] = 0; | |
1004 | } | |
7a8e76a3 | 1005 | cpu_buffer->write_stamp = *ts; |
bf41a158 SR |
1006 | /* let the caller know this was the commit */ |
1007 | ret = 1; | |
1008 | } else { | |
1009 | /* Darn, this is just wasted space */ | |
1010 | event->time_delta = 0; | |
1011 | event->array[0] = 0; | |
1012 | ret = 0; | |
7a8e76a3 SR |
1013 | } |
1014 | ||
bf41a158 SR |
1015 | *delta = 0; |
1016 | ||
1017 | return ret; | |
7a8e76a3 SR |
1018 | } |
1019 | ||
1020 | static struct ring_buffer_event * | |
1021 | rb_reserve_next_event(struct ring_buffer_per_cpu *cpu_buffer, | |
1022 | unsigned type, unsigned long length) | |
1023 | { | |
1024 | struct ring_buffer_event *event; | |
1025 | u64 ts, delta; | |
bf41a158 | 1026 | int commit = 0; |
7a8e76a3 | 1027 | |
bf41a158 | 1028 | again: |
7a8e76a3 SR |
1029 | ts = ring_buffer_time_stamp(cpu_buffer->cpu); |
1030 | ||
bf41a158 SR |
1031 | /* |
1032 | * Only the first commit can update the timestamp. | |
1033 | * Yes there is a race here. If an interrupt comes in | |
1034 | * just after the conditional and it traces too, then it | |
1035 | * will also check the deltas. More than one timestamp may | |
1036 | * also be made. But only the entry that did the actual | |
1037 | * commit will be something other than zero. | |
1038 | */ | |
1039 | if (cpu_buffer->tail_page == cpu_buffer->commit_page && | |
1040 | rb_page_write(cpu_buffer->tail_page) == | |
1041 | rb_commit_index(cpu_buffer)) { | |
1042 | ||
7a8e76a3 SR |
1043 | delta = ts - cpu_buffer->write_stamp; |
1044 | ||
bf41a158 SR |
1045 | /* make sure this delta is calculated here */ |
1046 | barrier(); | |
1047 | ||
1048 | /* Did the write stamp get updated already? */ | |
1049 | if (unlikely(ts < cpu_buffer->write_stamp)) | |
1050 | goto again; | |
1051 | ||
7a8e76a3 | 1052 | if (test_time_stamp(delta)) { |
7a8e76a3 | 1053 | |
bf41a158 SR |
1054 | commit = rb_add_time_stamp(cpu_buffer, &ts, &delta); |
1055 | ||
1056 | if (commit == -EBUSY) | |
7a8e76a3 | 1057 | return NULL; |
bf41a158 SR |
1058 | |
1059 | if (commit == -EAGAIN) | |
1060 | goto again; | |
1061 | ||
1062 | RB_WARN_ON(cpu_buffer, commit < 0); | |
7a8e76a3 | 1063 | } |
bf41a158 SR |
1064 | } else |
1065 | /* Non commits have zero deltas */ | |
7a8e76a3 | 1066 | delta = 0; |
7a8e76a3 SR |
1067 | |
1068 | event = __rb_reserve_next(cpu_buffer, type, length, &ts); | |
bf41a158 SR |
1069 | if (PTR_ERR(event) == -EAGAIN) |
1070 | goto again; | |
1071 | ||
1072 | if (!event) { | |
1073 | if (unlikely(commit)) | |
1074 | /* | |
1075 | * Ouch! We needed a timestamp and it was commited. But | |
1076 | * we didn't get our event reserved. | |
1077 | */ | |
1078 | rb_set_commit_to_write(cpu_buffer); | |
7a8e76a3 | 1079 | return NULL; |
bf41a158 | 1080 | } |
7a8e76a3 | 1081 | |
bf41a158 SR |
1082 | /* |
1083 | * If the timestamp was commited, make the commit our entry | |
1084 | * now so that we will update it when needed. | |
1085 | */ | |
1086 | if (commit) | |
1087 | rb_set_commit_event(cpu_buffer, event); | |
1088 | else if (!rb_is_commit(cpu_buffer, event)) | |
7a8e76a3 SR |
1089 | delta = 0; |
1090 | ||
1091 | event->time_delta = delta; | |
1092 | ||
1093 | return event; | |
1094 | } | |
1095 | ||
bf41a158 SR |
1096 | static DEFINE_PER_CPU(int, rb_need_resched); |
1097 | ||
7a8e76a3 SR |
1098 | /** |
1099 | * ring_buffer_lock_reserve - reserve a part of the buffer | |
1100 | * @buffer: the ring buffer to reserve from | |
1101 | * @length: the length of the data to reserve (excluding event header) | |
1102 | * @flags: a pointer to save the interrupt flags | |
1103 | * | |
1104 | * Returns a reseverd event on the ring buffer to copy directly to. | |
1105 | * The user of this interface will need to get the body to write into | |
1106 | * and can use the ring_buffer_event_data() interface. | |
1107 | * | |
1108 | * The length is the length of the data needed, not the event length | |
1109 | * which also includes the event header. | |
1110 | * | |
1111 | * Must be paired with ring_buffer_unlock_commit, unless NULL is returned. | |
1112 | * If NULL is returned, then nothing has been allocated or locked. | |
1113 | */ | |
1114 | struct ring_buffer_event * | |
1115 | ring_buffer_lock_reserve(struct ring_buffer *buffer, | |
1116 | unsigned long length, | |
1117 | unsigned long *flags) | |
1118 | { | |
1119 | struct ring_buffer_per_cpu *cpu_buffer; | |
1120 | struct ring_buffer_event *event; | |
bf41a158 | 1121 | int cpu, resched; |
7a8e76a3 SR |
1122 | |
