Commit | Line | Data |
---|---|---|
7a8e76a3 SR |
1 | /* |
2 | * Generic ring buffer | |
3 | * | |
4 | * Copyright (C) 2008 Steven Rostedt <srostedt@redhat.com> | |
5 | */ | |
6 | #include <linux/ring_buffer.h> | |
14131f2f | 7 | #include <linux/trace_clock.h> |
7a8e76a3 SR |
8 | #include <linux/spinlock.h> |
9 | #include <linux/debugfs.h> | |
10 | #include <linux/uaccess.h> | |
a81bd80a | 11 | #include <linux/hardirq.h> |
1744a21d | 12 | #include <linux/kmemcheck.h> |
7a8e76a3 SR |
13 | #include <linux/module.h> |
14 | #include <linux/percpu.h> | |
15 | #include <linux/mutex.h> | |
5a0e3ad6 | 16 | #include <linux/slab.h> |
7a8e76a3 SR |
17 | #include <linux/init.h> |
18 | #include <linux/hash.h> | |
19 | #include <linux/list.h> | |
554f786e | 20 | #include <linux/cpu.h> |
7a8e76a3 SR |
21 | #include <linux/fs.h> |
22 | ||
79615760 | 23 | #include <asm/local.h> |
182e9f5f SR |
24 | #include "trace.h" |
25 | ||
d1b182a8 SR |
26 | /* |
27 | * The ring buffer header is special. We must manually up keep it. | |
28 | */ | |
29 | int ring_buffer_print_entry_header(struct trace_seq *s) | |
30 | { | |
31 | int ret; | |
32 | ||
334d4169 LJ |
33 | ret = trace_seq_printf(s, "# compressed entry header\n"); |
34 | ret = trace_seq_printf(s, "\ttype_len : 5 bits\n"); | |
d1b182a8 SR |
35 | ret = trace_seq_printf(s, "\ttime_delta : 27 bits\n"); |
36 | ret = trace_seq_printf(s, "\tarray : 32 bits\n"); | |
37 | ret = trace_seq_printf(s, "\n"); | |
38 | ret = trace_seq_printf(s, "\tpadding : type == %d\n", | |
39 | RINGBUF_TYPE_PADDING); | |
40 | ret = trace_seq_printf(s, "\ttime_extend : type == %d\n", | |
41 | RINGBUF_TYPE_TIME_EXTEND); | |
334d4169 LJ |
42 | ret = trace_seq_printf(s, "\tdata max type_len == %d\n", |
43 | RINGBUF_TYPE_DATA_TYPE_LEN_MAX); | |
d1b182a8 SR |
44 | |
45 | return ret; | |
46 | } | |
47 | ||
5cc98548 SR |
48 | /* |
49 | * The ring buffer is made up of a list of pages. A separate list of pages is | |
50 | * allocated for each CPU. A writer may only write to a buffer that is | |
51 | * associated with the CPU it is currently executing on. A reader may read | |
52 | * from any per cpu buffer. | |
53 | * | |
54 | * The reader is special. For each per cpu buffer, the reader has its own | |
55 | * reader page. When a reader has read the entire reader page, this reader | |
56 | * page is swapped with another page in the ring buffer. | |
57 | * | |
58 | * Now, as long as the writer is off the reader page, the reader can do what | |
59 | * ever it wants with that page. The writer will never write to that page | |
60 | * again (as long as it is out of the ring buffer). | |
61 | * | |
62 | * Here's some silly ASCII art. | |
63 | * | |
64 | * +------+ | |
65 | * |reader| RING BUFFER | |
66 | * |page | | |
67 | * +------+ +---+ +---+ +---+ | |
68 | * | |-->| |-->| | | |
69 | * +---+ +---+ +---+ | |
70 | * ^ | | |
71 | * | | | |
72 | * +---------------+ | |
73 | * | |
74 | * | |
75 | * +------+ | |
76 | * |reader| RING BUFFER | |
77 | * |page |------------------v | |
78 | * +------+ +---+ +---+ +---+ | |
79 | * | |-->| |-->| | | |
80 | * +---+ +---+ +---+ | |
81 | * ^ | | |
82 | * | | | |
83 | * +---------------+ | |
84 | * | |
85 | * | |
86 | * +------+ | |
87 | * |reader| RING BUFFER | |
88 | * |page |------------------v | |
89 | * +------+ +---+ +---+ +---+ | |
90 | * ^ | |-->| |-->| | | |
91 | * | +---+ +---+ +---+ | |
92 | * | | | |
93 | * | | | |
94 | * +------------------------------+ | |
95 | * | |
96 | * | |
97 | * +------+ | |
98 | * |buffer| RING BUFFER | |
99 | * |page |------------------v | |
100 | * +------+ +---+ +---+ +---+ | |
101 | * ^ | | | |-->| | | |
102 | * | New +---+ +---+ +---+ | |
103 | * | Reader------^ | | |
104 | * | page | | |
105 | * +------------------------------+ | |
106 | * | |
107 | * | |
108 | * After we make this swap, the reader can hand this page off to the splice | |
109 | * code and be done with it. It can even allocate a new page if it needs to | |
110 | * and swap that into the ring buffer. | |
111 | * | |
112 | * We will be using cmpxchg soon to make all this lockless. | |
113 | * | |
114 | */ | |
115 | ||
033601a3 SR |
116 | /* |
117 | * A fast way to enable or disable all ring buffers is to | |
118 | * call tracing_on or tracing_off. Turning off the ring buffers | |
119 | * prevents all ring buffers from being recorded to. | |
120 | * Turning this switch on, makes it OK to write to the | |
121 | * ring buffer, if the ring buffer is enabled itself. | |
122 | * | |
123 | * There's three layers that must be on in order to write | |
124 | * to the ring buffer. | |
125 | * | |
126 | * 1) This global flag must be set. | |
127 | * 2) The ring buffer must be enabled for recording. | |
128 | * 3) The per cpu buffer must be enabled for recording. | |
129 | * | |
130 | * In case of an anomaly, this global flag has a bit set that | |
131 | * will permantly disable all ring buffers. | |
132 | */ | |
133 | ||
134 | /* | |
135 | * Global flag to disable all recording to ring buffers | |
136 | * This has two bits: ON, DISABLED | |
137 | * | |
138 | * ON DISABLED | |
139 | * ---- ---------- | |
140 | * 0 0 : ring buffers are off | |
141 | * 1 0 : ring buffers are on | |
142 | * X 1 : ring buffers are permanently disabled | |
143 | */ | |
144 | ||
145 | enum { | |
146 | RB_BUFFERS_ON_BIT = 0, | |
147 | RB_BUFFERS_DISABLED_BIT = 1, | |
148 | }; | |
149 | ||
150 | enum { | |
151 | RB_BUFFERS_ON = 1 << RB_BUFFERS_ON_BIT, | |
152 | RB_BUFFERS_DISABLED = 1 << RB_BUFFERS_DISABLED_BIT, | |
153 | }; | |
154 | ||
5e39841c | 155 | static unsigned long ring_buffer_flags __read_mostly = RB_BUFFERS_ON; |
a3583244 | 156 | |
474d32b6 SR |
157 | #define BUF_PAGE_HDR_SIZE offsetof(struct buffer_data_page, data) |
158 | ||
a3583244 SR |
159 | /** |
160 | * tracing_on - enable all tracing buffers | |
161 | * | |
162 | * This function enables all tracing buffers that may have been | |
163 | * disabled with tracing_off. | |
164 | */ | |
165 | void tracing_on(void) | |
166 | { | |
033601a3 | 167 | set_bit(RB_BUFFERS_ON_BIT, &ring_buffer_flags); |
a3583244 | 168 | } |
c4f50183 | 169 | EXPORT_SYMBOL_GPL(tracing_on); |
a3583244 SR |
170 | |
171 | /** | |
172 | * tracing_off - turn off all tracing buffers | |
173 | * | |
174 | * This function stops all tracing buffers from recording data. | |
175 | * It does not disable any overhead the tracers themselves may | |
176 | * be causing. This function simply causes all recording to | |
177 | * the ring buffers to fail. | |
178 | */ | |
179 | void tracing_off(void) | |
180 | { | |
033601a3 SR |
181 | clear_bit(RB_BUFFERS_ON_BIT, &ring_buffer_flags); |
182 | } | |
c4f50183 | 183 | EXPORT_SYMBOL_GPL(tracing_off); |
033601a3 SR |
184 | |
185 | /** | |
186 | * tracing_off_permanent - permanently disable ring buffers | |
187 | * | |
188 | * This function, once called, will disable all ring buffers | |
c3706f00 | 189 | * permanently. |
033601a3 SR |
190 | */ |
191 | void tracing_off_permanent(void) | |
192 | { | |
193 | set_bit(RB_BUFFERS_DISABLED_BIT, &ring_buffer_flags); | |
a3583244 SR |
194 | } |
195 | ||
988ae9d6 SR |
196 | /** |
197 | * tracing_is_on - show state of ring buffers enabled | |
198 | */ | |
199 | int tracing_is_on(void) | |
200 | { | |
201 | return ring_buffer_flags == RB_BUFFERS_ON; | |
202 | } | |
203 | EXPORT_SYMBOL_GPL(tracing_is_on); | |
204 | ||
e3d6bf0a | 205 | #define RB_EVNT_HDR_SIZE (offsetof(struct ring_buffer_event, array)) |
67d34724 | 206 | #define RB_ALIGNMENT 4U |
334d4169 | 207 | #define RB_MAX_SMALL_DATA (RB_ALIGNMENT * RINGBUF_TYPE_DATA_TYPE_LEN_MAX) |
c7b09308 | 208 | #define RB_EVNT_MIN_SIZE 8U /* two 32bit words */ |
334d4169 | 209 | |
2271048d SR |
210 | #if !defined(CONFIG_64BIT) || defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) |
211 | # define RB_FORCE_8BYTE_ALIGNMENT 0 | |
212 | # define RB_ARCH_ALIGNMENT RB_ALIGNMENT | |
213 | #else | |
214 | # define RB_FORCE_8BYTE_ALIGNMENT 1 | |
215 | # define RB_ARCH_ALIGNMENT 8U | |
216 | #endif | |
217 | ||
334d4169 LJ |
218 | /* define RINGBUF_TYPE_DATA for 'case RINGBUF_TYPE_DATA:' */ |
219 | #define RINGBUF_TYPE_DATA 0 ... RINGBUF_TYPE_DATA_TYPE_LEN_MAX | |
7a8e76a3 SR |
220 | |
221 | enum { | |
222 | RB_LEN_TIME_EXTEND = 8, | |
223 | RB_LEN_TIME_STAMP = 16, | |
224 | }; | |
225 | ||
69d1b839 SR |
226 | #define skip_time_extend(event) \ |
227 | ((struct ring_buffer_event *)((char *)event + RB_LEN_TIME_EXTEND)) | |
228 | ||
2d622719 TZ |
229 | static inline int rb_null_event(struct ring_buffer_event *event) |
230 | { | |
a1863c21 | 231 | return event->type_len == RINGBUF_TYPE_PADDING && !event->time_delta; |
2d622719 TZ |
232 | } |
233 | ||
234 | static void rb_event_set_padding(struct ring_buffer_event *event) | |
235 | { | |
a1863c21 | 236 | /* padding has a NULL time_delta */ |
334d4169 | 237 | event->type_len = RINGBUF_TYPE_PADDING; |
2d622719 TZ |
238 | event->time_delta = 0; |
239 | } | |
240 | ||
34a148bf | 241 | static unsigned |
2d622719 | 242 | rb_event_data_length(struct ring_buffer_event *event) |
7a8e76a3 SR |
243 | { |
244 | unsigned length; | |
245 | ||
334d4169 LJ |
246 | if (event->type_len) |
247 | length = event->type_len * RB_ALIGNMENT; | |
2d622719 TZ |
248 | else |
249 | length = event->array[0]; | |
250 | return length + RB_EVNT_HDR_SIZE; | |
251 | } | |
252 | ||
69d1b839 SR |
253 | /* |
254 | * Return the length of the given event. Will return | |
255 | * the length of the time extend if the event is a | |
256 | * time extend. | |
257 | */ | |
258 | static inline unsigned | |
2d622719 TZ |
259 | rb_event_length(struct ring_buffer_event *event) |
260 | { | |
334d4169 | 261 | switch (event->type_len) { |
7a8e76a3 | 262 | case RINGBUF_TYPE_PADDING: |
2d622719 TZ |
263 | if (rb_null_event(event)) |
264 | /* undefined */ | |
265 | return -1; | |
334d4169 | 266 | return event->array[0] + RB_EVNT_HDR_SIZE; |
7a8e76a3 SR |
267 | |
268 | case RINGBUF_TYPE_TIME_EXTEND: | |
269 | return RB_LEN_TIME_EXTEND; | |
270 | ||
271 | case RINGBUF_TYPE_TIME_STAMP: | |
272 | return RB_LEN_TIME_STAMP; | |
273 | ||
274 | case RINGBUF_TYPE_DATA: | |
2d622719 | 275 | return rb_event_data_length(event); |
7a8e76a3 SR |
276 | default: |
277 | BUG(); | |
278 | } | |
279 | /* not hit */ | |
280 | return 0; | |
281 | } | |
282 | ||
69d1b839 SR |
283 | /* |
284 | * Return total length of time extend and data, | |
285 | * or just the event length for all other events. | |
286 | */ | |
287 | static inline unsigned | |
288 | rb_event_ts_length(struct ring_buffer_event *event) | |
289 | { | |
290 | unsigned len = 0; | |
291 | ||
292 | if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) { | |
293 | /* time extends include the data event after it */ | |
294 | len = RB_LEN_TIME_EXTEND; | |
295 | event = skip_time_extend(event); | |
296 | } | |
297 | return len + rb_event_length(event); | |
298 | } | |
299 | ||
7a8e76a3 SR |
300 | /** |
301 | * ring_buffer_event_length - return the length of the event | |
302 | * @event: the event to get the length of | |
69d1b839 SR |
303 | * |
304 | * Returns the size of the data load of a data event. | |
305 | * If the event is something other than a data event, it | |
306 | * returns the size of the event itself. With the exception | |
307 | * of a TIME EXTEND, where it still returns the size of the | |
308 | * data load of the data event after it. | |
