1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Generic ring buffer
4 *
5 * Copyright (C) 2008 Steven Rostedt <srostedt@redhat.com>
6 */
7 #include <linux/trace_events.h>
8 #include <linux/ring_buffer.h>
9 #include <linux/trace_clock.h>
10 #include <linux/sched/clock.h>
11 #include <linux/trace_seq.h>
12 #include <linux/spinlock.h>
13 #include <linux/irq_work.h>
14 #include <linux/security.h>
15 #include <linux/uaccess.h>
16 #include <linux/hardirq.h>
17 #include <linux/kthread.h> /* for self test */
18 #include <linux/module.h>
19 #include <linux/percpu.h>
20 #include <linux/mutex.h>
21 #include <linux/delay.h>
22 #include <linux/slab.h>
23 #include <linux/init.h>
24 #include <linux/hash.h>
25 #include <linux/list.h>
26 #include <linux/cpu.h>
27 #include <linux/oom.h>
28
29 #include <asm/local.h>
30
31 static void update_pages_handler(struct work_struct *work);
32
33 /*
34 * The ring buffer header is special. We must manually up keep it.
35 */
ring_buffer_print_entry_header(struct trace_seq * s)36 int ring_buffer_print_entry_header(struct trace_seq *s)
37 {
38 trace_seq_puts(s, "# compressed entry header\n");
39 trace_seq_puts(s, "\ttype_len : 5 bits\n");
40 trace_seq_puts(s, "\ttime_delta : 27 bits\n");
41 trace_seq_puts(s, "\tarray : 32 bits\n");
42 trace_seq_putc(s, '\n');
43 trace_seq_printf(s, "\tpadding : type == %d\n",
44 RINGBUF_TYPE_PADDING);
45 trace_seq_printf(s, "\ttime_extend : type == %d\n",
46 RINGBUF_TYPE_TIME_EXTEND);
47 trace_seq_printf(s, "\ttime_stamp : type == %d\n",
48 RINGBUF_TYPE_TIME_STAMP);
49 trace_seq_printf(s, "\tdata max type_len == %d\n",
50 RINGBUF_TYPE_DATA_TYPE_LEN_MAX);
51
52 return !trace_seq_has_overflowed(s);
53 }
54
55 /*
56 * The ring buffer is made up of a list of pages. A separate list of pages is
57 * allocated for each CPU. A writer may only write to a buffer that is
58 * associated with the CPU it is currently executing on. A reader may read
59 * from any per cpu buffer.
60 *
61 * The reader is special. For each per cpu buffer, the reader has its own
62 * reader page. When a reader has read the entire reader page, this reader
63 * page is swapped with another page in the ring buffer.
64 *
65 * Now, as long as the writer is off the reader page, the reader can do what
66 * ever it wants with that page. The writer will never write to that page
67 * again (as long as it is out of the ring buffer).
68 *
69 * Here's some silly ASCII art.
70 *
71 * +------+
72 * |reader| RING BUFFER
73 * |page |
74 * +------+ +---+ +---+ +---+
75 * | |-->| |-->| |
76 * +---+ +---+ +---+
77 * ^ |
78 * | |
79 * +---------------+
80 *
81 *
82 * +------+
83 * |reader| RING BUFFER
84 * |page |------------------v
85 * +------+ +---+ +---+ +---+
86 * | |-->| |-->| |
87 * +---+ +---+ +---+
88 * ^ |
89 * | |
90 * +---------------+
91 *
92 *
93 * +------+
94 * |reader| RING BUFFER
95 * |page |------------------v
96 * +------+ +---+ +---+ +---+
97 * ^ | |-->| |-->| |
98 * | +---+ +---+ +---+
99 * | |
100 * | |
101 * +------------------------------+
102 *
103 *
104 * +------+
105 * |buffer| RING BUFFER
106 * |page |------------------v
107 * +------+ +---+ +---+ +---+
108 * ^ | | | |-->| |
109 * | New +---+ +---+ +---+
110 * | Reader------^ |
111 * | page |
112 * +------------------------------+
113 *
114 *
115 * After we make this swap, the reader can hand this page off to the splice
116 * code and be done with it. It can even allocate a new page if it needs to
117 * and swap that into the ring buffer.
118 *
119 * We will be using cmpxchg soon to make all this lockless.
120 *
121 */
122
123 /* Used for individual buffers (after the counter) */
124 #define RB_BUFFER_OFF (1 << 20)
125
126 #define BUF_PAGE_HDR_SIZE offsetof(struct buffer_data_page, data)
127
128 #define RB_EVNT_HDR_SIZE (offsetof(struct ring_buffer_event, array))
129 #define RB_ALIGNMENT 4U
130 #define RB_MAX_SMALL_DATA (RB_ALIGNMENT * RINGBUF_TYPE_DATA_TYPE_LEN_MAX)
131 #define RB_EVNT_MIN_SIZE 8U /* two 32bit words */
132 #define RB_ALIGN_DATA __aligned(RB_ALIGNMENT)
133
134 /* define RINGBUF_TYPE_DATA for 'case RINGBUF_TYPE_DATA:' */
135 #define RINGBUF_TYPE_DATA 0 ... RINGBUF_TYPE_DATA_TYPE_LEN_MAX
136
137 enum {
138 RB_LEN_TIME_EXTEND = 8,
139 RB_LEN_TIME_STAMP = 8,
140 };
141
142 #define skip_time_extend(event) \
143 ((struct ring_buffer_event *)((char *)event + RB_LEN_TIME_EXTEND))
144
145 #define extended_time(event) \
146 (event->type_len >= RINGBUF_TYPE_TIME_EXTEND)
147
rb_null_event(struct ring_buffer_event * event)148 static inline int rb_null_event(struct ring_buffer_event *event)
149 {
150 return event->type_len == RINGBUF_TYPE_PADDING && !event->time_delta;
151 }
152
rb_event_set_padding(struct ring_buffer_event * event)153 static void rb_event_set_padding(struct ring_buffer_event *event)
154 {
155 /* padding has a NULL time_delta */
156 event->type_len = RINGBUF_TYPE_PADDING;
157 event->time_delta = 0;
158 }
159
160 static unsigned
rb_event_data_length(struct ring_buffer_event * event)161 rb_event_data_length(struct ring_buffer_event *event)
162 {
163 unsigned length;
164
165 if (event->type_len)
166 length = event->type_len * RB_ALIGNMENT;
167 else
168 length = event->array[0];
169 return length + RB_EVNT_HDR_SIZE;
170 }
171
172 /*
173 * Return the length of the given event. Will return
174 * the length of the time extend if the event is a
175 * time extend.
176 */
177 static inline unsigned
rb_event_length(struct ring_buffer_event * event)178 rb_event_length(struct ring_buffer_event *event)
179 {
180 switch (event->type_len) {
181 case RINGBUF_TYPE_PADDING:
182 if (rb_null_event(event))
183 /* undefined */
184 return -1;
185 return event->array[0] + RB_EVNT_HDR_SIZE;
186
187 case RINGBUF_TYPE_TIME_EXTEND:
188 return RB_LEN_TIME_EXTEND;
189
190 case RINGBUF_TYPE_TIME_STAMP:
191 return RB_LEN_TIME_STAMP;
192
193 case RINGBUF_TYPE_DATA:
194 return rb_event_data_length(event);
195 default:
196 WARN_ON_ONCE(1);
197 }
198 /* not hit */
199 return 0;
200 }
201
202 /*
203 * Return total length of time extend and data,
204 * or just the event length for all other events.
205 */
206 static inline unsigned
rb_event_ts_length(struct ring_buffer_event * event)207 rb_event_ts_length(struct ring_buffer_event *event)
208 {
209 unsigned len = 0;
210
211 if (extended_time(event)) {
212 /* time extends include the data event after it */
213 len = RB_LEN_TIME_EXTEND;
214 event = skip_time_extend(event);
215 }
216 return len + rb_event_length(event);
217 }
218
219 /**
220 * ring_buffer_event_length - return the length of the event
221 * @event: the event to get the length of
222 *
223 * Returns the size of the data load of a data event.
224 * If the event is something other than a data event, it
225 * returns the size of the event itself. With the exception
226 * of a TIME EXTEND, where it still returns the size of the
227 * data load of the data event after it.
228 */
ring_buffer_event_length(struct ring_buffer_event * event)229 unsigned ring_buffer_event_length(struct ring_buffer_event *event)
230 {
231 unsigned length;
232
233 if (extended_time(event))
234 event = skip_time_extend(event);
235
236 length = rb_event_length(event);
237 if (event->type_len > RINGBUF_TYPE_DATA_TYPE_LEN_MAX)
238 return length;
239 length -= RB_EVNT_HDR_SIZE;
240 if (length > RB_MAX_SMALL_DATA + sizeof(event->array[0]))
241 length -= sizeof(event->array[0]);
242 return length;
243 }
244 EXPORT_SYMBOL_GPL(ring_buffer_event_length);
245
246 /* inline for ring buffer fast paths */
247 static __always_inline void *
rb_event_data(struct ring_buffer_event * event)248 rb_event_data(struct ring_buffer_event *event)
249 {
250 if (extended_time(event))
251 event = skip_time_extend(event);
252 WARN_ON_ONCE(event->type_len > RINGBUF_TYPE_DATA_TYPE_LEN_MAX);
253 /* If length is in len field, then array[0] has the data */
254 if (event->type_len)
255 return (void *)&event->array[0];
256 /* Otherwise length is in array[0] and array[1] has the data */
257 return (void *)&event->array[1];
258 }
259
260 /**
261 * ring_buffer_event_data - return the data of the event
262 * @event: the event to get the data from
263 */
ring_buffer_event_data(struct ring_buffer_event * event)264 void *ring_buffer_event_data(struct ring_buffer_event *event)
265 {
266 return rb_event_data(event);
267 }
268 EXPORT_SYMBOL_GPL(ring_buffer_event_data);
269
270 #define for_each_buffer_cpu(buffer, cpu) \
271 for_each_cpu(cpu, buffer->cpumask)
272
273 #define for_each_online_buffer_cpu(buffer, cpu) \
274 for_each_cpu_and(cpu, buffer->cpumask, cpu_online_mask)
275
276 #define TS_SHIFT 27
277 #define TS_MASK ((1ULL << TS_SHIFT) - 1)
278 #define TS_DELTA_TEST (~TS_MASK)
279
280 /**
281 * ring_buffer_event_time_stamp - return the event's extended timestamp
282 * @event: the event to get the timestamp of
283 *
284 * Returns the extended timestamp associated with a data event.
285 * An extended time_stamp is a 64-bit timestamp represented
286 * internally in a special way that makes the best use of space
287 * contained within a ring buffer event. This function decodes
288 * it and maps it to a straight u64 value.
289 */
ring_buffer_event_time_stamp(struct ring_buffer_event * event)290 u64 ring_buffer_event_time_stamp(struct ring_buffer_event *event)
291 {
292 u64 ts;
293
294 ts = event->array[0];
295 ts <<= TS_SHIFT;
296 ts += event->time_delta;
297
298 return ts;
299 }
300
301 /* Flag when events were overwritten */
302 #define RB_MISSED_EVENTS (1 << 31)
303 /* Missed count stored at end */
304 #define RB_MISSED_STORED (1 << 30)
305
306 struct buffer_data_page {
307 u64 time_stamp; /* page time stamp */
308 local_t commit; /* write committed index */
309 unsigned char data[] RB_ALIGN_DATA; /* data of buffer page */
310 };
311
312 /*
313 * Note, the buffer_page list must be first. The buffer pages
314 * are allocated in cache lines, which means that each buffer
315 * page will be at the beginning of a cache line, and thus
316 * the least significant bits will be zero. We use this to
317 * add flags in the list struct pointers, to make the ring buffer
318 * lockless.
319 */
320 struct buffer_page {
321 struct list_head list; /* list of buffer pages */
322 local_t write; /* index for next write */
323 unsigned read; /* index for next read */
324 local_t entries; /* entries on this page */
325 unsigned long real_end; /* real end of data */
326 struct buffer_data_page *page; /* Actual data page */
327 };
328
329 /*
330 * The buffer page counters, write and entries, must be reset
331 * atomically when crossing page boundaries. To synchronize this
332 * update, two counters are inserted into the number. One is
333 * the actual counter for the write position or count on the page.
334 *
335 * The other is a counter of updaters. Before an update happens
336 * the update partition of the counter is incremented. This will
337 * allow the updater to update the counter atomically.
338 *
339 * The counter is 20 bits, and the state data is 12.
340 */
341 #define RB_WRITE_MASK 0xfffff
342 #define RB_WRITE_INTCNT (1 << 20)
343
rb_init_page(struct buffer_data_page * bpage)344 static void rb_init_page(struct buffer_data_page *bpage)
345 {
346 local_set(&bpage->commit, 0);
347 }
348
349 /*
350 * Also stolen from mm/slob.c. Thanks to Mathieu Desnoyers for pointing
351 * this issue out.
352 */
free_buffer_page(struct buffer_page * bpage)353 static void free_buffer_page(struct buffer_page *bpage)
354 {
355 free_page((unsigned long)bpage->page);
356 kfree(bpage);
357 }
358
359 /*
360 * We need to fit the time_stamp delta into 27 bits.
361 */
test_time_stamp(u64 delta)362 static inline int test_time_stamp(u64 delta)
363 {
364 if (delta & TS_DELTA_TEST)
365 return 1;
366 return 0;
367 }
368
369 #define BUF_PAGE_SIZE (PAGE_SIZE - BUF_PAGE_HDR_SIZE)
370
371 /* Max payload is BUF_PAGE_SIZE - header (8bytes) */
372 #define BUF_MAX_DATA_SIZE (BUF_PAGE_SIZE - (sizeof(u32) * 2))
373
ring_buffer_print_page_header(struct trace_seq * s)374 int ring_buffer_print_page_header(struct trace_seq *s)
375 {
376 struct buffer_data_page field;
377
378 trace_seq_printf(s, "\tfield: u64 timestamp;\t"
379 "offset:0;\tsize:%u;\tsigned:%u;\n",
380 (unsigned int)sizeof(field.time_stamp),
381 (unsigned int)is_signed_type(u64));
382
383 trace_seq_printf(s, "\tfield: local_t commit;\t"
384 "offset:%u;\tsize:%u;\tsigned:%u;\n",
385 (unsigned int)offsetof(typeof(field), commit),
386 (unsigned int)sizeof(field.commit),
387 (unsigned int)is_signed_type(long));
388
389 trace_seq_printf(s, "\tfield: int overwrite;\t"
390 "offset:%u;\tsize:%u;\tsigned:%u;\n",
391 (unsigned int)offsetof(typeof(field), commit),
392 1,
393 (unsigned int)is_signed_type(long));
394
395 trace_seq_printf(s, "\tfield: char data;\t"
396 "offset:%u;\tsize:%u;\tsigned:%u;\n",
397 (unsigned int)offsetof(typeof(field), data),
398 (unsigned int)BUF_PAGE_SIZE,
399 (unsigned int)is_signed_type(char));
400
401 return !trace_seq_has_overflowed(s);
402 }
403
404 struct rb_irq_work {
405 struct irq_work work;
406 wait_queue_head_t waiters;
407 wait_queue_head_t full_waiters;
408 bool waiters_pending;
409 bool full_waiters_pending;
410 bool wakeup_full;
411 };
412
413 /*
414 * Structure to hold event state and handle nested events.
415 */
416 struct rb_event_info {
417 u64 ts;
418 u64 delta;
419 u64 before;
420 u64 after;
421 unsigned long length;
422 struct buffer_page *tail_page;
423 int add_timestamp;
424 };
425
426 /*
427 * Used for the add_timestamp
428 * NONE
429 * EXTEND - wants a time extend
430 * ABSOLUTE - the buffer requests all events to have absolute time stamps
431 * FORCE - force a full time stamp.
432 */
433 enum {
434 RB_ADD_STAMP_NONE = 0,
435 RB_ADD_STAMP_EXTEND = BIT(1),
436 RB_ADD_STAMP_ABSOLUTE = BIT(2),
437 RB_ADD_STAMP_FORCE = BIT(3)
438 };
439 /*
440 * Used for which event context the event is in.
441 * TRANSITION = 0
442 * NMI = 1
443 * IRQ = 2
444 * SOFTIRQ = 3
445 * NORMAL = 4
446 *
447 * See trace_recursive_lock() comment below for more details.
448 */
449 enum {
450 RB_CTX_TRANSITION,
451 RB_CTX_NMI,
452 RB_CTX_IRQ,
453 RB_CTX_SOFTIRQ,
454 RB_CTX_NORMAL,
455 RB_CTX_MAX
456 };
457
458 #if BITS_PER_LONG == 32
459 #define RB_TIME_32
460 #endif
461
462 /* To test on 64 bit machines */
463 //#define RB_TIME_32
464
465 #ifdef RB_TIME_32
466
467 struct rb_time_struct {
468 local_t cnt;
469 local_t top;
470 local_t bottom;
471 };
472 #else
473 #include <asm/local64.h>
474 struct rb_time_struct {
475 local64_t time;
476 };
477 #endif
478 typedef struct rb_time_struct rb_time_t;
479
480 /*
481 * head_page == tail_page && head == tail then buffer is empty.
482 */
483 struct ring_buffer_per_cpu {
484 int cpu;
485 atomic_t record_disabled;
486 atomic_t resize_disabled;
487 struct trace_buffer *buffer;
488 raw_spinlock_t reader_lock; /* serialize readers */
489 arch_spinlock_t lock;
490 struct lock_class_key lock_key;
491 struct buffer_data_page *free_page;
492 unsigned long nr_pages;
493 unsigned int current_context;
494 struct list_head *pages;
495 struct buffer_page *head_page; /* read from head */
496 struct buffer_page *tail_page; /* write to tail */
497 struct buffer_page *commit_page; /* committed pages */
498 struct buffer_page *reader_page;
499 unsigned long lost_events;
500 unsigned long last_overrun;
501 unsigned long nest;
502 local_t entries_bytes;
503 local_t entries;
504 local_t overrun;
505 local_t commit_overrun;
506 local_t dropped_events;
507 local_t committing;
508 local_t commits;
509 local_t pages_touched;
510 local_t pages_read;
511 long last_pages_touch;
512 size_t shortest_full;
513 unsigned long read;
514 unsigned long read_bytes;
515 rb_time_t write_stamp;
516 rb_time_t before_stamp;
517 u64 read_stamp;
518 /* ring buffer pages to update, > 0 to add, < 0 to remove */
519 long nr_pages_to_update;
520 struct list_head new_pages; /* new pages to add */
521 struct work_struct update_pages_work;
522 struct completion update_done;
523
524 struct rb_irq_work irq_work;
525 };
526
527 struct trace_buffer {
528 unsigned flags;
529 int cpus;
530 atomic_t record_disabled;
531 cpumask_var_t cpumask;
532
533 struct lock_class_key *reader_lock_key;
534
535 struct mutex mutex;
536
537 struct ring_buffer_per_cpu **buffers;
538
539 struct hlist_node node;
540 u64 (*clock)(void);
541
542 struct rb_irq_work irq_work;
543 bool time_stamp_abs;
544 };
545
546 struct ring_buffer_iter {
547 struct ring_buffer_per_cpu *cpu_buffer;
548 unsigned long head;
549 unsigned long next_event;
550 struct buffer_page *head_page;
551 struct buffer_page *cache_reader_page;
552 unsigned long cache_read;
553 u64 read_stamp;
554 u64 page_stamp;
555 struct ring_buffer_event *event;
556 int missed_events;
557 };
558
559 #ifdef RB_TIME_32
560
561 /*
562 * On 32 bit machines, local64_t is very expensive. As the ring
563 * buffer doesn't need all the features of a true 64 bit atomic,
564 * on 32 bit, it uses these functions (64 still uses local64_t).
565 *
566 * For the ring buffer, 64 bit required operations for the time is
567 * the following:
568 *
569 * - Only need 59 bits (uses 60 to make it even).
570 * - Reads may fail if it interrupted a modification of the time stamp.
571 * It will succeed if it did not interrupt another write even if
572 * the read itself is interrupted by a write.
573 * It returns whether it was successful or not.
574 *
575 * - Writes always succeed and will overwrite other writes and writes
576 * that were done by events interrupting the current write.
577 *
578 * - A write followed by a read of the same time stamp will always succeed,
579 * but may not contain the same value.
580 *
581 * - A cmpxchg will fail if it interrupted another write or cmpxchg.
582 * Other than that, it acts like a normal cmpxchg.
583 *
584 * The 60 bit time stamp is broken up by 30 bits in a top and bottom half
585 * (bottom being the least significant 30 bits of the 60 bit time stamp).
586 *
587 * The two most significant bits of each half holds a 2 bit counter (0-3).
588 * Each update will increment this counter by one.
589 * When reading the top and bottom, if the two counter bits match then the
590 * top and bottom together make a valid 60 bit number.
591 */
592 #define RB_TIME_SHIFT 30
593 #define RB_TIME_VAL_MASK ((1 << RB_TIME_SHIFT) - 1)
594
rb_time_cnt(unsigned long val)595 static inline int rb_time_cnt(unsigned long val)
596 {
597 return (val >> RB_TIME_SHIFT) & 3;
598 }
599
rb_time_val(unsigned long top,unsigned long bottom)600 static inline u64 rb_time_val(unsigned long top, unsigned long bottom)
601 {
602 u64 val;
603
604 val = top & RB_TIME_VAL_MASK;
605 val <<= RB_TIME_SHIFT;
606 val |= bottom & RB_TIME_VAL_MASK;
607
608 return val;
609 }
610
__rb_time_read(rb_time_t * t,u64 * ret,unsigned long * cnt)611 static inline bool __rb_time_read(rb_time_t *t, u64 *ret, unsigned long *cnt)
612 {
613 unsigned long top, bottom;
614 unsigned long c;
615
616 /*
617 * If the read is interrupted by a write, then the cnt will
618 * be different. Loop until both top and bottom have been read
619 * without interruption.
620 */
621 do {
622 c = local_read(&t->cnt);
623 top = local_read(&t->top);
624 bottom = local_read(&t->bottom);
625 } while (c != local_read(&t->cnt));
626
627 *cnt = rb_time_cnt(top);
628
629 /* If top and bottom counts don't match, this interrupted a write */
630 if (*cnt != rb_time_cnt(bottom))
631 return false;
632
633 *ret = rb_time_val(top, bottom);
634 return true;
635 }
636
rb_time_read(rb_time_t * t,u64 * ret)637 static bool rb_time_read(rb_time_t *t, u64 *ret)
638 {
639 unsigned long cnt;
640
641 return __rb_time_read(t, ret, &cnt);
642 }
643
rb_time_val_cnt(unsigned long val,unsigned long cnt)644 static inline unsigned long rb_time_val_cnt(unsigned long val, unsigned long cnt)
645 {
646 return (val & RB_TIME_VAL_MASK) | ((cnt & 3) << RB_TIME_SHIFT);
647 }
648
rb_time_split(u64 val,unsigned long * top,unsigned long * bottom)649 static inline void rb_time_split(u64 val, unsigned long *top, unsigned long *bottom)
650 {
651 *top = (unsigned long)((val >> RB_TIME_SHIFT) & RB_TIME_VAL_MASK);
652 *bottom = (unsigned long)(val & RB_TIME_VAL_MASK);
653 }
654
rb_time_val_set(local_t * t,unsigned long val,unsigned long cnt)655 static inline void rb_time_val_set(local_t *t, unsigned long val, unsigned long cnt)
656 {
657 val = rb_time_val_cnt(val, cnt);
658 local_set(t, val);
659 }
660
rb_time_set(rb_time_t * t,u64 val)661 static void rb_time_set(rb_time_t *t, u64 val)
662 {
663 unsigned long cnt, top, bottom;
664
665 rb_time_split(val, &top, &bottom);
666
667 /* Writes always succeed with a valid number even if it gets interrupted. */
668 do {
669 cnt = local_inc_return(&t->cnt);
670 rb_time_val_set(&t->top, top, cnt);
671 rb_time_val_set(&t->bottom, bottom, cnt);
672 } while (cnt != local_read(&t->cnt));
673 }
674
675 static inline bool
rb_time_read_cmpxchg(local_t * l,unsigned long expect,unsigned long set)676 rb_time_read_cmpxchg(local_t *l, unsigned long expect, unsigned long set)
677 {
678 unsigned long ret;
679
680 ret = local_cmpxchg(l, expect, set);
681 return ret == expect;
682 }
683
rb_time_cmpxchg(rb_time_t * t,u64 expect,u64 set)684 static int rb_time_cmpxchg(rb_time_t *t, u64 expect, u64 set)
685 {
686 unsigned long cnt, top, bottom;
687 unsigned long cnt2, top2, bottom2;
688 u64 val;
689
690 /* The cmpxchg always fails if it interrupted an update */
691 if (!__rb_time_read(t, &val, &cnt2))
692 return false;
693
694 if (val != expect)
695 return false;
696
697 cnt = local_read(&t->cnt);
698 if ((cnt & 3) != cnt2)
699 return false;
700
701 cnt2 = cnt + 1;
702
703 rb_time_split(val, &top, &bottom);
704 top = rb_time_val_cnt(top, cnt);
705 bottom = rb_time_val_cnt(bottom, cnt);
706
707 rb_time_split(set, &top2, &bottom2);
708 top2 = rb_time_val_cnt(top2, cnt2);
709 bottom2 = rb_time_val_cnt(bottom2, cnt2);
710
711 if (!rb_time_read_cmpxchg(&t->cnt, cnt, cnt2))
712 return false;
713 if (!rb_time_read_cmpxchg(&t->top, top, top2))
714 return false;
715 if (!rb_time_read_cmpxchg(&t->bottom, bottom, bottom2))
716 return false;
717 return true;
718 }
719
720 #else /* 64 bits */
721
722 /* local64_t always succeeds */
723
rb_time_read(rb_time_t * t,u64 * ret)724 static inline bool rb_time_read(rb_time_t *t, u64 *ret)
725 {
726 *ret = local64_read(&t->time);
727 return true;
728 }
rb_time_set(rb_time_t * t,u64 val)729 static void rb_time_set(rb_time_t *t, u64 val)
730 {
731 local64_set(&t->time, val);
732 }
733
rb_time_cmpxchg(rb_time_t * t,u64 expect,u64 set)734 static bool rb_time_cmpxchg(rb_time_t *t, u64 expect, u64 set)
735 {
736 u64 val;
737 val = local64_cmpxchg(&t->time, expect, set);
738 return val == expect;
739 }
740 #endif
741
742 /**
743 * ring_buffer_nr_pages - get the number of buffer pages in the ring buffer
744 * @buffer: The ring_buffer to get the number of pages from
745 * @cpu: The cpu of the ring_buffer to get the number of pages from
746 *
747 * Returns the number of pages used by a per_cpu buffer of the ring buffer.