1123 | if (atomic_read(&buffer->record_disabled)) | |
1124 | return NULL; | |
1125 | ||
bf41a158 | 1126 | /* If we are tracing schedule, we don't want to recurse */ |
182e9f5f | 1127 | resched = ftrace_preempt_disable(); |
bf41a158 | 1128 | |
7a8e76a3 SR |
1129 | cpu = raw_smp_processor_id(); |
1130 | ||
1131 | if (!cpu_isset(cpu, buffer->cpumask)) | |
d769041f | 1132 | goto out; |
7a8e76a3 SR |
1133 | |
1134 | cpu_buffer = buffer->buffers[cpu]; | |
7a8e76a3 SR |
1135 | |
1136 | if (atomic_read(&cpu_buffer->record_disabled)) | |
d769041f | 1137 | goto out; |
7a8e76a3 SR |
1138 | |
1139 | length = rb_calculate_event_length(length); | |
1140 | if (length > BUF_PAGE_SIZE) | |
bf41a158 | 1141 | goto out; |
7a8e76a3 SR |
1142 | |
1143 | event = rb_reserve_next_event(cpu_buffer, RINGBUF_TYPE_DATA, length); | |
1144 | if (!event) | |
d769041f | 1145 | goto out; |
7a8e76a3 | 1146 | |
bf41a158 SR |
1147 | /* |
1148 | * Need to store resched state on this cpu. | |
1149 | * Only the first needs to. | |
1150 | */ | |
1151 | ||
1152 | if (preempt_count() == 1) | |
1153 | per_cpu(rb_need_resched, cpu) = resched; | |
1154 | ||
7a8e76a3 SR |
1155 | return event; |
1156 | ||
d769041f | 1157 | out: |
182e9f5f | 1158 | ftrace_preempt_enable(resched); |
7a8e76a3 SR |
1159 | return NULL; |
1160 | } | |
1161 | ||
1162 | static void rb_commit(struct ring_buffer_per_cpu *cpu_buffer, | |
1163 | struct ring_buffer_event *event) | |
1164 | { | |
7a8e76a3 | 1165 | cpu_buffer->entries++; |
bf41a158 SR |
1166 | |
1167 | /* Only process further if we own the commit */ | |
1168 | if (!rb_is_commit(cpu_buffer, event)) | |
1169 | return; | |
1170 | ||
1171 | cpu_buffer->write_stamp += event->time_delta; | |
1172 | ||
1173 | rb_set_commit_to_write(cpu_buffer); | |
7a8e76a3 SR |
1174 | } |
1175 | ||
1176 | /** | |
1177 | * ring_buffer_unlock_commit - commit a reserved | |
1178 | * @buffer: The buffer to commit to | |
1179 | * @event: The event pointer to commit. | |
1180 | * @flags: the interrupt flags received from ring_buffer_lock_reserve. | |
1181 | * | |
1182 | * This commits the data to the ring buffer, and releases any locks held. | |
1183 | * | |
1184 | * Must be paired with ring_buffer_lock_reserve. | |
1185 | */ | |
1186 | int ring_buffer_unlock_commit(struct ring_buffer *buffer, | |
1187 | struct ring_buffer_event *event, | |
1188 | unsigned long flags) | |
1189 | { | |
1190 | struct ring_buffer_per_cpu *cpu_buffer; | |
1191 | int cpu = raw_smp_processor_id(); | |
1192 | ||
1193 | cpu_buffer = buffer->buffers[cpu]; | |
1194 | ||
7a8e76a3 SR |
1195 | rb_commit(cpu_buffer, event); |
1196 | ||
bf41a158 SR |
1197 | /* |
1198 | * Only the last preempt count needs to restore preemption. | |
1199 | */ | |
182e9f5f SR |
1200 | if (preempt_count() == 1) |
1201 | ftrace_preempt_enable(per_cpu(rb_need_resched, cpu)); | |
1202 | else | |
bf41a158 | 1203 | preempt_enable_no_resched_notrace(); |
7a8e76a3 SR |
1204 | |
1205 | return 0; | |
1206 | } | |
1207 | ||
1208 | /** | |
1209 | * ring_buffer_write - write data to the buffer without reserving | |
1210 | * @buffer: The ring buffer to write to. | |
1211 | * @length: The length of the data being written (excluding the event header) | |
1212 | * @data: The data to write to the buffer. | |
1213 | * | |
1214 | * This is like ring_buffer_lock_reserve and ring_buffer_unlock_commit as | |
1215 | * one function. If you already have the data to write to the buffer, it | |
1216 | * may be easier to simply call this function. | |
1217 | * | |
1218 | * Note, like ring_buffer_lock_reserve, the length is the length of the data | |
1219 | * and not the length of the event which would hold the header. | |
1220 | */ | |
1221 | int ring_buffer_write(struct ring_buffer *buffer, | |
1222 | unsigned long length, | |
1223 | void *data) | |
1224 | { | |
1225 | struct ring_buffer_per_cpu *cpu_buffer; | |
1226 | struct ring_buffer_event *event; | |
bf41a158 | 1227 | unsigned long event_length; |
7a8e76a3 SR |
1228 | void *body; |
1229 | int ret = -EBUSY; | |
bf41a158 | 1230 | int cpu, resched; |
7a8e76a3 SR |
1231 | |
1232 | if (atomic_read(&buffer->record_disabled)) | |
1233 | return -EBUSY; | |
1234 | ||
182e9f5f | 1235 | resched = ftrace_preempt_disable(); |
bf41a158 | 1236 | |
7a8e76a3 SR |
1237 | cpu = raw_smp_processor_id(); |
1238 | ||
1239 | if (!cpu_isset(cpu, buffer->cpumask)) | |
d769041f | 1240 | goto out; |
7a8e76a3 SR |
1241 | |
1242 | cpu_buffer = buffer->buffers[cpu]; | |
7a8e76a3 SR |
1243 | |
1244 | if (atomic_read(&cpu_buffer->record_disabled)) | |