7a8e76a3 SR |
309 | */ |
310 | unsigned ring_buffer_event_length(struct ring_buffer_event *event) | |
311 | { | |
69d1b839 SR |
312 | unsigned length; |
313 | ||
314 | if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) | |
315 | event = skip_time_extend(event); | |
316 | ||
317 | length = rb_event_length(event); | |
334d4169 | 318 | if (event->type_len > RINGBUF_TYPE_DATA_TYPE_LEN_MAX) |
465634ad RR |
319 | return length; |
320 | length -= RB_EVNT_HDR_SIZE; | |
321 | if (length > RB_MAX_SMALL_DATA + sizeof(event->array[0])) | |
322 | length -= sizeof(event->array[0]); | |
323 | return length; | |
7a8e76a3 | 324 | } |
c4f50183 | 325 | EXPORT_SYMBOL_GPL(ring_buffer_event_length); |
7a8e76a3 SR |
326 | |
327 | /* inline for ring buffer fast paths */ | |
34a148bf | 328 | static void * |
7a8e76a3 SR |
329 | rb_event_data(struct ring_buffer_event *event) |
330 | { | |
69d1b839 SR |
331 | if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) |
332 | event = skip_time_extend(event); | |
334d4169 | 333 | BUG_ON(event->type_len > RINGBUF_TYPE_DATA_TYPE_LEN_MAX); |
7a8e76a3 | 334 | /* If length is in len field, then array[0] has the data */ |
334d4169 | 335 | if (event->type_len) |
7a8e76a3 SR |
336 | return (void *)&event->array[0]; |
337 | /* Otherwise length is in array[0] and array[1] has the data */ | |
338 | return (void *)&event->array[1]; | |
339 | } | |
340 | ||
341 | /** | |
342 | * ring_buffer_event_data - return the data of the event | |
343 | * @event: the event to get the data from | |
344 | */ | |
345 | void *ring_buffer_event_data(struct ring_buffer_event *event) | |
346 | { | |
347 | return rb_event_data(event); | |
348 | } | |
c4f50183 | 349 | EXPORT_SYMBOL_GPL(ring_buffer_event_data); |
7a8e76a3 SR |
350 | |
351 | #define for_each_buffer_cpu(buffer, cpu) \ | |
9e01c1b7 | 352 | for_each_cpu(cpu, buffer->cpumask) |
7a8e76a3 SR |
353 | |
354 | #define TS_SHIFT 27 | |
355 | #define TS_MASK ((1ULL << TS_SHIFT) - 1) | |
356 | #define TS_DELTA_TEST (~TS_MASK) | |
357 | ||
66a8cb95 SR |
358 | /* Flag when events were overwritten */ |
359 | #define RB_MISSED_EVENTS (1 << 31) | |
ff0ff84a SR |
360 | /* Missed count stored at end */ |
361 | #define RB_MISSED_STORED (1 << 30) | |
66a8cb95 | 362 | |
abc9b56d | 363 | struct buffer_data_page { |
e4c2ce82 | 364 | u64 time_stamp; /* page time stamp */ |
c3706f00 | 365 | local_t commit; /* write committed index */ |
abc9b56d SR |
366 | unsigned char data[]; /* data of buffer page */ |
367 | }; | |
368 | ||
77ae365e SR |
369 | /* |
370 | * Note, the buffer_page list must be first. The buffer pages | |
371 | * are allocated in cache lines, which means that each buffer | |
372 | * page will be at the beginning of a cache line, and thus | |
373 | * the least significant bits will be zero. We use this to | |
374 | * add flags in the list struct pointers, to make the ring buffer | |
375 | * lockless. | |
376 | */ | |
abc9b56d | 377 | struct buffer_page { |
778c55d4 | 378 | struct list_head list; /* list of buffer pages */ |
abc9b56d | 379 | local_t write; /* index for next write */ |
6f807acd | 380 | unsigned read; /* index for next read */ |
778c55d4 | 381 | local_t entries; /* entries on this page */ |
ff0ff84a | 382 | unsigned long real_end; /* real end of data */ |
abc9b56d | 383 | struct buffer_data_page *page; /* Actual data page */ |
7a8e76a3 SR |
384 | }; |
385 | ||
77ae365e SR |
386 | /* |
387 | * The buffer page counters, write and entries, must be reset | |
388 | * atomically when crossing page boundaries. To synchronize this | |
389 | * update, two counters are inserted into the number. One is | |
390 | * the actual counter for the write position or count on the page. | |
391 | * | |
392 | * The other is a counter of updaters. Before an update happens | |
393 | * the update partition of the counter is incremented. This will | |
394 | * allow the updater to update the counter atomically. | |
395 | * | |
396 | * The counter is 20 bits, and the state data is 12. | |
397 | */ | |
398 | #define RB_WRITE_MASK 0xfffff | |
399 | #define RB_WRITE_INTCNT (1 << 20) | |
400 | ||
044fa782 | 401 | static void rb_init_page(struct buffer_data_page *bpage) |
abc9b56d | 402 | { |
044fa782 | 403 | local_set(&bpage->commit, 0); |
abc9b56d SR |
404 | } |
405 | ||
474d32b6 SR |
406 | /** |
407 | * ring_buffer_page_len - the size of data on the page. | |
408 | * @page: The page to read | |
409 | * | |
410 | * Returns the amount of data on the page, including buffer page header. | |
411 | */ | |
ef7a4a16 SR |
412 | size_t ring_buffer_page_len(void *page) |
413 | { | |
474d32b6 SR |
414 | return local_read(&((struct buffer_data_page *)page)->commit) |
415 | + BUF_PAGE_HDR_SIZE; | |
ef7a4a16 SR |
416 | } |
417 | ||
ed56829c SR |
418 | /* |
419 | * Also stolen from mm/slob.c. Thanks to Mathieu Desnoyers for pointing | |
420 | * this issue out. | |
421 | */ | |
34a148bf | 422 | static void free_buffer_page(struct buffer_page *bpage) |
ed56829c | 423 | { |
34a148bf | 424 | free_page((unsigned long)bpage->page); |
e4c2ce82 | 425 | kfree(bpage); |
ed56829c SR |
426 | } |
427 | ||
7a8e76a3 SR |
428 | /* |
429 | * We need to fit the time_stamp delta into 27 bits. | |
430 | */ | |
431 | static inline int test_time_stamp(u64 delta) | |
432 | { | |
433 | if (delta & TS_DELTA_TEST) | |
434 | return 1; | |
435 | return 0; | |
436 | } | |
437 | ||
474d32b6 | 438 | #define BUF_PAGE_SIZE (PAGE_SIZE - BUF_PAGE_HDR_SIZE) |
7a8e76a3 | 439 | |
be957c44 SR |
440 | /* Max payload is BUF_PAGE_SIZE - header (8bytes) */ |
441 | #define BUF_MAX_DATA_SIZE (BUF_PAGE_SIZE - (sizeof(u32) * 2)) | |
442 | ||
d1b182a8 SR |
443 | int ring_buffer_print_page_header(struct trace_seq *s) |
444 | { | |
445 | struct buffer_data_page field; | |
446 | int ret; | |
447 | ||
448 | ret = trace_seq_printf(s, "\tfield: u64 timestamp;\t" | |
26a50744 TZ |
449 | "offset:0;\tsize:%u;\tsigned:%u;\n", |
450 | (unsigned int)sizeof(field.time_stamp), | |
451 | (unsigned int)is_signed_type(u64)); | |
d1b182a8 SR |
452 | |
453 | ret = trace_seq_printf(s, "\tfield: local_t commit;\t" | |
26a50744 | 454 | "offset:%u;\tsize:%u;\tsigned:%u;\n", |
d1b182a8 | 455 | (unsigned int)offsetof(typeof(field), commit), |
26a50744 TZ |
456 | (unsigned int)sizeof(field.commit), |
457 | (unsigned int)is_signed_type(long)); | |
d1b182a8 | 458 | |
66a8cb95 SR |
459 | ret = trace_seq_printf(s, "\tfield: int overwrite;\t" |
460 | "offset:%u;\tsize:%u;\tsigned:%u;\n", | |
461 | (unsigned int)offsetof(typeof(field), commit), | |
462 | 1, | |
463 | (unsigned int)is_signed_type(long)); | |
464 | ||
d1b182a8 | 465 | ret = trace_seq_printf(s, "\tfield: char data;\t" |
26a50744 | 466 | "offset:%u;\tsize:%u;\tsigned:%u;\n", |
d1b182a8 | 467 | (unsigned int)offsetof(typeof(field), data), |
26a50744 TZ |
468 | (unsigned int)BUF_PAGE_SIZE, |
469 | (unsigned int)is_signed_type(char)); | |
d1b182a8 SR |
470 | |
471 | return ret; | |
472 | } | |
473 | ||
7a8e76a3 SR |
474 | /* |
475 | * head_page == tail_page && head == tail then buffer is empty. | |
476 | */ | |
477 | struct ring_buffer_per_cpu { | |
478 | int cpu; | |
985023de | 479 | atomic_t record_disabled; |
7a8e76a3 | 480 | struct ring_buffer *buffer; |
5389f6fa | 481 | raw_spinlock_t reader_lock; /* serialize readers */ |
445c8951 | 482 | arch_spinlock_t lock; |
7a8e76a3 | 483 | struct lock_class_key lock_key; |
3adc54fa | 484 | struct list_head *pages; |
6f807acd SR |
485 | struct buffer_page *head_page; /* read from head */ |
486 | struct buffer_page *tail_page; /* write to tail */ | |
c3706f00 | 487 | struct buffer_page *commit_page; /* committed pages */ |
d769041f | 488 | struct buffer_page *reader_page; |
66a8cb95 SR |
489 | unsigned long lost_events; |
490 | unsigned long last_overrun; | |
c64e148a | 491 | local_t entries_bytes; |
77ae365e SR |
492 | local_t commit_overrun; |
493 | local_t overrun; | |
e4906eff | 494 | local_t entries; |
fa743953 SR |
495 | local_t committing; |
496 | local_t commits; | |
77ae365e | 497 | unsigned long read; |
c64e148a | 498 | unsigned long read_bytes; |
7a8e76a3 SR |
499 | u64 write_stamp; |
500 | u64 read_stamp; | |
7a8e76a3 SR |
501 | }; |
502 | ||
503 | struct ring_buffer { | |
7a8e76a3 SR |
504 | unsigned pages; |
505 | unsigned flags; | |
506 | int cpus; | |
7a8e76a3 | 507 | atomic_t record_disabled; |
00f62f61 | 508 | cpumask_var_t cpumask; |
7a8e76a3 | 509 | |
1f8a6a10 PZ |
510 | struct lock_class_key *reader_lock_key; |
511 | ||
7a8e76a3 SR |
512 | struct mutex mutex; |
513 | ||
514 | struct ring_buffer_per_cpu **buffers; | |
554f786e | 515 | |
59222efe | 516 | #ifdef CONFIG_HOTPLUG_CPU |
554f786e SR |
517 | struct notifier_block cpu_notify; |
518 | #endif | |
37886f6a | 519 | u64 (*clock)(void); |
7a8e76a3 SR |
520 | }; |
521 | ||
522 | struct ring_buffer_iter { | |
523 | struct ring_buffer_per_cpu *cpu_buffer; | |
524 | unsigned long head; | |
525 | struct buffer_page *head_page; | |
492a74f4 SR |
526 | struct buffer_page *cache_reader_page; |
527 | unsigned long cache_read; | |
7a8e76a3 SR |
528 | u64 read_stamp; |
529 | }; | |
530 | ||
f536aafc | 531 | /* buffer may be either ring_buffer or ring_buffer_per_cpu */ |
077c5407 SR |
532 | #define RB_WARN_ON(b, cond) \ |
533 | ({ \ | |
534 | int _____ret = unlikely(cond); \ | |
535 | if (_____ret) { \ | |
536 | if (__same_type(*(b), struct ring_buffer_per_cpu)) { \ | |
537 | struct ring_buffer_per_cpu *__b = \ | |
538 | (void *)b; \ | |
539 | atomic_inc(&__b->buffer->record_disabled); \ | |
540 | } else \ | |
541 | atomic_inc(&b->record_disabled); \ | |
542 | WARN_ON(1); \ | |
543 | } \ | |
544 | _____ret; \ | |
3e89c7bb | 545 | }) |
f536aafc | 546 | |
37886f6a SR |
547 | /* Up this if you want to test the TIME_EXTENTS and normalization */ |
548 | #define DEBUG_SHIFT 0 | |
549 | ||
6d3f1e12 | 550 | static inline u64 rb_time_stamp(struct ring_buffer *buffer) |
88eb0125 SR |
551 | { |
552 | /* shift to debug/test normalization and TIME_EXTENTS */ | |
553 | return buffer->clock() << DEBUG_SHIFT; | |
554 | } | |
555 | ||
37886f6a SR |
556 | u64 ring_buffer_time_stamp(struct ring_buffer *buffer, int cpu) |
557 | { | |
558 | u64 time; | |
559 | ||
560 | preempt_disable_notrace(); | |
6d3f1e12 | 561 | time = rb_time_stamp(buffer); |
37886f6a SR |
562 | preempt_enable_no_resched_notrace(); |
563 | ||
564 | return time; | |
565 | } | |
566 | EXPORT_SYMBOL_GPL(ring_buffer_time_stamp); | |
567 | ||
568 | void ring_buffer_normalize_time_stamp(struct ring_buffer *buffer, | |
569 | int cpu, u64 *ts) | |
570 | { | |
571 | /* Just stupid testing the normalize function and deltas */ | |
572 | *ts >>= DEBUG_SHIFT; | |
573 | } | |
574 | EXPORT_SYMBOL_GPL(ring_buffer_normalize_time_stamp); | |
575 | ||
77ae365e SR |
576 | /* |
577 | * Making the ring buffer lockless makes things tricky. | |
578 | * Although writes only happen on the CPU that they are on, | |
579 | * and they only need to worry about interrupts. Reads can | |
580 | * happen on any CPU. | |
581 | * | |
582 | * The reader page is always off the ring buffer, but when the | |
583 | * reader finishes with a page, it needs to swap its page with | |
584 | * a new one from the buffer. The reader needs to take from | |
585 | * the head (writes go to the tail). But if a writer is in overwrite | |
586 | * mode and wraps, it must push the head page forward. | |
587 | * | |