748 */
ring_buffer_nr_pages(struct trace_buffer * buffer,int cpu)749 size_t ring_buffer_nr_pages(struct trace_buffer *buffer, int cpu)
750 {
751 return buffer->buffers[cpu]->nr_pages;
752 }
753
754 /**
755 * ring_buffer_nr_pages_dirty - get the number of used pages in the ring buffer
756 * @buffer: The ring_buffer to get the number of pages from
757 * @cpu: The cpu of the ring_buffer to get the number of pages from
758 *
759 * Returns the number of pages that have content in the ring buffer.
760 */
ring_buffer_nr_dirty_pages(struct trace_buffer * buffer,int cpu)761 size_t ring_buffer_nr_dirty_pages(struct trace_buffer *buffer, int cpu)
762 {
763 size_t read;
764 size_t cnt;
765
766 read = local_read(&buffer->buffers[cpu]->pages_read);
767 cnt = local_read(&buffer->buffers[cpu]->pages_touched);
768 /* The reader can read an empty page, but not more than that */
769 if (cnt < read) {
770 WARN_ON_ONCE(read > cnt + 1);
771 return 0;
772 }
773
774 return cnt - read;
775 }
776
777 /*
778 * rb_wake_up_waiters - wake up tasks waiting for ring buffer input
779 *
780 * Schedules a delayed work to wake up any task that is blocked on the
781 * ring buffer waiters queue.
782 */
rb_wake_up_waiters(struct irq_work * work)783 static void rb_wake_up_waiters(struct irq_work *work)
784 {
785 struct rb_irq_work *rbwork = container_of(work, struct rb_irq_work, work);
786
787 wake_up_all(&rbwork->waiters);
788 if (rbwork->wakeup_full) {
789 rbwork->wakeup_full = false;
790 wake_up_all(&rbwork->full_waiters);
791 }
792 }
793
794 /**
795 * ring_buffer_wait - wait for input to the ring buffer
796 * @buffer: buffer to wait on
797 * @cpu: the cpu buffer to wait on
798 * @full: wait until the percentage of pages are available, if @cpu != RING_BUFFER_ALL_CPUS
799 *
800 * If @cpu == RING_BUFFER_ALL_CPUS then the task will wake up as soon
801 * as data is added to any of the @buffer's cpu buffers. Otherwise
802 * it will wait for data to be added to a specific cpu buffer.
803 */
ring_buffer_wait(struct trace_buffer * buffer,int cpu,int full)804 int ring_buffer_wait(struct trace_buffer *buffer, int cpu, int full)
805 {
806 struct ring_buffer_per_cpu *cpu_buffer;
807 DEFINE_WAIT(wait);
808 struct rb_irq_work *work;
809 int ret = 0;
810
811 /*
812 * Depending on what the caller is waiting for, either any
813 * data in any cpu buffer, or a specific buffer, put the
814 * caller on the appropriate wait queue.
815 */
816 if (cpu == RING_BUFFER_ALL_CPUS) {
817 work = &buffer->irq_work;
818 /* Full only makes sense on per cpu reads */
819 full = 0;
820 } else {
821 if (!cpumask_test_cpu(cpu, buffer->cpumask))
822 return -ENODEV;
823 cpu_buffer = buffer->buffers[cpu];
824 work = &cpu_buffer->irq_work;
825 }
826
827
828 while (true) {
829 if (full)
830 prepare_to_wait(&work->full_waiters, &wait, TASK_INTERRUPTIBLE);
831 else
832 prepare_to_wait(&work->waiters, &wait, TASK_INTERRUPTIBLE);
833
834 /*
835 * The events can happen in critical sections where
836 * checking a work queue can cause deadlocks.
837 * After adding a task to the queue, this flag is set
838 * only to notify events to try to wake up the queue
839 * using irq_work.
840 *
841 * We don't clear it even if the buffer is no longer
842 * empty. The flag only causes the next event to run
843 * irq_work to do the work queue wake up. The worse
844 * that can happen if we race with !trace_empty() is that
845 * an event will cause an irq_work to try to wake up
846 * an empty queue.
847 *
848 * There's no reason to protect this flag either, as
849 * the work queue and irq_work logic will do the necessary
850 * synchronization for the wake ups. The only thing
851 * that is necessary is that the wake up happens after
852 * a task has been queued. It's OK for spurious wake ups.
853 */
854 if (full)
855 work->full_waiters_pending = true;
856 else
857 work->waiters_pending = true;
858
859 if (signal_pending(current)) {
860 ret = -EINTR;
861 break;
862 }
863
864 if (cpu == RING_BUFFER_ALL_CPUS && !ring_buffer_empty(buffer))
865 break;
866
867 if (cpu != RING_BUFFER_ALL_CPUS &&
868 !ring_buffer_empty_cpu(buffer, cpu)) {
869 unsigned long flags;
870 bool pagebusy;
871 size_t nr_pages;
872 size_t dirty;
873
874 if (!full)
875 break;
876
877 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
878 pagebusy = cpu_buffer->reader_page == cpu_buffer->commit_page;
879 nr_pages = cpu_buffer->nr_pages;
880 dirty = ring_buffer_nr_dirty_pages(buffer, cpu);
881 if (!cpu_buffer->shortest_full ||
882 cpu_buffer->shortest_full < full)
883 cpu_buffer->shortest_full = full;
884 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
885 if (!pagebusy &&
886 (!nr_pages || (dirty * 100) > full * nr_pages))
887 break;
888 }
889
890 schedule();
891 }
892
893 if (full)
894 finish_wait(&work->full_waiters, &wait);
895 else
896 finish_wait(&work->waiters, &wait);
897
898 return ret;
899 }
900
901 /**
902 * ring_buffer_poll_wait - poll on buffer input
903 * @buffer: buffer to wait on
904 * @cpu: the cpu buffer to wait on
905 * @filp: the file descriptor
906 * @poll_table: The poll descriptor
907 *
908 * If @cpu == RING_BUFFER_ALL_CPUS then the task will wake up as soon
909 * as data is added to any of the @buffer's cpu buffers. Otherwise
910 * it will wait for data to be added to a specific cpu buffer.
911 *
912 * Returns EPOLLIN | EPOLLRDNORM if data exists in the buffers,
913 * zero otherwise.
914 */
ring_buffer_poll_wait(struct trace_buffer * buffer,int cpu,struct file * filp,poll_table * poll_table)915 __poll_t ring_buffer_poll_wait(struct trace_buffer *buffer, int cpu,
916 struct file *filp, poll_table *poll_table)
917 {
918 struct ring_buffer_per_cpu *cpu_buffer;
919 struct rb_irq_work *work;
920
921 if (cpu == RING_BUFFER_ALL_CPUS)
922 work = &buffer->irq_work;
923 else {
924 if (!cpumask_test_cpu(cpu, buffer->cpumask))
925 return -EINVAL;
926
927 cpu_buffer = buffer->buffers[cpu];
928 work = &cpu_buffer->irq_work;
929 }
930
931 poll_wait(filp, &work->waiters, poll_table);
932 work->waiters_pending = true;
933 /*
934 * There's a tight race between setting the waiters_pending and
935 * checking if the ring buffer is empty. Once the waiters_pending bit
936 * is set, the next event will wake the task up, but we can get stuck
937 * if there's only a single event in.
938 *
939 * FIXME: Ideally, we need a memory barrier on the writer side as well,
940 * but adding a memory barrier to all events will cause too much of a
941 * performance hit in the fast path. We only need a memory barrier when
942 * the buffer goes from empty to having content. But as this race is
943 * extremely small, and it's not a problem if another event comes in, we
944 * will fix it later.
945 */
946 smp_mb();
947
948 if ((cpu == RING_BUFFER_ALL_CPUS && !ring_buffer_empty(buffer)) ||
949 (cpu != RING_BUFFER_ALL_CPUS && !ring_buffer_empty_cpu(buffer, cpu)))
950 return EPOLLIN | EPOLLRDNORM;
951 return 0;
952 }
953
954 /* buffer may be either ring_buffer or ring_buffer_per_cpu */
955 #define RB_WARN_ON(b, cond) \
956 ({ \
957 int _____ret = unlikely(cond); \
958 if (_____ret) { \
959 if (__same_type(*(b), struct ring_buffer_per_cpu)) { \
960 struct ring_buffer_per_cpu *__b = \
961 (void *)b; \
962 atomic_inc(&__b->buffer->record_disabled); \
963 } else \
964 atomic_inc(&b->record_disabled); \
965 WARN_ON(1); \
966 } \
967 _____ret; \
968 })
969
970 /* Up this if you want to test the TIME_EXTENTS and normalization */
971 #define DEBUG_SHIFT 0
972
rb_time_stamp(struct trace_buffer * buffer)973 static inline u64 rb_time_stamp(struct trace_buffer *buffer)
974 {
975 u64 ts;
976
977 /* Skip retpolines :-( */
978 if (IS_ENABLED(CONFIG_RETPOLINE) && likely(buffer->clock == trace_clock_local))
979 ts = trace_clock_local();
980 else
981 ts = buffer->clock();
982
983 /* shift to debug/test normalization and TIME_EXTENTS */
984 return ts << DEBUG_SHIFT;
985 }
986
ring_buffer_time_stamp(struct trace_buffer * buffer,int cpu)987 u64 ring_buffer_time_stamp(struct trace_buffer *buffer, int cpu)
988 {
989 u64 time;
990
991 preempt_disable_notrace();
992 time = rb_time_stamp(buffer);
993 preempt_enable_notrace();
994
995 return time;
996 }
997 EXPORT_SYMBOL_GPL(ring_buffer_time_stamp);
998
ring_buffer_normalize_time_stamp(struct trace_buffer * buffer,int cpu,u64 * ts)999 void ring_buffer_normalize_time_stamp(struct trace_buffer *buffer,
1000 int cpu, u64 *ts)
1001 {
1002 /* Just stupid testing the normalize function and deltas */
1003 *ts >>= DEBUG_SHIFT;
1004 }
1005 EXPORT_SYMBOL_GPL(ring_buffer_normalize_time_stamp);
1006
1007 /*
1008 * Making the ring buffer lockless makes things tricky.
1009 * Although writes only happen on the CPU that they are on,
1010 * and they only need to worry about interrupts. Reads can
1011 * happen on any CPU.
1012 *
1013 * The reader page is always off the ring buffer, but when the
1014 * reader finishes with a page, it needs to swap its page with
1015 * a new one from the buffer. The reader needs to take from
1016 * the head (writes go to the tail). But if a writer is in overwrite
1017 * mode and wraps, it must push the head page forward.
1018 *
1019 * Here lies the problem.
1020 *
1021 * The reader must be careful to replace only the head page, and
1022 * not another one. As described at the top of the file in the
1023 * ASCII art, the reader sets its old page to point to the next
1024 * page after head. It then sets the page after head to point to
1025 * the old reader page. But if the writer moves the head page
1026 * during this operation, the reader could end up with the tail.
1027 *
1028 * We use cmpxchg to help prevent this race. We also do something
1029 * special with the page before head. We set the LSB to 1.
1030 *
1031 * When the writer must push the page forward, it will clear the
1032 * bit that points to the head page, move the head, and then set
1033 * the bit that points to the new head page.
1034 *
1035 * We also don't want an interrupt coming in and moving the head
1036 * page on another writer. Thus we use the second LSB to catch
1037 * that too. Thus:
1038 *
1039 * head->list->prev->next bit 1 bit 0
1040 * ------- -------
1041 * Normal page 0 0
1042 * Points to head page 0 1
1043 * New head page 1 0
1044 *
1045 * Note we can not trust the prev pointer of the head page, because:
1046 *
1047 * +----+ +-----+ +-----+
1048 * | |------>| T |---X--->| N |
1049 * | |<------| | | |
1050 * +----+ +-----+ +-----+
1051 * ^ ^ |
1052 * | +-----+ | |
1053 * +----------| R |----------+ |
1054 * | |<-----------+
1055 * +-----+
1056 *
1057 * Key: ---X--> HEAD flag set in pointer
1058 * T Tail page
1059 * R Reader page
1060 * N Next page
1061 *
1062 * (see __rb_reserve_next() to see where this happens)
1063 *
1064 * What the above shows is that the reader just swapped out
1065 * the reader page with a page in the buffer, but before it
1066 * could make the new header point back to the new page added
1067 * it was preempted by a writer. The writer moved forward onto
1068 * the new page added by the reader and is about to move forward
1069 * again.
1070 *
1071 * You can see, it is legitimate for the previous pointer of
1072 * the head (or any page) not to point back to itself. But only
1073 * temporarily.
1074 */
1075
1076 #define RB_PAGE_NORMAL 0UL
1077 #define RB_PAGE_HEAD 1UL
1078 #define RB_PAGE_UPDATE 2UL
1079
1080
1081 #define RB_FLAG_MASK 3UL
1082
1083 /* PAGE_MOVED is not part of the mask */
1084 #define RB_PAGE_MOVED 4UL
1085
1086 /*
1087 * rb_list_head - remove any bit
1088 */
rb_list_head(struct list_head * list)1089 static struct list_head *rb_list_head(struct list_head *list)
1090 {
1091 unsigned long val = (unsigned long)list;
1092
1093 return (struct list_head *)(val & ~RB_FLAG_MASK);
1094 }
1095
1096 /*
1097 * rb_is_head_page - test if the given page is the head page
1098 *
1099 * Because the reader may move the head_page pointer, we can
1100 * not trust what the head page is (it may be pointing to
1101 * the reader page). But if the next page is a header page,
1102 * its flags will be non zero.
1103 */
1104 static inline int
rb_is_head_page(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * page,struct list_head * list)1105 rb_is_head_page(struct ring_buffer_per_cpu *cpu_buffer,
1106 struct buffer_page *page, struct list_head *list)
1107 {
1108 unsigned long val;
1109
1110 val = (unsigned long)list->next;
1111
1112 if ((val & ~RB_FLAG_MASK) != (unsigned long)&page->list)
1113 return RB_PAGE_MOVED;
1114
1115 return val & RB_FLAG_MASK;
1116 }
1117
1118 /*
1119 * rb_is_reader_page
1120 *
1121 * The unique thing about the reader page, is that, if the
1122 * writer is ever on it, the previous pointer never points
1123 * back to the reader page.
1124 */
rb_is_reader_page(struct buffer_page * page)1125 static bool rb_is_reader_page(struct buffer_page *page)
1126 {
1127 struct list_head *list = page->list.prev;
1128
1129 return rb_list_head(list->next) != &page->list;
1130 }
1131
1132 /*
1133 * rb_set_list_to_head - set a list_head to be pointing to head.
1134 */
rb_set_list_to_head(struct ring_buffer_per_cpu * cpu_buffer,struct list_head * list)1135 static void rb_set_list_to_head(struct ring_buffer_per_cpu *cpu_buffer,
1136 struct list_head *list)
1137 {
1138 unsigned long *ptr;
1139
1140 ptr = (unsigned long *)&list->next;
1141 *ptr |= RB_PAGE_HEAD;
1142 *ptr &= ~RB_PAGE_UPDATE;
1143 }
1144
1145 /*
1146 * rb_head_page_activate - sets up head page
1147 */
rb_head_page_activate(struct ring_buffer_per_cpu * cpu_buffer)1148 static void rb_head_page_activate(struct ring_buffer_per_cpu *cpu_buffer)
1149 {
1150 struct buffer_page *head;
1151
1152 head = cpu_buffer->head_page;
1153 if (!head)
1154 return;
1155
1156 /*
1157 * Set the previous list pointer to have the HEAD flag.
1158 */
1159 rb_set_list_to_head(cpu_buffer, head->list.prev);
1160 }
1161
rb_list_head_clear(struct list_head * list)1162 static void rb_list_head_clear(struct list_head *list)
1163 {
1164 unsigned long *ptr = (unsigned long *)&list->next;
1165
1166 *ptr &= ~RB_FLAG_MASK;
1167 }
1168
1169 /*
1170 * rb_head_page_deactivate - clears head page ptr (for free list)
1171 */
1172 static void
rb_head_page_deactivate(struct ring_buffer_per_cpu * cpu_buffer)1173 rb_head_page_deactivate(struct ring_buffer_per_cpu *cpu_buffer)
1174 {
1175 struct list_head *hd;
1176
1177 /* Go through the whole list and clear any pointers found. */
1178 rb_list_head_clear(cpu_buffer->pages);
1179
1180 list_for_each(hd, cpu_buffer->pages)
1181 rb_list_head_clear(hd);
1182 }
1183
rb_head_page_set(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * head,struct buffer_page * prev,int old_flag,int new_flag)1184 static int rb_head_page_set(struct ring_buffer_per_cpu *cpu_buffer,
1185 struct buffer_page *head,
1186 struct buffer_page *prev,
1187 int old_flag, int new_flag)
1188 {
1189 struct list_head *list;
1190 unsigned long val = (unsigned long)&head->list;
1191 unsigned long ret;
1192
1193 list = &prev->list;
1194
1195 val &= ~RB_FLAG_MASK;
1196
1197 ret = cmpxchg((unsigned long *)&list->next,
1198 val | old_flag, val | new_flag);
1199
1200 /* check if the reader took the page */
1201 if ((ret & ~RB_FLAG_MASK) != val)
1202 return RB_PAGE_MOVED;
1203
1204 return ret & RB_FLAG_MASK;
1205 }
1206
rb_head_page_set_update(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * head,struct buffer_page * prev,int old_flag)1207 static int rb_head_page_set_update(struct ring_buffer_per_cpu *cpu_buffer,
1208 struct buffer_page *head,
1209 struct buffer_page *prev,
1210 int old_flag)
1211 {
1212 return rb_head_page_set(cpu_buffer, head, prev,
1213 old_flag, RB_PAGE_UPDATE);
1214 }
1215
rb_head_page_set_head(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * head,struct buffer_page * prev,int old_flag)1216 static int rb_head_page_set_head(struct ring_buffer_per_cpu *cpu_buffer,
1217 struct buffer_page *head,
1218 struct buffer_page *prev,
1219 int old_flag)
1220 {
1221 return rb_head_page_set(cpu_buffer, head, prev,
1222 old_flag, RB_PAGE_HEAD);
1223 }
1224
rb_head_page_set_normal(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * head,struct buffer_page * prev,int old_flag)1225 static int rb_head_page_set_normal(struct ring_buffer_per_cpu *cpu_buffer,
1226 struct buffer_page *head,
1227 struct buffer_page *prev,
1228 int old_flag)
1229 {
1230 return rb_head_page_set(cpu_buffer, head, prev,
1231 old_flag, RB_PAGE_NORMAL);
1232 }
1233
rb_inc_page(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page ** bpage)1234 static inline void rb_inc_page(struct ring_buffer_per_cpu *cpu_buffer,
1235 struct buffer_page **bpage)
1236 {
1237 struct list_head *p = rb_list_head((*bpage)->list.next);
1238
1239 *bpage = list_entry(p, struct buffer_page, list);
1240 }
1241
1242 static struct buffer_page *
rb_set_head_page(struct ring_buffer_per_cpu * cpu_buffer)1243 rb_set_head_page(struct ring_buffer_per_cpu *cpu_buffer)
1244 {
1245 struct buffer_page *head;
1246 struct buffer_page *page;
1247 struct list_head *list;
1248 int i;
1249
1250 if (RB_WARN_ON(cpu_buffer, !cpu_buffer->head_page))
1251 return NULL;
1252
1253 /* sanity check */
1254 list = cpu_buffer->pages;
1255 if (RB_WARN_ON(cpu_buffer, rb_list_head(list->prev->next) != list))
1256 return NULL;
1257
1258 page = head = cpu_buffer->head_page;
1259 /*
1260 * It is possible that the writer moves the header behind
1261 * where we started, and we miss in one loop.
1262 * A second loop should grab the header, but we'll do
1263 * three loops just because I'm paranoid.
1264 */
1265 for (i = 0; i < 3; i++) {
1266 do {
1267 if (rb_is_head_page(cpu_buffer, page, page->list.prev)) {
1268 cpu_buffer->head_page = page;
1269 return page;
1270 }
1271 rb_inc_page(cpu_buffer, &page);
1272 } while (page != head);
1273 }
1274
1275 RB_WARN_ON(cpu_buffer, 1);
1276
1277 return NULL;
1278 }
1279
rb_head_page_replace(struct buffer_page * old,struct buffer_page * new)1280 static int rb_head_page_replace(struct buffer_page *old,
1281 struct buffer_page *new)
1282 {
1283 unsigned long *ptr = (unsigned long *)&old->list.prev->next;
1284 unsigned long val;
1285 unsigned long ret;
1286
1287 val = *ptr & ~RB_FLAG_MASK;
1288 val |= RB_PAGE_HEAD;
1289
1290 ret = cmpxchg(ptr, val, (unsigned long)&new->list);
1291
1292 return ret == val;
1293 }
1294
1295 /*
1296 * rb_tail_page_update - move the tail page forward
1297 */
rb_tail_page_update(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * tail_page,struct buffer_page * next_page)1298 static void rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer,
1299 struct buffer_page *tail_page,
1300 struct buffer_page *next_page)
1301 {
1302 unsigned long old_entries;
1303 unsigned long old_write;
1304
1305 /*
1306 * The tail page now needs to be moved forward.
1307 *
1308 * We need to reset the tail page, but without messing
1309 * with possible erasing of data brought in by interrupts
1310 * that have moved the tail page and are currently on it.
1311 *
1312 * We add a counter to the write field to denote this.
1313 */
1314 old_write = local_add_return(RB_WRITE_INTCNT, &next_page->write);
1315 old_entries = local_add_return(RB_WRITE_INTCNT, &next_page->entries);
1316
1317 local_inc(&cpu_buffer->pages_touched);
1318 /*
1319 * Just make sure we have seen our old_write and synchronize
1320 * with any interrupts that come in.
1321 */
1322 barrier();
1323
1324 /*
1325 * If the tail page is still the same as what we think
1326 * it is, then it is up to us to update the tail
1327 * pointer.
1328 */
1329 if (tail_page == READ_ONCE(cpu_buffer->tail_page)) {
1330 /* Zero the write counter */
1331 unsigned long val = old_write & ~RB_WRITE_MASK;
1332 unsigned long eval = old_entries & ~RB_WRITE_MASK;
1333
1334 /*
1335 * This will only succeed if an interrupt did
1336 * not come in and change it. In which case, we
1337 * do not want to modify it.
1338 *
1339 * We add (void) to let the compiler know that we do not care
1340 * about the return value of these functions. We use the
1341 * cmpxchg to only update if an interrupt did not already
1342 * do it for us. If the cmpxchg fails, we don't care.
1343 */
1344 (void)local_cmpxchg(&next_page->write, old_write, val);
1345 (void)local_cmpxchg(&next_page->entries, old_entries, eval);
1346
1347 /*
1348 * No need to worry about races with clearing out the commit.
1349 * it only can increment when a commit takes place. But that
1350 * only happens in the outer most nested commit.
1351 */
1352 local_set(&next_page->page->commit, 0);
1353
1354 /* Again, either we update tail_page or an interrupt does */
1355 (void)cmpxchg(&cpu_buffer->tail_page, tail_page, next_page);
1356 }
1357 }
1358
rb_check_bpage(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * bpage)1359 static int rb_check_bpage(struct ring_buffer_per_cpu *cpu_buffer,
1360 struct buffer_page *bpage)
1361 {
1362 unsigned long val = (unsigned long)bpage;
1363
1364 if (RB_WARN_ON(cpu_buffer, val & RB_FLAG_MASK))
1365 return 1;
1366
1367 return 0;
1368 }
1369
1370 /**
1371 * rb_check_list - make sure a pointer to a list has the last bits zero
1372 */
rb_check_list(struct ring_buffer_per_cpu * cpu_buffer,struct list_head * list)1373 static int rb_check_list(struct ring_buffer_per_cpu *cpu_buffer,
1374 struct list_head *list)
1375 {
1376 if (RB_WARN_ON(cpu_buffer, rb_list_head(list->prev) != list->prev))
1377 return 1;
1378 if (RB_WARN_ON(cpu_buffer, rb_list_head(list->next) != list->next))
1379 return 1;
1380 return 0;
1381 }
1382
1383 /**
1384 * rb_check_pages - integrity check of buffer pages
1385 * @cpu_buffer: CPU buffer with pages to test
1386 *
1387 * As a safety measure we check to make sure the data pages have not
1388 * been corrupted.
1389 */
rb_check_pages(struct ring_buffer_per_cpu * cpu_buffer)1390 static int rb_check_pages(struct ring_buffer_per_cpu *cpu_buffer)
1391 {
1392 struct list_head *head = cpu_buffer->pages;
1393 struct buffer_page *bpage, *tmp;
1394
1395 /* Reset the head page if it exists */
1396 if (cpu_buffer->head_page)
1397 rb_set_head_page(cpu_buffer);
1398
1399 rb_head_page_deactivate(cpu_buffer);
1400
1401 if (RB_WARN_ON(cpu_buffer, head->next->prev != head))
1402 return -1;
1403 if (RB_WARN_ON(cpu_buffer, head->prev->next != head))
1404 return -1;
1405
1406 if (rb_check_list(cpu_buffer, head))
1407 return -1;
1408
1409 list_for_each_entry_safe(bpage, tmp, head, list) {
1410 if (RB_WARN_ON(cpu_buffer,
1411 bpage->list.next->prev != &bpage->list))
1412 return -1;
1413 if (RB_WARN_ON(cpu_buffer,
1414 bpage->list.prev->next != &bpage->list))
1415 return -1;
1416 if (rb_check_list(cpu_buffer, &bpage->list))
1417 return -1;
1418 }
1419
1420 rb_head_page_activate(cpu_buffer);
1421
1422 return 0;
1423 }
1424
__rb_allocate_pages(long nr_pages,struct list_head * pages,int cpu)1425 static int __rb_allocate_pages(long nr_pages, struct list_head *pages, int cpu)
1426 {
1427 struct buffer_page *bpage, *tmp;
1428 bool user_thread = current->mm != NULL;
1429 gfp_t mflags;
1430 long i;
1431
1432 /*
1433 * Check if the available memory is there first.
1434 * Note, si_mem_available() only gives us a rough estimate of available
1435 * memory. It may not be accurate. But we don't care, we just want
1436 * to prevent doing any allocation when it is obvious that it is
1437 * not going to succeed.
1438 */
1439 i = si_mem_available();
1440 if (i < nr_pages)
1441 return -ENOMEM;
1442
1443 /*
1444 * __GFP_RETRY_MAYFAIL flag makes sure that the allocation fails
1445 * gracefully without invoking oom-killer and the system is not
1446 * destabilized.
1447 */
1448 mflags = GFP_KERNEL | __GFP_RETRY_MAYFAIL;
1449
1450 /*
1451 * If a user thread allocates too much, and si_mem_available()
1452 * reports there's enough memory, even though there is not.
1453 * Make sure the OOM killer kills this thread. This can happen
1454 * even with RETRY_MAYFAIL because another task may be doing
1455 * an allocation after this task has taken all memory.
1456 * This is the task the OOM killer needs to take out during this
1457 * loop, even if it was triggered by an allocation somewhere else.