1245 | goto out; | |
1246 | ||
1247 | event_length = rb_calculate_event_length(length); | |
1248 | event = rb_reserve_next_event(cpu_buffer, | |
1249 | RINGBUF_TYPE_DATA, event_length); | |
1250 | if (!event) | |
1251 | goto out; | |
1252 | ||
1253 | body = rb_event_data(event); | |
1254 | ||
1255 | memcpy(body, data, length); | |
1256 | ||
1257 | rb_commit(cpu_buffer, event); | |
1258 | ||
1259 | ret = 0; | |
1260 | out: | |
182e9f5f | 1261 | ftrace_preempt_enable(resched); |
7a8e76a3 SR |
1262 | |
1263 | return ret; | |
1264 | } | |
1265 | ||
bf41a158 SR |
1266 | static inline int rb_per_cpu_empty(struct ring_buffer_per_cpu *cpu_buffer) |
1267 | { | |
1268 | struct buffer_page *reader = cpu_buffer->reader_page; | |
1269 | struct buffer_page *head = cpu_buffer->head_page; | |
1270 | struct buffer_page *commit = cpu_buffer->commit_page; | |
1271 | ||
1272 | return reader->read == rb_page_commit(reader) && | |
1273 | (commit == reader || | |
1274 | (commit == head && | |
1275 | head->read == rb_page_commit(commit))); | |
1276 | } | |
1277 | ||
7a8e76a3 SR |
1278 | /** |
1279 | * ring_buffer_record_disable - stop all writes into the buffer | |
1280 | * @buffer: The ring buffer to stop writes to. | |
1281 | * | |
1282 | * This prevents all writes to the buffer. Any attempt to write | |
1283 | * to the buffer after this will fail and return NULL. | |
1284 | * | |
1285 | * The caller should call synchronize_sched() after this. | |
1286 | */ | |
1287 | void ring_buffer_record_disable(struct ring_buffer *buffer) | |
1288 | { | |
1289 | atomic_inc(&buffer->record_disabled); | |
1290 | } | |
1291 | ||
1292 | /** | |
1293 | * ring_buffer_record_enable - enable writes to the buffer | |
1294 | * @buffer: The ring buffer to enable writes | |
1295 | * | |
1296 | * Note, multiple disables will need the same number of enables | |
1297 | * to truely enable the writing (much like preempt_disable). | |
1298 | */ | |
1299 | void ring_buffer_record_enable(struct ring_buffer *buffer) | |
1300 | { | |
1301 | atomic_dec(&buffer->record_disabled); | |
1302 | } | |
1303 | ||
1304 | /** | |
1305 | * ring_buffer_record_disable_cpu - stop all writes into the cpu_buffer | |
1306 | * @buffer: The ring buffer to stop writes to. | |
1307 | * @cpu: The CPU buffer to stop | |
1308 | * | |
1309 | * This prevents all writes to the buffer. Any attempt to write | |
1310 | * to the buffer after this will fail and return NULL. | |
1311 | * | |
1312 | * The caller should call synchronize_sched() after this. | |
1313 | */ | |
1314 | void ring_buffer_record_disable_cpu(struct ring_buffer *buffer, int cpu) | |
1315 | { | |
1316 | struct ring_buffer_per_cpu *cpu_buffer; | |
1317 | ||
1318 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1319 | return; | |
1320 | ||
1321 | cpu_buffer = buffer->buffers[cpu]; | |
1322 | atomic_inc(&cpu_buffer->record_disabled); | |
1323 | } | |
1324 | ||
1325 | /** | |
1326 | * ring_buffer_record_enable_cpu - enable writes to the buffer | |
1327 | * @buffer: The ring buffer to enable writes | |
1328 | * @cpu: The CPU to enable. | |
1329 | * | |
1330 | * Note, multiple disables will need the same number of enables | |
1331 | * to truely enable the writing (much like preempt_disable). | |
1332 | */ | |
1333 | void ring_buffer_record_enable_cpu(struct ring_buffer *buffer, int cpu) | |
1334 | { | |
1335 | struct ring_buffer_per_cpu *cpu_buffer; | |
1336 | ||
1337 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1338 | return; | |
1339 | ||
1340 | cpu_buffer = buffer->buffers[cpu]; | |
1341 | atomic_dec(&cpu_buffer->record_disabled); | |
1342 | } | |
1343 | ||
1344 | /** | |
1345 | * ring_buffer_entries_cpu - get the number of entries in a cpu buffer | |
1346 | * @buffer: The ring buffer | |
1347 | * @cpu: The per CPU buffer to get the entries from. | |
1348 | */ | |
1349 | unsigned long ring_buffer_entries_cpu(struct ring_buffer *buffer, int cpu) | |
1350 | { | |
1351 | struct ring_buffer_per_cpu *cpu_buffer; | |
1352 | ||
1353 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1354 | return 0; | |
1355 | ||
1356 | cpu_buffer = buffer->buffers[cpu]; | |
1357 | return cpu_buffer->entries; | |
1358 | } | |
1359 | ||
1360 | /** | |
1361 | * ring_buffer_overrun_cpu - get the number of overruns in a cpu_buffer | |
1362 | * @buffer: The ring buffer | |
1363 | * @cpu: The per CPU buffer to get the number of overruns from | |
1364 | */ | |
1365 | unsigned long ring_buffer_overrun_cpu(struct ring_buffer *buffer, int cpu) | |