588 | * Here lies the problem. | |
589 | * | |
590 | * The reader must be careful to replace only the head page, and | |
591 | * not another one. As described at the top of the file in the | |
592 | * ASCII art, the reader sets its old page to point to the next | |
593 | * page after head. It then sets the page after head to point to | |
594 | * the old reader page. But if the writer moves the head page | |
595 | * during this operation, the reader could end up with the tail. | |
596 | * | |
597 | * We use cmpxchg to help prevent this race. We also do something | |
598 | * special with the page before head. We set the LSB to 1. | |
599 | * | |
600 | * When the writer must push the page forward, it will clear the | |
601 | * bit that points to the head page, move the head, and then set | |
602 | * the bit that points to the new head page. | |
603 | * | |
604 | * We also don't want an interrupt coming in and moving the head | |
605 | * page on another writer. Thus we use the second LSB to catch | |
606 | * that too. Thus: | |
607 | * | |
608 | * head->list->prev->next bit 1 bit 0 | |
609 | * ------- ------- | |
610 | * Normal page 0 0 | |
611 | * Points to head page 0 1 | |
612 | * New head page 1 0 | |
613 | * | |
614 | * Note we can not trust the prev pointer of the head page, because: | |
615 | * | |
616 | * +----+ +-----+ +-----+ | |
617 | * | |------>| T |---X--->| N | | |
618 | * | |<------| | | | | |
619 | * +----+ +-----+ +-----+ | |
620 | * ^ ^ | | |
621 | * | +-----+ | | | |
622 | * +----------| R |----------+ | | |
623 | * | |<-----------+ | |
624 | * +-----+ | |
625 | * | |
626 | * Key: ---X--> HEAD flag set in pointer | |
627 | * T Tail page | |
628 | * R Reader page | |
629 | * N Next page | |
630 | * | |
631 | * (see __rb_reserve_next() to see where this happens) | |
632 | * | |
633 | * What the above shows is that the reader just swapped out | |
634 | * the reader page with a page in the buffer, but before it | |
635 | * could make the new header point back to the new page added | |
636 | * it was preempted by a writer. The writer moved forward onto | |
637 | * the new page added by the reader and is about to move forward | |
638 | * again. | |
639 | * | |
640 | * You can see, it is legitimate for the previous pointer of | |
641 | * the head (or any page) not to point back to itself. But only | |
642 | * temporarially. | |
643 | */ | |
644 | ||
645 | #define RB_PAGE_NORMAL 0UL | |
646 | #define RB_PAGE_HEAD 1UL | |
647 | #define RB_PAGE_UPDATE 2UL | |
648 | ||
649 | ||
650 | #define RB_FLAG_MASK 3UL | |
651 | ||
652 | /* PAGE_MOVED is not part of the mask */ | |
653 | #define RB_PAGE_MOVED 4UL | |
654 | ||
655 | /* | |
656 | * rb_list_head - remove any bit | |
657 | */ | |
658 | static struct list_head *rb_list_head(struct list_head *list) | |
659 | { | |
660 | unsigned long val = (unsigned long)list; | |
661 | ||
662 | return (struct list_head *)(val & ~RB_FLAG_MASK); | |
663 | } | |
664 | ||
665 | /* | |
6d3f1e12 | 666 | * rb_is_head_page - test if the given page is the head page |
77ae365e SR |
667 | * |
668 | * Because the reader may move the head_page pointer, we can | |
669 | * not trust what the head page is (it may be pointing to | |
670 | * the reader page). But if the next page is a header page, | |
671 | * its flags will be non zero. | |
672 | */ | |
42b16b3f | 673 | static inline int |
77ae365e SR |
674 | rb_is_head_page(struct ring_buffer_per_cpu *cpu_buffer, |
675 | struct buffer_page *page, struct list_head *list) | |
676 | { | |
677 | unsigned long val; | |
678 | ||
679 | val = (unsigned long)list->next; | |
680 | ||
681 | if ((val & ~RB_FLAG_MASK) != (unsigned long)&page->list) | |
682 | return RB_PAGE_MOVED; | |
683 | ||
684 | return val & RB_FLAG_MASK; | |
685 | } | |
686 | ||
687 | /* | |
688 | * rb_is_reader_page | |
689 | * | |
690 | * The unique thing about the reader page, is that, if the | |
691 | * writer is ever on it, the previous pointer never points | |
692 | * back to the reader page. | |
693 | */ | |
694 | static int rb_is_reader_page(struct buffer_page *page) | |
695 | { | |
696 | struct list_head *list = page->list.prev; | |
697 | ||
698 | return rb_list_head(list->next) != &page->list; | |
699 | } | |
700 | ||
701 | /* | |
702 | * rb_set_list_to_head - set a list_head to be pointing to head. | |
703 | */ | |
704 | static void rb_set_list_to_head(struct ring_buffer_per_cpu *cpu_buffer, | |
705 | struct list_head *list) | |
706 | { | |
707 | unsigned long *ptr; | |
708 | ||
709 | ptr = (unsigned long *)&list->next; | |
710 | *ptr |= RB_PAGE_HEAD; | |
711 | *ptr &= ~RB_PAGE_UPDATE; | |
712 | } | |
713 | ||
714 | /* | |
715 | * rb_head_page_activate - sets up head page | |
716 | */ | |
717 | static void rb_head_page_activate(struct ring_buffer_per_cpu *cpu_buffer) | |
718 | { | |
719 | struct buffer_page *head; | |
720 | ||
721 | head = cpu_buffer->head_page; | |
722 | if (!head) | |
723 | return; | |
724 | ||
725 | /* | |
726 | * Set the previous list pointer to have the HEAD flag. | |
727 | */ | |
728 | rb_set_list_to_head(cpu_buffer, head->list.prev); | |
729 | } | |
730 | ||
731 | static void rb_list_head_clear(struct list_head *list) | |
732 | { | |
733 | unsigned long *ptr = (unsigned long *)&list->next; | |
734 | ||
735 | *ptr &= ~RB_FLAG_MASK; | |
736 | } | |
737 | ||
738 | /* | |
739 | * rb_head_page_dactivate - clears head page ptr (for free list) | |
740 | */ | |
741 | static void | |
742 | rb_head_page_deactivate(struct ring_buffer_per_cpu *cpu_buffer) | |
743 | { | |
744 | struct list_head *hd; | |
745 | ||
746 | /* Go through the whole list and clear any pointers found. */ | |
747 | rb_list_head_clear(cpu_buffer->pages); | |
748 | ||
749 | list_for_each(hd, cpu_buffer->pages) | |
750 | rb_list_head_clear(hd); | |
751 | } | |
752 | ||
753 | static int rb_head_page_set(struct ring_buffer_per_cpu *cpu_buffer, | |
754 | struct buffer_page *head, | |
755 | struct buffer_page *prev, | |
756 | int old_flag, int new_flag) | |
757 | { | |
758 | struct list_head *list; | |
759 | unsigned long val = (unsigned long)&head->list; | |
760 | unsigned long ret; | |
761 | ||
762 | list = &prev->list; | |
763 | ||
764 | val &= ~RB_FLAG_MASK; | |
765 | ||
08a40816 SR |
766 | ret = cmpxchg((unsigned long *)&list->next, |
767 | val | old_flag, val | new_flag); | |
77ae365e SR |
768 | |
769 | /* check if the reader took the page */ | |
770 | if ((ret & ~RB_FLAG_MASK) != val) | |
771 | return RB_PAGE_MOVED; | |
772 | ||
773 | return ret & RB_FLAG_MASK; | |
774 | } | |
775 | ||
776 | static int rb_head_page_set_update(struct ring_buffer_per_cpu *cpu_buffer, | |
777 | struct buffer_page *head, | |
778 | struct buffer_page *prev, | |
779 | int old_flag) | |
780 | { | |
781 | return rb_head_page_set(cpu_buffer, head, prev, | |
782 | old_flag, RB_PAGE_UPDATE); | |
783 | } | |
784 | ||
785 | static int rb_head_page_set_head(struct ring_buffer_per_cpu *cpu_buffer, | |
786 | struct buffer_page *head, | |
787 | struct buffer_page *prev, | |
788 | int old_flag) | |
789 | { | |
790 | return rb_head_page_set(cpu_buffer, head, prev, | |
791 | old_flag, RB_PAGE_HEAD); | |
792 | } | |
793 | ||
794 | static int rb_head_page_set_normal(struct ring_buffer_per_cpu *cpu_buffer, | |
795 | struct buffer_page *head, | |
796 | struct buffer_page *prev, | |
797 | int old_flag) | |
798 | { | |
799 | return rb_head_page_set(cpu_buffer, head, prev, | |
800 | old_flag, RB_PAGE_NORMAL); | |
801 | } | |
802 | ||
803 | static inline void rb_inc_page(struct ring_buffer_per_cpu *cpu_buffer, | |
804 | struct buffer_page **bpage) | |
805 | { | |
806 | struct list_head *p = rb_list_head((*bpage)->list.next); | |
807 | ||
808 | *bpage = list_entry(p, struct buffer_page, list); | |
809 | } | |
810 | ||
811 | static struct buffer_page * | |
812 | rb_set_head_page(struct ring_buffer_per_cpu *cpu_buffer) | |
813 | { | |
814 | struct buffer_page *head; | |
815 | struct buffer_page *page; | |
816 | struct list_head *list; | |
817 | int i; | |
818 | ||
819 | if (RB_WARN_ON(cpu_buffer, !cpu_buffer->head_page)) | |
820 | return NULL; | |
821 | ||
822 | /* sanity check */ | |
823 | list = cpu_buffer->pages; | |
824 | if (RB_WARN_ON(cpu_buffer, rb_list_head(list->prev->next) != list)) | |
825 | return NULL; | |
826 | ||
827 | page = head = cpu_buffer->head_page; | |
828 | /* | |
829 | * It is possible that the writer moves the header behind | |
830 | * where we started, and we miss in one loop. | |
831 | * A second loop should grab the header, but we'll do | |
832 | * three loops just because I'm paranoid. | |
833 | */ | |
834 | for (i = 0; i < 3; i++) { | |
835 | do { | |
836 | if (rb_is_head_page(cpu_buffer, page, page->list.prev)) { | |
837 | cpu_buffer->head_page = page; | |
838 | return page; | |
839 | } | |
840 | rb_inc_page(cpu_buffer, &page); | |
841 | } while (page != head); | |
842 | } | |
843 | ||
844 | RB_WARN_ON(cpu_buffer, 1); | |
845 | ||
846 | return NULL; | |
847 | } | |
848 | ||
849 | static int rb_head_page_replace(struct buffer_page *old, | |
850 | struct buffer_page *new) | |
851 | { | |
852 | unsigned long *ptr = (unsigned long *)&old->list.prev->next; | |
853 | unsigned long val; | |
854 | unsigned long ret; | |
855 | ||
856 | val = *ptr & ~RB_FLAG_MASK; | |
857 | val |= RB_PAGE_HEAD; | |
858 | ||
08a40816 | 859 | ret = cmpxchg(ptr, val, (unsigned long)&new->list); |
77ae365e SR |
860 | |
861 | return ret == val; | |
862 | } | |
863 | ||
864 | /* | |
865 | * rb_tail_page_update - move the tail page forward | |
866 | * | |
867 | * Returns 1 if moved tail page, 0 if someone else did. | |
868 | */ | |
869 | static int rb_tail_page_update(struct ring_buffer_per_cpu *cpu_buffer, | |
870 | struct buffer_page *tail_page, | |
871 | struct buffer_page *next_page) | |
872 | { | |
873 | struct buffer_page *old_tail; | |
874 | unsigned long old_entries; | |
875 | unsigned long old_write; | |
876 | int ret = 0; | |
877 | ||
878 | /* | |
879 | * The tail page now needs to be moved forward. | |
880 | * | |
881 | * We need to reset the tail page, but without messing | |
882 | * with possible erasing of data brought in by interrupts | |
883 | * that have moved the tail page and are currently on it. | |
884 | * | |
885 | * We add a counter to the write field to denote this. | |
886 | */ | |
887 | old_write = local_add_return(RB_WRITE_INTCNT, &next_page->write); | |
888 | old_entries = local_add_return(RB_WRITE_INTCNT, &next_page->entries); | |
889 | ||
890 | /* | |
891 | * Just make sure we have seen our old_write and synchronize | |
892 | * with any interrupts that come in. | |
893 | */ | |
894 | barrier(); | |
895 | ||
896 | /* | |
897 | * If the tail page is still the same as what we think | |
898 | * it is, then it is up to us to update the tail | |
899 | * pointer. | |
900 | */ | |
901 | if (tail_page == cpu_buffer->tail_page) { | |
902 | /* Zero the write counter */ | |
903 | unsigned long val = old_write & ~RB_WRITE_MASK; | |
904 | unsigned long eval = old_entries & ~RB_WRITE_MASK; | |
905 | ||
906 | /* | |
907 | * This will only succeed if an interrupt did | |
908 | * not come in and change it. In which case, we | |
909 | * do not want to modify it. | |
da706d8b LJ |
910 | * |
911 | * We add (void) to let the compiler know that we do not care | |
912 | * about the return value of these functions. We use the | |