1458 */
1459 if (user_thread)
1460 set_current_oom_origin();
1461 for (i = 0; i < nr_pages; i++) {
1462 struct page *page;
1463
1464 bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()),
1465 mflags, cpu_to_node(cpu));
1466 if (!bpage)
1467 goto free_pages;
1468
1469 list_add(&bpage->list, pages);
1470
1471 page = alloc_pages_node(cpu_to_node(cpu), mflags, 0);
1472 if (!page)
1473 goto free_pages;
1474 bpage->page = page_address(page);
1475 rb_init_page(bpage->page);
1476
1477 if (user_thread && fatal_signal_pending(current))
1478 goto free_pages;
1479 }
1480 if (user_thread)
1481 clear_current_oom_origin();
1482
1483 return 0;
1484
1485 free_pages:
1486 list_for_each_entry_safe(bpage, tmp, pages, list) {
1487 list_del_init(&bpage->list);
1488 free_buffer_page(bpage);
1489 }
1490 if (user_thread)
1491 clear_current_oom_origin();
1492
1493 return -ENOMEM;
1494 }
1495
rb_allocate_pages(struct ring_buffer_per_cpu * cpu_buffer,unsigned long nr_pages)1496 static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer,
1497 unsigned long nr_pages)
1498 {
1499 LIST_HEAD(pages);
1500
1501 WARN_ON(!nr_pages);
1502
1503 if (__rb_allocate_pages(nr_pages, &pages, cpu_buffer->cpu))
1504 return -ENOMEM;
1505
1506 /*
1507 * The ring buffer page list is a circular list that does not
1508 * start and end with a list head. All page list items point to
1509 * other pages.
1510 */
1511 cpu_buffer->pages = pages.next;
1512 list_del(&pages);
1513
1514 cpu_buffer->nr_pages = nr_pages;
1515
1516 rb_check_pages(cpu_buffer);
1517
1518 return 0;
1519 }
1520
1521 static struct ring_buffer_per_cpu *
rb_allocate_cpu_buffer(struct trace_buffer * buffer,long nr_pages,int cpu)1522 rb_allocate_cpu_buffer(struct trace_buffer *buffer, long nr_pages, int cpu)
1523 {
1524 struct ring_buffer_per_cpu *cpu_buffer;
1525 struct buffer_page *bpage;
1526 struct page *page;
1527 int ret;
1528
1529 cpu_buffer = kzalloc_node(ALIGN(sizeof(*cpu_buffer), cache_line_size()),
1530 GFP_KERNEL, cpu_to_node(cpu));
1531 if (!cpu_buffer)
1532 return NULL;
1533
1534 cpu_buffer->cpu = cpu;
1535 cpu_buffer->buffer = buffer;
1536 raw_spin_lock_init(&cpu_buffer->reader_lock);
1537 lockdep_set_class(&cpu_buffer->reader_lock, buffer->reader_lock_key);
1538 cpu_buffer->lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
1539 INIT_WORK(&cpu_buffer->update_pages_work, update_pages_handler);
1540 init_completion(&cpu_buffer->update_done);
1541 init_irq_work(&cpu_buffer->irq_work.work, rb_wake_up_waiters);
1542 init_waitqueue_head(&cpu_buffer->irq_work.waiters);
1543 init_waitqueue_head(&cpu_buffer->irq_work.full_waiters);
1544
1545 bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()),
1546 GFP_KERNEL, cpu_to_node(cpu));
1547 if (!bpage)
1548 goto fail_free_buffer;
1549
1550 rb_check_bpage(cpu_buffer, bpage);
1551
1552 cpu_buffer->reader_page = bpage;
1553 page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0);
1554 if (!page)
1555 goto fail_free_reader;
1556 bpage->page = page_address(page);
1557 rb_init_page(bpage->page);
1558
1559 INIT_LIST_HEAD(&cpu_buffer->reader_page->list);
1560 INIT_LIST_HEAD(&cpu_buffer->new_pages);
1561
1562 ret = rb_allocate_pages(cpu_buffer, nr_pages);
1563 if (ret < 0)
1564 goto fail_free_reader;
1565
1566 cpu_buffer->head_page
1567 = list_entry(cpu_buffer->pages, struct buffer_page, list);
1568 cpu_buffer->tail_page = cpu_buffer->commit_page = cpu_buffer->head_page;
1569
1570 rb_head_page_activate(cpu_buffer);
1571
1572 return cpu_buffer;
1573
1574 fail_free_reader:
1575 free_buffer_page(cpu_buffer->reader_page);
1576
1577 fail_free_buffer:
1578 kfree(cpu_buffer);
1579 return NULL;
1580 }
1581
rb_free_cpu_buffer(struct ring_buffer_per_cpu * cpu_buffer)1582 static void rb_free_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer)
1583 {
1584 struct list_head *head = cpu_buffer->pages;
1585 struct buffer_page *bpage, *tmp;
1586
1587 free_buffer_page(cpu_buffer->reader_page);
1588
1589 rb_head_page_deactivate(cpu_buffer);
1590
1591 if (head) {
1592 list_for_each_entry_safe(bpage, tmp, head, list) {
1593 list_del_init(&bpage->list);
1594 free_buffer_page(bpage);
1595 }
1596 bpage = list_entry(head, struct buffer_page, list);
1597 free_buffer_page(bpage);
1598 }
1599
1600 kfree(cpu_buffer);
1601 }
1602
1603 /**
1604 * __ring_buffer_alloc - allocate a new ring_buffer
1605 * @size: the size in bytes per cpu that is needed.
1606 * @flags: attributes to set for the ring buffer.
1607 * @key: ring buffer reader_lock_key.
1608 *
1609 * Currently the only flag that is available is the RB_FL_OVERWRITE
1610 * flag. This flag means that the buffer will overwrite old data
1611 * when the buffer wraps. If this flag is not set, the buffer will
1612 * drop data when the tail hits the head.
1613 */
__ring_buffer_alloc(unsigned long size,unsigned flags,struct lock_class_key * key)1614 struct trace_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags,
1615 struct lock_class_key *key)
1616 {
1617 struct trace_buffer *buffer;
1618 long nr_pages;
1619 int bsize;
1620 int cpu;
1621 int ret;
1622
1623 /* keep it in its own cache line */
1624 buffer = kzalloc(ALIGN(sizeof(*buffer), cache_line_size()),
1625 GFP_KERNEL);
1626 if (!buffer)
1627 return NULL;
1628
1629 if (!zalloc_cpumask_var(&buffer->cpumask, GFP_KERNEL))
1630 goto fail_free_buffer;
1631
1632 nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE);
1633 buffer->flags = flags;
1634 buffer->clock = trace_clock_local;
1635 buffer->reader_lock_key = key;
1636
1637 init_irq_work(&buffer->irq_work.work, rb_wake_up_waiters);
1638 init_waitqueue_head(&buffer->irq_work.waiters);
1639
1640 /* need at least two pages */
1641 if (nr_pages < 2)
1642 nr_pages = 2;
1643
1644 buffer->cpus = nr_cpu_ids;
1645
1646 bsize = sizeof(void *) * nr_cpu_ids;
1647 buffer->buffers = kzalloc(ALIGN(bsize, cache_line_size()),
1648 GFP_KERNEL);
1649 if (!buffer->buffers)
1650 goto fail_free_cpumask;
1651
1652 cpu = raw_smp_processor_id();
1653 cpumask_set_cpu(cpu, buffer->cpumask);
1654 buffer->buffers[cpu] = rb_allocate_cpu_buffer(buffer, nr_pages, cpu);
1655 if (!buffer->buffers[cpu])
1656 goto fail_free_buffers;
1657
1658 ret = cpuhp_state_add_instance(CPUHP_TRACE_RB_PREPARE, &buffer->node);
1659 if (ret < 0)
1660 goto fail_free_buffers;
1661
1662 mutex_init(&buffer->mutex);
1663
1664 return buffer;
1665
1666 fail_free_buffers:
1667 for_each_buffer_cpu(buffer, cpu) {
1668 if (buffer->buffers[cpu])
1669 rb_free_cpu_buffer(buffer->buffers[cpu]);
1670 }
1671 kfree(buffer->buffers);
1672
1673 fail_free_cpumask:
1674 free_cpumask_var(buffer->cpumask);
1675
1676 fail_free_buffer:
1677 kfree(buffer);
1678 return NULL;
1679 }
1680 EXPORT_SYMBOL_GPL(__ring_buffer_alloc);
1681
1682 /**
1683 * ring_buffer_free - free a ring buffer.
1684 * @buffer: the buffer to free.
1685 */
1686 void
ring_buffer_free(struct trace_buffer * buffer)1687 ring_buffer_free(struct trace_buffer *buffer)
1688 {
1689 int cpu;
1690
1691 cpuhp_state_remove_instance(CPUHP_TRACE_RB_PREPARE, &buffer->node);
1692
1693 for_each_buffer_cpu(buffer, cpu)
1694 rb_free_cpu_buffer(buffer->buffers[cpu]);
1695
1696 kfree(buffer->buffers);
1697 free_cpumask_var(buffer->cpumask);
1698
1699 kfree(buffer);
1700 }
1701 EXPORT_SYMBOL_GPL(ring_buffer_free);
1702
ring_buffer_set_clock(struct trace_buffer * buffer,u64 (* clock)(void))1703 void ring_buffer_set_clock(struct trace_buffer *buffer,
1704 u64 (*clock)(void))
1705 {
1706 buffer->clock = clock;
1707 }
1708
ring_buffer_set_time_stamp_abs(struct trace_buffer * buffer,bool abs)1709 void ring_buffer_set_time_stamp_abs(struct trace_buffer *buffer, bool abs)
1710 {
1711 buffer->time_stamp_abs = abs;
1712 }
1713
ring_buffer_time_stamp_abs(struct trace_buffer * buffer)1714 bool ring_buffer_time_stamp_abs(struct trace_buffer *buffer)
1715 {
1716 return buffer->time_stamp_abs;
1717 }
1718
1719 static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer);
1720
rb_page_entries(struct buffer_page * bpage)1721 static inline unsigned long rb_page_entries(struct buffer_page *bpage)
1722 {
1723 return local_read(&bpage->entries) & RB_WRITE_MASK;
1724 }
1725
rb_page_write(struct buffer_page * bpage)1726 static inline unsigned long rb_page_write(struct buffer_page *bpage)
1727 {
1728 return local_read(&bpage->write) & RB_WRITE_MASK;
1729 }
1730
1731 static int
rb_remove_pages(struct ring_buffer_per_cpu * cpu_buffer,unsigned long nr_pages)1732 rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned long nr_pages)
1733 {
1734 struct list_head *tail_page, *to_remove, *next_page;
1735 struct buffer_page *to_remove_page, *tmp_iter_page;
1736 struct buffer_page *last_page, *first_page;
1737 unsigned long nr_removed;
1738 unsigned long head_bit;
1739 int page_entries;
1740
1741 head_bit = 0;
1742
1743 raw_spin_lock_irq(&cpu_buffer->reader_lock);
1744 atomic_inc(&cpu_buffer->record_disabled);
1745 /*
1746 * We don't race with the readers since we have acquired the reader
1747 * lock. We also don't race with writers after disabling recording.
1748 * This makes it easy to figure out the first and the last page to be
1749 * removed from the list. We unlink all the pages in between including
1750 * the first and last pages. This is done in a busy loop so that we
1751 * lose the least number of traces.
1752 * The pages are freed after we restart recording and unlock readers.
1753 */
1754 tail_page = &cpu_buffer->tail_page->list;
1755
1756 /*
1757 * tail page might be on reader page, we remove the next page
1758 * from the ring buffer
1759 */
1760 if (cpu_buffer->tail_page == cpu_buffer->reader_page)
1761 tail_page = rb_list_head(tail_page->next);
1762 to_remove = tail_page;
1763
1764 /* start of pages to remove */
1765 first_page = list_entry(rb_list_head(to_remove->next),
1766 struct buffer_page, list);
1767
1768 for (nr_removed = 0; nr_removed < nr_pages; nr_removed++) {
1769 to_remove = rb_list_head(to_remove)->next;
1770 head_bit |= (unsigned long)to_remove & RB_PAGE_HEAD;
1771 }
1772
1773 next_page = rb_list_head(to_remove)->next;
1774
1775 /*
1776 * Now we remove all pages between tail_page and next_page.
1777 * Make sure that we have head_bit value preserved for the
1778 * next page
1779 */
1780 tail_page->next = (struct list_head *)((unsigned long)next_page |
1781 head_bit);
1782 next_page = rb_list_head(next_page);
1783 next_page->prev = tail_page;
1784
1785 /* make sure pages points to a valid page in the ring buffer */
1786 cpu_buffer->pages = next_page;
1787
1788 /* update head page */
1789 if (head_bit)
1790 cpu_buffer->head_page = list_entry(next_page,
1791 struct buffer_page, list);
1792
1793 /*
1794 * change read pointer to make sure any read iterators reset
1795 * themselves
1796 */
1797 cpu_buffer->read = 0;
1798
1799 /* pages are removed, resume tracing and then free the pages */
1800 atomic_dec(&cpu_buffer->record_disabled);
1801 raw_spin_unlock_irq(&cpu_buffer->reader_lock);
1802
1803 RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages));
1804
1805 /* last buffer page to remove */
1806 last_page = list_entry(rb_list_head(to_remove), struct buffer_page,
1807 list);
1808 tmp_iter_page = first_page;
1809
1810 do {
1811 cond_resched();
1812
1813 to_remove_page = tmp_iter_page;
1814 rb_inc_page(cpu_buffer, &tmp_iter_page);
1815
1816 /* update the counters */
1817 page_entries = rb_page_entries(to_remove_page);
1818 if (page_entries) {
1819 /*
1820 * If something was added to this page, it was full
1821 * since it is not the tail page. So we deduct the
1822 * bytes consumed in ring buffer from here.
1823 * Increment overrun to account for the lost events.
1824 */
1825 local_add(page_entries, &cpu_buffer->overrun);
1826 local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes);
1827 }
1828
1829 /*
1830 * We have already removed references to this list item, just
1831 * free up the buffer_page and its page
1832 */
1833 free_buffer_page(to_remove_page);
1834 nr_removed--;
1835
1836 } while (to_remove_page != last_page);
1837
1838 RB_WARN_ON(cpu_buffer, nr_removed);
1839
1840 return nr_removed == 0;
1841 }
1842
1843 static int
rb_insert_pages(struct ring_buffer_per_cpu * cpu_buffer)1844 rb_insert_pages(struct ring_buffer_per_cpu *cpu_buffer)
1845 {
1846 struct list_head *pages = &cpu_buffer->new_pages;
1847 int retries, success;
1848
1849 raw_spin_lock_irq(&cpu_buffer->reader_lock);
1850 /*
1851 * We are holding the reader lock, so the reader page won't be swapped
1852 * in the ring buffer. Now we are racing with the writer trying to
1853 * move head page and the tail page.
1854 * We are going to adapt the reader page update process where:
1855 * 1. We first splice the start and end of list of new pages between
1856 * the head page and its previous page.
1857 * 2. We cmpxchg the prev_page->next to point from head page to the
1858 * start of new pages list.
1859 * 3. Finally, we update the head->prev to the end of new list.
1860 *
1861 * We will try this process 10 times, to make sure that we don't keep
1862 * spinning.
1863 */
1864 retries = 10;
1865 success = 0;
1866 while (retries--) {
1867 struct list_head *head_page, *prev_page, *r;
1868 struct list_head *last_page, *first_page;
1869 struct list_head *head_page_with_bit;
1870
1871 head_page = &rb_set_head_page(cpu_buffer)->list;
1872 if (!head_page)
1873 break;
1874 prev_page = head_page->prev;
1875
1876 first_page = pages->next;
1877 last_page = pages->prev;
1878
1879 head_page_with_bit = (struct list_head *)
1880 ((unsigned long)head_page | RB_PAGE_HEAD);
1881
1882 last_page->next = head_page_with_bit;
1883 first_page->prev = prev_page;
1884
1885 r = cmpxchg(&prev_page->next, head_page_with_bit, first_page);
1886
1887 if (r == head_page_with_bit) {
1888 /*
1889 * yay, we replaced the page pointer to our new list,
1890 * now, we just have to update to head page's prev
1891 * pointer to point to end of list
1892 */
1893 head_page->prev = last_page;
1894 success = 1;
1895 break;
1896 }
1897 }
1898
1899 if (success)
1900 INIT_LIST_HEAD(pages);
1901 /*
1902 * If we weren't successful in adding in new pages, warn and stop
1903 * tracing
1904 */
1905 RB_WARN_ON(cpu_buffer, !success);
1906 raw_spin_unlock_irq(&cpu_buffer->reader_lock);
1907
1908 /* free pages if they weren't inserted */
1909 if (!success) {
1910 struct buffer_page *bpage, *tmp;
1911 list_for_each_entry_safe(bpage, tmp, &cpu_buffer->new_pages,
1912 list) {
1913 list_del_init(&bpage->list);
1914 free_buffer_page(bpage);
1915 }
1916 }
1917 return success;
1918 }
1919
rb_update_pages(struct ring_buffer_per_cpu * cpu_buffer)1920 static void rb_update_pages(struct ring_buffer_per_cpu *cpu_buffer)
1921 {
1922 int success;
1923
1924 if (cpu_buffer->nr_pages_to_update > 0)
1925 success = rb_insert_pages(cpu_buffer);
1926 else
1927 success = rb_remove_pages(cpu_buffer,
1928 -cpu_buffer->nr_pages_to_update);
1929
1930 if (success)
1931 cpu_buffer->nr_pages += cpu_buffer->nr_pages_to_update;
1932 }
1933
update_pages_handler(struct work_struct * work)1934 static void update_pages_handler(struct work_struct *work)
1935 {
1936 struct ring_buffer_per_cpu *cpu_buffer = container_of(work,
1937 struct ring_buffer_per_cpu, update_pages_work);
1938 rb_update_pages(cpu_buffer);
1939 complete(&cpu_buffer->update_done);
1940 }
1941
1942 /**
1943 * ring_buffer_resize - resize the ring buffer
1944 * @buffer: the buffer to resize.
1945 * @size: the new size.
1946 * @cpu_id: the cpu buffer to resize
1947 *
1948 * Minimum size is 2 * BUF_PAGE_SIZE.
1949 *
1950 * Returns 0 on success and < 0 on failure.
1951 */
ring_buffer_resize(struct trace_buffer * buffer,unsigned long size,int cpu_id)1952 int ring_buffer_resize(struct trace_buffer *buffer, unsigned long size,
1953 int cpu_id)
1954 {
1955 struct ring_buffer_per_cpu *cpu_buffer;
1956 unsigned long nr_pages;
1957 int cpu, err;
1958
1959 /*
1960 * Always succeed at resizing a non-existent buffer:
1961 */
1962 if (!buffer)
1963 return 0;
1964
1965 /* Make sure the requested buffer exists */
1966 if (cpu_id != RING_BUFFER_ALL_CPUS &&
1967 !cpumask_test_cpu(cpu_id, buffer->cpumask))
1968 return 0;
1969
1970 nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE);
1971
1972 /* we need a minimum of two pages */
1973 if (nr_pages < 2)
1974 nr_pages = 2;
1975
1976 size = nr_pages * BUF_PAGE_SIZE;
1977
1978 /* prevent another thread from changing buffer sizes */
1979 mutex_lock(&buffer->mutex);
1980
1981
1982 if (cpu_id == RING_BUFFER_ALL_CPUS) {
1983 /*
1984 * Don't succeed if resizing is disabled, as a reader might be
1985 * manipulating the ring buffer and is expecting a sane state while
1986 * this is true.
1987 */
1988 for_each_buffer_cpu(buffer, cpu) {
1989 cpu_buffer = buffer->buffers[cpu];
1990 if (atomic_read(&cpu_buffer->resize_disabled)) {
1991 err = -EBUSY;
1992 goto out_err_unlock;
1993 }
1994 }
1995
1996 /* calculate the pages to update */
1997 for_each_buffer_cpu(buffer, cpu) {
1998 cpu_buffer = buffer->buffers[cpu];
1999
2000 cpu_buffer->nr_pages_to_update = nr_pages -
2001 cpu_buffer->nr_pages;
2002 /*
2003 * nothing more to do for removing pages or no update
2004 */
2005 if (cpu_buffer->nr_pages_to_update <= 0)
2006 continue;
2007 /*
2008 * to add pages, make sure all new pages can be
2009 * allocated without receiving ENOMEM
2010 */
2011 INIT_LIST_HEAD(&cpu_buffer->new_pages);
2012 if (__rb_allocate_pages(cpu_buffer->nr_pages_to_update,
2013 &cpu_buffer->new_pages, cpu)) {
2014 /* not enough memory for new pages */
2015 err = -ENOMEM;
2016 goto out_err;
2017 }
2018 }
2019
2020 get_online_cpus();
2021 /*
2022 * Fire off all the required work handlers
2023 * We can't schedule on offline CPUs, but it's not necessary
2024 * since we can change their buffer sizes without any race.
2025 */
2026 for_each_buffer_cpu(buffer, cpu) {
2027 cpu_buffer = buffer->buffers[cpu];
2028 if (!cpu_buffer->nr_pages_to_update)
2029 continue;
2030
2031 /* Can't run something on an offline CPU. */
2032 if (!cpu_online(cpu)) {
2033 rb_update_pages(cpu_buffer);
2034 cpu_buffer->nr_pages_to_update = 0;
2035 } else {
2036 schedule_work_on(cpu,
2037 &cpu_buffer->update_pages_work);
2038 }
2039 }
2040
2041 /* wait for all the updates to complete */
2042 for_each_buffer_cpu(buffer, cpu) {
2043 cpu_buffer = buffer->buffers[cpu];
2044 if (!cpu_buffer->nr_pages_to_update)
2045 continue;
2046
2047 if (cpu_online(cpu))
2048 wait_for_completion(&cpu_buffer->update_done);
2049 cpu_buffer->nr_pages_to_update = 0;
2050 }
2051
2052 put_online_cpus();
2053 } else {
2054 /* Make sure this CPU has been initialized */
2055 if (!cpumask_test_cpu(cpu_id, buffer->cpumask))
2056 goto out;
2057
2058 cpu_buffer = buffer->buffers[cpu_id];
2059
2060 if (nr_pages == cpu_buffer->nr_pages)
2061 goto out;
2062
2063 /*
2064 * Don't succeed if resizing is disabled, as a reader might be
2065 * manipulating the ring buffer and is expecting a sane state while
2066 * this is true.
2067 */
2068 if (atomic_read(&cpu_buffer->resize_disabled)) {
2069 err = -EBUSY;
2070 goto out_err_unlock;
2071 }
2072
2073 cpu_buffer->nr_pages_to_update = nr_pages -
2074 cpu_buffer->nr_pages;
2075
2076 INIT_LIST_HEAD(&cpu_buffer->new_pages);
2077 if (cpu_buffer->nr_pages_to_update > 0 &&
2078 __rb_allocate_pages(cpu_buffer->nr_pages_to_update,
2079 &cpu_buffer->new_pages, cpu_id)) {
2080 err = -ENOMEM;
2081 goto out_err;
2082 }
2083
2084 get_online_cpus();
2085
2086 /* Can't run something on an offline CPU. */
2087 if (!cpu_online(cpu_id))
2088 rb_update_pages(cpu_buffer);
2089 else {
2090 schedule_work_on(cpu_id,
2091 &cpu_buffer->update_pages_work);
2092 wait_for_completion(&cpu_buffer->update_done);
2093 }
2094
2095 cpu_buffer->nr_pages_to_update = 0;
2096 put_online_cpus();
2097 }
2098
2099 out:
2100 /*
2101 * The ring buffer resize can happen with the ring buffer
2102 * enabled, so that the update disturbs the tracing as little
2103 * as possible. But if the buffer is disabled, we do not need
2104 * to worry about that, and we can take the time to verify
2105 * that the buffer is not corrupt.
2106 */
2107 if (atomic_read(&buffer->record_disabled)) {
2108 atomic_inc(&buffer->record_disabled);
2109 /*
2110 * Even though the buffer was disabled, we must make sure
2111 * that it is truly disabled before calling rb_check_pages.
2112 * There could have been a race between checking
2113 * record_disable and incrementing it.
2114 */
2115 synchronize_rcu();
2116 for_each_buffer_cpu(buffer, cpu) {
2117 cpu_buffer = buffer->buffers[cpu];
2118 rb_check_pages(cpu_buffer);
2119 }
2120 atomic_dec(&buffer->record_disabled);
2121 }
2122
2123 mutex_unlock(&buffer->mutex);
2124 return 0;
2125
2126 out_err:
2127 for_each_buffer_cpu(buffer, cpu) {
2128 struct buffer_page *bpage, *tmp;
2129
2130 cpu_buffer = buffer->buffers[cpu];
2131 cpu_buffer->nr_pages_to_update = 0;
2132
2133 if (list_empty(&cpu_buffer->new_pages))
2134 continue;
2135
2136 list_for_each_entry_safe(bpage, tmp, &cpu_buffer->new_pages,
2137 list) {
2138 list_del_init(&bpage->list);
2139 free_buffer_page(bpage);
2140 }
2141 }
2142 out_err_unlock:
2143 mutex_unlock(&buffer->mutex);
2144 return err;
2145 }
2146 EXPORT_SYMBOL_GPL(ring_buffer_resize);
2147
ring_buffer_change_overwrite(struct trace_buffer * buffer,int val)2148 void ring_buffer_change_overwrite(struct trace_buffer *buffer, int val)
2149 {
2150 mutex_lock(&buffer->mutex);
2151 if (val)
2152 buffer->flags |= RB_FL_OVERWRITE;
2153 else
2154 buffer->flags &= ~RB_FL_OVERWRITE;
2155 mutex_unlock(&buffer->mutex);
2156 }
2157 EXPORT_SYMBOL_GPL(ring_buffer_change_overwrite);
2158
__rb_page_index(struct buffer_page * bpage,unsigned index)2159 static __always_inline void *__rb_page_index(struct buffer_page *bpage, unsigned index)
2160 {
2161 return bpage->page->data + index;
2162 }
2163
2164 static __always_inline struct ring_buffer_event *
rb_reader_event(struct ring_buffer_per_cpu * cpu_buffer)2165 rb_reader_event(struct ring_buffer_per_cpu *cpu_buffer)
2166 {
2167 return __rb_page_index(cpu_buffer->reader_page,
2168 cpu_buffer->reader_page->read);
2169 }
2170
rb_page_commit(struct buffer_page * bpage)2171 static __always_inline unsigned rb_page_commit(struct buffer_page *bpage)
2172 {
2173 return local_read(&bpage->page->commit);
2174 }
2175
2176 static struct ring_buffer_event *
rb_iter_head_event(struct ring_buffer_iter * iter)2177 rb_iter_head_event(struct ring_buffer_iter *iter)
2178 {
2179 struct ring_buffer_event *event;
2180 struct buffer_page *iter_head_page = iter->head_page;
2181 unsigned long commit;
2182 unsigned length;
2183
2184 if (iter->head != iter->next_event)
2185 return iter->event;
2186
2187 /*
2188 * When the writer goes across pages, it issues a cmpxchg which
2189 * is a mb(), which will synchronize with the rmb here.