1366 | { | |
1367 | struct ring_buffer_per_cpu *cpu_buffer; | |
1368 | ||
1369 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1370 | return 0; | |
1371 | ||
1372 | cpu_buffer = buffer->buffers[cpu]; | |
1373 | return cpu_buffer->overrun; | |
1374 | } | |
1375 | ||
1376 | /** | |
1377 | * ring_buffer_entries - get the number of entries in a buffer | |
1378 | * @buffer: The ring buffer | |
1379 | * | |
1380 | * Returns the total number of entries in the ring buffer | |
1381 | * (all CPU entries) | |
1382 | */ | |
1383 | unsigned long ring_buffer_entries(struct ring_buffer *buffer) | |
1384 | { | |
1385 | struct ring_buffer_per_cpu *cpu_buffer; | |
1386 | unsigned long entries = 0; | |
1387 | int cpu; | |
1388 | ||
1389 | /* if you care about this being correct, lock the buffer */ | |
1390 | for_each_buffer_cpu(buffer, cpu) { | |
1391 | cpu_buffer = buffer->buffers[cpu]; | |
1392 | entries += cpu_buffer->entries; | |
1393 | } | |
1394 | ||
1395 | return entries; | |
1396 | } | |
1397 | ||
1398 | /** | |
1399 | * ring_buffer_overrun_cpu - get the number of overruns in buffer | |
1400 | * @buffer: The ring buffer | |
1401 | * | |
1402 | * Returns the total number of overruns in the ring buffer | |
1403 | * (all CPU entries) | |
1404 | */ | |
1405 | unsigned long ring_buffer_overruns(struct ring_buffer *buffer) | |
1406 | { | |
1407 | struct ring_buffer_per_cpu *cpu_buffer; | |
1408 | unsigned long overruns = 0; | |
1409 | int cpu; | |
1410 | ||
1411 | /* if you care about this being correct, lock the buffer */ | |
1412 | for_each_buffer_cpu(buffer, cpu) { | |
1413 | cpu_buffer = buffer->buffers[cpu]; | |
1414 | overruns += cpu_buffer->overrun; | |
1415 | } | |
1416 | ||
1417 | return overruns; | |
1418 | } | |
1419 | ||
1420 | /** | |
1421 | * ring_buffer_iter_reset - reset an iterator | |
1422 | * @iter: The iterator to reset | |
1423 | * | |
1424 | * Resets the iterator, so that it will start from the beginning | |
1425 | * again. | |
1426 | */ | |
1427 | void ring_buffer_iter_reset(struct ring_buffer_iter *iter) | |
1428 | { | |
1429 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
1430 | ||
d769041f SR |
1431 | /* Iterator usage is expected to have record disabled */ |
1432 | if (list_empty(&cpu_buffer->reader_page->list)) { | |
1433 | iter->head_page = cpu_buffer->head_page; | |
6f807acd | 1434 | iter->head = cpu_buffer->head_page->read; |
d769041f SR |
1435 | } else { |
1436 | iter->head_page = cpu_buffer->reader_page; | |
6f807acd | 1437 | iter->head = cpu_buffer->reader_page->read; |
d769041f SR |
1438 | } |
1439 | if (iter->head) | |
1440 | iter->read_stamp = cpu_buffer->read_stamp; | |
1441 | else | |
1442 | iter->read_stamp = iter->head_page->time_stamp; | |
7a8e76a3 SR |
1443 | } |
1444 | ||
1445 | /** | |
1446 | * ring_buffer_iter_empty - check if an iterator has no more to read | |
1447 | * @iter: The iterator to check | |
1448 | */ | |
1449 | int ring_buffer_iter_empty(struct ring_buffer_iter *iter) | |
1450 | { | |
1451 | struct ring_buffer_per_cpu *cpu_buffer; | |
1452 | ||
1453 | cpu_buffer = iter->cpu_buffer; | |
1454 | ||
bf41a158 SR |
1455 | return iter->head_page == cpu_buffer->commit_page && |
1456 | iter->head == rb_commit_index(cpu_buffer); | |
7a8e76a3 SR |
1457 | } |
1458 | ||
1459 | static void | |
1460 | rb_update_read_stamp(struct ring_buffer_per_cpu *cpu_buffer, | |
1461 | struct ring_buffer_event *event) | |
1462 | { | |
1463 | u64 delta; | |
1464 | ||
1465 | switch (event->type) { | |
1466 | case RINGBUF_TYPE_PADDING: | |
1467 | return; | |
1468 | ||
1469 | case RINGBUF_TYPE_TIME_EXTEND: | |
1470 | delta = event->array[0]; | |
1471 | delta <<= TS_SHIFT; | |
1472 | delta += event->time_delta; | |
1473 | cpu_buffer->read_stamp += delta; | |
1474 | return; | |
1475 | ||
1476 | case RINGBUF_TYPE_TIME_STAMP: | |
1477 | /* FIXME: not implemented */ | |
1478 | return; | |
1479 | ||
1480 | case RINGBUF_TYPE_DATA: | |
1481 | cpu_buffer->read_stamp += event->time_delta; | |
1482 | return; | |
1483 | ||
1484 | default: | |
1485 | BUG(); | |
1486 | } | |
1487 | return; | |
1488 | } | |
1489 | ||
1490 | static void | |
1491 | rb_update_iter_read_stamp(struct ring_buffer_iter *iter, | |
1492 | struct ring_buffer_event *event) | |
1493 | { | |
1494 | u64 delta; | |
1495 | ||
1496 | switch (event->type) { | |
1497 | case RINGBUF_TYPE_PADDING: | |
1498 | return; | |
1499 | ||
1500 | case RINGBUF_TYPE_TIME_EXTEND: | |
1501 | delta = event->array[0]; | |
1502 | delta <<= TS_SHIFT; | |