913 | * cmpxchg to only update if an interrupt did not already | |
914 | * do it for us. If the cmpxchg fails, we don't care. | |
77ae365e | 915 | */ |
da706d8b LJ |
916 | (void)local_cmpxchg(&next_page->write, old_write, val); |
917 | (void)local_cmpxchg(&next_page->entries, old_entries, eval); | |
77ae365e SR |
918 | |
919 | /* | |
920 | * No need to worry about races with clearing out the commit. | |
921 | * it only can increment when a commit takes place. But that | |
922 | * only happens in the outer most nested commit. | |
923 | */ | |
924 | local_set(&next_page->page->commit, 0); | |
925 | ||
926 | old_tail = cmpxchg(&cpu_buffer->tail_page, | |
927 | tail_page, next_page); | |
928 | ||
929 | if (old_tail == tail_page) | |
930 | ret = 1; | |
931 | } | |
932 | ||
933 | return ret; | |
934 | } | |
935 | ||
936 | static int rb_check_bpage(struct ring_buffer_per_cpu *cpu_buffer, | |
937 | struct buffer_page *bpage) | |
938 | { | |
939 | unsigned long val = (unsigned long)bpage; | |
940 | ||
941 | if (RB_WARN_ON(cpu_buffer, val & RB_FLAG_MASK)) | |
942 | return 1; | |
943 | ||
944 | return 0; | |
945 | } | |
946 | ||
947 | /** | |
948 | * rb_check_list - make sure a pointer to a list has the last bits zero | |
949 | */ | |
950 | static int rb_check_list(struct ring_buffer_per_cpu *cpu_buffer, | |
951 | struct list_head *list) | |
952 | { | |
953 | if (RB_WARN_ON(cpu_buffer, rb_list_head(list->prev) != list->prev)) | |
954 | return 1; | |
955 | if (RB_WARN_ON(cpu_buffer, rb_list_head(list->next) != list->next)) | |
956 | return 1; | |
957 | return 0; | |
958 | } | |
959 | ||
7a8e76a3 SR |
960 | /** |
961 | * check_pages - integrity check of buffer pages | |
962 | * @cpu_buffer: CPU buffer with pages to test | |
963 | * | |
c3706f00 | 964 | * As a safety measure we check to make sure the data pages have not |
7a8e76a3 SR |
965 | * been corrupted. |
966 | */ | |
967 | static int rb_check_pages(struct ring_buffer_per_cpu *cpu_buffer) | |
968 | { | |
3adc54fa | 969 | struct list_head *head = cpu_buffer->pages; |
044fa782 | 970 | struct buffer_page *bpage, *tmp; |
7a8e76a3 | 971 | |
77ae365e SR |
972 | rb_head_page_deactivate(cpu_buffer); |
973 | ||
3e89c7bb SR |
974 | if (RB_WARN_ON(cpu_buffer, head->next->prev != head)) |
975 | return -1; | |
976 | if (RB_WARN_ON(cpu_buffer, head->prev->next != head)) | |
977 | return -1; | |
7a8e76a3 | 978 | |
77ae365e SR |
979 | if (rb_check_list(cpu_buffer, head)) |
980 | return -1; | |
981 | ||
044fa782 | 982 | list_for_each_entry_safe(bpage, tmp, head, list) { |
3e89c7bb | 983 | if (RB_WARN_ON(cpu_buffer, |
044fa782 | 984 | bpage->list.next->prev != &bpage->list)) |
3e89c7bb SR |
985 | return -1; |
986 | if (RB_WARN_ON(cpu_buffer, | |
044fa782 | 987 | bpage->list.prev->next != &bpage->list)) |
3e89c7bb | 988 | return -1; |
77ae365e SR |
989 | if (rb_check_list(cpu_buffer, &bpage->list)) |
990 | return -1; | |
7a8e76a3 SR |
991 | } |
992 | ||
77ae365e SR |
993 | rb_head_page_activate(cpu_buffer); |
994 | ||
7a8e76a3 SR |
995 | return 0; |
996 | } | |
997 | ||
7a8e76a3 SR |
998 | static int rb_allocate_pages(struct ring_buffer_per_cpu *cpu_buffer, |
999 | unsigned nr_pages) | |
1000 | { | |
044fa782 | 1001 | struct buffer_page *bpage, *tmp; |
7a8e76a3 SR |
1002 | LIST_HEAD(pages); |
1003 | unsigned i; | |
1004 | ||
3adc54fa SR |
1005 | WARN_ON(!nr_pages); |
1006 | ||
7a8e76a3 | 1007 | for (i = 0; i < nr_pages; i++) { |
7ea59064 | 1008 | struct page *page; |
d7ec4bfe VN |
1009 | /* |
1010 | * __GFP_NORETRY flag makes sure that the allocation fails | |
1011 | * gracefully without invoking oom-killer and the system is | |
1012 | * not destabilized. | |
1013 | */ | |
044fa782 | 1014 | bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), |
d7ec4bfe VN |
1015 | GFP_KERNEL | __GFP_NORETRY, |
1016 | cpu_to_node(cpu_buffer->cpu)); | |
044fa782 | 1017 | if (!bpage) |
e4c2ce82 | 1018 | goto free_pages; |
77ae365e SR |
1019 | |
1020 | rb_check_bpage(cpu_buffer, bpage); | |
1021 | ||
044fa782 | 1022 | list_add(&bpage->list, &pages); |
e4c2ce82 | 1023 | |
7ea59064 | 1024 | page = alloc_pages_node(cpu_to_node(cpu_buffer->cpu), |
d7ec4bfe | 1025 | GFP_KERNEL | __GFP_NORETRY, 0); |
7ea59064 | 1026 | if (!page) |
7a8e76a3 | 1027 | goto free_pages; |
7ea59064 | 1028 | bpage->page = page_address(page); |
044fa782 | 1029 | rb_init_page(bpage->page); |
7a8e76a3 SR |
1030 | } |
1031 | ||
3adc54fa SR |
1032 | /* |
1033 | * The ring buffer page list is a circular list that does not | |
1034 | * start and end with a list head. All page list items point to | |
1035 | * other pages. | |
1036 | */ | |
1037 | cpu_buffer->pages = pages.next; | |
1038 | list_del(&pages); | |
7a8e76a3 SR |
1039 | |
1040 | rb_check_pages(cpu_buffer); | |
1041 | ||
1042 | return 0; | |
1043 | ||
1044 | free_pages: | |
044fa782 SR |
1045 | list_for_each_entry_safe(bpage, tmp, &pages, list) { |
1046 | list_del_init(&bpage->list); | |
1047 | free_buffer_page(bpage); | |
7a8e76a3 SR |
1048 | } |
1049 | return -ENOMEM; | |
1050 | } | |
1051 | ||
1052 | static struct ring_buffer_per_cpu * | |
1053 | rb_allocate_cpu_buffer(struct ring_buffer *buffer, int cpu) | |
1054 | { | |
1055 | struct ring_buffer_per_cpu *cpu_buffer; | |
044fa782 | 1056 | struct buffer_page *bpage; |
7ea59064 | 1057 | struct page *page; |
7a8e76a3 SR |
1058 | int ret; |
1059 | ||
1060 | cpu_buffer = kzalloc_node(ALIGN(sizeof(*cpu_buffer), cache_line_size()), | |
1061 | GFP_KERNEL, cpu_to_node(cpu)); | |
1062 | if (!cpu_buffer) | |
1063 | return NULL; | |
1064 | ||
1065 | cpu_buffer->cpu = cpu; | |
1066 | cpu_buffer->buffer = buffer; | |
5389f6fa | 1067 | raw_spin_lock_init(&cpu_buffer->reader_lock); |
1f8a6a10 | 1068 | lockdep_set_class(&cpu_buffer->reader_lock, buffer->reader_lock_key); |
edc35bd7 | 1069 | cpu_buffer->lock = (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED; |
7a8e76a3 | 1070 | |
044fa782 | 1071 | bpage = kzalloc_node(ALIGN(sizeof(*bpage), cache_line_size()), |
e4c2ce82 | 1072 | GFP_KERNEL, cpu_to_node(cpu)); |
044fa782 | 1073 | if (!bpage) |
e4c2ce82 SR |
1074 | goto fail_free_buffer; |
1075 | ||
77ae365e SR |
1076 | rb_check_bpage(cpu_buffer, bpage); |
1077 | ||
044fa782 | 1078 | cpu_buffer->reader_page = bpage; |
7ea59064 VN |
1079 | page = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, 0); |
1080 | if (!page) | |
e4c2ce82 | 1081 | goto fail_free_reader; |
7ea59064 | 1082 | bpage->page = page_address(page); |
044fa782 | 1083 | rb_init_page(bpage->page); |
e4c2ce82 | 1084 | |
d769041f | 1085 | INIT_LIST_HEAD(&cpu_buffer->reader_page->list); |
d769041f | 1086 | |
7a8e76a3 SR |
1087 | ret = rb_allocate_pages(cpu_buffer, buffer->pages); |
1088 | if (ret < 0) | |
d769041f | 1089 | goto fail_free_reader; |
7a8e76a3 SR |
1090 | |
1091 | cpu_buffer->head_page | |
3adc54fa | 1092 | = list_entry(cpu_buffer->pages, struct buffer_page, list); |
bf41a158 | 1093 | cpu_buffer->tail_page = cpu_buffer->commit_page = cpu_buffer->head_page; |
7a8e76a3 | 1094 | |
77ae365e SR |
1095 | rb_head_page_activate(cpu_buffer); |
1096 | ||
7a8e76a3 SR |
1097 | return cpu_buffer; |
1098 | ||
d769041f SR |
1099 | fail_free_reader: |
1100 | free_buffer_page(cpu_buffer->reader_page); | |
1101 | ||
7a8e76a3 SR |
1102 | fail_free_buffer: |
1103 | kfree(cpu_buffer); | |
1104 | return NULL; | |
1105 | } | |
1106 | ||
1107 | static void rb_free_cpu_buffer(struct ring_buffer_per_cpu *cpu_buffer) | |
1108 | { | |
3adc54fa | 1109 | struct list_head *head = cpu_buffer->pages; |
044fa782 | 1110 | struct buffer_page *bpage, *tmp; |
7a8e76a3 | 1111 | |
d769041f SR |
1112 | free_buffer_page(cpu_buffer->reader_page); |
1113 | ||
77ae365e SR |
1114 | rb_head_page_deactivate(cpu_buffer); |
1115 | ||
3adc54fa SR |
1116 | if (head) { |
1117 | list_for_each_entry_safe(bpage, tmp, head, list) { | |
1118 | list_del_init(&bpage->list); | |
1119 | free_buffer_page(bpage); | |
1120 | } | |
1121 | bpage = list_entry(head, struct buffer_page, list); | |
044fa782 | 1122 | free_buffer_page(bpage); |
7a8e76a3 | 1123 | } |
3adc54fa | 1124 | |
7a8e76a3 SR |
1125 | kfree(cpu_buffer); |
1126 | } | |
1127 | ||
59222efe | 1128 | #ifdef CONFIG_HOTPLUG_CPU |
09c9e84d FW |
1129 | static int rb_cpu_notify(struct notifier_block *self, |
1130 | unsigned long action, void *hcpu); | |
554f786e SR |
1131 | #endif |
1132 | ||
7a8e76a3 SR |
1133 | /** |
1134 | * ring_buffer_alloc - allocate a new ring_buffer | |
68814b58 | 1135 | * @size: the size in bytes per cpu that is needed. |
7a8e76a3 SR |
1136 | * @flags: attributes to set for the ring buffer. |
1137 | * | |
1138 | * Currently the only flag that is available is the RB_FL_OVERWRITE | |
1139 | * flag. This flag means that the buffer will overwrite old data | |
1140 | * when the buffer wraps. If this flag is not set, the buffer will | |
1141 | * drop data when the tail hits the head. | |
1142 | */ | |
1f8a6a10 PZ |
1143 | struct ring_buffer *__ring_buffer_alloc(unsigned long size, unsigned flags, |
1144 | struct lock_class_key *key) | |
7a8e76a3 SR |
1145 | { |
1146 | struct ring_buffer *buffer; | |
1147 | int bsize; | |
1148 | int cpu; | |
1149 | ||
1150 | /* keep it in its own cache line */ | |
1151 | buffer = kzalloc(ALIGN(sizeof(*buffer), cache_line_size()), | |
1152 | GFP_KERNEL); | |
1153 | if (!buffer) | |
1154 | return NULL; | |
1155 | ||
9e01c1b7 RR |
1156 | if (!alloc_cpumask_var(&buffer->cpumask, GFP_KERNEL)) |
1157 | goto fail_free_buffer; | |
1158 | ||
7a8e76a3 SR |
1159 | buffer->pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE); |
1160 | buffer->flags = flags; | |
37886f6a | 1161 | buffer->clock = trace_clock_local; |
1f8a6a10 | 1162 | buffer->reader_lock_key = key; |
7a8e76a3 SR |
1163 | |
1164 | /* need at least two pages */ | |
5f78abee SR |
1165 | if (buffer->pages < 2) |
1166 | buffer->pages = 2; | |
7a8e76a3 | 1167 | |
3bf832ce FW |
1168 | /* |
1169 | * In case of non-hotplug cpu, if the ring-buffer is allocated | |
1170 | * in early initcall, it will not be notified of secondary cpus. | |
1171 | * In that off case, we need to allocate for all possible cpus. | |
1172 | */ | |
1173 | #ifdef CONFIG_HOTPLUG_CPU | |
554f786e SR |
1174 | get_online_cpus(); |
1175 | cpumask_copy(buffer->cpumask, cpu_online_mask); | |
3bf832ce FW |
1176 | #else |
1177 | cpumask_copy(buffer->cpumask, cpu_possible_mask); | |
1178 | #endif | |
7a8e76a3 SR |
1179 | buffer->cpus = nr_cpu_ids; |
1180 | ||
1181 | bsize = sizeof(void *) * nr_cpu_ids; | |
1182 | buffer->buffers = kzalloc(ALIGN(bsize, cache_line_size()), | |
1183 | GFP_KERNEL); | |
1184 | if (!buffer->buffers) | |
9e01c1b7 | 1185 | goto fail_free_cpumask; |
7a8e76a3 SR |
1186 | |
1187 | for_each_buffer_cpu(buffer, cpu) { | |
1188 | buffer->buffers[cpu] = | |
1189 | rb_allocate_cpu_buffer(buffer, cpu); | |
1190 | if (!buffer->buffers[cpu]) | |
1191 | goto fail_free_buffers; | |
1192 | } | |
1193 | ||
59222efe | 1194 | #ifdef CONFIG_HOTPLUG_CPU |
554f786e SR |
1195 | buffer->cpu_notify.notifier_call = rb_cpu_notify; |
1196 | buffer->cpu_notify.priority = 0; | |
1197 | register_cpu_notifier(&buffer->cpu_notify); | |
1198 | #endif | |
1199 | ||
1200 | put_online_cpus(); | |
7a8e76a3 SR |
1201 | mutex_init(&buffer->mutex); |
1202 | ||
1203 | return buffer; | |
1204 | ||
1205 | fail_free_buffers: | |
1206 | for_each_buffer_cpu(buffer, cpu) { | |
1207 | if (buffer->buffers[cpu]) | |
1208 | rb_free_cpu_buffer(buffer->buffers[cpu]); | |