2190 * (see rb_tail_page_update() and __rb_reserve_next())
2191 */
2192 commit = rb_page_commit(iter_head_page);
2193 smp_rmb();
2194 event = __rb_page_index(iter_head_page, iter->head);
2195 length = rb_event_length(event);
2196
2197 /*
2198 * READ_ONCE() doesn't work on functions and we don't want the
2199 * compiler doing any crazy optimizations with length.
2200 */
2201 barrier();
2202
2203 if ((iter->head + length) > commit || length > BUF_MAX_DATA_SIZE)
2204 /* Writer corrupted the read? */
2205 goto reset;
2206
2207 memcpy(iter->event, event, length);
2208 /*
2209 * If the page stamp is still the same after this rmb() then the
2210 * event was safely copied without the writer entering the page.
2211 */
2212 smp_rmb();
2213
2214 /* Make sure the page didn't change since we read this */
2215 if (iter->page_stamp != iter_head_page->page->time_stamp ||
2216 commit > rb_page_commit(iter_head_page))
2217 goto reset;
2218
2219 iter->next_event = iter->head + length;
2220 return iter->event;
2221 reset:
2222 /* Reset to the beginning */
2223 iter->page_stamp = iter->read_stamp = iter->head_page->page->time_stamp;
2224 iter->head = 0;
2225 iter->next_event = 0;
2226 iter->missed_events = 1;
2227 return NULL;
2228 }
2229
2230 /* Size is determined by what has been committed */
rb_page_size(struct buffer_page * bpage)2231 static __always_inline unsigned rb_page_size(struct buffer_page *bpage)
2232 {
2233 return rb_page_commit(bpage);
2234 }
2235
2236 static __always_inline unsigned
rb_commit_index(struct ring_buffer_per_cpu * cpu_buffer)2237 rb_commit_index(struct ring_buffer_per_cpu *cpu_buffer)
2238 {
2239 return rb_page_commit(cpu_buffer->commit_page);
2240 }
2241
2242 static __always_inline unsigned
rb_event_index(struct ring_buffer_event * event)2243 rb_event_index(struct ring_buffer_event *event)
2244 {
2245 unsigned long addr = (unsigned long)event;
2246
2247 return (addr & ~PAGE_MASK) - BUF_PAGE_HDR_SIZE;
2248 }
2249
rb_inc_iter(struct ring_buffer_iter * iter)2250 static void rb_inc_iter(struct ring_buffer_iter *iter)
2251 {
2252 struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
2253
2254 /*
2255 * The iterator could be on the reader page (it starts there).
2256 * But the head could have moved, since the reader was
2257 * found. Check for this case and assign the iterator
2258 * to the head page instead of next.
2259 */
2260 if (iter->head_page == cpu_buffer->reader_page)
2261 iter->head_page = rb_set_head_page(cpu_buffer);
2262 else
2263 rb_inc_page(cpu_buffer, &iter->head_page);
2264
2265 iter->page_stamp = iter->read_stamp = iter->head_page->page->time_stamp;
2266 iter->head = 0;
2267 iter->next_event = 0;
2268 }
2269
2270 /*
2271 * rb_handle_head_page - writer hit the head page
2272 *
2273 * Returns: +1 to retry page
2274 * 0 to continue
2275 * -1 on error
2276 */
2277 static int
rb_handle_head_page(struct ring_buffer_per_cpu * cpu_buffer,struct buffer_page * tail_page,struct buffer_page * next_page)2278 rb_handle_head_page(struct ring_buffer_per_cpu *cpu_buffer,
2279 struct buffer_page *tail_page,
2280 struct buffer_page *next_page)
2281 {
2282 struct buffer_page *new_head;
2283 int entries;
2284 int type;
2285 int ret;
2286
2287 entries = rb_page_entries(next_page);
2288
2289 /*
2290 * The hard part is here. We need to move the head
2291 * forward, and protect against both readers on
2292 * other CPUs and writers coming in via interrupts.
2293 */
2294 type = rb_head_page_set_update(cpu_buffer, next_page, tail_page,
2295 RB_PAGE_HEAD);
2296
2297 /*
2298 * type can be one of four:
2299 * NORMAL - an interrupt already moved it for us
2300 * HEAD - we are the first to get here.
2301 * UPDATE - we are the interrupt interrupting
2302 * a current move.
2303 * MOVED - a reader on another CPU moved the next
2304 * pointer to its reader page. Give up
2305 * and try again.
2306 */
2307
2308 switch (type) {
2309 case RB_PAGE_HEAD:
2310 /*
2311 * We changed the head to UPDATE, thus
2312 * it is our responsibility to update
2313 * the counters.
2314 */
2315 local_add(entries, &cpu_buffer->overrun);
2316 local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes);
2317
2318 /*
2319 * The entries will be zeroed out when we move the
2320 * tail page.
2321 */
2322
2323 /* still more to do */
2324 break;
2325
2326 case RB_PAGE_UPDATE:
2327 /*
2328 * This is an interrupt that interrupt the
2329 * previous update. Still more to do.
2330 */
2331 break;
2332 case RB_PAGE_NORMAL:
2333 /*
2334 * An interrupt came in before the update
2335 * and processed this for us.
2336 * Nothing left to do.
2337 */
2338 return 1;
2339 case RB_PAGE_MOVED:
2340 /*
2341 * The reader is on another CPU and just did
2342 * a swap with our next_page.
2343 * Try again.
2344 */
2345 return 1;
2346 default:
2347 RB_WARN_ON(cpu_buffer, 1); /* WTF??? */
2348 return -1;
2349 }
2350
2351 /*
2352 * Now that we are here, the old head pointer is
2353 * set to UPDATE. This will keep the reader from
2354 * swapping the head page with the reader page.
2355 * The reader (on another CPU) will spin till
2356 * we are finished.
2357 *
2358 * We just need to protect against interrupts
2359 * doing the job. We will set the next pointer
2360 * to HEAD. After that, we set the old pointer
2361 * to NORMAL, but only if it was HEAD before.
2362 * otherwise we are an interrupt, and only
2363 * want the outer most commit to reset it.
2364 */
2365 new_head = next_page;
2366 rb_inc_page(cpu_buffer, &new_head);
2367
2368 ret = rb_head_page_set_head(cpu_buffer, new_head, next_page,
2369 RB_PAGE_NORMAL);
2370
2371 /*
2372 * Valid returns are:
2373 * HEAD - an interrupt came in and already set it.
2374 * NORMAL - One of two things:
2375 * 1) We really set it.
2376 * 2) A bunch of interrupts came in and moved
2377 * the page forward again.
2378 */
2379 switch (ret) {
2380 case RB_PAGE_HEAD:
2381 case RB_PAGE_NORMAL:
2382 /* OK */
2383 break;
2384 default:
2385 RB_WARN_ON(cpu_buffer, 1);
2386 return -1;
2387 }
2388
2389 /*
2390 * It is possible that an interrupt came in,
2391 * set the head up, then more interrupts came in
2392 * and moved it again. When we get back here,
2393 * the page would have been set to NORMAL but we
2394 * just set it back to HEAD.
2395 *
2396 * How do you detect this? Well, if that happened
2397 * the tail page would have moved.
2398 */
2399 if (ret == RB_PAGE_NORMAL) {
2400 struct buffer_page *buffer_tail_page;
2401
2402 buffer_tail_page = READ_ONCE(cpu_buffer->tail_page);
2403 /*
2404 * If the tail had moved passed next, then we need
2405 * to reset the pointer.
2406 */
2407 if (buffer_tail_page != tail_page &&
2408 buffer_tail_page != next_page)
2409 rb_head_page_set_normal(cpu_buffer, new_head,
2410 next_page,
2411 RB_PAGE_HEAD);
2412 }
2413
2414 /*
2415 * If this was the outer most commit (the one that
2416 * changed the original pointer from HEAD to UPDATE),
2417 * then it is up to us to reset it to NORMAL.
2418 */
2419 if (type == RB_PAGE_HEAD) {
2420 ret = rb_head_page_set_normal(cpu_buffer, next_page,
2421 tail_page,
2422 RB_PAGE_UPDATE);
2423 if (RB_WARN_ON(cpu_buffer,
2424 ret != RB_PAGE_UPDATE))
2425 return -1;
2426 }
2427
2428 return 0;
2429 }
2430
2431 static inline void
rb_reset_tail(struct ring_buffer_per_cpu * cpu_buffer,unsigned long tail,struct rb_event_info * info)2432 rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer,
2433 unsigned long tail, struct rb_event_info *info)
2434 {
2435 struct buffer_page *tail_page = info->tail_page;
2436 struct ring_buffer_event *event;
2437 unsigned long length = info->length;
2438
2439 /*
2440 * Only the event that crossed the page boundary
2441 * must fill the old tail_page with padding.
2442 */
2443 if (tail >= BUF_PAGE_SIZE) {
2444 /*
2445 * If the page was filled, then we still need
2446 * to update the real_end. Reset it to zero
2447 * and the reader will ignore it.
2448 */
2449 if (tail == BUF_PAGE_SIZE)
2450 tail_page->real_end = 0;
2451
2452 local_sub(length, &tail_page->write);
2453 return;
2454 }
2455
2456 event = __rb_page_index(tail_page, tail);
2457
2458 /* account for padding bytes */
2459 local_add(BUF_PAGE_SIZE - tail, &cpu_buffer->entries_bytes);
2460
2461 /*
2462 * Save the original length to the meta data.
2463 * This will be used by the reader to add lost event
2464 * counter.
2465 */
2466 tail_page->real_end = tail;
2467
2468 /*
2469 * If this event is bigger than the minimum size, then
2470 * we need to be careful that we don't subtract the
2471 * write counter enough to allow another writer to slip
2472 * in on this page.
2473 * We put in a discarded commit instead, to make sure
2474 * that this space is not used again.
2475 *
2476 * If we are less than the minimum size, we don't need to
2477 * worry about it.
2478 */
2479 if (tail > (BUF_PAGE_SIZE - RB_EVNT_MIN_SIZE)) {
2480 /* No room for any events */
2481
2482 /* Mark the rest of the page with padding */
2483 rb_event_set_padding(event);
2484
2485 /* Set the write back to the previous setting */
2486 local_sub(length, &tail_page->write);
2487 return;
2488 }
2489
2490 /* Put in a discarded event */
2491 event->array[0] = (BUF_PAGE_SIZE - tail) - RB_EVNT_HDR_SIZE;
2492 event->type_len = RINGBUF_TYPE_PADDING;
2493 /* time delta must be non zero */
2494 event->time_delta = 1;
2495
2496 /* Set write to end of buffer */
2497 length = (tail + length) - BUF_PAGE_SIZE;
2498 local_sub(length, &tail_page->write);
2499 }
2500
2501 static inline void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer);
2502
2503 /*
2504 * This is the slow path, force gcc not to inline it.
2505 */
2506 static noinline struct ring_buffer_event *
rb_move_tail(struct ring_buffer_per_cpu * cpu_buffer,unsigned long tail,struct rb_event_info * info)2507 rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer,
2508 unsigned long tail, struct rb_event_info *info)
2509 {
2510 struct buffer_page *tail_page = info->tail_page;
2511 struct buffer_page *commit_page = cpu_buffer->commit_page;
2512 struct trace_buffer *buffer = cpu_buffer->buffer;
2513 struct buffer_page *next_page;
2514 int ret;
2515
2516 next_page = tail_page;
2517
2518 rb_inc_page(cpu_buffer, &next_page);
2519
2520 /*
2521 * If for some reason, we had an interrupt storm that made
2522 * it all the way around the buffer, bail, and warn
2523 * about it.
2524 */
2525 if (unlikely(next_page == commit_page)) {
2526 local_inc(&cpu_buffer->commit_overrun);
2527 goto out_reset;
2528 }
2529
2530 /*
2531 * This is where the fun begins!
2532 *
2533 * We are fighting against races between a reader that
2534 * could be on another CPU trying to swap its reader
2535 * page with the buffer head.
2536 *
2537 * We are also fighting against interrupts coming in and
2538 * moving the head or tail on us as well.
2539 *
2540 * If the next page is the head page then we have filled
2541 * the buffer, unless the commit page is still on the
2542 * reader page.
2543 */
2544 if (rb_is_head_page(cpu_buffer, next_page, &tail_page->list)) {
2545
2546 /*
2547 * If the commit is not on the reader page, then
2548 * move the header page.
2549 */
2550 if (!rb_is_reader_page(cpu_buffer->commit_page)) {
2551 /*
2552 * If we are not in overwrite mode,
2553 * this is easy, just stop here.
2554 */
2555 if (!(buffer->flags & RB_FL_OVERWRITE)) {
2556 local_inc(&cpu_buffer->dropped_events);
2557 goto out_reset;
2558 }
2559
2560 ret = rb_handle_head_page(cpu_buffer,
2561 tail_page,
2562 next_page);
2563 if (ret < 0)
2564 goto out_reset;
2565 if (ret)
2566 goto out_again;
2567 } else {
2568 /*
2569 * We need to be careful here too. The
2570 * commit page could still be on the reader
2571 * page. We could have a small buffer, and
2572 * have filled up the buffer with events
2573 * from interrupts and such, and wrapped.
2574 *
2575 * Note, if the tail page is also the on the
2576 * reader_page, we let it move out.
2577 */
2578 if (unlikely((cpu_buffer->commit_page !=
2579 cpu_buffer->tail_page) &&
2580 (cpu_buffer->commit_page ==
2581 cpu_buffer->reader_page))) {
2582 local_inc(&cpu_buffer->commit_overrun);
2583 goto out_reset;
2584 }
2585 }
2586 }
2587
2588 rb_tail_page_update(cpu_buffer, tail_page, next_page);
2589
2590 out_again:
2591
2592 rb_reset_tail(cpu_buffer, tail, info);
2593
2594 /* Commit what we have for now. */
2595 rb_end_commit(cpu_buffer);
2596 /* rb_end_commit() decs committing */
2597 local_inc(&cpu_buffer->committing);
2598
2599 /* fail and let the caller try again */
2600 return ERR_PTR(-EAGAIN);
2601
2602 out_reset:
2603 /* reset write */
2604 rb_reset_tail(cpu_buffer, tail, info);
2605
2606 return NULL;
2607 }
2608
2609 /* Slow path */
2610 static struct ring_buffer_event *
rb_add_time_stamp(struct ring_buffer_event * event,u64 delta,bool abs)2611 rb_add_time_stamp(struct ring_buffer_event *event, u64 delta, bool abs)
2612 {
2613 if (abs)
2614 event->type_len = RINGBUF_TYPE_TIME_STAMP;
2615 else
2616 event->type_len = RINGBUF_TYPE_TIME_EXTEND;
2617
2618 /* Not the first event on the page, or not delta? */
2619 if (abs || rb_event_index(event)) {
2620 event->time_delta = delta & TS_MASK;
2621 event->array[0] = delta >> TS_SHIFT;
2622 } else {
2623 /* nope, just zero it */
2624 event->time_delta = 0;
2625 event->array[0] = 0;
2626 }
2627
2628 return skip_time_extend(event);
2629 }
2630
2631 static inline bool rb_event_is_commit(struct ring_buffer_per_cpu *cpu_buffer,
2632 struct ring_buffer_event *event);
2633
2634 #ifndef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
sched_clock_stable(void)2635 static inline bool sched_clock_stable(void)
2636 {
2637 return true;
2638 }
2639 #endif
2640
2641 static void
rb_check_timestamp(struct ring_buffer_per_cpu * cpu_buffer,struct rb_event_info * info)2642 rb_check_timestamp(struct ring_buffer_per_cpu *cpu_buffer,
2643 struct rb_event_info *info)
2644 {
2645 u64 write_stamp;
2646
2647 WARN_ONCE(1, "Delta way too big! %llu ts=%llu before=%llu after=%llu write stamp=%llu\n%s",
2648 (unsigned long long)info->delta,
2649 (unsigned long long)info->ts,
2650 (unsigned long long)info->before,
2651 (unsigned long long)info->after,
2652 (unsigned long long)(rb_time_read(&cpu_buffer->write_stamp, &write_stamp) ? write_stamp : 0),
2653 sched_clock_stable() ? "" :
2654 "If you just came from a suspend/resume,\n"
2655 "please switch to the trace global clock:\n"
2656 " echo global > /sys/kernel/debug/tracing/trace_clock\n"
2657 "or add trace_clock=global to the kernel command line\n");
2658 }
2659
rb_add_timestamp(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event ** event,struct rb_event_info * info,u64 * delta,unsigned int * length)2660 static void rb_add_timestamp(struct ring_buffer_per_cpu *cpu_buffer,
2661 struct ring_buffer_event **event,
2662 struct rb_event_info *info,
2663 u64 *delta,
2664 unsigned int *length)
2665 {
2666 bool abs = info->add_timestamp &
2667 (RB_ADD_STAMP_FORCE | RB_ADD_STAMP_ABSOLUTE);
2668
2669 if (unlikely(info->delta > (1ULL << 59))) {
2670 /* did the clock go backwards */
2671 if (info->before == info->after && info->before > info->ts) {
2672 /* not interrupted */
2673 static int once;
2674
2675 /*
2676 * This is possible with a recalibrating of the TSC.
2677 * Do not produce a call stack, but just report it.
2678 */
2679 if (!once) {
2680 once++;
2681 pr_warn("Ring buffer clock went backwards: %llu -> %llu\n",
2682 info->before, info->ts);
2683 }
2684 } else
2685 rb_check_timestamp(cpu_buffer, info);
2686 if (!abs)
2687 info->delta = 0;
2688 }
2689 *event = rb_add_time_stamp(*event, info->delta, abs);
2690 *length -= RB_LEN_TIME_EXTEND;
2691 *delta = 0;
2692 }
2693
2694 /**
2695 * rb_update_event - update event type and data
2696 * @cpu_buffer: The per cpu buffer of the @event
2697 * @event: the event to update
2698 * @info: The info to update the @event with (contains length and delta)
2699 *
2700 * Update the type and data fields of the @event. The length
2701 * is the actual size that is written to the ring buffer,
2702 * and with this, we can determine what to place into the
2703 * data field.
2704 */
2705 static void
rb_update_event(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event * event,struct rb_event_info * info)2706 rb_update_event(struct ring_buffer_per_cpu *cpu_buffer,
2707 struct ring_buffer_event *event,
2708 struct rb_event_info *info)
2709 {
2710 unsigned length = info->length;
2711 u64 delta = info->delta;
2712
2713 /*
2714 * If we need to add a timestamp, then we
2715 * add it to the start of the reserved space.
2716 */
2717 if (unlikely(info->add_timestamp))
2718 rb_add_timestamp(cpu_buffer, &event, info, &delta, &length);
2719
2720 event->time_delta = delta;
2721 length -= RB_EVNT_HDR_SIZE;
2722 if (length > RB_MAX_SMALL_DATA) {
2723 event->type_len = 0;
2724 event->array[0] = length;
2725 } else
2726 event->type_len = DIV_ROUND_UP(length, RB_ALIGNMENT);
2727 }
2728
rb_calculate_event_length(unsigned length)2729 static unsigned rb_calculate_event_length(unsigned length)
2730 {
2731 struct ring_buffer_event event; /* Used only for sizeof array */
2732
2733 /* zero length can cause confusions */
2734 if (!length)
2735 length++;
2736
2737 if (length > RB_MAX_SMALL_DATA)
2738 length += sizeof(event.array[0]);
2739
2740 length += RB_EVNT_HDR_SIZE;
2741 length = ALIGN(length, RB_ALIGNMENT);
2742
2743 /*
2744 * In case the time delta is larger than the 27 bits for it
2745 * in the header, we need to add a timestamp. If another
2746 * event comes in when trying to discard this one to increase
2747 * the length, then the timestamp will be added in the allocated
2748 * space of this event. If length is bigger than the size needed
2749 * for the TIME_EXTEND, then padding has to be used. The events
2750 * length must be either RB_LEN_TIME_EXTEND, or greater than or equal
2751 * to RB_LEN_TIME_EXTEND + 8, as 8 is the minimum size for padding.
2752 * As length is a multiple of 4, we only need to worry if it
2753 * is 12 (RB_LEN_TIME_EXTEND + 4).
2754 */
2755 if (length == RB_LEN_TIME_EXTEND + RB_ALIGNMENT)
2756 length += RB_ALIGNMENT;
2757
2758 return length;
2759 }
2760
2761 static __always_inline bool
rb_event_is_commit(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event * event)2762 rb_event_is_commit(struct ring_buffer_per_cpu *cpu_buffer,
2763 struct ring_buffer_event *event)
2764 {
2765 unsigned long addr = (unsigned long)event;
2766 unsigned long index;
2767
2768 index = rb_event_index(event);
2769 addr &= PAGE_MASK;
2770
2771 return cpu_buffer->commit_page->page == (void *)addr &&
2772 rb_commit_index(cpu_buffer) == index;
2773 }
2774
rb_time_delta(struct ring_buffer_event * event)2775 static u64 rb_time_delta(struct ring_buffer_event *event)
2776 {
2777 switch (event->type_len) {
2778 case RINGBUF_TYPE_PADDING:
2779 return 0;
2780
2781 case RINGBUF_TYPE_TIME_EXTEND:
2782 return ring_buffer_event_time_stamp(event);
2783
2784 case RINGBUF_TYPE_TIME_STAMP:
2785 return 0;
2786
2787 case RINGBUF_TYPE_DATA:
2788 return event->time_delta;
2789 default:
2790 return 0;
2791 }
2792 }
2793
2794 static inline int
rb_try_to_discard(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event * event)2795 rb_try_to_discard(struct ring_buffer_per_cpu *cpu_buffer,
2796 struct ring_buffer_event *event)
2797 {
2798 unsigned long new_index, old_index;
2799 struct buffer_page *bpage;
2800 unsigned long index;
2801 unsigned long addr;
2802 u64 write_stamp;
2803 u64 delta;
2804
2805 new_index = rb_event_index(event);
2806 old_index = new_index + rb_event_ts_length(event);
2807 addr = (unsigned long)event;
2808 addr &= PAGE_MASK;
2809
2810 bpage = READ_ONCE(cpu_buffer->tail_page);
2811
2812 delta = rb_time_delta(event);
2813
2814 if (!rb_time_read(&cpu_buffer->write_stamp, &write_stamp))
2815 return 0;
2816
2817 /* Make sure the write stamp is read before testing the location */
2818 barrier();
2819
2820 if (bpage->page == (void *)addr && rb_page_write(bpage) == old_index) {
2821 unsigned long write_mask =
2822 local_read(&bpage->write) & ~RB_WRITE_MASK;
2823 unsigned long event_length = rb_event_length(event);
2824
2825 /* Something came in, can't discard */
2826 if (!rb_time_cmpxchg(&cpu_buffer->write_stamp,
2827 write_stamp, write_stamp - delta))
2828 return 0;
2829
2830 /*
2831 * If an event were to come in now, it would see that the
2832 * write_stamp and the before_stamp are different, and assume
2833 * that this event just added itself before updating
2834 * the write stamp. The interrupting event will fix the
2835 * write stamp for us, and use the before stamp as its delta.
2836 */
2837
2838 /*
2839 * This is on the tail page. It is possible that
2840 * a write could come in and move the tail page
2841 * and write to the next page. That is fine
2842 * because we just shorten what is on this page.
2843 */
2844 old_index += write_mask;
2845 new_index += write_mask;
2846 index = local_cmpxchg(&bpage->write, old_index, new_index);
2847 if (index == old_index) {
2848 /* update counters */
2849 local_sub(event_length, &cpu_buffer->entries_bytes);
2850 return 1;
2851 }
2852 }
2853
2854 /* could not discard */
2855 return 0;
2856 }
2857
rb_start_commit(struct ring_buffer_per_cpu * cpu_buffer)2858 static void rb_start_commit(struct ring_buffer_per_cpu *cpu_buffer)
2859 {
2860 local_inc(&cpu_buffer->committing);
2861 local_inc(&cpu_buffer->commits);
2862 }
2863
2864 static __always_inline void
rb_set_commit_to_write(struct ring_buffer_per_cpu * cpu_buffer)2865 rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer)
2866 {
2867 unsigned long max_count;
2868
2869 /*
2870 * We only race with interrupts and NMIs on this CPU.
2871 * If we own the commit event, then we can commit
2872 * all others that interrupted us, since the interruptions
2873 * are in stack format (they finish before they come
2874 * back to us). This allows us to do a simple loop to
2875 * assign the commit to the tail.
2876 */
2877 again:
2878 max_count = cpu_buffer->nr_pages * 100;
2879
2880 while (cpu_buffer->commit_page != READ_ONCE(cpu_buffer->tail_page)) {
2881 if (RB_WARN_ON(cpu_buffer, !(--max_count)))
2882 return;
2883 if (RB_WARN_ON(cpu_buffer,
2884 rb_is_reader_page(cpu_buffer->tail_page)))
2885 return;
2886 local_set(&cpu_buffer->commit_page->page->commit,
2887 rb_page_write(cpu_buffer->commit_page));
2888 rb_inc_page(cpu_buffer, &cpu_buffer->commit_page);
2889 /* add barrier to keep gcc from optimizing too much */
2890 barrier();
2891 }
2892 while (rb_commit_index(cpu_buffer) !=
2893 rb_page_write(cpu_buffer->commit_page)) {
2894
2895 local_set(&cpu_buffer->commit_page->page->commit,
2896 rb_page_write(cpu_buffer->commit_page));
2897 RB_WARN_ON(cpu_buffer,
2898 local_read(&cpu_buffer->commit_page->page->commit) &
2899 ~RB_WRITE_MASK);
2900 barrier();
2901 }
2902
2903 /* again, keep gcc from optimizing */
2904 barrier();
2905
2906 /*
2907 * If an interrupt came in just after the first while loop
2908 * and pushed the tail page forward, we will be left with
2909 * a dangling commit that will never go forward.
2910 */
2911 if (unlikely(cpu_buffer->commit_page != READ_ONCE(cpu_buffer->tail_page)))
2912 goto again;
2913 }
2914
rb_end_commit(struct ring_buffer_per_cpu * cpu_buffer)2915 static __always_inline void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer)
2916 {
2917 unsigned long commits;
2918
2919 if (RB_WARN_ON(cpu_buffer,
2920 !local_read(&cpu_buffer->committing)))
2921 return;
2922
2923 again:
2924 commits = local_read(&cpu_buffer->commits);
2925 /* synchronize with interrupts */
2926 barrier();
2927 if (local_read(&cpu_buffer->committing) == 1)
2928 rb_set_commit_to_write(cpu_buffer);
2929
2930 local_dec(&cpu_buffer->committing);
2931
2932 /* synchronize with interrupts */
2933 barrier();
2934
2935 /*
2936 * Need to account for interrupts coming in between the
2937 * updating of the commit page and the clearing of the
2938 * committing counter.