1503 | delta += event->time_delta; | |
1504 | iter->read_stamp += delta; | |
1505 | return; | |
1506 | ||
1507 | case RINGBUF_TYPE_TIME_STAMP: | |
1508 | /* FIXME: not implemented */ | |
1509 | return; | |
1510 | ||
1511 | case RINGBUF_TYPE_DATA: | |
1512 | iter->read_stamp += event->time_delta; | |
1513 | return; | |
1514 | ||
1515 | default: | |
1516 | BUG(); | |
1517 | } | |
1518 | return; | |
1519 | } | |
1520 | ||
d769041f SR |
1521 | static struct buffer_page * |
1522 | rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer) | |
7a8e76a3 | 1523 | { |
d769041f SR |
1524 | struct buffer_page *reader = NULL; |
1525 | unsigned long flags; | |
1526 | ||
1527 | spin_lock_irqsave(&cpu_buffer->lock, flags); | |
1528 | ||
1529 | again: | |
1530 | reader = cpu_buffer->reader_page; | |
1531 | ||
1532 | /* If there's more to read, return this page */ | |
bf41a158 | 1533 | if (cpu_buffer->reader_page->read < rb_page_size(reader)) |
d769041f SR |
1534 | goto out; |
1535 | ||
1536 | /* Never should we have an index greater than the size */ | |
bf41a158 SR |
1537 | RB_WARN_ON(cpu_buffer, |
1538 | cpu_buffer->reader_page->read > rb_page_size(reader)); | |
d769041f SR |
1539 | |
1540 | /* check if we caught up to the tail */ | |
1541 | reader = NULL; | |
bf41a158 | 1542 | if (cpu_buffer->commit_page == cpu_buffer->reader_page) |
d769041f | 1543 | goto out; |
7a8e76a3 SR |
1544 | |
1545 | /* | |
d769041f SR |
1546 | * Splice the empty reader page into the list around the head. |
1547 | * Reset the reader page to size zero. | |
7a8e76a3 | 1548 | */ |
7a8e76a3 | 1549 | |
d769041f SR |
1550 | reader = cpu_buffer->head_page; |
1551 | cpu_buffer->reader_page->list.next = reader->list.next; | |
1552 | cpu_buffer->reader_page->list.prev = reader->list.prev; | |
bf41a158 SR |
1553 | |
1554 | local_set(&cpu_buffer->reader_page->write, 0); | |
1555 | local_set(&cpu_buffer->reader_page->commit, 0); | |
7a8e76a3 | 1556 | |
d769041f SR |
1557 | /* Make the reader page now replace the head */ |
1558 | reader->list.prev->next = &cpu_buffer->reader_page->list; | |
1559 | reader->list.next->prev = &cpu_buffer->reader_page->list; | |
7a8e76a3 SR |
1560 | |
1561 | /* | |
d769041f SR |
1562 | * If the tail is on the reader, then we must set the head |
1563 | * to the inserted page, otherwise we set it one before. | |
7a8e76a3 | 1564 | */ |
d769041f | 1565 | cpu_buffer->head_page = cpu_buffer->reader_page; |
7a8e76a3 | 1566 | |
bf41a158 | 1567 | if (cpu_buffer->commit_page != reader) |
d769041f SR |
1568 | rb_inc_page(cpu_buffer, &cpu_buffer->head_page); |
1569 | ||
1570 | /* Finally update the reader page to the new head */ | |
1571 | cpu_buffer->reader_page = reader; | |
1572 | rb_reset_reader_page(cpu_buffer); | |
1573 | ||
1574 | goto again; | |
1575 | ||
1576 | out: | |
1577 | spin_unlock_irqrestore(&cpu_buffer->lock, flags); | |
1578 | ||
1579 | return reader; | |
1580 | } | |
1581 | ||
1582 | static void rb_advance_reader(struct ring_buffer_per_cpu *cpu_buffer) | |
1583 | { | |
1584 | struct ring_buffer_event *event; | |
1585 | struct buffer_page *reader; | |
1586 | unsigned length; | |
1587 | ||
1588 | reader = rb_get_reader_page(cpu_buffer); | |
7a8e76a3 | 1589 | |
d769041f SR |
1590 | /* This function should not be called when buffer is empty */ |
1591 | BUG_ON(!reader); | |
7a8e76a3 | 1592 | |
d769041f SR |
1593 | event = rb_reader_event(cpu_buffer); |
1594 | ||
1595 | if (event->type == RINGBUF_TYPE_DATA) | |
1596 | cpu_buffer->entries--; | |
1597 | ||
1598 | rb_update_read_stamp(cpu_buffer, event); | |
1599 | ||
1600 | length = rb_event_length(event); | |
6f807acd | 1601 | cpu_buffer->reader_page->read += length; |
7a8e76a3 SR |
1602 | } |
1603 | ||
1604 | static void rb_advance_iter(struct ring_buffer_iter *iter) | |
1605 | { | |
1606 | struct ring_buffer *buffer; | |
1607 | struct ring_buffer_per_cpu *cpu_buffer; | |
1608 | struct ring_buffer_event *event; | |
1609 | unsigned length; | |
1610 | ||
1611 | cpu_buffer = iter->cpu_buffer; | |
1612 | buffer = cpu_buffer->buffer; | |
1613 | ||
1614 | /* | |
1615 | * Check if we are at the end of the buffer. | |
1616 | */ | |
bf41a158 SR |
1617 | if (iter->head >= rb_page_size(iter->head_page)) { |
1618 | BUG_ON(iter->head_page == cpu_buffer->commit_page); | |
d769041f | 1619 | rb_inc_iter(iter); |
7a8e76a3 SR |
1620 | return; |
1621 | } | |
1622 | ||
1623 | event = rb_iter_head_event(iter); | |
1624 | ||
1625 | length = rb_event_length(event); | |
1626 | ||
1627 | /* | |