1209 | } | |
1210 | kfree(buffer->buffers); | |
1211 | ||
9e01c1b7 RR |
1212 | fail_free_cpumask: |
1213 | free_cpumask_var(buffer->cpumask); | |
554f786e | 1214 | put_online_cpus(); |
9e01c1b7 | 1215 | |
7a8e76a3 SR |
1216 | fail_free_buffer: |
1217 | kfree(buffer); | |
1218 | return NULL; | |
1219 | } | |
1f8a6a10 | 1220 | EXPORT_SYMBOL_GPL(__ring_buffer_alloc); |
7a8e76a3 SR |
1221 | |
1222 | /** | |
1223 | * ring_buffer_free - free a ring buffer. | |
1224 | * @buffer: the buffer to free. | |
1225 | */ | |
1226 | void | |
1227 | ring_buffer_free(struct ring_buffer *buffer) | |
1228 | { | |
1229 | int cpu; | |
1230 | ||
554f786e SR |
1231 | get_online_cpus(); |
1232 | ||
59222efe | 1233 | #ifdef CONFIG_HOTPLUG_CPU |
554f786e SR |
1234 | unregister_cpu_notifier(&buffer->cpu_notify); |
1235 | #endif | |
1236 | ||
7a8e76a3 SR |
1237 | for_each_buffer_cpu(buffer, cpu) |
1238 | rb_free_cpu_buffer(buffer->buffers[cpu]); | |
1239 | ||
554f786e SR |
1240 | put_online_cpus(); |
1241 | ||
bd3f0221 | 1242 | kfree(buffer->buffers); |
9e01c1b7 RR |
1243 | free_cpumask_var(buffer->cpumask); |
1244 | ||
7a8e76a3 SR |
1245 | kfree(buffer); |
1246 | } | |
c4f50183 | 1247 | EXPORT_SYMBOL_GPL(ring_buffer_free); |
7a8e76a3 | 1248 | |
37886f6a SR |
1249 | void ring_buffer_set_clock(struct ring_buffer *buffer, |
1250 | u64 (*clock)(void)) | |
1251 | { | |
1252 | buffer->clock = clock; | |
1253 | } | |
1254 | ||
7a8e76a3 SR |
1255 | static void rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer); |
1256 | ||
1257 | static void | |
1258 | rb_remove_pages(struct ring_buffer_per_cpu *cpu_buffer, unsigned nr_pages) | |
1259 | { | |
044fa782 | 1260 | struct buffer_page *bpage; |
7a8e76a3 SR |
1261 | struct list_head *p; |
1262 | unsigned i; | |
1263 | ||
5389f6fa | 1264 | raw_spin_lock_irq(&cpu_buffer->reader_lock); |
77ae365e SR |
1265 | rb_head_page_deactivate(cpu_buffer); |
1266 | ||
7a8e76a3 | 1267 | for (i = 0; i < nr_pages; i++) { |
3adc54fa | 1268 | if (RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages))) |
292f60c0 | 1269 | goto out; |
3adc54fa | 1270 | p = cpu_buffer->pages->next; |
044fa782 SR |
1271 | bpage = list_entry(p, struct buffer_page, list); |
1272 | list_del_init(&bpage->list); | |
1273 | free_buffer_page(bpage); | |
7a8e76a3 | 1274 | } |
3adc54fa | 1275 | if (RB_WARN_ON(cpu_buffer, list_empty(cpu_buffer->pages))) |
292f60c0 | 1276 | goto out; |
7a8e76a3 SR |
1277 | |
1278 | rb_reset_cpu(cpu_buffer); | |
7a8e76a3 SR |
1279 | rb_check_pages(cpu_buffer); |
1280 | ||
292f60c0 | 1281 | out: |
5389f6fa | 1282 | raw_spin_unlock_irq(&cpu_buffer->reader_lock); |
7a8e76a3 SR |
1283 | } |
1284 | ||
1285 | static void | |
1286 | rb_insert_pages(struct ring_buffer_per_cpu *cpu_buffer, | |
1287 | struct list_head *pages, unsigned nr_pages) | |
1288 | { | |
044fa782 | 1289 | struct buffer_page *bpage; |
7a8e76a3 SR |
1290 | struct list_head *p; |
1291 | unsigned i; | |
1292 | ||
5389f6fa | 1293 | raw_spin_lock_irq(&cpu_buffer->reader_lock); |
77ae365e SR |
1294 | rb_head_page_deactivate(cpu_buffer); |
1295 | ||
7a8e76a3 | 1296 | for (i = 0; i < nr_pages; i++) { |
3e89c7bb | 1297 | if (RB_WARN_ON(cpu_buffer, list_empty(pages))) |
292f60c0 | 1298 | goto out; |
7a8e76a3 | 1299 | p = pages->next; |
044fa782 SR |
1300 | bpage = list_entry(p, struct buffer_page, list); |
1301 | list_del_init(&bpage->list); | |
3adc54fa | 1302 | list_add_tail(&bpage->list, cpu_buffer->pages); |
7a8e76a3 SR |
1303 | } |
1304 | rb_reset_cpu(cpu_buffer); | |
7a8e76a3 SR |
1305 | rb_check_pages(cpu_buffer); |
1306 | ||
292f60c0 | 1307 | out: |
5389f6fa | 1308 | raw_spin_unlock_irq(&cpu_buffer->reader_lock); |
7a8e76a3 SR |
1309 | } |
1310 | ||
1311 | /** | |
1312 | * ring_buffer_resize - resize the ring buffer | |
1313 | * @buffer: the buffer to resize. | |
1314 | * @size: the new size. | |
1315 | * | |
7a8e76a3 SR |
1316 | * Minimum size is 2 * BUF_PAGE_SIZE. |
1317 | * | |
1318 | * Returns -1 on failure. | |
1319 | */ | |
1320 | int ring_buffer_resize(struct ring_buffer *buffer, unsigned long size) | |
1321 | { | |
1322 | struct ring_buffer_per_cpu *cpu_buffer; | |
1323 | unsigned nr_pages, rm_pages, new_pages; | |
044fa782 | 1324 | struct buffer_page *bpage, *tmp; |
7a8e76a3 | 1325 | unsigned long buffer_size; |
7a8e76a3 SR |
1326 | LIST_HEAD(pages); |
1327 | int i, cpu; | |
1328 | ||
ee51a1de IM |
1329 | /* |
1330 | * Always succeed at resizing a non-existent buffer: | |
1331 | */ | |
1332 | if (!buffer) | |
1333 | return size; | |
1334 | ||
7a8e76a3 SR |
1335 | size = DIV_ROUND_UP(size, BUF_PAGE_SIZE); |
1336 | size *= BUF_PAGE_SIZE; | |
1337 | buffer_size = buffer->pages * BUF_PAGE_SIZE; | |
1338 | ||
1339 | /* we need a minimum of two pages */ | |
1340 | if (size < BUF_PAGE_SIZE * 2) | |
1341 | size = BUF_PAGE_SIZE * 2; | |
1342 | ||
1343 | if (size == buffer_size) | |
1344 | return size; | |
1345 | ||
18421015 SR |
1346 | atomic_inc(&buffer->record_disabled); |
1347 | ||
1348 | /* Make sure all writers are done with this buffer. */ | |
1349 | synchronize_sched(); | |
1350 | ||
7a8e76a3 | 1351 | mutex_lock(&buffer->mutex); |
554f786e | 1352 | get_online_cpus(); |
7a8e76a3 SR |
1353 | |
1354 | nr_pages = DIV_ROUND_UP(size, BUF_PAGE_SIZE); | |
1355 | ||
1356 | if (size < buffer_size) { | |
1357 | ||
1358 | /* easy case, just free pages */ | |
554f786e SR |
1359 | if (RB_WARN_ON(buffer, nr_pages >= buffer->pages)) |
1360 | goto out_fail; | |
7a8e76a3 SR |
1361 | |
1362 | rm_pages = buffer->pages - nr_pages; | |
1363 | ||
1364 | for_each_buffer_cpu(buffer, cpu) { | |
1365 | cpu_buffer = buffer->buffers[cpu]; | |
1366 | rb_remove_pages(cpu_buffer, rm_pages); | |
1367 | } | |
1368 | goto out; | |
1369 | } | |
1370 | ||
1371 | /* | |
1372 | * This is a bit more difficult. We only want to add pages | |
1373 | * when we can allocate enough for all CPUs. We do this | |
1374 | * by allocating all the pages and storing them on a local | |
1375 | * link list. If we succeed in our allocation, then we | |
1376 | * add these pages to the cpu_buffers. Otherwise we just free | |
1377 | * them all and return -ENOMEM; | |
1378 | */ | |
554f786e SR |
1379 | if (RB_WARN_ON(buffer, nr_pages <= buffer->pages)) |
1380 | goto out_fail; | |
f536aafc | 1381 | |
7a8e76a3 SR |
1382 | new_pages = nr_pages - buffer->pages; |
1383 | ||
1384 | for_each_buffer_cpu(buffer, cpu) { | |
1385 | for (i = 0; i < new_pages; i++) { | |
7ea59064 | 1386 | struct page *page; |
d7ec4bfe VN |
1387 | /* |
1388 | * __GFP_NORETRY flag makes sure that the allocation | |
1389 | * fails gracefully without invoking oom-killer and | |
1390 | * the system is not destabilized. | |
1391 | */ | |
044fa782 | 1392 | bpage = kzalloc_node(ALIGN(sizeof(*bpage), |
e4c2ce82 | 1393 | cache_line_size()), |
d7ec4bfe VN |
1394 | GFP_KERNEL | __GFP_NORETRY, |
1395 | cpu_to_node(cpu)); | |
044fa782 | 1396 | if (!bpage) |
e4c2ce82 | 1397 | goto free_pages; |
044fa782 | 1398 | list_add(&bpage->list, &pages); |
d7ec4bfe VN |
1399 | page = alloc_pages_node(cpu_to_node(cpu), |
1400 | GFP_KERNEL | __GFP_NORETRY, 0); | |
7ea59064 | 1401 | if (!page) |
7a8e76a3 | 1402 | goto free_pages; |
7ea59064 | 1403 | bpage->page = page_address(page); |
044fa782 | 1404 | rb_init_page(bpage->page); |
7a8e76a3 SR |
1405 | } |
1406 | } | |
1407 | ||
1408 | for_each_buffer_cpu(buffer, cpu) { | |
1409 | cpu_buffer = buffer->buffers[cpu]; | |
1410 | rb_insert_pages(cpu_buffer, &pages, new_pages); | |
1411 | } | |
1412 | ||
554f786e SR |
1413 | if (RB_WARN_ON(buffer, !list_empty(&pages))) |
1414 | goto out_fail; | |
7a8e76a3 SR |
1415 | |
1416 | out: | |
1417 | buffer->pages = nr_pages; | |
554f786e | 1418 | put_online_cpus(); |
7a8e76a3 SR |
1419 | mutex_unlock(&buffer->mutex); |
1420 | ||
18421015 SR |
1421 | atomic_dec(&buffer->record_disabled); |
1422 | ||
7a8e76a3 SR |
1423 | return size; |
1424 | ||
1425 | free_pages: | |
044fa782 SR |
1426 | list_for_each_entry_safe(bpage, tmp, &pages, list) { |
1427 | list_del_init(&bpage->list); | |
1428 | free_buffer_page(bpage); | |
7a8e76a3 | 1429 | } |
554f786e | 1430 | put_online_cpus(); |
641d2f63 | 1431 | mutex_unlock(&buffer->mutex); |
18421015 | 1432 | atomic_dec(&buffer->record_disabled); |
7a8e76a3 | 1433 | return -ENOMEM; |
554f786e SR |
1434 | |
1435 | /* | |
1436 | * Something went totally wrong, and we are too paranoid | |
1437 | * to even clean up the mess. | |
1438 | */ | |
1439 | out_fail: | |
1440 | put_online_cpus(); | |
1441 | mutex_unlock(&buffer->mutex); | |
18421015 | 1442 | atomic_dec(&buffer->record_disabled); |
554f786e | 1443 | return -1; |
7a8e76a3 | 1444 | } |
c4f50183 | 1445 | EXPORT_SYMBOL_GPL(ring_buffer_resize); |
7a8e76a3 | 1446 | |
750912fa DS |
1447 | void ring_buffer_change_overwrite(struct ring_buffer *buffer, int val) |
1448 | { | |
1449 | mutex_lock(&buffer->mutex); | |
1450 | if (val) | |
1451 | buffer->flags |= RB_FL_OVERWRITE; | |
1452 | else | |
1453 | buffer->flags &= ~RB_FL_OVERWRITE; | |
1454 | mutex_unlock(&buffer->mutex); | |
1455 | } | |
1456 | EXPORT_SYMBOL_GPL(ring_buffer_change_overwrite); | |
1457 | ||
8789a9e7 | 1458 | static inline void * |
044fa782 | 1459 | __rb_data_page_index(struct buffer_data_page *bpage, unsigned index) |
8789a9e7 | 1460 | { |
044fa782 | 1461 | return bpage->data + index; |
8789a9e7 SR |
1462 | } |
1463 | ||
044fa782 | 1464 | static inline void *__rb_page_index(struct buffer_page *bpage, unsigned index) |
7a8e76a3 | 1465 | { |
044fa782 | 1466 | return bpage->page->data + index; |
7a8e76a3 SR |
1467 | } |
1468 | ||
1469 | static inline struct ring_buffer_event * | |
d769041f | 1470 | rb_reader_event(struct ring_buffer_per_cpu *cpu_buffer) |
7a8e76a3 | 1471 | { |
6f807acd SR |
1472 | return __rb_page_index(cpu_buffer->reader_page, |
1473 | cpu_buffer->reader_page->read); | |
1474 | } | |
1475 | ||
7a8e76a3 SR |
1476 | static inline struct ring_buffer_event * |
1477 | rb_iter_head_event(struct ring_buffer_iter *iter) | |
1478 | { | |
6f807acd | 1479 | return __rb_page_index(iter->head_page, iter->head); |
7a8e76a3 SR |
1480 | } |
1481 | ||
77ae365e | 1482 | static inline unsigned long rb_page_write(struct buffer_page *bpage) |
bf41a158 | 1483 | { |
77ae365e | 1484 | return local_read(&bpage->write) & RB_WRITE_MASK; |
bf41a158 SR |
1485 | } |
1486 | ||
1487 | static inline unsigned rb_page_commit(struct buffer_page *bpage) | |
1488 | { | |
abc9b56d | 1489 | return local_read(&bpage->page->commit); |
bf41a158 SR |
1490 | } |
1491 | ||
77ae365e SR |
1492 | static inline unsigned long rb_page_entries(struct buffer_page *bpage) |
1493 | { | |
1494 | return local_read(&bpage->entries) & RB_WRITE_MASK; | |
1495 | } | |
1496 | ||
25985edc | 1497 | /* Size is determined by what has been committed */ |
bf41a158 SR |
1498 | static inline unsigned rb_page_size(struct buffer_page *bpage) |
1499 | { | |
1500 | return rb_page_commit(bpage); | |
1501 | } | |
1502 | ||
1503 | static inline unsigned | |
1504 | rb_commit_index(struct ring_buffer_per_cpu *cpu_buffer) | |
1505 | { | |
1506 | return rb_page_commit(cpu_buffer->commit_page); | |
1507 | } | |
1508 | ||
bf41a158 SR |
1509 | static inline unsigned |
1510 | rb_event_index(struct ring_buffer_event *event) | |
1511 | { | |
1512 | unsigned long addr = (unsigned long)event; | |
1513 | ||
22f470f8 | 1514 | return (addr & ~PAGE_MASK) - BUF_PAGE_HDR_SIZE; |
bf41a158 SR |
1515 | } |