2939 */
2940 if (unlikely(local_read(&cpu_buffer->commits) != commits) &&
2941 !local_read(&cpu_buffer->committing)) {
2942 local_inc(&cpu_buffer->committing);
2943 goto again;
2944 }
2945 }
2946
rb_event_discard(struct ring_buffer_event * event)2947 static inline void rb_event_discard(struct ring_buffer_event *event)
2948 {
2949 if (extended_time(event))
2950 event = skip_time_extend(event);
2951
2952 /* array[0] holds the actual length for the discarded event */
2953 event->array[0] = rb_event_data_length(event) - RB_EVNT_HDR_SIZE;
2954 event->type_len = RINGBUF_TYPE_PADDING;
2955 /* time delta must be non zero */
2956 if (!event->time_delta)
2957 event->time_delta = 1;
2958 }
2959
rb_commit(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event * event)2960 static void rb_commit(struct ring_buffer_per_cpu *cpu_buffer,
2961 struct ring_buffer_event *event)
2962 {
2963 local_inc(&cpu_buffer->entries);
2964 rb_end_commit(cpu_buffer);
2965 }
2966
2967 static __always_inline void
rb_wakeups(struct trace_buffer * buffer,struct ring_buffer_per_cpu * cpu_buffer)2968 rb_wakeups(struct trace_buffer *buffer, struct ring_buffer_per_cpu *cpu_buffer)
2969 {
2970 size_t nr_pages;
2971 size_t dirty;
2972 size_t full;
2973
2974 if (buffer->irq_work.waiters_pending) {
2975 buffer->irq_work.waiters_pending = false;
2976 /* irq_work_queue() supplies it's own memory barriers */
2977 irq_work_queue(&buffer->irq_work.work);
2978 }
2979
2980 if (cpu_buffer->irq_work.waiters_pending) {
2981 cpu_buffer->irq_work.waiters_pending = false;
2982 /* irq_work_queue() supplies it's own memory barriers */
2983 irq_work_queue(&cpu_buffer->irq_work.work);
2984 }
2985
2986 if (cpu_buffer->last_pages_touch == local_read(&cpu_buffer->pages_touched))
2987 return;
2988
2989 if (cpu_buffer->reader_page == cpu_buffer->commit_page)
2990 return;
2991
2992 if (!cpu_buffer->irq_work.full_waiters_pending)
2993 return;
2994
2995 cpu_buffer->last_pages_touch = local_read(&cpu_buffer->pages_touched);
2996
2997 full = cpu_buffer->shortest_full;
2998 nr_pages = cpu_buffer->nr_pages;
2999 dirty = ring_buffer_nr_dirty_pages(buffer, cpu_buffer->cpu);
3000 if (full && nr_pages && (dirty * 100) <= full * nr_pages)
3001 return;
3002
3003 cpu_buffer->irq_work.wakeup_full = true;
3004 cpu_buffer->irq_work.full_waiters_pending = false;
3005 /* irq_work_queue() supplies it's own memory barriers */
3006 irq_work_queue(&cpu_buffer->irq_work.work);
3007 }
3008
3009 /*
3010 * The lock and unlock are done within a preempt disable section.
3011 * The current_context per_cpu variable can only be modified
3012 * by the current task between lock and unlock. But it can
3013 * be modified more than once via an interrupt. To pass this
3014 * information from the lock to the unlock without having to
3015 * access the 'in_interrupt()' functions again (which do show
3016 * a bit of overhead in something as critical as function tracing,
3017 * we use a bitmask trick.
3018 *
3019 * bit 1 = NMI context
3020 * bit 2 = IRQ context
3021 * bit 3 = SoftIRQ context
3022 * bit 4 = normal context.
3023 *
3024 * This works because this is the order of contexts that can
3025 * preempt other contexts. A SoftIRQ never preempts an IRQ
3026 * context.
3027 *
3028 * When the context is determined, the corresponding bit is
3029 * checked and set (if it was set, then a recursion of that context
3030 * happened).
3031 *
3032 * On unlock, we need to clear this bit. To do so, just subtract
3033 * 1 from the current_context and AND it to itself.
3034 *
3035 * (binary)
3036 * 101 - 1 = 100
3037 * 101 & 100 = 100 (clearing bit zero)
3038 *
3039 * 1010 - 1 = 1001
3040 * 1010 & 1001 = 1000 (clearing bit 1)
3041 *
3042 * The least significant bit can be cleared this way, and it
3043 * just so happens that it is the same bit corresponding to
3044 * the current context.
3045 *
3046 * Now the TRANSITION bit breaks the above slightly. The TRANSITION bit
3047 * is set when a recursion is detected at the current context, and if
3048 * the TRANSITION bit is already set, it will fail the recursion.
3049 * This is needed because there's a lag between the changing of
3050 * interrupt context and updating the preempt count. In this case,
3051 * a false positive will be found. To handle this, one extra recursion
3052 * is allowed, and this is done by the TRANSITION bit. If the TRANSITION
3053 * bit is already set, then it is considered a recursion and the function
3054 * ends. Otherwise, the TRANSITION bit is set, and that bit is returned.
3055 *
3056 * On the trace_recursive_unlock(), the TRANSITION bit will be the first
3057 * to be cleared. Even if it wasn't the context that set it. That is,
3058 * if an interrupt comes in while NORMAL bit is set and the ring buffer
3059 * is called before preempt_count() is updated, since the check will
3060 * be on the NORMAL bit, the TRANSITION bit will then be set. If an
3061 * NMI then comes in, it will set the NMI bit, but when the NMI code
3062 * does the trace_recursive_unlock() it will clear the TRANSTION bit
3063 * and leave the NMI bit set. But this is fine, because the interrupt
3064 * code that set the TRANSITION bit will then clear the NMI bit when it
3065 * calls trace_recursive_unlock(). If another NMI comes in, it will
3066 * set the TRANSITION bit and continue.
3067 *
3068 * Note: The TRANSITION bit only handles a single transition between context.
3069 */
3070
3071 static __always_inline int
trace_recursive_lock(struct ring_buffer_per_cpu * cpu_buffer)3072 trace_recursive_lock(struct ring_buffer_per_cpu *cpu_buffer)
3073 {
3074 unsigned int val = cpu_buffer->current_context;
3075 unsigned long pc = preempt_count();
3076 int bit;
3077
3078 if (!(pc & (NMI_MASK | HARDIRQ_MASK | SOFTIRQ_OFFSET)))
3079 bit = RB_CTX_NORMAL;
3080 else
3081 bit = pc & NMI_MASK ? RB_CTX_NMI :
3082 pc & HARDIRQ_MASK ? RB_CTX_IRQ : RB_CTX_SOFTIRQ;
3083
3084 if (unlikely(val & (1 << (bit + cpu_buffer->nest)))) {
3085 /*
3086 * It is possible that this was called by transitioning
3087 * between interrupt context, and preempt_count() has not
3088 * been updated yet. In this case, use the TRANSITION bit.
3089 */
3090 bit = RB_CTX_TRANSITION;
3091 if (val & (1 << (bit + cpu_buffer->nest)))
3092 return 1;
3093 }
3094
3095 val |= (1 << (bit + cpu_buffer->nest));
3096 cpu_buffer->current_context = val;
3097
3098 return 0;
3099 }
3100
3101 static __always_inline void
trace_recursive_unlock(struct ring_buffer_per_cpu * cpu_buffer)3102 trace_recursive_unlock(struct ring_buffer_per_cpu *cpu_buffer)
3103 {
3104 cpu_buffer->current_context &=
3105 cpu_buffer->current_context - (1 << cpu_buffer->nest);
3106 }
3107
3108 /* The recursive locking above uses 5 bits */
3109 #define NESTED_BITS 5
3110
3111 /**
3112 * ring_buffer_nest_start - Allow to trace while nested
3113 * @buffer: The ring buffer to modify
3114 *
3115 * The ring buffer has a safety mechanism to prevent recursion.
3116 * But there may be a case where a trace needs to be done while
3117 * tracing something else. In this case, calling this function
3118 * will allow this function to nest within a currently active
3119 * ring_buffer_lock_reserve().
3120 *
3121 * Call this function before calling another ring_buffer_lock_reserve() and
3122 * call ring_buffer_nest_end() after the nested ring_buffer_unlock_commit().
3123 */
ring_buffer_nest_start(struct trace_buffer * buffer)3124 void ring_buffer_nest_start(struct trace_buffer *buffer)
3125 {
3126 struct ring_buffer_per_cpu *cpu_buffer;
3127 int cpu;
3128
3129 /* Enabled by ring_buffer_nest_end() */
3130 preempt_disable_notrace();
3131 cpu = raw_smp_processor_id();
3132 cpu_buffer = buffer->buffers[cpu];
3133 /* This is the shift value for the above recursive locking */
3134 cpu_buffer->nest += NESTED_BITS;
3135 }
3136
3137 /**
3138 * ring_buffer_nest_end - Allow to trace while nested
3139 * @buffer: The ring buffer to modify
3140 *
3141 * Must be called after ring_buffer_nest_start() and after the
3142 * ring_buffer_unlock_commit().
3143 */
ring_buffer_nest_end(struct trace_buffer * buffer)3144 void ring_buffer_nest_end(struct trace_buffer *buffer)
3145 {
3146 struct ring_buffer_per_cpu *cpu_buffer;
3147 int cpu;
3148
3149 /* disabled by ring_buffer_nest_start() */
3150 cpu = raw_smp_processor_id();
3151 cpu_buffer = buffer->buffers[cpu];
3152 /* This is the shift value for the above recursive locking */
3153 cpu_buffer->nest -= NESTED_BITS;
3154 preempt_enable_notrace();
3155 }
3156
3157 /**
3158 * ring_buffer_unlock_commit - commit a reserved
3159 * @buffer: The buffer to commit to
3160 * @event: The event pointer to commit.
3161 *
3162 * This commits the data to the ring buffer, and releases any locks held.
3163 *
3164 * Must be paired with ring_buffer_lock_reserve.
3165 */
ring_buffer_unlock_commit(struct trace_buffer * buffer,struct ring_buffer_event * event)3166 int ring_buffer_unlock_commit(struct trace_buffer *buffer,
3167 struct ring_buffer_event *event)
3168 {
3169 struct ring_buffer_per_cpu *cpu_buffer;
3170 int cpu = raw_smp_processor_id();
3171
3172 cpu_buffer = buffer->buffers[cpu];
3173
3174 rb_commit(cpu_buffer, event);
3175
3176 rb_wakeups(buffer, cpu_buffer);
3177
3178 trace_recursive_unlock(cpu_buffer);
3179
3180 preempt_enable_notrace();
3181
3182 return 0;
3183 }
3184 EXPORT_SYMBOL_GPL(ring_buffer_unlock_commit);
3185
3186 static struct ring_buffer_event *
__rb_reserve_next(struct ring_buffer_per_cpu * cpu_buffer,struct rb_event_info * info)3187 __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer,
3188 struct rb_event_info *info)
3189 {
3190 struct ring_buffer_event *event;
3191 struct buffer_page *tail_page;
3192 unsigned long tail, write, w;
3193 bool a_ok;
3194 bool b_ok;
3195
3196 /* Don't let the compiler play games with cpu_buffer->tail_page */
3197 tail_page = info->tail_page = READ_ONCE(cpu_buffer->tail_page);
3198
3199 /*A*/ w = local_read(&tail_page->write) & RB_WRITE_MASK;
3200 barrier();
3201 b_ok = rb_time_read(&cpu_buffer->before_stamp, &info->before);
3202 a_ok = rb_time_read(&cpu_buffer->write_stamp, &info->after);
3203 barrier();
3204 info->ts = rb_time_stamp(cpu_buffer->buffer);
3205
3206 if ((info->add_timestamp & RB_ADD_STAMP_ABSOLUTE)) {
3207 info->delta = info->ts;
3208 } else {
3209 /*
3210 * If interrupting an event time update, we may need an
3211 * absolute timestamp.
3212 * Don't bother if this is the start of a new page (w == 0).
3213 */
3214 if (unlikely(!a_ok || !b_ok || (info->before != info->after && w))) {
3215 info->add_timestamp |= RB_ADD_STAMP_FORCE | RB_ADD_STAMP_EXTEND;
3216 info->length += RB_LEN_TIME_EXTEND;
3217 } else {
3218 info->delta = info->ts - info->after;
3219 if (unlikely(test_time_stamp(info->delta))) {
3220 info->add_timestamp |= RB_ADD_STAMP_EXTEND;
3221 info->length += RB_LEN_TIME_EXTEND;
3222 }
3223 }
3224 }
3225
3226 /*B*/ rb_time_set(&cpu_buffer->before_stamp, info->ts);
3227
3228 /*C*/ write = local_add_return(info->length, &tail_page->write);
3229
3230 /* set write to only the index of the write */
3231 write &= RB_WRITE_MASK;
3232
3233 tail = write - info->length;
3234
3235 /* See if we shot pass the end of this buffer page */
3236 if (unlikely(write > BUF_PAGE_SIZE)) {
3237 /* before and after may now different, fix it up*/
3238 b_ok = rb_time_read(&cpu_buffer->before_stamp, &info->before);
3239 a_ok = rb_time_read(&cpu_buffer->write_stamp, &info->after);
3240 if (a_ok && b_ok && info->before != info->after)
3241 (void)rb_time_cmpxchg(&cpu_buffer->before_stamp,
3242 info->before, info->after);
3243 return rb_move_tail(cpu_buffer, tail, info);
3244 }
3245
3246 if (likely(tail == w)) {
3247 u64 save_before;
3248 bool s_ok;
3249
3250 /* Nothing interrupted us between A and C */
3251 /*D*/ rb_time_set(&cpu_buffer->write_stamp, info->ts);
3252 barrier();
3253 /*E*/ s_ok = rb_time_read(&cpu_buffer->before_stamp, &save_before);
3254 RB_WARN_ON(cpu_buffer, !s_ok);
3255 if (likely(!(info->add_timestamp &
3256 (RB_ADD_STAMP_FORCE | RB_ADD_STAMP_ABSOLUTE))))
3257 /* This did not interrupt any time update */
3258 info->delta = info->ts - info->after;
3259 else
3260 /* Just use full timestamp for inerrupting event */
3261 info->delta = info->ts;
3262 barrier();
3263 if (unlikely(info->ts != save_before)) {
3264 /* SLOW PATH - Interrupted between C and E */
3265
3266 a_ok = rb_time_read(&cpu_buffer->write_stamp, &info->after);
3267 RB_WARN_ON(cpu_buffer, !a_ok);
3268
3269 /* Write stamp must only go forward */
3270 if (save_before > info->after) {
3271 /*
3272 * We do not care about the result, only that
3273 * it gets updated atomically.
3274 */
3275 (void)rb_time_cmpxchg(&cpu_buffer->write_stamp,
3276 info->after, save_before);
3277 }
3278 }
3279 } else {
3280 u64 ts;
3281 /* SLOW PATH - Interrupted between A and C */
3282 a_ok = rb_time_read(&cpu_buffer->write_stamp, &info->after);
3283 /* Was interrupted before here, write_stamp must be valid */
3284 RB_WARN_ON(cpu_buffer, !a_ok);
3285 ts = rb_time_stamp(cpu_buffer->buffer);
3286 barrier();
3287 /*E*/ if (write == (local_read(&tail_page->write) & RB_WRITE_MASK) &&
3288 info->after < ts &&
3289 rb_time_cmpxchg(&cpu_buffer->write_stamp,
3290 info->after, ts)) {
3291 /* Nothing came after this event between C and E */
3292 info->delta = ts - info->after;
3293 info->ts = ts;
3294 } else {
3295 /*
3296 * Interrupted beween C and E:
3297 * Lost the previous events time stamp. Just set the
3298 * delta to zero, and this will be the same time as
3299 * the event this event interrupted. And the events that
3300 * came after this will still be correct (as they would
3301 * have built their delta on the previous event.
3302 */
3303 info->delta = 0;
3304 }
3305 info->add_timestamp &= ~RB_ADD_STAMP_FORCE;
3306 }
3307
3308 /*
3309 * If this is the first commit on the page, then it has the same
3310 * timestamp as the page itself.
3311 */
3312 if (unlikely(!tail && !(info->add_timestamp &
3313 (RB_ADD_STAMP_FORCE | RB_ADD_STAMP_ABSOLUTE))))
3314 info->delta = 0;
3315
3316 /* We reserved something on the buffer */
3317
3318 event = __rb_page_index(tail_page, tail);
3319 rb_update_event(cpu_buffer, event, info);
3320
3321 local_inc(&tail_page->entries);
3322
3323 /*
3324 * If this is the first commit on the page, then update
3325 * its timestamp.
3326 */
3327 if (unlikely(!tail))
3328 tail_page->page->time_stamp = info->ts;
3329
3330 /* account for these added bytes */
3331 local_add(info->length, &cpu_buffer->entries_bytes);
3332
3333 return event;
3334 }
3335
3336 static __always_inline struct ring_buffer_event *
rb_reserve_next_event(struct trace_buffer * buffer,struct ring_buffer_per_cpu * cpu_buffer,unsigned long length)3337 rb_reserve_next_event(struct trace_buffer *buffer,
3338 struct ring_buffer_per_cpu *cpu_buffer,
3339 unsigned long length)
3340 {
3341 struct ring_buffer_event *event;
3342 struct rb_event_info info;
3343 int nr_loops = 0;
3344 int add_ts_default;
3345
3346 rb_start_commit(cpu_buffer);
3347 /* The commit page can not change after this */
3348
3349 #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP
3350 /*
3351 * Due to the ability to swap a cpu buffer from a buffer
3352 * it is possible it was swapped before we committed.
3353 * (committing stops a swap). We check for it here and
3354 * if it happened, we have to fail the write.
3355 */
3356 barrier();
3357 if (unlikely(READ_ONCE(cpu_buffer->buffer) != buffer)) {
3358 local_dec(&cpu_buffer->committing);
3359 local_dec(&cpu_buffer->commits);
3360 return NULL;
3361 }
3362 #endif
3363
3364 info.length = rb_calculate_event_length(length);
3365
3366 if (ring_buffer_time_stamp_abs(cpu_buffer->buffer)) {
3367 add_ts_default = RB_ADD_STAMP_ABSOLUTE;
3368 info.length += RB_LEN_TIME_EXTEND;
3369 } else {
3370 add_ts_default = RB_ADD_STAMP_NONE;
3371 }
3372
3373 again:
3374 info.add_timestamp = add_ts_default;
3375 info.delta = 0;
3376
3377 /*
3378 * We allow for interrupts to reenter here and do a trace.
3379 * If one does, it will cause this original code to loop
3380 * back here. Even with heavy interrupts happening, this
3381 * should only happen a few times in a row. If this happens
3382 * 1000 times in a row, there must be either an interrupt
3383 * storm or we have something buggy.
3384 * Bail!
3385 */
3386 if (RB_WARN_ON(cpu_buffer, ++nr_loops > 1000))
3387 goto out_fail;
3388
3389 event = __rb_reserve_next(cpu_buffer, &info);
3390
3391 if (unlikely(PTR_ERR(event) == -EAGAIN)) {
3392 if (info.add_timestamp & (RB_ADD_STAMP_FORCE | RB_ADD_STAMP_EXTEND))
3393 info.length -= RB_LEN_TIME_EXTEND;
3394 goto again;
3395 }
3396
3397 if (likely(event))
3398 return event;
3399 out_fail:
3400 rb_end_commit(cpu_buffer);
3401 return NULL;
3402 }
3403
3404 /**
3405 * ring_buffer_lock_reserve - reserve a part of the buffer
3406 * @buffer: the ring buffer to reserve from
3407 * @length: the length of the data to reserve (excluding event header)
3408 *
3409 * Returns a reserved event on the ring buffer to copy directly to.
3410 * The user of this interface will need to get the body to write into
3411 * and can use the ring_buffer_event_data() interface.
3412 *
3413 * The length is the length of the data needed, not the event length
3414 * which also includes the event header.
3415 *
3416 * Must be paired with ring_buffer_unlock_commit, unless NULL is returned.
3417 * If NULL is returned, then nothing has been allocated or locked.
3418 */
3419 struct ring_buffer_event *
ring_buffer_lock_reserve(struct trace_buffer * buffer,unsigned long length)3420 ring_buffer_lock_reserve(struct trace_buffer *buffer, unsigned long length)
3421 {
3422 struct ring_buffer_per_cpu *cpu_buffer;
3423 struct ring_buffer_event *event;
3424 int cpu;
3425
3426 /* If we are tracing schedule, we don't want to recurse */
3427 preempt_disable_notrace();
3428
3429 if (unlikely(atomic_read(&buffer->record_disabled)))
3430 goto out;
3431
3432 cpu = raw_smp_processor_id();
3433
3434 if (unlikely(!cpumask_test_cpu(cpu, buffer->cpumask)))
3435 goto out;
3436
3437 cpu_buffer = buffer->buffers[cpu];
3438
3439 if (unlikely(atomic_read(&cpu_buffer->record_disabled)))
3440 goto out;
3441
3442 if (unlikely(length > BUF_MAX_DATA_SIZE))
3443 goto out;
3444
3445 if (unlikely(trace_recursive_lock(cpu_buffer)))
3446 goto out;
3447
3448 event = rb_reserve_next_event(buffer, cpu_buffer, length);
3449 if (!event)
3450 goto out_unlock;
3451
3452 return event;
3453
3454 out_unlock:
3455 trace_recursive_unlock(cpu_buffer);
3456 out:
3457 preempt_enable_notrace();
3458 return NULL;
3459 }
3460 EXPORT_SYMBOL_GPL(ring_buffer_lock_reserve);
3461
3462 /*
3463 * Decrement the entries to the page that an event is on.
3464 * The event does not even need to exist, only the pointer
3465 * to the page it is on. This may only be called before the commit
3466 * takes place.
3467 */
3468 static inline void
rb_decrement_entry(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event * event)3469 rb_decrement_entry(struct ring_buffer_per_cpu *cpu_buffer,
3470 struct ring_buffer_event *event)
3471 {
3472 unsigned long addr = (unsigned long)event;
3473 struct buffer_page *bpage = cpu_buffer->commit_page;
3474 struct buffer_page *start;
3475
3476 addr &= PAGE_MASK;
3477
3478 /* Do the likely case first */
3479 if (likely(bpage->page == (void *)addr)) {
3480 local_dec(&bpage->entries);
3481 return;
3482 }
3483
3484 /*
3485 * Because the commit page may be on the reader page we
3486 * start with the next page and check the end loop there.
3487 */
3488 rb_inc_page(cpu_buffer, &bpage);
3489 start = bpage;
3490 do {
3491 if (bpage->page == (void *)addr) {
3492 local_dec(&bpage->entries);
3493 return;
3494 }
3495 rb_inc_page(cpu_buffer, &bpage);
3496 } while (bpage != start);
3497
3498 /* commit not part of this buffer?? */
3499 RB_WARN_ON(cpu_buffer, 1);
3500 }
3501
3502 /**
3503 * ring_buffer_commit_discard - discard an event that has not been committed
3504 * @buffer: the ring buffer
3505 * @event: non committed event to discard
3506 *
3507 * Sometimes an event that is in the ring buffer needs to be ignored.
3508 * This function lets the user discard an event in the ring buffer
3509 * and then that event will not be read later.
3510 *
3511 * This function only works if it is called before the item has been
3512 * committed. It will try to free the event from the ring buffer
3513 * if another event has not been added behind it.
3514 *
3515 * If another event has been added behind it, it will set the event
3516 * up as discarded, and perform the commit.
3517 *
3518 * If this function is called, do not call ring_buffer_unlock_commit on
3519 * the event.
3520 */
ring_buffer_discard_commit(struct trace_buffer * buffer,struct ring_buffer_event * event)3521 void ring_buffer_discard_commit(struct trace_buffer *buffer,
3522 struct ring_buffer_event *event)
3523 {
3524 struct ring_buffer_per_cpu *cpu_buffer;
3525 int cpu;
3526
3527 /* The event is discarded regardless */
3528 rb_event_discard(event);
3529
3530 cpu = smp_processor_id();
3531 cpu_buffer = buffer->buffers[cpu];
3532
3533 /*
3534 * This must only be called if the event has not been
3535 * committed yet. Thus we can assume that preemption
3536 * is still disabled.
3537 */
3538 RB_WARN_ON(buffer, !local_read(&cpu_buffer->committing));
3539
3540 rb_decrement_entry(cpu_buffer, event);
3541 if (rb_try_to_discard(cpu_buffer, event))
3542 goto out;
3543
3544 out:
3545 rb_end_commit(cpu_buffer);
3546
3547 trace_recursive_unlock(cpu_buffer);
3548
3549 preempt_enable_notrace();
3550
3551 }
3552 EXPORT_SYMBOL_GPL(ring_buffer_discard_commit);
3553
3554 /**
3555 * ring_buffer_write - write data to the buffer without reserving
3556 * @buffer: The ring buffer to write to.
3557 * @length: The length of the data being written (excluding the event header)
3558 * @data: The data to write to the buffer.
3559 *
3560 * This is like ring_buffer_lock_reserve and ring_buffer_unlock_commit as
3561 * one function. If you already have the data to write to the buffer, it
3562 * may be easier to simply call this function.
3563 *
3564 * Note, like ring_buffer_lock_reserve, the length is the length of the data
3565 * and not the length of the event which would hold the header.
3566 */
ring_buffer_write(struct trace_buffer * buffer,unsigned long length,void * data)3567 int ring_buffer_write(struct trace_buffer *buffer,
3568 unsigned long length,
3569 void *data)
3570 {
3571 struct ring_buffer_per_cpu *cpu_buffer;
3572 struct ring_buffer_event *event;
3573 void *body;
3574 int ret = -EBUSY;
3575 int cpu;
3576
3577 preempt_disable_notrace();
3578
3579 if (atomic_read(&buffer->record_disabled))
3580 goto out;
3581
3582 cpu = raw_smp_processor_id();
3583
3584 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3585 goto out;
3586
3587 cpu_buffer = buffer->buffers[cpu];
3588
3589 if (atomic_read(&cpu_buffer->record_disabled))
3590 goto out;
3591
3592 if (length > BUF_MAX_DATA_SIZE)
3593 goto out;
3594
3595 if (unlikely(trace_recursive_lock(cpu_buffer)))
3596 goto out;
3597
3598 event = rb_reserve_next_event(buffer, cpu_buffer, length);
3599 if (!event)
3600 goto out_unlock;
3601
3602 body = rb_event_data(event);
3603
3604 memcpy(body, data, length);
3605
3606 rb_commit(cpu_buffer, event);
3607
3608 rb_wakeups(buffer, cpu_buffer);
3609
3610 ret = 0;
3611
3612 out_unlock:
3613 trace_recursive_unlock(cpu_buffer);
3614
3615 out:
3616 preempt_enable_notrace();
3617
3618 return ret;
3619 }
3620 EXPORT_SYMBOL_GPL(ring_buffer_write);
3621
rb_per_cpu_empty(struct ring_buffer_per_cpu * cpu_buffer)3622 static bool rb_per_cpu_empty(struct ring_buffer_per_cpu *cpu_buffer)
3623 {
3624 struct buffer_page *reader = cpu_buffer->reader_page;
3625 struct buffer_page *head = rb_set_head_page(cpu_buffer);
3626 struct buffer_page *commit = cpu_buffer->commit_page;
3627
3628 /* In case of error, head will be NULL */
3629 if (unlikely(!head))
3630 return true;
3631
3632 return reader->read == rb_page_commit(reader) &&
3633 (commit == reader ||
3634 (commit == head &&
3635 head->read == rb_page_commit(commit)));
3636 }
3637
3638 /**
3639 * ring_buffer_record_disable - stop all writes into the buffer
3640 * @buffer: The ring buffer to stop writes to.