1628 | * This should not be called to advance the header if we are | |
1629 | * at the tail of the buffer. | |
1630 | */ | |
bf41a158 SR |
1631 | BUG_ON((iter->head_page == cpu_buffer->commit_page) && |
1632 | (iter->head + length > rb_commit_index(cpu_buffer))); | |
7a8e76a3 SR |
1633 | |
1634 | rb_update_iter_read_stamp(iter, event); | |
1635 | ||
1636 | iter->head += length; | |
1637 | ||
1638 | /* check for end of page padding */ | |
bf41a158 SR |
1639 | if ((iter->head >= rb_page_size(iter->head_page)) && |
1640 | (iter->head_page != cpu_buffer->commit_page)) | |
7a8e76a3 SR |
1641 | rb_advance_iter(iter); |
1642 | } | |
1643 | ||
1644 | /** | |
1645 | * ring_buffer_peek - peek at the next event to be read | |
1646 | * @buffer: The ring buffer to read | |
1647 | * @cpu: The cpu to peak at | |
1648 | * @ts: The timestamp counter of this event. | |
1649 | * | |
1650 | * This will return the event that will be read next, but does | |
1651 | * not consume the data. | |
1652 | */ | |
1653 | struct ring_buffer_event * | |
1654 | ring_buffer_peek(struct ring_buffer *buffer, int cpu, u64 *ts) | |
1655 | { | |
1656 | struct ring_buffer_per_cpu *cpu_buffer; | |
1657 | struct ring_buffer_event *event; | |
d769041f | 1658 | struct buffer_page *reader; |
7a8e76a3 SR |
1659 | |
1660 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1661 | return NULL; | |
1662 | ||
1663 | cpu_buffer = buffer->buffers[cpu]; | |
1664 | ||
1665 | again: | |
d769041f SR |
1666 | reader = rb_get_reader_page(cpu_buffer); |
1667 | if (!reader) | |
7a8e76a3 SR |
1668 | return NULL; |
1669 | ||
d769041f | 1670 | event = rb_reader_event(cpu_buffer); |
7a8e76a3 SR |
1671 | |
1672 | switch (event->type) { | |
1673 | case RINGBUF_TYPE_PADDING: | |
bf41a158 | 1674 | RB_WARN_ON(cpu_buffer, 1); |
d769041f SR |
1675 | rb_advance_reader(cpu_buffer); |
1676 | return NULL; | |
7a8e76a3 SR |
1677 | |
1678 | case RINGBUF_TYPE_TIME_EXTEND: | |
1679 | /* Internal data, OK to advance */ | |
d769041f | 1680 | rb_advance_reader(cpu_buffer); |
7a8e76a3 SR |
1681 | goto again; |
1682 | ||
1683 | case RINGBUF_TYPE_TIME_STAMP: | |
1684 | /* FIXME: not implemented */ | |
d769041f | 1685 | rb_advance_reader(cpu_buffer); |
7a8e76a3 SR |
1686 | goto again; |
1687 | ||
1688 | case RINGBUF_TYPE_DATA: | |
1689 | if (ts) { | |
1690 | *ts = cpu_buffer->read_stamp + event->time_delta; | |
1691 | ring_buffer_normalize_time_stamp(cpu_buffer->cpu, ts); | |
1692 | } | |
1693 | return event; | |
1694 | ||
1695 | default: | |
1696 | BUG(); | |
1697 | } | |
1698 | ||
1699 | return NULL; | |
1700 | } | |
1701 | ||
1702 | /** | |
1703 | * ring_buffer_iter_peek - peek at the next event to be read | |
1704 | * @iter: The ring buffer iterator | |
1705 | * @ts: The timestamp counter of this event. | |
1706 | * | |
1707 | * This will return the event that will be read next, but does | |
1708 | * not increment the iterator. | |
1709 | */ | |
1710 | struct ring_buffer_event * | |
1711 | ring_buffer_iter_peek(struct ring_buffer_iter *iter, u64 *ts) | |
1712 | { | |
1713 | struct ring_buffer *buffer; | |
1714 | struct ring_buffer_per_cpu *cpu_buffer; | |
1715 | struct ring_buffer_event *event; | |
1716 | ||
1717 | if (ring_buffer_iter_empty(iter)) | |
1718 | return NULL; | |
1719 | ||
1720 | cpu_buffer = iter->cpu_buffer; | |
1721 | buffer = cpu_buffer->buffer; | |
1722 | ||
1723 | again: | |
1724 | if (rb_per_cpu_empty(cpu_buffer)) | |
1725 | return NULL; | |
1726 | ||
1727 | event = rb_iter_head_event(iter); | |
1728 | ||
1729 | switch (event->type) { | |
1730 | case RINGBUF_TYPE_PADDING: | |
d769041f | 1731 | rb_inc_iter(iter); |
7a8e76a3 SR |
1732 | goto again; |
1733 | ||
1734 | case RINGBUF_TYPE_TIME_EXTEND: | |
1735 | /* Internal data, OK to advance */ | |
1736 | rb_advance_iter(iter); | |
1737 | goto again; | |
1738 | ||
1739 | case RINGBUF_TYPE_TIME_STAMP: | |
1740 | /* FIXME: not implemented */ | |
1741 | rb_advance_iter(iter); | |
1742 | goto again; | |
1743 | ||
1744 | case RINGBUF_TYPE_DATA: | |
1745 | if (ts) { | |
1746 | *ts = iter->read_stamp + event->time_delta; | |
1747 | ring_buffer_normalize_time_stamp(cpu_buffer->cpu, ts); | |
1748 | } | |
1749 | return event; | |
1750 | ||
1751 | default: | |
1752 | BUG(); | |
1753 | } | |
1754 | ||
1755 | return NULL; | |
1756 | } | |
1757 | ||
1758 | /** | |
1759 | * ring_buffer_consume - return an event and consume it | |
1760 | * @buffer: The ring buffer to get the next event from | |
1761 | * | |