1516 | ||
0f0c85fc | 1517 | static inline int |
fa743953 SR |
1518 | rb_event_is_commit(struct ring_buffer_per_cpu *cpu_buffer, |
1519 | struct ring_buffer_event *event) | |
bf41a158 SR |
1520 | { |
1521 | unsigned long addr = (unsigned long)event; | |
1522 | unsigned long index; | |
1523 | ||
1524 | index = rb_event_index(event); | |
1525 | addr &= PAGE_MASK; | |
1526 | ||
1527 | return cpu_buffer->commit_page->page == (void *)addr && | |
1528 | rb_commit_index(cpu_buffer) == index; | |
1529 | } | |
1530 | ||
34a148bf | 1531 | static void |
bf41a158 | 1532 | rb_set_commit_to_write(struct ring_buffer_per_cpu *cpu_buffer) |
7a8e76a3 | 1533 | { |
77ae365e SR |
1534 | unsigned long max_count; |
1535 | ||
bf41a158 SR |
1536 | /* |
1537 | * We only race with interrupts and NMIs on this CPU. | |
1538 | * If we own the commit event, then we can commit | |
1539 | * all others that interrupted us, since the interruptions | |
1540 | * are in stack format (they finish before they come | |
1541 | * back to us). This allows us to do a simple loop to | |
1542 | * assign the commit to the tail. | |
1543 | */ | |
a8ccf1d6 | 1544 | again: |
77ae365e SR |
1545 | max_count = cpu_buffer->buffer->pages * 100; |
1546 | ||
bf41a158 | 1547 | while (cpu_buffer->commit_page != cpu_buffer->tail_page) { |
77ae365e SR |
1548 | if (RB_WARN_ON(cpu_buffer, !(--max_count))) |
1549 | return; | |
1550 | if (RB_WARN_ON(cpu_buffer, | |
1551 | rb_is_reader_page(cpu_buffer->tail_page))) | |
1552 | return; | |
1553 | local_set(&cpu_buffer->commit_page->page->commit, | |
1554 | rb_page_write(cpu_buffer->commit_page)); | |
bf41a158 | 1555 | rb_inc_page(cpu_buffer, &cpu_buffer->commit_page); |
abc9b56d SR |
1556 | cpu_buffer->write_stamp = |
1557 | cpu_buffer->commit_page->page->time_stamp; | |
bf41a158 SR |
1558 | /* add barrier to keep gcc from optimizing too much */ |
1559 | barrier(); | |
1560 | } | |
1561 | while (rb_commit_index(cpu_buffer) != | |
1562 | rb_page_write(cpu_buffer->commit_page)) { | |
77ae365e SR |
1563 | |
1564 | local_set(&cpu_buffer->commit_page->page->commit, | |
1565 | rb_page_write(cpu_buffer->commit_page)); | |
1566 | RB_WARN_ON(cpu_buffer, | |
1567 | local_read(&cpu_buffer->commit_page->page->commit) & | |
1568 | ~RB_WRITE_MASK); | |
bf41a158 SR |
1569 | barrier(); |
1570 | } | |
a8ccf1d6 SR |
1571 | |
1572 | /* again, keep gcc from optimizing */ | |
1573 | barrier(); | |
1574 | ||
1575 | /* | |
1576 | * If an interrupt came in just after the first while loop | |
1577 | * and pushed the tail page forward, we will be left with | |
1578 | * a dangling commit that will never go forward. | |
1579 | */ | |
1580 | if (unlikely(cpu_buffer->commit_page != cpu_buffer->tail_page)) | |
1581 | goto again; | |
7a8e76a3 SR |
1582 | } |
1583 | ||
d769041f | 1584 | static void rb_reset_reader_page(struct ring_buffer_per_cpu *cpu_buffer) |
7a8e76a3 | 1585 | { |
abc9b56d | 1586 | cpu_buffer->read_stamp = cpu_buffer->reader_page->page->time_stamp; |
6f807acd | 1587 | cpu_buffer->reader_page->read = 0; |
d769041f SR |
1588 | } |
1589 | ||
34a148bf | 1590 | static void rb_inc_iter(struct ring_buffer_iter *iter) |
d769041f SR |
1591 | { |
1592 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
1593 | ||
1594 | /* | |
1595 | * The iterator could be on the reader page (it starts there). | |
1596 | * But the head could have moved, since the reader was | |
1597 | * found. Check for this case and assign the iterator | |
1598 | * to the head page instead of next. | |
1599 | */ | |
1600 | if (iter->head_page == cpu_buffer->reader_page) | |
77ae365e | 1601 | iter->head_page = rb_set_head_page(cpu_buffer); |
d769041f SR |
1602 | else |
1603 | rb_inc_page(cpu_buffer, &iter->head_page); | |
1604 | ||
abc9b56d | 1605 | iter->read_stamp = iter->head_page->page->time_stamp; |
7a8e76a3 SR |
1606 | iter->head = 0; |
1607 | } | |
1608 | ||
69d1b839 SR |
1609 | /* Slow path, do not inline */ |
1610 | static noinline struct ring_buffer_event * | |
1611 | rb_add_time_stamp(struct ring_buffer_event *event, u64 delta) | |
1612 | { | |
1613 | event->type_len = RINGBUF_TYPE_TIME_EXTEND; | |
1614 | ||
1615 | /* Not the first event on the page? */ | |
1616 | if (rb_event_index(event)) { | |
1617 | event->time_delta = delta & TS_MASK; | |
1618 | event->array[0] = delta >> TS_SHIFT; | |
1619 | } else { | |
1620 | /* nope, just zero it */ | |
1621 | event->time_delta = 0; | |
1622 | event->array[0] = 0; | |
1623 | } | |
1624 | ||
1625 | return skip_time_extend(event); | |
1626 | } | |
1627 | ||
7a8e76a3 SR |
1628 | /** |
1629 | * ring_buffer_update_event - update event type and data | |
1630 | * @event: the even to update | |
1631 | * @type: the type of event | |
1632 | * @length: the size of the event field in the ring buffer | |
1633 | * | |
1634 | * Update the type and data fields of the event. The length | |
1635 | * is the actual size that is written to the ring buffer, | |
1636 | * and with this, we can determine what to place into the | |
1637 | * data field. | |
1638 | */ | |
34a148bf | 1639 | static void |
69d1b839 SR |
1640 | rb_update_event(struct ring_buffer_per_cpu *cpu_buffer, |
1641 | struct ring_buffer_event *event, unsigned length, | |
1642 | int add_timestamp, u64 delta) | |
7a8e76a3 | 1643 | { |
69d1b839 SR |
1644 | /* Only a commit updates the timestamp */ |
1645 | if (unlikely(!rb_event_is_commit(cpu_buffer, event))) | |
1646 | delta = 0; | |
7a8e76a3 | 1647 | |
69d1b839 SR |
1648 | /* |
1649 | * If we need to add a timestamp, then we | |
1650 | * add it to the start of the resevered space. | |
1651 | */ | |
1652 | if (unlikely(add_timestamp)) { | |
1653 | event = rb_add_time_stamp(event, delta); | |
1654 | length -= RB_LEN_TIME_EXTEND; | |
1655 | delta = 0; | |
7a8e76a3 | 1656 | } |
69d1b839 SR |
1657 | |
1658 | event->time_delta = delta; | |
1659 | length -= RB_EVNT_HDR_SIZE; | |
1660 | if (length > RB_MAX_SMALL_DATA || RB_FORCE_8BYTE_ALIGNMENT) { | |
1661 | event->type_len = 0; | |
1662 | event->array[0] = length; | |
1663 | } else | |
1664 | event->type_len = DIV_ROUND_UP(length, RB_ALIGNMENT); | |
7a8e76a3 SR |
1665 | } |
1666 | ||
77ae365e SR |
1667 | /* |
1668 | * rb_handle_head_page - writer hit the head page | |
1669 | * | |
1670 | * Returns: +1 to retry page | |
1671 | * 0 to continue | |
1672 | * -1 on error | |
1673 | */ | |
1674 | static int | |
1675 | rb_handle_head_page(struct ring_buffer_per_cpu *cpu_buffer, | |
1676 | struct buffer_page *tail_page, | |
1677 | struct buffer_page *next_page) | |
1678 | { | |
1679 | struct buffer_page *new_head; | |
1680 | int entries; | |
1681 | int type; | |
1682 | int ret; | |
1683 | ||
1684 | entries = rb_page_entries(next_page); | |
1685 | ||
1686 | /* | |
1687 | * The hard part is here. We need to move the head | |
1688 | * forward, and protect against both readers on | |
1689 | * other CPUs and writers coming in via interrupts. | |
1690 | */ | |
1691 | type = rb_head_page_set_update(cpu_buffer, next_page, tail_page, | |
1692 | RB_PAGE_HEAD); | |
1693 | ||
1694 | /* | |
1695 | * type can be one of four: | |
1696 | * NORMAL - an interrupt already moved it for us | |
1697 | * HEAD - we are the first to get here. | |
1698 | * UPDATE - we are the interrupt interrupting | |
1699 | * a current move. | |
1700 | * MOVED - a reader on another CPU moved the next | |
1701 | * pointer to its reader page. Give up | |
1702 | * and try again. | |
1703 | */ | |
1704 | ||
1705 | switch (type) { | |
1706 | case RB_PAGE_HEAD: | |
1707 | /* | |
1708 | * We changed the head to UPDATE, thus | |
1709 | * it is our responsibility to update | |
1710 | * the counters. | |
1711 | */ | |
1712 | local_add(entries, &cpu_buffer->overrun); | |
c64e148a | 1713 | local_sub(BUF_PAGE_SIZE, &cpu_buffer->entries_bytes); |
77ae365e SR |
1714 | |
1715 | /* | |
1716 | * The entries will be zeroed out when we move the | |
1717 | * tail page. | |
1718 | */ | |
1719 | ||
1720 | /* still more to do */ | |
1721 | break; | |
1722 | ||
1723 | case RB_PAGE_UPDATE: | |
1724 | /* | |
1725 | * This is an interrupt that interrupt the | |
1726 | * previous update. Still more to do. | |
1727 | */ | |
1728 | break; | |
1729 | case RB_PAGE_NORMAL: | |
1730 | /* | |
1731 | * An interrupt came in before the update | |
1732 | * and processed this for us. | |
1733 | * Nothing left to do. | |
1734 | */ | |
1735 | return 1; | |
1736 | case RB_PAGE_MOVED: | |
1737 | /* | |
1738 | * The reader is on another CPU and just did | |
1739 | * a swap with our next_page. | |
1740 | * Try again. | |
1741 | */ | |
1742 | return 1; | |
1743 | default: | |
1744 | RB_WARN_ON(cpu_buffer, 1); /* WTF??? */ | |
1745 | return -1; | |
1746 | } | |
1747 | ||
1748 | /* | |
1749 | * Now that we are here, the old head pointer is | |
1750 | * set to UPDATE. This will keep the reader from | |
1751 | * swapping the head page with the reader page. | |
1752 | * The reader (on another CPU) will spin till | |
1753 | * we are finished. | |
1754 | * | |
1755 | * We just need to protect against interrupts | |
1756 | * doing the job. We will set the next pointer | |
1757 | * to HEAD. After that, we set the old pointer | |
1758 | * to NORMAL, but only if it was HEAD before. | |
1759 | * otherwise we are an interrupt, and only | |
1760 | * want the outer most commit to reset it. | |
1761 | */ | |
1762 | new_head = next_page; | |
1763 | rb_inc_page(cpu_buffer, &new_head); | |
1764 | ||
1765 | ret = rb_head_page_set_head(cpu_buffer, new_head, next_page, | |
1766 | RB_PAGE_NORMAL); | |
1767 | ||
1768 | /* | |
1769 | * Valid returns are: | |
1770 | * HEAD - an interrupt came in and already set it. | |
1771 | * NORMAL - One of two things: | |
1772 | * 1) We really set it. | |
1773 | * 2) A bunch of interrupts came in and moved | |
1774 | * the page forward again. | |
1775 | */ | |
1776 | switch (ret) { | |
1777 | case RB_PAGE_HEAD: | |
1778 | case RB_PAGE_NORMAL: | |
1779 | /* OK */ | |
1780 | break; | |
1781 | default: | |
1782 | RB_WARN_ON(cpu_buffer, 1); | |
1783 | return -1; | |
1784 | } | |
1785 | ||
1786 | /* | |
1787 | * It is possible that an interrupt came in, | |
1788 | * set the head up, then more interrupts came in | |
1789 | * and moved it again. When we get back here, | |
1790 | * the page would have been set to NORMAL but we | |
1791 | * just set it back to HEAD. | |
1792 | * | |
1793 | * How do you detect this? Well, if that happened | |
1794 | * the tail page would have moved. | |
1795 | */ | |
1796 | if (ret == RB_PAGE_NORMAL) { | |
1797 | /* | |
1798 | * If the tail had moved passed next, then we need | |
1799 | * to reset the pointer. | |
1800 | */ | |
1801 | if (cpu_buffer->tail_page != tail_page && | |
1802 | cpu_buffer->tail_page != next_page) | |
1803 | rb_head_page_set_normal(cpu_buffer, new_head, | |
1804 | next_page, | |
1805 | RB_PAGE_HEAD); | |
1806 | } | |
1807 | ||
1808 | /* | |
1809 | * If this was the outer most commit (the one that | |
1810 | * changed the original pointer from HEAD to UPDATE), | |
1811 | * then it is up to us to reset it to NORMAL. | |
1812 | */ | |
1813 | if (type == RB_PAGE_HEAD) { | |
1814 | ret = rb_head_page_set_normal(cpu_buffer, next_page, | |
1815 | tail_page, | |
1816 | RB_PAGE_UPDATE); | |
1817 | if (RB_WARN_ON(cpu_buffer, | |
1818 | ret != RB_PAGE_UPDATE)) | |
1819 | return -1; | |
1820 | } | |
1821 | ||
1822 | return 0; | |
1823 | } | |