3641 *
3642 * This prevents all writes to the buffer. Any attempt to write
3643 * to the buffer after this will fail and return NULL.
3644 *
3645 * The caller should call synchronize_rcu() after this.
3646 */
ring_buffer_record_disable(struct trace_buffer * buffer)3647 void ring_buffer_record_disable(struct trace_buffer *buffer)
3648 {
3649 atomic_inc(&buffer->record_disabled);
3650 }
3651 EXPORT_SYMBOL_GPL(ring_buffer_record_disable);
3652
3653 /**
3654 * ring_buffer_record_enable - enable writes to the buffer
3655 * @buffer: The ring buffer to enable writes
3656 *
3657 * Note, multiple disables will need the same number of enables
3658 * to truly enable the writing (much like preempt_disable).
3659 */
ring_buffer_record_enable(struct trace_buffer * buffer)3660 void ring_buffer_record_enable(struct trace_buffer *buffer)
3661 {
3662 atomic_dec(&buffer->record_disabled);
3663 }
3664 EXPORT_SYMBOL_GPL(ring_buffer_record_enable);
3665
3666 /**
3667 * ring_buffer_record_off - stop all writes into the buffer
3668 * @buffer: The ring buffer to stop writes to.
3669 *
3670 * This prevents all writes to the buffer. Any attempt to write
3671 * to the buffer after this will fail and return NULL.
3672 *
3673 * This is different than ring_buffer_record_disable() as
3674 * it works like an on/off switch, where as the disable() version
3675 * must be paired with a enable().
3676 */
ring_buffer_record_off(struct trace_buffer * buffer)3677 void ring_buffer_record_off(struct trace_buffer *buffer)
3678 {
3679 unsigned int rd;
3680 unsigned int new_rd;
3681
3682 do {
3683 rd = atomic_read(&buffer->record_disabled);
3684 new_rd = rd | RB_BUFFER_OFF;
3685 } while (atomic_cmpxchg(&buffer->record_disabled, rd, new_rd) != rd);
3686 }
3687 EXPORT_SYMBOL_GPL(ring_buffer_record_off);
3688
3689 /**
3690 * ring_buffer_record_on - restart writes into the buffer
3691 * @buffer: The ring buffer to start writes to.
3692 *
3693 * This enables all writes to the buffer that was disabled by
3694 * ring_buffer_record_off().
3695 *
3696 * This is different than ring_buffer_record_enable() as
3697 * it works like an on/off switch, where as the enable() version
3698 * must be paired with a disable().
3699 */
ring_buffer_record_on(struct trace_buffer * buffer)3700 void ring_buffer_record_on(struct trace_buffer *buffer)
3701 {
3702 unsigned int rd;
3703 unsigned int new_rd;
3704
3705 do {
3706 rd = atomic_read(&buffer->record_disabled);
3707 new_rd = rd & ~RB_BUFFER_OFF;
3708 } while (atomic_cmpxchg(&buffer->record_disabled, rd, new_rd) != rd);
3709 }
3710 EXPORT_SYMBOL_GPL(ring_buffer_record_on);
3711
3712 /**
3713 * ring_buffer_record_is_on - return true if the ring buffer can write
3714 * @buffer: The ring buffer to see if write is enabled
3715 *
3716 * Returns true if the ring buffer is in a state that it accepts writes.
3717 */
ring_buffer_record_is_on(struct trace_buffer * buffer)3718 bool ring_buffer_record_is_on(struct trace_buffer *buffer)
3719 {
3720 return !atomic_read(&buffer->record_disabled);
3721 }
3722
3723 /**
3724 * ring_buffer_record_is_set_on - return true if the ring buffer is set writable
3725 * @buffer: The ring buffer to see if write is set enabled
3726 *
3727 * Returns true if the ring buffer is set writable by ring_buffer_record_on().
3728 * Note that this does NOT mean it is in a writable state.
3729 *
3730 * It may return true when the ring buffer has been disabled by
3731 * ring_buffer_record_disable(), as that is a temporary disabling of
3732 * the ring buffer.
3733 */
ring_buffer_record_is_set_on(struct trace_buffer * buffer)3734 bool ring_buffer_record_is_set_on(struct trace_buffer *buffer)
3735 {
3736 return !(atomic_read(&buffer->record_disabled) & RB_BUFFER_OFF);
3737 }
3738
3739 /**
3740 * ring_buffer_record_disable_cpu - stop all writes into the cpu_buffer
3741 * @buffer: The ring buffer to stop writes to.
3742 * @cpu: The CPU buffer to stop
3743 *
3744 * This prevents all writes to the buffer. Any attempt to write
3745 * to the buffer after this will fail and return NULL.
3746 *
3747 * The caller should call synchronize_rcu() after this.
3748 */
ring_buffer_record_disable_cpu(struct trace_buffer * buffer,int cpu)3749 void ring_buffer_record_disable_cpu(struct trace_buffer *buffer, int cpu)
3750 {
3751 struct ring_buffer_per_cpu *cpu_buffer;
3752
3753 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3754 return;
3755
3756 cpu_buffer = buffer->buffers[cpu];
3757 atomic_inc(&cpu_buffer->record_disabled);
3758 }
3759 EXPORT_SYMBOL_GPL(ring_buffer_record_disable_cpu);
3760
3761 /**
3762 * ring_buffer_record_enable_cpu - enable writes to the buffer
3763 * @buffer: The ring buffer to enable writes
3764 * @cpu: The CPU to enable.
3765 *
3766 * Note, multiple disables will need the same number of enables
3767 * to truly enable the writing (much like preempt_disable).
3768 */
ring_buffer_record_enable_cpu(struct trace_buffer * buffer,int cpu)3769 void ring_buffer_record_enable_cpu(struct trace_buffer *buffer, int cpu)
3770 {
3771 struct ring_buffer_per_cpu *cpu_buffer;
3772
3773 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3774 return;
3775
3776 cpu_buffer = buffer->buffers[cpu];
3777 atomic_dec(&cpu_buffer->record_disabled);
3778 }
3779 EXPORT_SYMBOL_GPL(ring_buffer_record_enable_cpu);
3780
3781 /*
3782 * The total entries in the ring buffer is the running counter
3783 * of entries entered into the ring buffer, minus the sum of
3784 * the entries read from the ring buffer and the number of
3785 * entries that were overwritten.
3786 */
3787 static inline unsigned long
rb_num_of_entries(struct ring_buffer_per_cpu * cpu_buffer)3788 rb_num_of_entries(struct ring_buffer_per_cpu *cpu_buffer)
3789 {
3790 return local_read(&cpu_buffer->entries) -
3791 (local_read(&cpu_buffer->overrun) + cpu_buffer->read);
3792 }
3793
3794 /**
3795 * ring_buffer_oldest_event_ts - get the oldest event timestamp from the buffer
3796 * @buffer: The ring buffer
3797 * @cpu: The per CPU buffer to read from.
3798 */
ring_buffer_oldest_event_ts(struct trace_buffer * buffer,int cpu)3799 u64 ring_buffer_oldest_event_ts(struct trace_buffer *buffer, int cpu)
3800 {
3801 unsigned long flags;
3802 struct ring_buffer_per_cpu *cpu_buffer;
3803 struct buffer_page *bpage;
3804 u64 ret = 0;
3805
3806 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3807 return 0;
3808
3809 cpu_buffer = buffer->buffers[cpu];
3810 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
3811 /*
3812 * if the tail is on reader_page, oldest time stamp is on the reader
3813 * page
3814 */
3815 if (cpu_buffer->tail_page == cpu_buffer->reader_page)
3816 bpage = cpu_buffer->reader_page;
3817 else
3818 bpage = rb_set_head_page(cpu_buffer);
3819 if (bpage)
3820 ret = bpage->page->time_stamp;
3821 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
3822
3823 return ret;
3824 }
3825 EXPORT_SYMBOL_GPL(ring_buffer_oldest_event_ts);
3826
3827 /**
3828 * ring_buffer_bytes_cpu - get the number of bytes consumed in a cpu buffer
3829 * @buffer: The ring buffer
3830 * @cpu: The per CPU buffer to read from.
3831 */
ring_buffer_bytes_cpu(struct trace_buffer * buffer,int cpu)3832 unsigned long ring_buffer_bytes_cpu(struct trace_buffer *buffer, int cpu)
3833 {
3834 struct ring_buffer_per_cpu *cpu_buffer;
3835 unsigned long ret;
3836
3837 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3838 return 0;
3839
3840 cpu_buffer = buffer->buffers[cpu];
3841 ret = local_read(&cpu_buffer->entries_bytes) - cpu_buffer->read_bytes;
3842
3843 return ret;
3844 }
3845 EXPORT_SYMBOL_GPL(ring_buffer_bytes_cpu);
3846
3847 /**
3848 * ring_buffer_entries_cpu - get the number of entries in a cpu buffer
3849 * @buffer: The ring buffer
3850 * @cpu: The per CPU buffer to get the entries from.
3851 */
ring_buffer_entries_cpu(struct trace_buffer * buffer,int cpu)3852 unsigned long ring_buffer_entries_cpu(struct trace_buffer *buffer, int cpu)
3853 {
3854 struct ring_buffer_per_cpu *cpu_buffer;
3855
3856 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3857 return 0;
3858
3859 cpu_buffer = buffer->buffers[cpu];
3860
3861 return rb_num_of_entries(cpu_buffer);
3862 }
3863 EXPORT_SYMBOL_GPL(ring_buffer_entries_cpu);
3864
3865 /**
3866 * ring_buffer_overrun_cpu - get the number of overruns caused by the ring
3867 * buffer wrapping around (only if RB_FL_OVERWRITE is on).
3868 * @buffer: The ring buffer
3869 * @cpu: The per CPU buffer to get the number of overruns from
3870 */
ring_buffer_overrun_cpu(struct trace_buffer * buffer,int cpu)3871 unsigned long ring_buffer_overrun_cpu(struct trace_buffer *buffer, int cpu)
3872 {
3873 struct ring_buffer_per_cpu *cpu_buffer;
3874 unsigned long ret;
3875
3876 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3877 return 0;
3878
3879 cpu_buffer = buffer->buffers[cpu];
3880 ret = local_read(&cpu_buffer->overrun);
3881
3882 return ret;
3883 }
3884 EXPORT_SYMBOL_GPL(ring_buffer_overrun_cpu);
3885
3886 /**
3887 * ring_buffer_commit_overrun_cpu - get the number of overruns caused by
3888 * commits failing due to the buffer wrapping around while there are uncommitted
3889 * events, such as during an interrupt storm.
3890 * @buffer: The ring buffer
3891 * @cpu: The per CPU buffer to get the number of overruns from
3892 */
3893 unsigned long
ring_buffer_commit_overrun_cpu(struct trace_buffer * buffer,int cpu)3894 ring_buffer_commit_overrun_cpu(struct trace_buffer *buffer, int cpu)
3895 {
3896 struct ring_buffer_per_cpu *cpu_buffer;
3897 unsigned long ret;
3898
3899 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3900 return 0;
3901
3902 cpu_buffer = buffer->buffers[cpu];
3903 ret = local_read(&cpu_buffer->commit_overrun);
3904
3905 return ret;
3906 }
3907 EXPORT_SYMBOL_GPL(ring_buffer_commit_overrun_cpu);
3908
3909 /**
3910 * ring_buffer_dropped_events_cpu - get the number of dropped events caused by
3911 * the ring buffer filling up (only if RB_FL_OVERWRITE is off).
3912 * @buffer: The ring buffer
3913 * @cpu: The per CPU buffer to get the number of overruns from
3914 */
3915 unsigned long
ring_buffer_dropped_events_cpu(struct trace_buffer * buffer,int cpu)3916 ring_buffer_dropped_events_cpu(struct trace_buffer *buffer, int cpu)
3917 {
3918 struct ring_buffer_per_cpu *cpu_buffer;
3919 unsigned long ret;
3920
3921 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3922 return 0;
3923
3924 cpu_buffer = buffer->buffers[cpu];
3925 ret = local_read(&cpu_buffer->dropped_events);
3926
3927 return ret;
3928 }
3929 EXPORT_SYMBOL_GPL(ring_buffer_dropped_events_cpu);
3930
3931 /**
3932 * ring_buffer_read_events_cpu - get the number of events successfully read
3933 * @buffer: The ring buffer
3934 * @cpu: The per CPU buffer to get the number of events read
3935 */
3936 unsigned long
ring_buffer_read_events_cpu(struct trace_buffer * buffer,int cpu)3937 ring_buffer_read_events_cpu(struct trace_buffer *buffer, int cpu)
3938 {
3939 struct ring_buffer_per_cpu *cpu_buffer;
3940
3941 if (!cpumask_test_cpu(cpu, buffer->cpumask))
3942 return 0;
3943
3944 cpu_buffer = buffer->buffers[cpu];
3945 return cpu_buffer->read;
3946 }
3947 EXPORT_SYMBOL_GPL(ring_buffer_read_events_cpu);
3948
3949 /**
3950 * ring_buffer_entries - get the number of entries in a buffer
3951 * @buffer: The ring buffer
3952 *
3953 * Returns the total number of entries in the ring buffer
3954 * (all CPU entries)
3955 */
ring_buffer_entries(struct trace_buffer * buffer)3956 unsigned long ring_buffer_entries(struct trace_buffer *buffer)
3957 {
3958 struct ring_buffer_per_cpu *cpu_buffer;
3959 unsigned long entries = 0;
3960 int cpu;
3961
3962 /* if you care about this being correct, lock the buffer */
3963 for_each_buffer_cpu(buffer, cpu) {
3964 cpu_buffer = buffer->buffers[cpu];
3965 entries += rb_num_of_entries(cpu_buffer);
3966 }
3967
3968 return entries;
3969 }
3970 EXPORT_SYMBOL_GPL(ring_buffer_entries);
3971
3972 /**
3973 * ring_buffer_overruns - get the number of overruns in buffer
3974 * @buffer: The ring buffer
3975 *
3976 * Returns the total number of overruns in the ring buffer
3977 * (all CPU entries)
3978 */
ring_buffer_overruns(struct trace_buffer * buffer)3979 unsigned long ring_buffer_overruns(struct trace_buffer *buffer)
3980 {
3981 struct ring_buffer_per_cpu *cpu_buffer;
3982 unsigned long overruns = 0;
3983 int cpu;
3984
3985 /* if you care about this being correct, lock the buffer */
3986 for_each_buffer_cpu(buffer, cpu) {
3987 cpu_buffer = buffer->buffers[cpu];
3988 overruns += local_read(&cpu_buffer->overrun);
3989 }
3990
3991 return overruns;
3992 }
3993 EXPORT_SYMBOL_GPL(ring_buffer_overruns);
3994
rb_iter_reset(struct ring_buffer_iter * iter)3995 static void rb_iter_reset(struct ring_buffer_iter *iter)
3996 {
3997 struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
3998
3999 /* Iterator usage is expected to have record disabled */
4000 iter->head_page = cpu_buffer->reader_page;
4001 iter->head = cpu_buffer->reader_page->read;
4002 iter->next_event = iter->head;
4003
4004 iter->cache_reader_page = iter->head_page;
4005 iter->cache_read = cpu_buffer->read;
4006
4007 if (iter->head) {
4008 iter->read_stamp = cpu_buffer->read_stamp;
4009 iter->page_stamp = cpu_buffer->reader_page->page->time_stamp;
4010 } else {
4011 iter->read_stamp = iter->head_page->page->time_stamp;
4012 iter->page_stamp = iter->read_stamp;
4013 }
4014 }
4015
4016 /**
4017 * ring_buffer_iter_reset - reset an iterator
4018 * @iter: The iterator to reset
4019 *
4020 * Resets the iterator, so that it will start from the beginning
4021 * again.
4022 */
ring_buffer_iter_reset(struct ring_buffer_iter * iter)4023 void ring_buffer_iter_reset(struct ring_buffer_iter *iter)
4024 {
4025 struct ring_buffer_per_cpu *cpu_buffer;
4026 unsigned long flags;
4027
4028 if (!iter)
4029 return;
4030
4031 cpu_buffer = iter->cpu_buffer;
4032
4033 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
4034 rb_iter_reset(iter);
4035 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
4036 }
4037 EXPORT_SYMBOL_GPL(ring_buffer_iter_reset);
4038
4039 /**
4040 * ring_buffer_iter_empty - check if an iterator has no more to read
4041 * @iter: The iterator to check
4042 */
ring_buffer_iter_empty(struct ring_buffer_iter * iter)4043 int ring_buffer_iter_empty(struct ring_buffer_iter *iter)
4044 {
4045 struct ring_buffer_per_cpu *cpu_buffer;
4046 struct buffer_page *reader;
4047 struct buffer_page *head_page;
4048 struct buffer_page *commit_page;
4049 struct buffer_page *curr_commit_page;
4050 unsigned commit;
4051 u64 curr_commit_ts;
4052 u64 commit_ts;
4053
4054 cpu_buffer = iter->cpu_buffer;
4055 reader = cpu_buffer->reader_page;
4056 head_page = cpu_buffer->head_page;
4057 commit_page = cpu_buffer->commit_page;
4058 commit_ts = commit_page->page->time_stamp;
4059
4060 /*
4061 * When the writer goes across pages, it issues a cmpxchg which
4062 * is a mb(), which will synchronize with the rmb here.
4063 * (see rb_tail_page_update())
4064 */
4065 smp_rmb();
4066 commit = rb_page_commit(commit_page);
4067 /* We want to make sure that the commit page doesn't change */
4068 smp_rmb();
4069
4070 /* Make sure commit page didn't change */
4071 curr_commit_page = READ_ONCE(cpu_buffer->commit_page);
4072 curr_commit_ts = READ_ONCE(curr_commit_page->page->time_stamp);
4073
4074 /* If the commit page changed, then there's more data */
4075 if (curr_commit_page != commit_page ||
4076 curr_commit_ts != commit_ts)
4077 return 0;
4078
4079 /* Still racy, as it may return a false positive, but that's OK */
4080 return ((iter->head_page == commit_page && iter->head >= commit) ||
4081 (iter->head_page == reader && commit_page == head_page &&
4082 head_page->read == commit &&
4083 iter->head == rb_page_commit(cpu_buffer->reader_page)));
4084 }
4085 EXPORT_SYMBOL_GPL(ring_buffer_iter_empty);
4086
4087 static void
rb_update_read_stamp(struct ring_buffer_per_cpu * cpu_buffer,struct ring_buffer_event * event)4088 rb_update_read_stamp(struct ring_buffer_per_cpu *cpu_buffer,
4089 struct ring_buffer_event *event)
4090 {
4091 u64 delta;
4092
4093 switch (event->type_len) {
4094 case RINGBUF_TYPE_PADDING:
4095 return;
4096
4097 case RINGBUF_TYPE_TIME_EXTEND:
4098 delta = ring_buffer_event_time_stamp(event);
4099 cpu_buffer->read_stamp += delta;
4100 return;
4101
4102 case RINGBUF_TYPE_TIME_STAMP:
4103 delta = ring_buffer_event_time_stamp(event);
4104 cpu_buffer->read_stamp = delta;
4105 return;
4106
4107 case RINGBUF_TYPE_DATA:
4108 cpu_buffer->read_stamp += event->time_delta;
4109 return;
4110
4111 default:
4112 RB_WARN_ON(cpu_buffer, 1);
4113 }
4114 return;
4115 }
4116
4117 static void
rb_update_iter_read_stamp(struct ring_buffer_iter * iter,struct ring_buffer_event * event)4118 rb_update_iter_read_stamp(struct ring_buffer_iter *iter,
4119 struct ring_buffer_event *event)
4120 {
4121 u64 delta;
4122
4123 switch (event->type_len) {
4124 case RINGBUF_TYPE_PADDING:
4125 return;
4126
4127 case RINGBUF_TYPE_TIME_EXTEND:
4128 delta = ring_buffer_event_time_stamp(event);
4129 iter->read_stamp += delta;
4130 return;
4131
4132 case RINGBUF_TYPE_TIME_STAMP:
4133 delta = ring_buffer_event_time_stamp(event);
4134 iter->read_stamp = delta;
4135 return;
4136
4137 case RINGBUF_TYPE_DATA:
4138 iter->read_stamp += event->time_delta;
4139 return;
4140
4141 default:
4142 RB_WARN_ON(iter->cpu_buffer, 1);
4143 }
4144 return;
4145 }
4146
4147 static struct buffer_page *
rb_get_reader_page(struct ring_buffer_per_cpu * cpu_buffer)4148 rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer)
4149 {
4150 struct buffer_page *reader = NULL;
4151 unsigned long overwrite;
4152 unsigned long flags;
4153 int nr_loops = 0;
4154 int ret;
4155
4156 local_irq_save(flags);
4157 arch_spin_lock(&cpu_buffer->lock);
4158
4159 again:
4160 /*
4161 * This should normally only loop twice. But because the
4162 * start of the reader inserts an empty page, it causes
4163 * a case where we will loop three times. There should be no
4164 * reason to loop four times (that I know of).
4165 */
4166 if (RB_WARN_ON(cpu_buffer, ++nr_loops > 3)) {
4167 reader = NULL;
4168 goto out;
4169 }
4170
4171 reader = cpu_buffer->reader_page;
4172
4173 /* If there's more to read, return this page */
4174 if (cpu_buffer->reader_page->read < rb_page_size(reader))
4175 goto out;
4176
4177 /* Never should we have an index greater than the size */
4178 if (RB_WARN_ON(cpu_buffer,
4179 cpu_buffer->reader_page->read > rb_page_size(reader)))
4180 goto out;
4181
4182 /* check if we caught up to the tail */
4183 reader = NULL;
4184 if (cpu_buffer->commit_page == cpu_buffer->reader_page)
4185 goto out;
4186
4187 /* Don't bother swapping if the ring buffer is empty */
4188 if (rb_num_of_entries(cpu_buffer) == 0)
4189 goto out;
4190
4191 /*
4192 * Reset the reader page to size zero.
4193 */
4194 local_set(&cpu_buffer->reader_page->write, 0);
4195 local_set(&cpu_buffer->reader_page->entries, 0);
4196 local_set(&cpu_buffer->reader_page->page->commit, 0);
4197 cpu_buffer->reader_page->real_end = 0;
4198
4199 spin:
4200 /*
4201 * Splice the empty reader page into the list around the head.
4202 */
4203 reader = rb_set_head_page(cpu_buffer);
4204 if (!reader)
4205 goto out;
4206 cpu_buffer->reader_page->list.next = rb_list_head(reader->list.next);
4207 cpu_buffer->reader_page->list.prev = reader->list.prev;
4208
4209 /*
4210 * cpu_buffer->pages just needs to point to the buffer, it
4211 * has no specific buffer page to point to. Lets move it out
4212 * of our way so we don't accidentally swap it.
4213 */
4214 cpu_buffer->pages = reader->list.prev;
4215
4216 /* The reader page will be pointing to the new head */
4217 rb_set_list_to_head(cpu_buffer, &cpu_buffer->reader_page->list);
4218
4219 /*
4220 * We want to make sure we read the overruns after we set up our
4221 * pointers to the next object. The writer side does a
4222 * cmpxchg to cross pages which acts as the mb on the writer
4223 * side. Note, the reader will constantly fail the swap
4224 * while the writer is updating the pointers, so this
4225 * guarantees that the overwrite recorded here is the one we
4226 * want to compare with the last_overrun.
4227 */
4228 smp_mb();
4229 overwrite = local_read(&(cpu_buffer->overrun));
4230
4231 /*
4232 * Here's the tricky part.
4233 *
4234 * We need to move the pointer past the header page.
4235 * But we can only do that if a writer is not currently
4236 * moving it. The page before the header page has the
4237 * flag bit '1' set if it is pointing to the page we want.
4238 * but if the writer is in the process of moving it
4239 * than it will be '2' or already moved '0'.
4240 */
4241
4242 ret = rb_head_page_replace(reader, cpu_buffer->reader_page);
4243
4244 /*
4245 * If we did not convert it, then we must try again.
4246 */
4247 if (!ret)
4248 goto spin;
4249
4250 /*
4251 * Yay! We succeeded in replacing the page.
4252 *
4253 * Now make the new head point back to the reader page.
4254 */
4255 rb_list_head(reader->list.next)->prev = &cpu_buffer->reader_page->list;
4256 rb_inc_page(cpu_buffer, &cpu_buffer->head_page);
4257
4258 local_inc(&cpu_buffer->pages_read);
4259
4260 /* Finally update the reader page to the new head */
4261 cpu_buffer->reader_page = reader;
4262 cpu_buffer->reader_page->read = 0;
4263
4264 if (overwrite != cpu_buffer->last_overrun) {
4265 cpu_buffer->lost_events = overwrite - cpu_buffer->last_overrun;
4266 cpu_buffer->last_overrun = overwrite;
4267 }
4268
4269 goto again;
4270
4271 out:
4272 /* Update the read_stamp on the first event */
4273 if (reader && reader->read == 0)
4274 cpu_buffer->read_stamp = reader->page->time_stamp;
4275
4276 arch_spin_unlock(&cpu_buffer->lock);
4277 local_irq_restore(flags);
4278
4279 return reader;
4280 }
4281
rb_advance_reader(struct ring_buffer_per_cpu * cpu_buffer)4282 static void rb_advance_reader(struct ring_buffer_per_cpu *cpu_buffer)
4283 {
4284 struct ring_buffer_event *event;
4285 struct buffer_page *reader;
4286 unsigned length;
4287
4288 reader = rb_get_reader_page(cpu_buffer);
4289
4290 /* This function should not be called when buffer is empty */
4291 if (RB_WARN_ON(cpu_buffer, !reader))
4292 return;
4293
4294 event = rb_reader_event(cpu_buffer);
4295
4296 if (event->type_len <= RINGBUF_TYPE_DATA_TYPE_LEN_MAX)
4297 cpu_buffer->read++;
4298
4299 rb_update_read_stamp(cpu_buffer, event);
4300
4301 length = rb_event_length(event);
4302 cpu_buffer->reader_page->read += length;
4303 }
4304
rb_advance_iter(struct ring_buffer_iter * iter)4305 static void rb_advance_iter(struct ring_buffer_iter *iter)
4306 {
4307 struct ring_buffer_per_cpu *cpu_buffer;
4308
4309 cpu_buffer = iter->cpu_buffer;
4310
4311 /* If head == next_event then we need to jump to the next event */
4312 if (iter->head == iter->next_event) {
4313 /* If the event gets overwritten again, there's nothing to do */
4314 if (rb_iter_head_event(iter) == NULL)
4315 return;
4316 }
4317
4318 iter->head = iter->next_event;
4319
4320 /*
4321 * Check if we are at the end of the buffer.