1762 | * Returns the next event in the ring buffer, and that event is consumed. | |
1763 | * Meaning, that sequential reads will keep returning a different event, | |
1764 | * and eventually empty the ring buffer if the producer is slower. | |
1765 | */ | |
1766 | struct ring_buffer_event * | |
1767 | ring_buffer_consume(struct ring_buffer *buffer, int cpu, u64 *ts) | |
1768 | { | |
1769 | struct ring_buffer_per_cpu *cpu_buffer; | |
1770 | struct ring_buffer_event *event; | |
1771 | ||
1772 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1773 | return NULL; | |
1774 | ||
1775 | event = ring_buffer_peek(buffer, cpu, ts); | |
1776 | if (!event) | |
1777 | return NULL; | |
1778 | ||
1779 | cpu_buffer = buffer->buffers[cpu]; | |
d769041f | 1780 | rb_advance_reader(cpu_buffer); |
7a8e76a3 SR |
1781 | |
1782 | return event; | |
1783 | } | |
1784 | ||
1785 | /** | |
1786 | * ring_buffer_read_start - start a non consuming read of the buffer | |
1787 | * @buffer: The ring buffer to read from | |
1788 | * @cpu: The cpu buffer to iterate over | |
1789 | * | |
1790 | * This starts up an iteration through the buffer. It also disables | |
1791 | * the recording to the buffer until the reading is finished. | |
1792 | * This prevents the reading from being corrupted. This is not | |
1793 | * a consuming read, so a producer is not expected. | |
1794 | * | |
1795 | * Must be paired with ring_buffer_finish. | |
1796 | */ | |
1797 | struct ring_buffer_iter * | |
1798 | ring_buffer_read_start(struct ring_buffer *buffer, int cpu) | |
1799 | { | |
1800 | struct ring_buffer_per_cpu *cpu_buffer; | |
1801 | struct ring_buffer_iter *iter; | |
d769041f | 1802 | unsigned long flags; |
7a8e76a3 SR |
1803 | |
1804 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1805 | return NULL; | |
1806 | ||
1807 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); | |
1808 | if (!iter) | |
1809 | return NULL; | |
1810 | ||
1811 | cpu_buffer = buffer->buffers[cpu]; | |
1812 | ||
1813 | iter->cpu_buffer = cpu_buffer; | |
1814 | ||
1815 | atomic_inc(&cpu_buffer->record_disabled); | |
1816 | synchronize_sched(); | |
1817 | ||
d769041f SR |
1818 | spin_lock_irqsave(&cpu_buffer->lock, flags); |
1819 | ring_buffer_iter_reset(iter); | |
1820 | spin_unlock_irqrestore(&cpu_buffer->lock, flags); | |
7a8e76a3 SR |
1821 | |
1822 | return iter; | |
1823 | } | |
1824 | ||
1825 | /** | |
1826 | * ring_buffer_finish - finish reading the iterator of the buffer | |
1827 | * @iter: The iterator retrieved by ring_buffer_start | |
1828 | * | |
1829 | * This re-enables the recording to the buffer, and frees the | |
1830 | * iterator. | |
1831 | */ | |
1832 | void | |
1833 | ring_buffer_read_finish(struct ring_buffer_iter *iter) | |
1834 | { | |
1835 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
1836 | ||
1837 | atomic_dec(&cpu_buffer->record_disabled); | |
1838 | kfree(iter); | |
1839 | } | |
1840 | ||
1841 | /** | |
1842 | * ring_buffer_read - read the next item in the ring buffer by the iterator | |
1843 | * @iter: The ring buffer iterator | |
1844 | * @ts: The time stamp of the event read. | |
1845 | * | |
1846 | * This reads the next event in the ring buffer and increments the iterator. | |
1847 | */ | |
1848 | struct ring_buffer_event * | |
1849 | ring_buffer_read(struct ring_buffer_iter *iter, u64 *ts) | |
1850 | { | |
1851 | struct ring_buffer_event *event; | |
1852 | ||
1853 | event = ring_buffer_iter_peek(iter, ts); | |
1854 | if (!event) | |
1855 | return NULL; | |
1856 | ||
1857 | rb_advance_iter(iter); | |
1858 | ||
1859 | return event; | |
1860 | } | |
1861 | ||
1862 | /** | |
1863 | * ring_buffer_size - return the size of the ring buffer (in bytes) | |
1864 | * @buffer: The ring buffer. | |
1865 | */ | |
1866 | unsigned long ring_buffer_size(struct ring_buffer *buffer) | |
1867 | { | |
1868 | return BUF_PAGE_SIZE * buffer->pages; | |
1869 | } | |
1870 | ||
1871 | static void | |
1872 | rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) | |
1873 | { | |
1874 | cpu_buffer->head_page | |
1875 | = list_entry(cpu_buffer->pages.next, struct buffer_page, list); | |
bf41a158 SR |
1876 | local_set(&cpu_buffer->head_page->write, 0); |
1877 | local_set(&cpu_buffer->head_page->commit, 0); | |
d769041f | 1878 | |
6f807acd | 1879 | cpu_buffer->head_page->read = 0; |
bf41a158 SR |
1880 | |
1881 | cpu_buffer->tail_page = cpu_buffer->head_page; | |
1882 | cpu_buffer->commit_page = cpu_buffer->head_page; | |
1883 | ||
1884 | INIT_LIST_HEAD(&cpu_buffer->reader_page->list); | |