1824 | ||
34a148bf | 1825 | static unsigned rb_calculate_event_length(unsigned length) |
7a8e76a3 SR |
1826 | { |
1827 | struct ring_buffer_event event; /* Used only for sizeof array */ | |
1828 | ||
1829 | /* zero length can cause confusions */ | |
1830 | if (!length) | |
1831 | length = 1; | |
1832 | ||
2271048d | 1833 | if (length > RB_MAX_SMALL_DATA || RB_FORCE_8BYTE_ALIGNMENT) |
7a8e76a3 SR |
1834 | length += sizeof(event.array[0]); |
1835 | ||
1836 | length += RB_EVNT_HDR_SIZE; | |
2271048d | 1837 | length = ALIGN(length, RB_ARCH_ALIGNMENT); |
7a8e76a3 SR |
1838 | |
1839 | return length; | |
1840 | } | |
1841 | ||
c7b09308 SR |
1842 | static inline void |
1843 | rb_reset_tail(struct ring_buffer_per_cpu *cpu_buffer, | |
1844 | struct buffer_page *tail_page, | |
1845 | unsigned long tail, unsigned long length) | |
1846 | { | |
1847 | struct ring_buffer_event *event; | |
1848 | ||
1849 | /* | |
1850 | * Only the event that crossed the page boundary | |
1851 | * must fill the old tail_page with padding. | |
1852 | */ | |
1853 | if (tail >= BUF_PAGE_SIZE) { | |
b3230c8b SR |
1854 | /* |
1855 | * If the page was filled, then we still need | |
1856 | * to update the real_end. Reset it to zero | |
1857 | * and the reader will ignore it. | |
1858 | */ | |
1859 | if (tail == BUF_PAGE_SIZE) | |
1860 | tail_page->real_end = 0; | |
1861 | ||
c7b09308 SR |
1862 | local_sub(length, &tail_page->write); |
1863 | return; | |
1864 | } | |
1865 | ||
1866 | event = __rb_page_index(tail_page, tail); | |
b0b7065b | 1867 | kmemcheck_annotate_bitfield(event, bitfield); |
c7b09308 | 1868 | |
c64e148a VN |
1869 | /* account for padding bytes */ |
1870 | local_add(BUF_PAGE_SIZE - tail, &cpu_buffer->entries_bytes); | |
1871 | ||
ff0ff84a SR |
1872 | /* |
1873 | * Save the original length to the meta data. | |
1874 | * This will be used by the reader to add lost event | |
1875 | * counter. | |
1876 | */ | |
1877 | tail_page->real_end = tail; | |
1878 | ||
c7b09308 SR |
1879 | /* |
1880 | * If this event is bigger than the minimum size, then | |
1881 | * we need to be careful that we don't subtract the | |
1882 | * write counter enough to allow another writer to slip | |
1883 | * in on this page. | |
1884 | * We put in a discarded commit instead, to make sure | |
1885 | * that this space is not used again. | |
1886 | * | |
1887 | * If we are less than the minimum size, we don't need to | |
1888 | * worry about it. | |
1889 | */ | |
1890 | if (tail > (BUF_PAGE_SIZE - RB_EVNT_MIN_SIZE)) { | |
1891 | /* No room for any events */ | |
1892 | ||
1893 | /* Mark the rest of the page with padding */ | |
1894 | rb_event_set_padding(event); | |
1895 | ||
1896 | /* Set the write back to the previous setting */ | |
1897 | local_sub(length, &tail_page->write); | |
1898 | return; | |
1899 | } | |
1900 | ||
1901 | /* Put in a discarded event */ | |
1902 | event->array[0] = (BUF_PAGE_SIZE - tail) - RB_EVNT_HDR_SIZE; | |
1903 | event->type_len = RINGBUF_TYPE_PADDING; | |
1904 | /* time delta must be non zero */ | |
1905 | event->time_delta = 1; | |
c7b09308 SR |
1906 | |
1907 | /* Set write to end of buffer */ | |
1908 | length = (tail + length) - BUF_PAGE_SIZE; | |
1909 | local_sub(length, &tail_page->write); | |
1910 | } | |
6634ff26 | 1911 | |
747e94ae SR |
1912 | /* |
1913 | * This is the slow path, force gcc not to inline it. | |
1914 | */ | |
1915 | static noinline struct ring_buffer_event * | |
6634ff26 SR |
1916 | rb_move_tail(struct ring_buffer_per_cpu *cpu_buffer, |
1917 | unsigned long length, unsigned long tail, | |
e8bc43e8 | 1918 | struct buffer_page *tail_page, u64 ts) |
7a8e76a3 | 1919 | { |
5a50e33c | 1920 | struct buffer_page *commit_page = cpu_buffer->commit_page; |
7a8e76a3 | 1921 | struct ring_buffer *buffer = cpu_buffer->buffer; |
77ae365e SR |
1922 | struct buffer_page *next_page; |
1923 | int ret; | |
aa20ae84 SR |
1924 | |
1925 | next_page = tail_page; | |
1926 | ||
aa20ae84 SR |
1927 | rb_inc_page(cpu_buffer, &next_page); |
1928 | ||
aa20ae84 SR |
1929 | /* |
1930 | * If for some reason, we had an interrupt storm that made | |
1931 | * it all the way around the buffer, bail, and warn | |
1932 | * about it. | |
1933 | */ | |
1934 | if (unlikely(next_page == commit_page)) { | |
77ae365e | 1935 | local_inc(&cpu_buffer->commit_overrun); |
aa20ae84 SR |
1936 | goto out_reset; |
1937 | } | |
1938 | ||
77ae365e SR |
1939 | /* |
1940 | * This is where the fun begins! | |
1941 | * | |
1942 | * We are fighting against races between a reader that | |
1943 | * could be on another CPU trying to swap its reader | |
1944 | * page with the buffer head. | |
1945 | * | |
1946 | * We are also fighting against interrupts coming in and | |
1947 | * moving the head or tail on us as well. | |
1948 | * | |
1949 | * If the next page is the head page then we have filled | |
1950 | * the buffer, unless the commit page is still on the | |
1951 | * reader page. | |
1952 | */ | |
1953 | if (rb_is_head_page(cpu_buffer, next_page, &tail_page->list)) { | |
aa20ae84 | 1954 | |
77ae365e SR |
1955 | /* |
1956 | * If the commit is not on the reader page, then | |
1957 | * move the header page. | |
1958 | */ | |
1959 | if (!rb_is_reader_page(cpu_buffer->commit_page)) { | |
1960 | /* | |
1961 | * If we are not in overwrite mode, | |
1962 | * this is easy, just stop here. | |
1963 | */ | |
1964 | if (!(buffer->flags & RB_FL_OVERWRITE)) | |
1965 | goto out_reset; | |
1966 | ||
1967 | ret = rb_handle_head_page(cpu_buffer, | |
1968 | tail_page, | |
1969 | next_page); | |
1970 | if (ret < 0) | |
1971 | goto out_reset; | |
1972 | if (ret) | |
1973 | goto out_again; | |
1974 | } else { | |
1975 | /* | |
1976 | * We need to be careful here too. The | |
1977 | * commit page could still be on the reader | |
1978 | * page. We could have a small buffer, and | |
1979 | * have filled up the buffer with events | |
1980 | * from interrupts and such, and wrapped. | |
1981 | * | |
1982 | * Note, if the tail page is also the on the | |
1983 | * reader_page, we let it move out. | |
1984 | */ | |
1985 | if (unlikely((cpu_buffer->commit_page != | |
1986 | cpu_buffer->tail_page) && | |
1987 | (cpu_buffer->commit_page == | |
1988 | cpu_buffer->reader_page))) { | |
1989 | local_inc(&cpu_buffer->commit_overrun); | |
1990 | goto out_reset; | |
1991 | } | |
aa20ae84 SR |
1992 | } |
1993 | } | |
1994 | ||
77ae365e SR |
1995 | ret = rb_tail_page_update(cpu_buffer, tail_page, next_page); |
1996 | if (ret) { | |
1997 | /* | |
1998 | * Nested commits always have zero deltas, so | |
1999 | * just reread the time stamp | |
2000 | */ | |
e8bc43e8 SR |
2001 | ts = rb_time_stamp(buffer); |
2002 | next_page->page->time_stamp = ts; | |
aa20ae84 SR |
2003 | } |
2004 | ||
77ae365e | 2005 | out_again: |
aa20ae84 | 2006 | |
77ae365e | 2007 | rb_reset_tail(cpu_buffer, tail_page, tail, length); |
aa20ae84 SR |
2008 | |
2009 | /* fail and let the caller try again */ | |
2010 | return ERR_PTR(-EAGAIN); | |
2011 | ||
45141d46 | 2012 | out_reset: |
6f3b3440 | 2013 | /* reset write */ |
c7b09308 | 2014 | rb_reset_tail(cpu_buffer, tail_page, tail, length); |
6f3b3440 | 2015 | |
bf41a158 | 2016 | return NULL; |
7a8e76a3 SR |
2017 | } |
2018 | ||
6634ff26 SR |
2019 | static struct ring_buffer_event * |
2020 | __rb_reserve_next(struct ring_buffer_per_cpu *cpu_buffer, | |
69d1b839 SR |
2021 | unsigned long length, u64 ts, |
2022 | u64 delta, int add_timestamp) | |
6634ff26 | 2023 | { |
5a50e33c | 2024 | struct buffer_page *tail_page; |
6634ff26 SR |
2025 | struct ring_buffer_event *event; |
2026 | unsigned long tail, write; | |
2027 | ||
69d1b839 SR |
2028 | /* |
2029 | * If the time delta since the last event is too big to | |
2030 | * hold in the time field of the event, then we append a | |
2031 | * TIME EXTEND event ahead of the data event. | |
2032 | */ | |
2033 | if (unlikely(add_timestamp)) | |
2034 | length += RB_LEN_TIME_EXTEND; | |
2035 | ||
6634ff26 SR |
2036 | tail_page = cpu_buffer->tail_page; |
2037 | write = local_add_return(length, &tail_page->write); | |
77ae365e SR |
2038 | |
2039 | /* set write to only the index of the write */ | |
2040 | write &= RB_WRITE_MASK; | |
6634ff26 SR |
2041 | tail = write - length; |
2042 | ||
2043 | /* See if we shot pass the end of this buffer page */ | |
747e94ae | 2044 | if (unlikely(write > BUF_PAGE_SIZE)) |
6634ff26 | 2045 | return rb_move_tail(cpu_buffer, length, tail, |
5a50e33c | 2046 | tail_page, ts); |
6634ff26 SR |
2047 | |
2048 | /* We reserved something on the buffer */ | |
2049 | ||
6634ff26 | 2050 | event = __rb_page_index(tail_page, tail); |
1744a21d | 2051 | kmemcheck_annotate_bitfield(event, bitfield); |
69d1b839 | 2052 | rb_update_event(cpu_buffer, event, length, add_timestamp, delta); |
6634ff26 | 2053 | |
69d1b839 | 2054 | local_inc(&tail_page->entries); |
6634ff26 SR |
2055 | |
2056 | /* | |
fa743953 SR |
2057 | * If this is the first commit on the page, then update |
2058 | * its timestamp. | |
6634ff26 | 2059 | */ |
fa743953 | 2060 | if (!tail) |
e8bc43e8 | 2061 | tail_page->page->time_stamp = ts; |
6634ff26 | 2062 | |
c64e148a VN |
2063 | /* account for these added bytes */ |
2064 | local_add(length, &cpu_buffer->entries_bytes); | |
2065 | ||
6634ff26 SR |
2066 | return event; |
2067 | } | |
2068 | ||
edd813bf SR |
2069 | static inline int |
2070 | rb_try_to_discard(struct ring_buffer_per_cpu *cpu_buffer, | |
2071 | struct ring_buffer_event *event) | |
2072 | { | |
2073 | unsigned long new_index, old_index; | |
2074 | struct buffer_page *bpage; | |
2075 | unsigned long index; | |
2076 | unsigned long addr; | |
2077 | ||
2078 | new_index = rb_event_index(event); | |
69d1b839 | 2079 | old_index = new_index + rb_event_ts_length(event); |
edd813bf SR |
2080 | addr = (unsigned long)event; |
2081 | addr &= PAGE_MASK; | |
2082 | ||
2083 | bpage = cpu_buffer->tail_page; | |
2084 | ||
2085 | if (bpage->page == (void *)addr && rb_page_write(bpage) == old_index) { | |
77ae365e SR |
2086 | unsigned long write_mask = |
2087 | local_read(&bpage->write) & ~RB_WRITE_MASK; | |
c64e148a | 2088 | unsigned long event_length = rb_event_length(event); |
edd813bf SR |
2089 | /* |
2090 | * This is on the tail page. It is possible that | |
2091 | * a write could come in and move the tail page | |
2092 | * and write to the next page. That is fine | |
2093 | * because we just shorten what is on this page. | |
2094 | */ | |
77ae365e SR |
2095 | old_index += write_mask; |
2096 | new_index += write_mask; | |
edd813bf | 2097 | index = local_cmpxchg(&bpage->write, old_index, new_index); |
c64e148a VN |
2098 | if (index == old_index) { |
2099 | /* update counters */ | |
2100 | local_sub(event_length, &cpu_buffer->entries_bytes); | |
edd813bf | 2101 | return 1; |
c64e148a | 2102 | } |
edd813bf SR |
2103 | } |
2104 | ||
2105 | /* could not discard */ | |
2106 | return 0; | |
2107 | } | |
2108 | ||
fa743953 SR |
2109 | static void rb_start_commit(struct ring_buffer_per_cpu *cpu_buffer) |
2110 | { | |
2111 | local_inc(&cpu_buffer->committing); | |
2112 | local_inc(&cpu_buffer->commits); | |
2113 | } | |
2114 | ||
d9abde21 | 2115 | static inline void rb_end_commit(struct ring_buffer_per_cpu *cpu_buffer) |
fa743953 SR |
2116 | { |
2117 | unsigned long commits; | |
2118 | ||
2119 | if (RB_WARN_ON(cpu_buffer, | |
2120 | !local_read(&cpu_buffer->committing))) | |
2121 | return; | |
2122 | ||