4322 */
4323 if (iter->next_event >= rb_page_size(iter->head_page)) {
4324 /* discarded commits can make the page empty */
4325 if (iter->head_page == cpu_buffer->commit_page)
4326 return;
4327 rb_inc_iter(iter);
4328 return;
4329 }
4330
4331 rb_update_iter_read_stamp(iter, iter->event);
4332 }
4333
rb_lost_events(struct ring_buffer_per_cpu * cpu_buffer)4334 static int rb_lost_events(struct ring_buffer_per_cpu *cpu_buffer)
4335 {
4336 return cpu_buffer->lost_events;
4337 }
4338
4339 static struct ring_buffer_event *
rb_buffer_peek(struct ring_buffer_per_cpu * cpu_buffer,u64 * ts,unsigned long * lost_events)4340 rb_buffer_peek(struct ring_buffer_per_cpu *cpu_buffer, u64 *ts,
4341 unsigned long *lost_events)
4342 {
4343 struct ring_buffer_event *event;
4344 struct buffer_page *reader;
4345 int nr_loops = 0;
4346
4347 if (ts)
4348 *ts = 0;
4349 again:
4350 /*
4351 * We repeat when a time extend is encountered.
4352 * Since the time extend is always attached to a data event,
4353 * we should never loop more than once.
4354 * (We never hit the following condition more than twice).
4355 */
4356 if (RB_WARN_ON(cpu_buffer, ++nr_loops > 2))
4357 return NULL;
4358
4359 reader = rb_get_reader_page(cpu_buffer);
4360 if (!reader)
4361 return NULL;
4362
4363 event = rb_reader_event(cpu_buffer);
4364
4365 switch (event->type_len) {
4366 case RINGBUF_TYPE_PADDING:
4367 if (rb_null_event(event))
4368 RB_WARN_ON(cpu_buffer, 1);
4369 /*
4370 * Because the writer could be discarding every
4371 * event it creates (which would probably be bad)
4372 * if we were to go back to "again" then we may never
4373 * catch up, and will trigger the warn on, or lock
4374 * the box. Return the padding, and we will release
4375 * the current locks, and try again.
4376 */
4377 return event;
4378
4379 case RINGBUF_TYPE_TIME_EXTEND:
4380 /* Internal data, OK to advance */
4381 rb_advance_reader(cpu_buffer);
4382 goto again;
4383
4384 case RINGBUF_TYPE_TIME_STAMP:
4385 if (ts) {
4386 *ts = ring_buffer_event_time_stamp(event);
4387 ring_buffer_normalize_time_stamp(cpu_buffer->buffer,
4388 cpu_buffer->cpu, ts);
4389 }
4390 /* Internal data, OK to advance */
4391 rb_advance_reader(cpu_buffer);
4392 goto again;
4393
4394 case RINGBUF_TYPE_DATA:
4395 if (ts && !(*ts)) {
4396 *ts = cpu_buffer->read_stamp + event->time_delta;
4397 ring_buffer_normalize_time_stamp(cpu_buffer->buffer,
4398 cpu_buffer->cpu, ts);
4399 }
4400 if (lost_events)
4401 *lost_events = rb_lost_events(cpu_buffer);
4402 return event;
4403
4404 default:
4405 RB_WARN_ON(cpu_buffer, 1);
4406 }
4407
4408 return NULL;
4409 }
4410 EXPORT_SYMBOL_GPL(ring_buffer_peek);
4411
4412 static struct ring_buffer_event *
rb_iter_peek(struct ring_buffer_iter * iter,u64 * ts)4413 rb_iter_peek(struct ring_buffer_iter *iter, u64 *ts)
4414 {
4415 struct trace_buffer *buffer;
4416 struct ring_buffer_per_cpu *cpu_buffer;
4417 struct ring_buffer_event *event;
4418 int nr_loops = 0;
4419
4420 if (ts)
4421 *ts = 0;
4422
4423 cpu_buffer = iter->cpu_buffer;
4424 buffer = cpu_buffer->buffer;
4425
4426 /*
4427 * Check if someone performed a consuming read to
4428 * the buffer. A consuming read invalidates the iterator
4429 * and we need to reset the iterator in this case.
4430 */
4431 if (unlikely(iter->cache_read != cpu_buffer->read ||
4432 iter->cache_reader_page != cpu_buffer->reader_page))
4433 rb_iter_reset(iter);
4434
4435 again:
4436 if (ring_buffer_iter_empty(iter))
4437 return NULL;
4438
4439 /*
4440 * As the writer can mess with what the iterator is trying
4441 * to read, just give up if we fail to get an event after
4442 * three tries. The iterator is not as reliable when reading
4443 * the ring buffer with an active write as the consumer is.
4444 * Do not warn if the three failures is reached.
4445 */
4446 if (++nr_loops > 3)
4447 return NULL;
4448
4449 if (rb_per_cpu_empty(cpu_buffer))
4450 return NULL;
4451
4452 if (iter->head >= rb_page_size(iter->head_page)) {
4453 rb_inc_iter(iter);
4454 goto again;
4455 }
4456
4457 event = rb_iter_head_event(iter);
4458 if (!event)
4459 goto again;
4460
4461 switch (event->type_len) {
4462 case RINGBUF_TYPE_PADDING:
4463 if (rb_null_event(event)) {
4464 rb_inc_iter(iter);
4465 goto again;
4466 }
4467 rb_advance_iter(iter);
4468 return event;
4469
4470 case RINGBUF_TYPE_TIME_EXTEND:
4471 /* Internal data, OK to advance */
4472 rb_advance_iter(iter);
4473 goto again;
4474
4475 case RINGBUF_TYPE_TIME_STAMP:
4476 if (ts) {
4477 *ts = ring_buffer_event_time_stamp(event);
4478 ring_buffer_normalize_time_stamp(cpu_buffer->buffer,
4479 cpu_buffer->cpu, ts);
4480 }
4481 /* Internal data, OK to advance */
4482 rb_advance_iter(iter);
4483 goto again;
4484
4485 case RINGBUF_TYPE_DATA:
4486 if (ts && !(*ts)) {
4487 *ts = iter->read_stamp + event->time_delta;
4488 ring_buffer_normalize_time_stamp(buffer,
4489 cpu_buffer->cpu, ts);
4490 }
4491 return event;
4492
4493 default:
4494 RB_WARN_ON(cpu_buffer, 1);
4495 }
4496
4497 return NULL;
4498 }
4499 EXPORT_SYMBOL_GPL(ring_buffer_iter_peek);
4500
rb_reader_lock(struct ring_buffer_per_cpu * cpu_buffer)4501 static inline bool rb_reader_lock(struct ring_buffer_per_cpu *cpu_buffer)
4502 {
4503 if (likely(!in_nmi())) {
4504 raw_spin_lock(&cpu_buffer->reader_lock);
4505 return true;
4506 }
4507
4508 /*
4509 * If an NMI die dumps out the content of the ring buffer
4510 * trylock must be used to prevent a deadlock if the NMI
4511 * preempted a task that holds the ring buffer locks. If
4512 * we get the lock then all is fine, if not, then continue
4513 * to do the read, but this can corrupt the ring buffer,
4514 * so it must be permanently disabled from future writes.
4515 * Reading from NMI is a oneshot deal.
4516 */
4517 if (raw_spin_trylock(&cpu_buffer->reader_lock))
4518 return true;
4519
4520 /* Continue without locking, but disable the ring buffer */
4521 atomic_inc(&cpu_buffer->record_disabled);
4522 return false;
4523 }
4524
4525 static inline void
rb_reader_unlock(struct ring_buffer_per_cpu * cpu_buffer,bool locked)4526 rb_reader_unlock(struct ring_buffer_per_cpu *cpu_buffer, bool locked)
4527 {
4528 if (likely(locked))
4529 raw_spin_unlock(&cpu_buffer->reader_lock);
4530 return;
4531 }
4532
4533 /**
4534 * ring_buffer_peek - peek at the next event to be read
4535 * @buffer: The ring buffer to read
4536 * @cpu: The cpu to peak at
4537 * @ts: The timestamp counter of this event.
4538 * @lost_events: a variable to store if events were lost (may be NULL)
4539 *
4540 * This will return the event that will be read next, but does
4541 * not consume the data.
4542 */
4543 struct ring_buffer_event *
ring_buffer_peek(struct trace_buffer * buffer,int cpu,u64 * ts,unsigned long * lost_events)4544 ring_buffer_peek(struct trace_buffer *buffer, int cpu, u64 *ts,
4545 unsigned long *lost_events)
4546 {
4547 struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
4548 struct ring_buffer_event *event;
4549 unsigned long flags;
4550 bool dolock;
4551
4552 if (!cpumask_test_cpu(cpu, buffer->cpumask))
4553 return NULL;
4554
4555 again:
4556 local_irq_save(flags);
4557 dolock = rb_reader_lock(cpu_buffer);
4558 event = rb_buffer_peek(cpu_buffer, ts, lost_events);
4559 if (event && event->type_len == RINGBUF_TYPE_PADDING)
4560 rb_advance_reader(cpu_buffer);
4561 rb_reader_unlock(cpu_buffer, dolock);
4562 local_irq_restore(flags);
4563
4564 if (event && event->type_len == RINGBUF_TYPE_PADDING)
4565 goto again;
4566
4567 return event;
4568 }
4569
4570 /** ring_buffer_iter_dropped - report if there are dropped events
4571 * @iter: The ring buffer iterator
4572 *
4573 * Returns true if there was dropped events since the last peek.
4574 */
ring_buffer_iter_dropped(struct ring_buffer_iter * iter)4575 bool ring_buffer_iter_dropped(struct ring_buffer_iter *iter)
4576 {
4577 bool ret = iter->missed_events != 0;
4578
4579 iter->missed_events = 0;
4580 return ret;
4581 }
4582 EXPORT_SYMBOL_GPL(ring_buffer_iter_dropped);
4583
4584 /**
4585 * ring_buffer_iter_peek - peek at the next event to be read
4586 * @iter: The ring buffer iterator
4587 * @ts: The timestamp counter of this event.
4588 *
4589 * This will return the event that will be read next, but does
4590 * not increment the iterator.
4591 */
4592 struct ring_buffer_event *
ring_buffer_iter_peek(struct ring_buffer_iter * iter,u64 * ts)4593 ring_buffer_iter_peek(struct ring_buffer_iter *iter, u64 *ts)
4594 {
4595 struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
4596 struct ring_buffer_event *event;
4597 unsigned long flags;
4598
4599 again:
4600 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
4601 event = rb_iter_peek(iter, ts);
4602 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
4603
4604 if (event && event->type_len == RINGBUF_TYPE_PADDING)
4605 goto again;
4606
4607 return event;
4608 }
4609
4610 /**
4611 * ring_buffer_consume - return an event and consume it
4612 * @buffer: The ring buffer to get the next event from
4613 * @cpu: the cpu to read the buffer from
4614 * @ts: a variable to store the timestamp (may be NULL)
4615 * @lost_events: a variable to store if events were lost (may be NULL)
4616 *
4617 * Returns the next event in the ring buffer, and that event is consumed.
4618 * Meaning, that sequential reads will keep returning a different event,
4619 * and eventually empty the ring buffer if the producer is slower.
4620 */
4621 struct ring_buffer_event *
ring_buffer_consume(struct trace_buffer * buffer,int cpu,u64 * ts,unsigned long * lost_events)4622 ring_buffer_consume(struct trace_buffer *buffer, int cpu, u64 *ts,
4623 unsigned long *lost_events)
4624 {
4625 struct ring_buffer_per_cpu *cpu_buffer;
4626 struct ring_buffer_event *event = NULL;
4627 unsigned long flags;
4628 bool dolock;
4629
4630 again:
4631 /* might be called in atomic */
4632 preempt_disable();
4633
4634 if (!cpumask_test_cpu(cpu, buffer->cpumask))
4635 goto out;
4636
4637 cpu_buffer = buffer->buffers[cpu];
4638 local_irq_save(flags);
4639 dolock = rb_reader_lock(cpu_buffer);
4640
4641 event = rb_buffer_peek(cpu_buffer, ts, lost_events);
4642 if (event) {
4643 cpu_buffer->lost_events = 0;
4644 rb_advance_reader(cpu_buffer);
4645 }
4646
4647 rb_reader_unlock(cpu_buffer, dolock);
4648 local_irq_restore(flags);
4649
4650 out:
4651 preempt_enable();
4652
4653 if (event && event->type_len == RINGBUF_TYPE_PADDING)
4654 goto again;
4655
4656 return event;
4657 }
4658 EXPORT_SYMBOL_GPL(ring_buffer_consume);
4659
4660 /**
4661 * ring_buffer_read_prepare - Prepare for a non consuming read of the buffer
4662 * @buffer: The ring buffer to read from
4663 * @cpu: The cpu buffer to iterate over
4664 * @flags: gfp flags to use for memory allocation
4665 *
4666 * This performs the initial preparations necessary to iterate
4667 * through the buffer. Memory is allocated, buffer recording
4668 * is disabled, and the iterator pointer is returned to the caller.
4669 *
4670 * Disabling buffer recording prevents the reading from being
4671 * corrupted. This is not a consuming read, so a producer is not
4672 * expected.
4673 *
4674 * After a sequence of ring_buffer_read_prepare calls, the user is
4675 * expected to make at least one call to ring_buffer_read_prepare_sync.
4676 * Afterwards, ring_buffer_read_start is invoked to get things going
4677 * for real.
4678 *
4679 * This overall must be paired with ring_buffer_read_finish.
4680 */
4681 struct ring_buffer_iter *
ring_buffer_read_prepare(struct trace_buffer * buffer,int cpu,gfp_t flags)4682 ring_buffer_read_prepare(struct trace_buffer *buffer, int cpu, gfp_t flags)
4683 {
4684 struct ring_buffer_per_cpu *cpu_buffer;
4685 struct ring_buffer_iter *iter;
4686
4687 if (!cpumask_test_cpu(cpu, buffer->cpumask))
4688 return NULL;
4689
4690 iter = kzalloc(sizeof(*iter), flags);
4691 if (!iter)
4692 return NULL;
4693
4694 iter->event = kmalloc(BUF_MAX_DATA_SIZE, flags);
4695 if (!iter->event) {
4696 kfree(iter);
4697 return NULL;
4698 }
4699
4700 cpu_buffer = buffer->buffers[cpu];
4701
4702 iter->cpu_buffer = cpu_buffer;
4703
4704 atomic_inc(&cpu_buffer->resize_disabled);
4705
4706 return iter;
4707 }
4708 EXPORT_SYMBOL_GPL(ring_buffer_read_prepare);
4709
4710 /**
4711 * ring_buffer_read_prepare_sync - Synchronize a set of prepare calls
4712 *
4713 * All previously invoked ring_buffer_read_prepare calls to prepare
4714 * iterators will be synchronized. Afterwards, read_buffer_read_start
4715 * calls on those iterators are allowed.
4716 */
4717 void
ring_buffer_read_prepare_sync(void)4718 ring_buffer_read_prepare_sync(void)
4719 {
4720 synchronize_rcu();
4721 }
4722 EXPORT_SYMBOL_GPL(ring_buffer_read_prepare_sync);
4723
4724 /**
4725 * ring_buffer_read_start - start a non consuming read of the buffer
4726 * @iter: The iterator returned by ring_buffer_read_prepare
4727 *
4728 * This finalizes the startup of an iteration through the buffer.
4729 * The iterator comes from a call to ring_buffer_read_prepare and
4730 * an intervening ring_buffer_read_prepare_sync must have been
4731 * performed.
4732 *
4733 * Must be paired with ring_buffer_read_finish.
4734 */
4735 void
ring_buffer_read_start(struct ring_buffer_iter * iter)4736 ring_buffer_read_start(struct ring_buffer_iter *iter)
4737 {
4738 struct ring_buffer_per_cpu *cpu_buffer;
4739 unsigned long flags;
4740
4741 if (!iter)
4742 return;
4743
4744 cpu_buffer = iter->cpu_buffer;
4745
4746 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
4747 arch_spin_lock(&cpu_buffer->lock);
4748 rb_iter_reset(iter);
4749 arch_spin_unlock(&cpu_buffer->lock);
4750 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
4751 }
4752 EXPORT_SYMBOL_GPL(ring_buffer_read_start);
4753
4754 /**
4755 * ring_buffer_read_finish - finish reading the iterator of the buffer
4756 * @iter: The iterator retrieved by ring_buffer_start
4757 *
4758 * This re-enables the recording to the buffer, and frees the
4759 * iterator.
4760 */
4761 void
ring_buffer_read_finish(struct ring_buffer_iter * iter)4762 ring_buffer_read_finish(struct ring_buffer_iter *iter)
4763 {
4764 struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
4765 unsigned long flags;
4766
4767 /*
4768 * Ring buffer is disabled from recording, here's a good place
4769 * to check the integrity of the ring buffer.
4770 * Must prevent readers from trying to read, as the check
4771 * clears the HEAD page and readers require it.
4772 */
4773 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
4774 rb_check_pages(cpu_buffer);
4775 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
4776
4777 atomic_dec(&cpu_buffer->resize_disabled);
4778 kfree(iter->event);
4779 kfree(iter);
4780 }
4781 EXPORT_SYMBOL_GPL(ring_buffer_read_finish);
4782
4783 /**
4784 * ring_buffer_iter_advance - advance the iterator to the next location
4785 * @iter: The ring buffer iterator
4786 *
4787 * Move the location of the iterator such that the next read will
4788 * be the next location of the iterator.
4789 */
ring_buffer_iter_advance(struct ring_buffer_iter * iter)4790 void ring_buffer_iter_advance(struct ring_buffer_iter *iter)
4791 {
4792 struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer;
4793 unsigned long flags;
4794
4795 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
4796
4797 rb_advance_iter(iter);
4798
4799 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
4800 }
4801 EXPORT_SYMBOL_GPL(ring_buffer_iter_advance);
4802
4803 /**
4804 * ring_buffer_size - return the size of the ring buffer (in bytes)
4805 * @buffer: The ring buffer.
4806 * @cpu: The CPU to get ring buffer size from.
4807 */
ring_buffer_size(struct trace_buffer * buffer,int cpu)4808 unsigned long ring_buffer_size(struct trace_buffer *buffer, int cpu)
4809 {
4810 /*
4811 * Earlier, this method returned
4812 * BUF_PAGE_SIZE * buffer->nr_pages
4813 * Since the nr_pages field is now removed, we have converted this to
4814 * return the per cpu buffer value.
4815 */
4816 if (!cpumask_test_cpu(cpu, buffer->cpumask))
4817 return 0;
4818
4819 return BUF_PAGE_SIZE * buffer->buffers[cpu]->nr_pages;
4820 }
4821 EXPORT_SYMBOL_GPL(ring_buffer_size);
4822
4823 static void
rb_reset_cpu(struct ring_buffer_per_cpu * cpu_buffer)4824 rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer)
4825 {
4826 rb_head_page_deactivate(cpu_buffer);
4827
4828 cpu_buffer->head_page
4829 = list_entry(cpu_buffer->pages, struct buffer_page, list);
4830 local_set(&cpu_buffer->head_page->write, 0);
4831 local_set(&cpu_buffer->head_page->entries, 0);
4832 local_set(&cpu_buffer->head_page->page->commit, 0);
4833
4834 cpu_buffer->head_page->read = 0;
4835
4836 cpu_buffer->tail_page = cpu_buffer->head_page;
4837 cpu_buffer->commit_page = cpu_buffer->head_page;
4838
4839 INIT_LIST_HEAD(&cpu_buffer->reader_page->list);
4840 INIT_LIST_HEAD(&cpu_buffer->new_pages);
4841 local_set(&cpu_buffer->reader_page->write, 0);
4842 local_set(&cpu_buffer->reader_page->entries, 0);
4843 local_set(&cpu_buffer->reader_page->page->commit, 0);
4844 cpu_buffer->reader_page->read = 0;
4845
4846 local_set(&cpu_buffer->entries_bytes, 0);
4847 local_set(&cpu_buffer->overrun, 0);
4848 local_set(&cpu_buffer->commit_overrun, 0);
4849 local_set(&cpu_buffer->dropped_events, 0);
4850 local_set(&cpu_buffer->entries, 0);
4851 local_set(&cpu_buffer->committing, 0);
4852 local_set(&cpu_buffer->commits, 0);
4853 local_set(&cpu_buffer->pages_touched, 0);
4854 local_set(&cpu_buffer->pages_read, 0);
4855 cpu_buffer->last_pages_touch = 0;
4856 cpu_buffer->shortest_full = 0;
4857 cpu_buffer->read = 0;
4858 cpu_buffer->read_bytes = 0;
4859
4860 rb_time_set(&cpu_buffer->write_stamp, 0);
4861 rb_time_set(&cpu_buffer->before_stamp, 0);
4862
4863 cpu_buffer->lost_events = 0;
4864 cpu_buffer->last_overrun = 0;
4865
4866 rb_head_page_activate(cpu_buffer);
4867 }
4868
4869 /* Must have disabled the cpu buffer then done a synchronize_rcu */
reset_disabled_cpu_buffer(struct ring_buffer_per_cpu * cpu_buffer)4870 static void reset_disabled_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer)
4871 {
4872 unsigned long flags;
4873
4874 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
4875
4876 if (RB_WARN_ON(cpu_buffer, local_read(&cpu_buffer->committing)))
4877 goto out;
4878
4879 arch_spin_lock(&cpu_buffer->lock);
4880
4881 rb_reset_cpu(cpu_buffer);
4882
4883 arch_spin_unlock(&cpu_buffer->lock);
4884
4885 out:
4886 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
4887 }
4888
4889 /**
4890 * ring_buffer_reset_cpu - reset a ring buffer per CPU buffer
4891 * @buffer: The ring buffer to reset a per cpu buffer of
4892 * @cpu: The CPU buffer to be reset
4893 */
ring_buffer_reset_cpu(struct trace_buffer * buffer,int cpu)4894 void ring_buffer_reset_cpu(struct trace_buffer *buffer, int cpu)
4895 {
4896 struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
4897
4898 if (!cpumask_test_cpu(cpu, buffer->cpumask))
4899 return;
4900
4901 /* prevent another thread from changing buffer sizes */
4902 mutex_lock(&buffer->mutex);
4903
4904 atomic_inc(&cpu_buffer->resize_disabled);
4905 atomic_inc(&cpu_buffer->record_disabled);
4906
4907 /* Make sure all commits have finished */
4908 synchronize_rcu();
4909
4910 reset_disabled_cpu_buffer(cpu_buffer);
4911
4912 atomic_dec(&cpu_buffer->record_disabled);
4913 atomic_dec(&cpu_buffer->resize_disabled);
4914
4915 mutex_unlock(&buffer->mutex);
4916 }
4917 EXPORT_SYMBOL_GPL(ring_buffer_reset_cpu);
4918
4919 /**
4920 * ring_buffer_reset_cpu - reset a ring buffer per CPU buffer
4921 * @buffer: The ring buffer to reset a per cpu buffer of
4922 * @cpu: The CPU buffer to be reset
4923 */
ring_buffer_reset_online_cpus(struct trace_buffer * buffer)4924 void ring_buffer_reset_online_cpus(struct trace_buffer *buffer)
4925 {
4926 struct ring_buffer_per_cpu *cpu_buffer;
4927 int cpu;
4928
4929 /* prevent another thread from changing buffer sizes */
4930 mutex_lock(&buffer->mutex);
4931
4932 for_each_online_buffer_cpu(buffer, cpu) {
4933 cpu_buffer = buffer->buffers[cpu];
4934
4935 atomic_inc(&cpu_buffer->resize_disabled);
4936 atomic_inc(&cpu_buffer->record_disabled);
4937 }
4938
4939 /* Make sure all commits have finished */
4940 synchronize_rcu();
4941
4942 for_each_online_buffer_cpu(buffer, cpu) {
4943 cpu_buffer = buffer->buffers[cpu];
4944
4945 reset_disabled_cpu_buffer(cpu_buffer);
4946
4947 atomic_dec(&cpu_buffer->record_disabled);
4948 atomic_dec(&cpu_buffer->resize_disabled);
4949 }
4950
4951 mutex_unlock(&buffer->mutex);
4952 }
4953
4954 /**
4955 * ring_buffer_reset - reset a ring buffer
4956 * @buffer: The ring buffer to reset all cpu buffers
4957 */
ring_buffer_reset(struct trace_buffer * buffer)4958 void ring_buffer_reset(struct trace_buffer *buffer)
4959 {
4960 struct ring_buffer_per_cpu *cpu_buffer;
4961 int cpu;
4962
4963 for_each_buffer_cpu(buffer, cpu) {
4964 cpu_buffer = buffer->buffers[cpu];
4965
4966 atomic_inc(&cpu_buffer->resize_disabled);
4967 atomic_inc(&cpu_buffer->record_disabled);
4968 }
4969
4970 /* Make sure all commits have finished */
4971 synchronize_rcu();
4972
4973 for_each_buffer_cpu(buffer, cpu) {
4974 cpu_buffer = buffer->buffers[cpu];
4975
4976 reset_disabled_cpu_buffer(cpu_buffer);
4977
4978 atomic_dec(&cpu_buffer->record_disabled);
4979 atomic_dec(&cpu_buffer->resize_disabled);
4980 }
4981 }
4982 EXPORT_SYMBOL_GPL(ring_buffer_reset);
4983
4984 /**
4985 * rind_buffer_empty - is the ring buffer empty?
4986 * @buffer: The ring buffer to test
4987 */
ring_buffer_empty(struct trace_buffer * buffer)4988 bool ring_buffer_empty(struct trace_buffer *buffer)
4989 {
4990 struct ring_buffer_per_cpu *cpu_buffer;
4991 unsigned long flags;
4992 bool dolock;
4993 int cpu;
4994 int ret;
4995
4996 /* yes this is racy, but if you don't like the race, lock the buffer */
4997 for_each_buffer_cpu(buffer, cpu) {
4998 cpu_buffer = buffer->buffers[cpu];
4999 local_irq_save(flags);
5000 dolock = rb_reader_lock(cpu_buffer);
5001 ret = rb_per_cpu_empty(cpu_buffer);
5002 rb_reader_unlock(cpu_buffer, dolock);
5003 local_irq_restore(flags);
5004
5005 if (!ret)
5006 return false;
5007 }
5008
5009 return true;
5010 }
5011 EXPORT_SYMBOL_GPL(ring_buffer_empty);
5012
5013 /**
5014 * ring_buffer_empty_cpu - is a cpu buffer of a ring buffer empty?