1885 | local_set(&cpu_buffer->reader_page->write, 0); | |
1886 | local_set(&cpu_buffer->reader_page->commit, 0); | |
6f807acd | 1887 | cpu_buffer->reader_page->read = 0; |
7a8e76a3 | 1888 | |
7a8e76a3 SR |
1889 | cpu_buffer->overrun = 0; |
1890 | cpu_buffer->entries = 0; | |
1891 | } | |
1892 | ||
1893 | /** | |
1894 | * ring_buffer_reset_cpu - reset a ring buffer per CPU buffer | |
1895 | * @buffer: The ring buffer to reset a per cpu buffer of | |
1896 | * @cpu: The CPU buffer to be reset | |
1897 | */ | |
1898 | void ring_buffer_reset_cpu(struct ring_buffer *buffer, int cpu) | |
1899 | { | |
1900 | struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu]; | |
1901 | unsigned long flags; | |
1902 | ||
1903 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1904 | return; | |
1905 | ||
d769041f | 1906 | spin_lock_irqsave(&cpu_buffer->lock, flags); |
7a8e76a3 SR |
1907 | |
1908 | rb_reset_cpu(cpu_buffer); | |
1909 | ||
d769041f | 1910 | spin_unlock_irqrestore(&cpu_buffer->lock, flags); |
7a8e76a3 SR |
1911 | } |
1912 | ||
1913 | /** | |
1914 | * ring_buffer_reset - reset a ring buffer | |
1915 | * @buffer: The ring buffer to reset all cpu buffers | |
1916 | */ | |
1917 | void ring_buffer_reset(struct ring_buffer *buffer) | |
1918 | { | |
7a8e76a3 SR |
1919 | int cpu; |
1920 | ||
7a8e76a3 | 1921 | for_each_buffer_cpu(buffer, cpu) |
d769041f | 1922 | ring_buffer_reset_cpu(buffer, cpu); |
7a8e76a3 SR |
1923 | } |
1924 | ||
1925 | /** | |
1926 | * rind_buffer_empty - is the ring buffer empty? | |
1927 | * @buffer: The ring buffer to test | |
1928 | */ | |
1929 | int ring_buffer_empty(struct ring_buffer *buffer) | |
1930 | { | |
1931 | struct ring_buffer_per_cpu *cpu_buffer; | |
1932 | int cpu; | |
1933 | ||
1934 | /* yes this is racy, but if you don't like the race, lock the buffer */ | |
1935 | for_each_buffer_cpu(buffer, cpu) { | |
1936 | cpu_buffer = buffer->buffers[cpu]; | |
1937 | if (!rb_per_cpu_empty(cpu_buffer)) | |
1938 | return 0; | |
1939 | } | |
1940 | return 1; | |
1941 | } | |
1942 | ||
1943 | /** | |
1944 | * ring_buffer_empty_cpu - is a cpu buffer of a ring buffer empty? | |
1945 | * @buffer: The ring buffer | |
1946 | * @cpu: The CPU buffer to test | |
1947 | */ | |
1948 | int ring_buffer_empty_cpu(struct ring_buffer *buffer, int cpu) | |
1949 | { | |
1950 | struct ring_buffer_per_cpu *cpu_buffer; | |
1951 | ||
1952 | if (!cpu_isset(cpu, buffer->cpumask)) | |
1953 | return 1; | |
1954 | ||
1955 | cpu_buffer = buffer->buffers[cpu]; | |
1956 | return rb_per_cpu_empty(cpu_buffer); | |
1957 | } | |
1958 | ||
1959 | /** | |
1960 | * ring_buffer_swap_cpu - swap a CPU buffer between two ring buffers | |
1961 | * @buffer_a: One buffer to swap with | |
1962 | * @buffer_b: The other buffer to swap with | |
1963 | * | |
1964 | * This function is useful for tracers that want to take a "snapshot" | |
1965 | * of a CPU buffer and has another back up buffer lying around. | |
1966 | * it is expected that the tracer handles the cpu buffer not being | |
1967 | * used at the moment. | |
1968 | */ | |
1969 | int ring_buffer_swap_cpu(struct ring_buffer *buffer_a, | |
1970 | struct ring_buffer *buffer_b, int cpu) | |
1971 | { | |
1972 | struct ring_buffer_per_cpu *cpu_buffer_a; | |
1973 | struct ring_buffer_per_cpu *cpu_buffer_b; | |
1974 | ||
1975 | if (!cpu_isset(cpu, buffer_a->cpumask) || | |
1976 | !cpu_isset(cpu, buffer_b->cpumask)) | |
1977 | return -EINVAL; | |
1978 | ||
1979 | /* At least make sure the two buffers are somewhat the same */ | |
1980 | if (buffer_a->size != buffer_b->size || | |
1981 | buffer_a->pages != buffer_b->pages) | |
1982 | return -EINVAL; | |
1983 | ||
1984 | cpu_buffer_a = buffer_a->buffers[cpu]; | |
1985 | cpu_buffer_b = buffer_b->buffers[cpu]; | |
1986 | ||
1987 | /* | |
1988 | * We can't do a synchronize_sched here because this | |
1989 | * function can be called in atomic context. | |
1990 | * Normally this will be called from the same CPU as cpu. | |
1991 | * If not it's up to the caller to protect this. | |
1992 | */ | |
1993 | atomic_inc(&cpu_buffer_a->record_disabled); | |
1994 | atomic_inc(&cpu_buffer_b->record_disabled); | |
1995 | ||
1996 | buffer_a->buffers[cpu] = cpu_buffer_b; | |
1997 | buffer_b->buffers[cpu] = cpu_buffer_a; | |
1998 | ||
1999 | cpu_buffer_b->buffer = buffer_a; | |
2000 | cpu_buffer_a->buffer = buffer_b; | |
2001 | ||
2002 | atomic_dec(&cpu_buffer_a->record_disabled); | |
2003 | atomic_dec(&cpu_buffer_b->record_disabled); | |
2004 | ||
2005 | return 0; | |
2006 | } | |
2007 |