5015 * @buffer: The ring buffer
5016 * @cpu: The CPU buffer to test
5017 */
ring_buffer_empty_cpu(struct trace_buffer * buffer,int cpu)5018 bool ring_buffer_empty_cpu(struct trace_buffer *buffer, int cpu)
5019 {
5020 struct ring_buffer_per_cpu *cpu_buffer;
5021 unsigned long flags;
5022 bool dolock;
5023 int ret;
5024
5025 if (!cpumask_test_cpu(cpu, buffer->cpumask))
5026 return true;
5027
5028 cpu_buffer = buffer->buffers[cpu];
5029 local_irq_save(flags);
5030 dolock = rb_reader_lock(cpu_buffer);
5031 ret = rb_per_cpu_empty(cpu_buffer);
5032 rb_reader_unlock(cpu_buffer, dolock);
5033 local_irq_restore(flags);
5034
5035 return ret;
5036 }
5037 EXPORT_SYMBOL_GPL(ring_buffer_empty_cpu);
5038
5039 #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP
5040 /**
5041 * ring_buffer_swap_cpu - swap a CPU buffer between two ring buffers
5042 * @buffer_a: One buffer to swap with
5043 * @buffer_b: The other buffer to swap with
5044 * @cpu: the CPU of the buffers to swap
5045 *
5046 * This function is useful for tracers that want to take a "snapshot"
5047 * of a CPU buffer and has another back up buffer lying around.
5048 * it is expected that the tracer handles the cpu buffer not being
5049 * used at the moment.
5050 */
ring_buffer_swap_cpu(struct trace_buffer * buffer_a,struct trace_buffer * buffer_b,int cpu)5051 int ring_buffer_swap_cpu(struct trace_buffer *buffer_a,
5052 struct trace_buffer *buffer_b, int cpu)
5053 {
5054 struct ring_buffer_per_cpu *cpu_buffer_a;
5055 struct ring_buffer_per_cpu *cpu_buffer_b;
5056 int ret = -EINVAL;
5057
5058 if (!cpumask_test_cpu(cpu, buffer_a->cpumask) ||
5059 !cpumask_test_cpu(cpu, buffer_b->cpumask))
5060 goto out;
5061
5062 cpu_buffer_a = buffer_a->buffers[cpu];
5063 cpu_buffer_b = buffer_b->buffers[cpu];
5064
5065 /* At least make sure the two buffers are somewhat the same */
5066 if (cpu_buffer_a->nr_pages != cpu_buffer_b->nr_pages)
5067 goto out;
5068
5069 ret = -EAGAIN;
5070
5071 if (atomic_read(&buffer_a->record_disabled))
5072 goto out;
5073
5074 if (atomic_read(&buffer_b->record_disabled))
5075 goto out;
5076
5077 if (atomic_read(&cpu_buffer_a->record_disabled))
5078 goto out;
5079
5080 if (atomic_read(&cpu_buffer_b->record_disabled))
5081 goto out;
5082
5083 /*
5084 * We can't do a synchronize_rcu here because this
5085 * function can be called in atomic context.
5086 * Normally this will be called from the same CPU as cpu.
5087 * If not it's up to the caller to protect this.
5088 */
5089 atomic_inc(&cpu_buffer_a->record_disabled);
5090 atomic_inc(&cpu_buffer_b->record_disabled);
5091
5092 ret = -EBUSY;
5093 if (local_read(&cpu_buffer_a->committing))
5094 goto out_dec;
5095 if (local_read(&cpu_buffer_b->committing))
5096 goto out_dec;
5097
5098 buffer_a->buffers[cpu] = cpu_buffer_b;
5099 buffer_b->buffers[cpu] = cpu_buffer_a;
5100
5101 cpu_buffer_b->buffer = buffer_a;
5102 cpu_buffer_a->buffer = buffer_b;
5103
5104 ret = 0;
5105
5106 out_dec:
5107 atomic_dec(&cpu_buffer_a->record_disabled);
5108 atomic_dec(&cpu_buffer_b->record_disabled);
5109 out:
5110 return ret;
5111 }
5112 EXPORT_SYMBOL_GPL(ring_buffer_swap_cpu);
5113 #endif /* CONFIG_RING_BUFFER_ALLOW_SWAP */
5114
5115 /**
5116 * ring_buffer_alloc_read_page - allocate a page to read from buffer
5117 * @buffer: the buffer to allocate for.
5118 * @cpu: the cpu buffer to allocate.
5119 *
5120 * This function is used in conjunction with ring_buffer_read_page.
5121 * When reading a full page from the ring buffer, these functions
5122 * can be used to speed up the process. The calling function should
5123 * allocate a few pages first with this function. Then when it
5124 * needs to get pages from the ring buffer, it passes the result
5125 * of this function into ring_buffer_read_page, which will swap
5126 * the page that was allocated, with the read page of the buffer.
5127 *
5128 * Returns:
5129 * The page allocated, or ERR_PTR
5130 */
ring_buffer_alloc_read_page(struct trace_buffer * buffer,int cpu)5131 void *ring_buffer_alloc_read_page(struct trace_buffer *buffer, int cpu)
5132 {
5133 struct ring_buffer_per_cpu *cpu_buffer;
5134 struct buffer_data_page *bpage = NULL;
5135 unsigned long flags;
5136 struct page *page;
5137
5138 if (!cpumask_test_cpu(cpu, buffer->cpumask))
5139 return ERR_PTR(-ENODEV);
5140
5141 cpu_buffer = buffer->buffers[cpu];
5142 local_irq_save(flags);
5143 arch_spin_lock(&cpu_buffer->lock);
5144
5145 if (cpu_buffer->free_page) {
5146 bpage = cpu_buffer->free_page;
5147 cpu_buffer->free_page = NULL;
5148 }
5149
5150 arch_spin_unlock(&cpu_buffer->lock);
5151 local_irq_restore(flags);
5152
5153 if (bpage)
5154 goto out;
5155
5156 page = alloc_pages_node(cpu_to_node(cpu),
5157 GFP_KERNEL | __GFP_NORETRY, 0);
5158 if (!page)
5159 return ERR_PTR(-ENOMEM);
5160
5161 bpage = page_address(page);
5162
5163 out:
5164 rb_init_page(bpage);
5165
5166 return bpage;
5167 }
5168 EXPORT_SYMBOL_GPL(ring_buffer_alloc_read_page);
5169
5170 /**
5171 * ring_buffer_free_read_page - free an allocated read page
5172 * @buffer: the buffer the page was allocate for
5173 * @cpu: the cpu buffer the page came from
5174 * @data: the page to free
5175 *
5176 * Free a page allocated from ring_buffer_alloc_read_page.
5177 */
ring_buffer_free_read_page(struct trace_buffer * buffer,int cpu,void * data)5178 void ring_buffer_free_read_page(struct trace_buffer *buffer, int cpu, void *data)
5179 {
5180 struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
5181 struct buffer_data_page *bpage = data;
5182 struct page *page = virt_to_page(bpage);
5183 unsigned long flags;
5184
5185 /* If the page is still in use someplace else, we can't reuse it */
5186 if (page_ref_count(page) > 1)
5187 goto out;
5188
5189 local_irq_save(flags);
5190 arch_spin_lock(&cpu_buffer->lock);
5191
5192 if (!cpu_buffer->free_page) {
5193 cpu_buffer->free_page = bpage;
5194 bpage = NULL;
5195 }
5196
5197 arch_spin_unlock(&cpu_buffer->lock);
5198 local_irq_restore(flags);
5199
5200 out:
5201 free_page((unsigned long)bpage);
5202 }
5203 EXPORT_SYMBOL_GPL(ring_buffer_free_read_page);
5204
5205 /**
5206 * ring_buffer_read_page - extract a page from the ring buffer
5207 * @buffer: buffer to extract from
5208 * @data_page: the page to use allocated from ring_buffer_alloc_read_page
5209 * @len: amount to extract
5210 * @cpu: the cpu of the buffer to extract
5211 * @full: should the extraction only happen when the page is full.
5212 *
5213 * This function will pull out a page from the ring buffer and consume it.
5214 * @data_page must be the address of the variable that was returned
5215 * from ring_buffer_alloc_read_page. This is because the page might be used
5216 * to swap with a page in the ring buffer.
5217 *
5218 * for example:
5219 * rpage = ring_buffer_alloc_read_page(buffer, cpu);
5220 * if (IS_ERR(rpage))
5221 * return PTR_ERR(rpage);
5222 * ret = ring_buffer_read_page(buffer, &rpage, len, cpu, 0);
5223 * if (ret >= 0)
5224 * process_page(rpage, ret);
5225 *
5226 * When @full is set, the function will not return true unless
5227 * the writer is off the reader page.
5228 *
5229 * Note: it is up to the calling functions to handle sleeps and wakeups.
5230 * The ring buffer can be used anywhere in the kernel and can not
5231 * blindly call wake_up. The layer that uses the ring buffer must be
5232 * responsible for that.
5233 *
5234 * Returns:
5235 * >=0 if data has been transferred, returns the offset of consumed data.
5236 * <0 if no data has been transferred.
5237 */
ring_buffer_read_page(struct trace_buffer * buffer,void ** data_page,size_t len,int cpu,int full)5238 int ring_buffer_read_page(struct trace_buffer *buffer,
5239 void **data_page, size_t len, int cpu, int full)
5240 {
5241 struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
5242 struct ring_buffer_event *event;
5243 struct buffer_data_page *bpage;
5244 struct buffer_page *reader;
5245 unsigned long missed_events;
5246 unsigned long flags;
5247 unsigned int commit;
5248 unsigned int read;
5249 u64 save_timestamp;
5250 int ret = -1;
5251
5252 if (!cpumask_test_cpu(cpu, buffer->cpumask))
5253 goto out;
5254
5255 /*
5256 * If len is not big enough to hold the page header, then
5257 * we can not copy anything.
5258 */
5259 if (len <= BUF_PAGE_HDR_SIZE)
5260 goto out;
5261
5262 len -= BUF_PAGE_HDR_SIZE;
5263
5264 if (!data_page)
5265 goto out;
5266
5267 bpage = *data_page;
5268 if (!bpage)
5269 goto out;
5270
5271 raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags);
5272
5273 reader = rb_get_reader_page(cpu_buffer);
5274 if (!reader)
5275 goto out_unlock;
5276
5277 event = rb_reader_event(cpu_buffer);
5278
5279 read = reader->read;
5280 commit = rb_page_commit(reader);
5281
5282 /* Check if any events were dropped */
5283 missed_events = cpu_buffer->lost_events;
5284
5285 /*
5286 * If this page has been partially read or
5287 * if len is not big enough to read the rest of the page or
5288 * a writer is still on the page, then
5289 * we must copy the data from the page to the buffer.
5290 * Otherwise, we can simply swap the page with the one passed in.
5291 */
5292 if (read || (len < (commit - read)) ||
5293 cpu_buffer->reader_page == cpu_buffer->commit_page) {
5294 struct buffer_data_page *rpage = cpu_buffer->reader_page->page;
5295 unsigned int rpos = read;
5296 unsigned int pos = 0;
5297 unsigned int size;
5298
5299 if (full)
5300 goto out_unlock;
5301
5302 if (len > (commit - read))
5303 len = (commit - read);
5304
5305 /* Always keep the time extend and data together */
5306 size = rb_event_ts_length(event);
5307
5308 if (len < size)
5309 goto out_unlock;
5310
5311 /* save the current timestamp, since the user will need it */
5312 save_timestamp = cpu_buffer->read_stamp;
5313
5314 /* Need to copy one event at a time */
5315 do {
5316 /* We need the size of one event, because
5317 * rb_advance_reader only advances by one event,
5318 * whereas rb_event_ts_length may include the size of
5319 * one or two events.
5320 * We have already ensured there's enough space if this
5321 * is a time extend. */
5322 size = rb_event_length(event);
5323 memcpy(bpage->data + pos, rpage->data + rpos, size);
5324
5325 len -= size;
5326
5327 rb_advance_reader(cpu_buffer);
5328 rpos = reader->read;
5329 pos += size;
5330
5331 if (rpos >= commit)
5332 break;
5333
5334 event = rb_reader_event(cpu_buffer);
5335 /* Always keep the time extend and data together */
5336 size = rb_event_ts_length(event);
5337 } while (len >= size);
5338
5339 /* update bpage */
5340 local_set(&bpage->commit, pos);
5341 bpage->time_stamp = save_timestamp;
5342
5343 /* we copied everything to the beginning */
5344 read = 0;
5345 } else {
5346 /* update the entry counter */
5347 cpu_buffer->read += rb_page_entries(reader);
5348 cpu_buffer->read_bytes += BUF_PAGE_SIZE;
5349
5350 /* swap the pages */
5351 rb_init_page(bpage);
5352 bpage = reader->page;
5353 reader->page = *data_page;
5354 local_set(&reader->write, 0);
5355 local_set(&reader->entries, 0);
5356 reader->read = 0;
5357 *data_page = bpage;
5358
5359 /*
5360 * Use the real_end for the data size,
5361 * This gives us a chance to store the lost events
5362 * on the page.
5363 */
5364 if (reader->real_end)
5365 local_set(&bpage->commit, reader->real_end);
5366 }
5367 ret = read;
5368
5369 cpu_buffer->lost_events = 0;
5370
5371 commit = local_read(&bpage->commit);
5372 /*
5373 * Set a flag in the commit field if we lost events
5374 */
5375 if (missed_events) {
5376 /* If there is room at the end of the page to save the
5377 * missed events, then record it there.
5378 */
5379 if (BUF_PAGE_SIZE - commit >= sizeof(missed_events)) {
5380 memcpy(&bpage->data[commit], &missed_events,
5381 sizeof(missed_events));
5382 local_add(RB_MISSED_STORED, &bpage->commit);
5383 commit += sizeof(missed_events);
5384 }
5385 local_add(RB_MISSED_EVENTS, &bpage->commit);
5386 }
5387
5388 /*
5389 * This page may be off to user land. Zero it out here.
5390 */
5391 if (commit < BUF_PAGE_SIZE)
5392 memset(&bpage->data[commit], 0, BUF_PAGE_SIZE - commit);
5393
5394 out_unlock:
5395 raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags);
5396
5397 out:
5398 return ret;
5399 }
5400 EXPORT_SYMBOL_GPL(ring_buffer_read_page);
5401
5402 /*
5403 * We only allocate new buffers, never free them if the CPU goes down.
5404 * If we were to free the buffer, then the user would lose any trace that was in
5405 * the buffer.
5406 */
trace_rb_cpu_prepare(unsigned int cpu,struct hlist_node * node)5407 int trace_rb_cpu_prepare(unsigned int cpu, struct hlist_node *node)
5408 {
5409 struct trace_buffer *buffer;
5410 long nr_pages_same;
5411 int cpu_i;
5412 unsigned long nr_pages;
5413
5414 buffer = container_of(node, struct trace_buffer, node);
5415 if (cpumask_test_cpu(cpu, buffer->cpumask))
5416 return 0;
5417
5418 nr_pages = 0;
5419 nr_pages_same = 1;
5420 /* check if all cpu sizes are same */
5421 for_each_buffer_cpu(buffer, cpu_i) {
5422 /* fill in the size from first enabled cpu */
5423 if (nr_pages == 0)
5424 nr_pages = buffer->buffers[cpu_i]->nr_pages;
5425 if (nr_pages != buffer->buffers[cpu_i]->nr_pages) {
5426 nr_pages_same = 0;
5427 break;
5428 }
5429 }
5430 /* allocate minimum pages, user can later expand it */
5431 if (!nr_pages_same)
5432 nr_pages = 2;
5433 buffer->buffers[cpu] =
5434 rb_allocate_cpu_buffer(buffer, nr_pages, cpu);
5435 if (!buffer->buffers[cpu]) {
5436 WARN(1, "failed to allocate ring buffer on CPU %u\n",
5437 cpu);
5438 return -ENOMEM;
5439 }
5440 smp_wmb();
5441 cpumask_set_cpu(cpu, buffer->cpumask);
5442 return 0;
5443 }
5444
5445 #ifdef CONFIG_RING_BUFFER_STARTUP_TEST
5446 /*
5447 * This is a basic integrity check of the ring buffer.
5448 * Late in the boot cycle this test will run when configured in.
5449 * It will kick off a thread per CPU that will go into a loop
5450 * writing to the per cpu ring buffer various sizes of data.
5451 * Some of the data will be large items, some small.
5452 *
5453 * Another thread is created that goes into a spin, sending out
5454 * IPIs to the other CPUs to also write into the ring buffer.
5455 * this is to test the nesting ability of the buffer.
5456 *
5457 * Basic stats are recorded and reported. If something in the
5458 * ring buffer should happen that's not expected, a big warning
5459 * is displayed and all ring buffers are disabled.
5460 */
5461 static struct task_struct *rb_threads[NR_CPUS] __initdata;
5462
5463 struct rb_test_data {
5464 struct trace_buffer *buffer;
5465 unsigned long events;
5466 unsigned long bytes_written;
5467 unsigned long bytes_alloc;
5468 unsigned long bytes_dropped;
5469 unsigned long events_nested;
5470 unsigned long bytes_written_nested;
5471 unsigned long bytes_alloc_nested;
5472 unsigned long bytes_dropped_nested;
5473 int min_size_nested;
5474 int max_size_nested;
5475 int max_size;
5476 int min_size;
5477 int cpu;
5478 int cnt;
5479 };
5480
5481 static struct rb_test_data rb_data[NR_CPUS] __initdata;
5482
5483 /* 1 meg per cpu */
5484 #define RB_TEST_BUFFER_SIZE 1048576
5485
5486 static char rb_string[] __initdata =
5487 "abcdefghijklmnopqrstuvwxyz1234567890!@#$%^&*()?+\\"
5488 "?+|:';\",.<>/?abcdefghijklmnopqrstuvwxyz1234567890"
5489 "!@#$%^&*()?+\\?+|:';\",.<>/?abcdefghijklmnopqrstuv";
5490
5491 static bool rb_test_started __initdata;
5492
5493 struct rb_item {
5494 int size;
5495 char str[];
5496 };
5497
rb_write_something(struct rb_test_data * data,bool nested)5498 static __init int rb_write_something(struct rb_test_data *data, bool nested)
5499 {
5500 struct ring_buffer_event *event;
5501 struct rb_item *item;
5502 bool started;
5503 int event_len;
5504 int size;
5505 int len;
5506 int cnt;
5507
5508 /* Have nested writes different that what is written */
5509 cnt = data->cnt + (nested ? 27 : 0);
5510
5511 /* Multiply cnt by ~e, to make some unique increment */
5512 size = (cnt * 68 / 25) % (sizeof(rb_string) - 1);
5513
5514 len = size + sizeof(struct rb_item);
5515
5516 started = rb_test_started;
5517 /* read rb_test_started before checking buffer enabled */
5518 smp_rmb();
5519
5520 event = ring_buffer_lock_reserve(data->buffer, len);
5521 if (!event) {
5522 /* Ignore dropped events before test starts. */
5523 if (started) {
5524 if (nested)
5525 data->bytes_dropped += len;
5526 else
5527 data->bytes_dropped_nested += len;
5528 }
5529 return len;
5530 }
5531
5532 event_len = ring_buffer_event_length(event);
5533
5534 if (RB_WARN_ON(data->buffer, event_len < len))
5535 goto out;
5536
5537 item = ring_buffer_event_data(event);
5538 item->size = size;
5539 memcpy(item->str, rb_string, size);
5540
5541 if (nested) {
5542 data->bytes_alloc_nested += event_len;
5543 data->bytes_written_nested += len;
5544 data->events_nested++;
5545 if (!data->min_size_nested || len < data->min_size_nested)
5546 data->min_size_nested = len;
5547 if (len > data->max_size_nested)
5548 data->max_size_nested = len;
5549 } else {
5550 data->bytes_alloc += event_len;
5551 data->bytes_written += len;
5552 data->events++;
5553 if (!data->min_size || len < data->min_size)
5554 data->max_size = len;
5555 if (len > data->max_size)
5556 data->max_size = len;
5557 }
5558
5559 out:
5560 ring_buffer_unlock_commit(data->buffer, event);
5561
5562 return 0;
5563 }
5564
rb_test(void * arg)5565 static __init int rb_test(void *arg)
5566 {
5567 struct rb_test_data *data = arg;
5568
5569 while (!kthread_should_stop()) {
5570 rb_write_something(data, false);
5571 data->cnt++;
5572
5573 set_current_state(TASK_INTERRUPTIBLE);
5574 /* Now sleep between a min of 100-300us and a max of 1ms */
5575 usleep_range(((data->cnt % 3) + 1) * 100, 1000);
5576 }
5577
5578 return 0;
5579 }
5580
rb_ipi(void * ignore)5581 static __init void rb_ipi(void *ignore)
5582 {
5583 struct rb_test_data *data;
5584 int cpu = smp_processor_id();
5585
5586 data = &rb_data[cpu];
5587 rb_write_something(data, true);
5588 }
5589
rb_hammer_test(void * arg)5590 static __init int rb_hammer_test(void *arg)
5591 {
5592 while (!kthread_should_stop()) {
5593
5594 /* Send an IPI to all cpus to write data! */
5595 smp_call_function(rb_ipi, NULL, 1);
5596 /* No sleep, but for non preempt, let others run */
5597 schedule();
5598 }
5599
5600 return 0;
5601 }
5602
test_ringbuffer(void)5603 static __init int test_ringbuffer(void)
5604 {
5605 struct task_struct *rb_hammer;
5606 struct trace_buffer *buffer;
5607 int cpu;
5608 int ret = 0;
5609
5610 if (security_locked_down(LOCKDOWN_TRACEFS)) {
5611 pr_warn("Lockdown is enabled, skipping ring buffer tests\n");
5612 return 0;
5613 }
5614
5615 pr_info("Running ring buffer tests...\n");
5616
5617 buffer = ring_buffer_alloc(RB_TEST_BUFFER_SIZE, RB_FL_OVERWRITE);
5618 if (WARN_ON(!buffer))
5619 return 0;
5620
5621 /* Disable buffer so that threads can't write to it yet */
5622 ring_buffer_record_off(buffer);
5623
5624 for_each_online_cpu(cpu) {
5625 rb_data[cpu].buffer = buffer;
5626 rb_data[cpu].cpu = cpu;
5627 rb_data[cpu].cnt = cpu;
5628 rb_threads[cpu] = kthread_create(rb_test, &rb_data[cpu],
5629 "rbtester/%d", cpu);
5630 if (WARN_ON(IS_ERR(rb_threads[cpu]))) {
5631 pr_cont("FAILED\n");
5632 ret = PTR_ERR(rb_threads[cpu]);
5633 goto out_free;
5634 }
5635
5636 kthread_bind(rb_threads[cpu], cpu);
5637 wake_up_process(rb_threads[cpu]);
5638 }
5639
5640 /* Now create the rb hammer! */
5641 rb_hammer = kthread_run(rb_hammer_test, NULL, "rbhammer");
5642 if (WARN_ON(IS_ERR(rb_hammer))) {
5643 pr_cont("FAILED\n");
5644 ret = PTR_ERR(rb_hammer);
5645 goto out_free;
5646 }
5647
5648 ring_buffer_record_on(buffer);
5649 /*
5650 * Show buffer is enabled before setting rb_test_started.
5651 * Yes there's a small race window where events could be
5652 * dropped and the thread wont catch it. But when a ring
5653 * buffer gets enabled, there will always be some kind of
5654 * delay before other CPUs see it. Thus, we don't care about
5655 * those dropped events. We care about events dropped after
5656 * the threads see that the buffer is active.
5657 */
5658 smp_wmb();
5659 rb_test_started = true;
5660
5661 set_current_state(TASK_INTERRUPTIBLE);
5662 /* Just run for 10 seconds */;
5663 schedule_timeout(10 * HZ);
5664
5665 kthread_stop(rb_hammer);
5666
5667 out_free:
5668 for_each_online_cpu(cpu) {
5669 if (!rb_threads[cpu])
5670 break;
5671 kthread_stop(rb_threads[cpu]);
5672 }
5673 if (ret) {
5674 ring_buffer_free(buffer);
5675 return ret;
5676 }
5677
5678 /* Report! */
5679 pr_info("finished\n");
5680 for_each_online_cpu(cpu) {
5681 struct ring_buffer_event *event;
5682 struct rb_test_data *data = &rb_data[cpu];
5683 struct rb_item *item;
5684 unsigned long total_events;
5685 unsigned long total_dropped;
5686 unsigned long total_written;
5687 unsigned long total_alloc;
5688 unsigned long total_read = 0;
5689 unsigned long total_size = 0;
5690 unsigned long total_len = 0;
5691 unsigned long total_lost = 0;
5692 unsigned long lost;
5693 int big_event_size;
5694 int small_event_size;
5695
5696 ret = -1;
5697
5698 total_events = data->events + data->events_nested;
5699 total_written = data->bytes_written + data->bytes_written_nested;
5700 total_alloc = data->bytes_alloc + data->bytes_alloc_nested;
5701 total_dropped = data->bytes_dropped + data->bytes_dropped_nested;
5702
5703 big_event_size = data->max_size + data->max_size_nested;
5704 small_event_size = data->min_size + data->min_size_nested;
5705
5706 pr_info("CPU %d:\n", cpu);
5707 pr_info(" events: %ld\n", total_events);
5708 pr_info(" dropped bytes: %ld\n", total_dropped);
5709 pr_info(" alloced bytes: %ld\n", total_alloc);
5710 pr_info(" written bytes: %ld\n", total_written);
5711 pr_info(" biggest event: %d\n", big_event_size);
5712 pr_info(" smallest event: %d\n", small_event_size);
5713
5714 if (RB_WARN_ON(buffer, total_dropped))
5715 break;
5716
5717 ret = 0;
5718
5719 while ((event = ring_buffer_consume(buffer, cpu, NULL, &lost))) {
5720 total_lost += lost;
5721 item = ring_buffer_event_data(event);
5722 total_len += ring_buffer_event_length(event);
5723 total_size += item->size + sizeof(struct rb_item);
5724 if (memcmp(&item->str[0], rb_string, item->size) != 0) {
5725 pr_info("FAILED!\n");
5726 pr_info("buffer had: %.*s\n", item->size, item->str);
5727 pr_info("expected: %.*s\n", item->size, rb_string);
5728 RB_WARN_ON(buffer, 1);
5729 ret = -1;
5730 break;
5731 }
5732 total_read++;
5733 }
5734 if (ret)
5735 break;
5736
5737 ret = -1;
5738
5739 pr_info(" read events: %ld\n", total_read);
5740 pr_info(" lost events: %ld\n", total_lost);
5741 pr_info(" total events: %ld\n", total_lost + total_read);
5742 pr_info(" recorded len bytes: %ld\n", total_len);
5743 pr_info(" recorded size bytes: %ld\n", total_size);
5744 if (total_lost)
5745 pr_info(" With dropped events, record len and size may not match\n"
5746 " alloced and written from above\n");
5747 if (!total_lost) {
5748 if (RB_WARN_ON(buffer, total_len != total_alloc ||
5749 total_size != total_written))
5750 break;
5751 }
5752 if (RB_WARN_ON(buffer, total_lost + total_read != total_events))
5753 break;
5754
5755 ret = 0;
5756 }
5757 if (!ret)
5758 pr_info("Ring buffer PASSED!\n");
5759
5760 ring_buffer_free(buffer);
5761 return 0;
5762 }
5763
5764 late_initcall(test_ringbuffer);
5765 #endif /* CONFIG_RING_BUFFER_STARTUP_TEST */
5766