2123 | again: | |
2124 | commits = local_read(&cpu_buffer->commits); | |
2125 | /* synchronize with interrupts */ | |
2126 | barrier(); | |
2127 | if (local_read(&cpu_buffer->committing) == 1) | |
2128 | rb_set_commit_to_write(cpu_buffer); | |
2129 | ||
2130 | local_dec(&cpu_buffer->committing); | |
2131 | ||
2132 | /* synchronize with interrupts */ | |
2133 | barrier(); | |
2134 | ||
2135 | /* | |
2136 | * Need to account for interrupts coming in between the | |
2137 | * updating of the commit page and the clearing of the | |
2138 | * committing counter. | |
2139 | */ | |
2140 | if (unlikely(local_read(&cpu_buffer->commits) != commits) && | |
2141 | !local_read(&cpu_buffer->committing)) { | |
2142 | local_inc(&cpu_buffer->committing); | |
2143 | goto again; | |
2144 | } | |
2145 | } | |
2146 | ||
7a8e76a3 | 2147 | static struct ring_buffer_event * |
62f0b3eb SR |
2148 | rb_reserve_next_event(struct ring_buffer *buffer, |
2149 | struct ring_buffer_per_cpu *cpu_buffer, | |
1cd8d735 | 2150 | unsigned long length) |
7a8e76a3 SR |
2151 | { |
2152 | struct ring_buffer_event *event; | |
69d1b839 | 2153 | u64 ts, delta; |
818e3dd3 | 2154 | int nr_loops = 0; |
69d1b839 | 2155 | int add_timestamp; |
140ff891 | 2156 | u64 diff; |
7a8e76a3 | 2157 | |
fa743953 SR |
2158 | rb_start_commit(cpu_buffer); |
2159 | ||
85bac32c | 2160 | #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP |
62f0b3eb SR |
2161 | /* |
2162 | * Due to the ability to swap a cpu buffer from a buffer | |
2163 | * it is possible it was swapped before we committed. | |
2164 | * (committing stops a swap). We check for it here and | |
2165 | * if it happened, we have to fail the write. | |
2166 | */ | |
2167 | barrier(); | |
2168 | if (unlikely(ACCESS_ONCE(cpu_buffer->buffer) != buffer)) { | |
2169 | local_dec(&cpu_buffer->committing); | |
2170 | local_dec(&cpu_buffer->commits); | |
2171 | return NULL; | |
2172 | } | |
85bac32c | 2173 | #endif |
62f0b3eb | 2174 | |
be957c44 | 2175 | length = rb_calculate_event_length(length); |
bf41a158 | 2176 | again: |
69d1b839 SR |
2177 | add_timestamp = 0; |
2178 | delta = 0; | |
2179 | ||
818e3dd3 SR |
2180 | /* |
2181 | * We allow for interrupts to reenter here and do a trace. | |
2182 | * If one does, it will cause this original code to loop | |
2183 | * back here. Even with heavy interrupts happening, this | |
2184 | * should only happen a few times in a row. If this happens | |
2185 | * 1000 times in a row, there must be either an interrupt | |
2186 | * storm or we have something buggy. | |
2187 | * Bail! | |
2188 | */ | |
3e89c7bb | 2189 | if (RB_WARN_ON(cpu_buffer, ++nr_loops > 1000)) |
fa743953 | 2190 | goto out_fail; |
818e3dd3 | 2191 | |
6d3f1e12 | 2192 | ts = rb_time_stamp(cpu_buffer->buffer); |
140ff891 | 2193 | diff = ts - cpu_buffer->write_stamp; |
7a8e76a3 | 2194 | |
140ff891 SR |
2195 | /* make sure this diff is calculated here */ |
2196 | barrier(); | |
bf41a158 | 2197 | |
140ff891 SR |
2198 | /* Did the write stamp get updated already? */ |
2199 | if (likely(ts >= cpu_buffer->write_stamp)) { | |
168b6b1d SR |
2200 | delta = diff; |
2201 | if (unlikely(test_time_stamp(delta))) { | |
31274d72 JO |
2202 | int local_clock_stable = 1; |
2203 | #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK | |
2204 | local_clock_stable = sched_clock_stable; | |
2205 | #endif | |
69d1b839 | 2206 | WARN_ONCE(delta > (1ULL << 59), |
31274d72 | 2207 | KERN_WARNING "Delta way too big! %llu ts=%llu write stamp = %llu\n%s", |
69d1b839 SR |
2208 | (unsigned long long)delta, |
2209 | (unsigned long long)ts, | |
31274d72 JO |
2210 | (unsigned long long)cpu_buffer->write_stamp, |
2211 | local_clock_stable ? "" : | |
2212 | "If you just came from a suspend/resume,\n" | |
2213 | "please switch to the trace global clock:\n" | |
2214 | " echo global > /sys/kernel/debug/tracing/trace_clock\n"); | |
69d1b839 | 2215 | add_timestamp = 1; |
7a8e76a3 | 2216 | } |
168b6b1d | 2217 | } |
7a8e76a3 | 2218 | |
69d1b839 SR |
2219 | event = __rb_reserve_next(cpu_buffer, length, ts, |
2220 | delta, add_timestamp); | |
168b6b1d | 2221 | if (unlikely(PTR_ERR(event) == -EAGAIN)) |
bf41a158 SR |
2222 | goto again; |
2223 | ||
fa743953 SR |
2224 | if (!event) |
2225 | goto out_fail; | |
7a8e76a3 | 2226 | |
7a8e76a3 | 2227 | return event; |
fa743953 SR |
2228 | |
2229 | out_fail: | |
2230 | rb_end_commit(cpu_buffer); | |
2231 | return NULL; | |
7a8e76a3 SR |
2232 | } |
2233 | ||
1155de47 PM |
2234 | #ifdef CONFIG_TRACING |
2235 | ||
aa18efb2 | 2236 | #define TRACE_RECURSIVE_DEPTH 16 |
261842b7 | 2237 | |
d9abde21 SR |
2238 | /* Keep this code out of the fast path cache */ |
2239 | static noinline void trace_recursive_fail(void) | |
261842b7 | 2240 | { |
aa18efb2 SR |
2241 | /* Disable all tracing before we do anything else */ |
2242 | tracing_off_permanent(); | |
261842b7 | 2243 | |
7d7d2b80 | 2244 | printk_once(KERN_WARNING "Tracing recursion: depth[%ld]:" |
aa18efb2 | 2245 | "HC[%lu]:SC[%lu]:NMI[%lu]\n", |
b1cff0ad | 2246 | trace_recursion_buffer(), |
aa18efb2 SR |
2247 | hardirq_count() >> HARDIRQ_SHIFT, |
2248 | softirq_count() >> SOFTIRQ_SHIFT, | |
2249 | in_nmi()); | |
261842b7 | 2250 | |
aa18efb2 | 2251 | WARN_ON_ONCE(1); |
d9abde21 SR |
2252 | } |
2253 | ||
2254 | static inline int trace_recursive_lock(void) | |
2255 | { | |
b1cff0ad | 2256 | trace_recursion_inc(); |
d9abde21 | 2257 | |
b1cff0ad | 2258 | if (likely(trace_recursion_buffer() < TRACE_RECURSIVE_DEPTH)) |
d9abde21 SR |
2259 | return 0; |
2260 | ||
2261 | trace_recursive_fail(); | |
2262 | ||
aa18efb2 | 2263 | return -1; |
261842b7 SR |
2264 | } |
2265 | ||
d9abde21 | 2266 | static inline void trace_recursive_unlock(void) |
261842b7 | 2267 | { |
b1cff0ad | 2268 | WARN_ON_ONCE(!trace_recursion_buffer()); |
261842b7 | 2269 | |
b1cff0ad | 2270 | trace_recursion_dec(); |
261842b7 SR |
2271 | } |
2272 | ||
1155de47 PM |
2273 | #else |
2274 | ||
2275 | #define trace_recursive_lock() (0) | |
2276 | #define trace_recursive_unlock() do { } while (0) | |
2277 | ||
2278 | #endif | |
2279 | ||
7a8e76a3 SR |
2280 | /** |
2281 | * ring_buffer_lock_reserve - reserve a part of the buffer | |
2282 | * @buffer: the ring buffer to reserve from | |
2283 | * @length: the length of the data to reserve (excluding event header) | |
7a8e76a3 SR |
2284 | * |
2285 | * Returns a reseverd event on the ring buffer to copy directly to. | |
2286 | * The user of this interface will need to get the body to write into | |
2287 | * and can use the ring_buffer_event_data() interface. | |
2288 | * | |
2289 | * The length is the length of the data needed, not the event length | |
2290 | * which also includes the event header. | |
2291 | * | |
2292 | * Must be paired with ring_buffer_unlock_commit, unless NULL is returned. | |
2293 | * If NULL is returned, then nothing has been allocated or locked. | |
2294 | */ | |
2295 | struct ring_buffer_event * | |
0a987751 | 2296 | ring_buffer_lock_reserve(struct ring_buffer *buffer, unsigned long length) |
7a8e76a3 SR |
2297 | { |
2298 | struct ring_buffer_per_cpu *cpu_buffer; | |
2299 | struct ring_buffer_event *event; | |
5168ae50 | 2300 | int cpu; |
7a8e76a3 | 2301 | |
033601a3 | 2302 | if (ring_buffer_flags != RB_BUFFERS_ON) |
a3583244 SR |
2303 | return NULL; |
2304 | ||
bf41a158 | 2305 | /* If we are tracing schedule, we don't want to recurse */ |
5168ae50 | 2306 | preempt_disable_notrace(); |
bf41a158 | 2307 | |
52fbe9cd LJ |
2308 | if (atomic_read(&buffer->record_disabled)) |
2309 | goto out_nocheck; | |
2310 | ||
261842b7 SR |
2311 | if (trace_recursive_lock()) |
2312 | goto out_nocheck; | |
2313 | ||
7a8e76a3 SR |
2314 | cpu = raw_smp_processor_id(); |
2315 | ||
9e01c1b7 | 2316 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
d769041f | 2317 | goto out; |
7a8e76a3 SR |
2318 | |
2319 | cpu_buffer = buffer->buffers[cpu]; | |
7a8e76a3 SR |
2320 | |
2321 | if (atomic_read(&cpu_buffer->record_disabled)) | |
d769041f | 2322 | goto out; |
7a8e76a3 | 2323 | |
be957c44 | 2324 | if (length > BUF_MAX_DATA_SIZE) |
bf41a158 | 2325 | goto out; |
7a8e76a3 | 2326 | |
62f0b3eb | 2327 | event = rb_reserve_next_event(buffer, cpu_buffer, length); |
7a8e76a3 | 2328 | if (!event) |
d769041f | 2329 | goto out; |
7a8e76a3 SR |
2330 | |
2331 | return event; | |
2332 | ||
d769041f | 2333 | out: |
261842b7 SR |
2334 | trace_recursive_unlock(); |
2335 | ||
2336 | out_nocheck: | |
5168ae50 | 2337 | preempt_enable_notrace(); |
7a8e76a3 SR |
2338 | return NULL; |
2339 | } | |
c4f50183 | 2340 | EXPORT_SYMBOL_GPL(ring_buffer_lock_reserve); |
7a8e76a3 | 2341 | |
a1863c21 SR |
2342 | static void |
2343 | rb_update_write_stamp(struct ring_buffer_per_cpu *cpu_buffer, | |
7a8e76a3 SR |
2344 | struct ring_buffer_event *event) |
2345 | { | |
69d1b839 SR |
2346 | u64 delta; |
2347 | ||
fa743953 SR |
2348 | /* |
2349 | * The event first in the commit queue updates the | |
2350 | * time stamp. | |
2351 | */ | |
69d1b839 SR |
2352 | if (rb_event_is_commit(cpu_buffer, event)) { |
2353 | /* | |
2354 | * A commit event that is first on a page | |
2355 | * updates the write timestamp with the page stamp | |
2356 | */ | |
2357 | if (!rb_event_index(event)) | |
2358 | cpu_buffer->write_stamp = | |
2359 | cpu_buffer->commit_page->page->time_stamp; | |
2360 | else if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) { | |
2361 | delta = event->array[0]; | |
2362 | delta <<= TS_SHIFT; | |
2363 | delta += event->time_delta; | |
2364 | cpu_buffer->write_stamp += delta; | |
2365 | } else | |
2366 | cpu_buffer->write_stamp += event->time_delta; | |
2367 | } | |
a1863c21 | 2368 | } |
bf41a158 | 2369 | |
a1863c21 SR |
2370 | static void rb_commit(struct ring_buffer_per_cpu *cpu_buffer, |
2371 | struct ring_buffer_event *event) | |
2372 | { | |
2373 | local_inc(&cpu_buffer->entries); | |
2374 | rb_update_write_stamp(cpu_buffer, event); | |
fa743953 | 2375 | rb_end_commit(cpu_buffer); |
7a8e76a3 SR |
2376 | } |
2377 | ||
2378 | /** | |
2379 | * ring_buffer_unlock_commit - commit a reserved | |
2380 | * @buffer: The buffer to commit to | |
2381 | * @event: The event pointer to commit. | |
7a8e76a3 SR |
2382 | * |
2383 | * This commits the data to the ring buffer, and releases any locks held. | |
2384 | * | |
2385 | * Must be paired with ring_buffer_lock_reserve. | |
2386 | */ | |
2387 | int ring_buffer_unlock_commit(struct ring_buffer *buffer, | |
0a987751 | 2388 | struct ring_buffer_event *event) |
7a8e76a3 SR |
2389 | { |
2390 | struct ring_buffer_per_cpu *cpu_buffer; | |
2391 | int cpu = raw_smp_processor_id(); | |
2392 | ||
2393 | cpu_buffer = buffer->buffers[cpu]; | |
2394 | ||
7a8e76a3 SR |
2395 | rb_commit(cpu_buffer, event); |
2396 | ||
261842b7 SR |
2397 | trace_recursive_unlock(); |
2398 | ||
5168ae50 | 2399 | preempt_enable_notrace(); |
7a8e76a3 SR |
2400 | |
2401 | return 0; | |
2402 | } | |
c4f50183 | 2403 | EXPORT_SYMBOL_GPL(ring_buffer_unlock_commit); |
7a8e76a3 | 2404 | |
f3b9aae1 FW |
2405 | static inline void rb_event_discard(struct ring_buffer_event *event) |
2406 | { | |
69d1b839 SR |
2407 | if (event->type_len == RINGBUF_TYPE_TIME_EXTEND) |
2408 | event = skip_time_extend(event); | |
2409 | ||
334d4169 LJ |
2410 | /* array[0] holds the actual length for the discarded event */ |
2411 | event->array[0] = rb_event_data_length(event) - RB_EVNT_HDR_SIZE; | |
2412 | event->type_len = RINGBUF_TYPE_PADDING; | |
f3b9aae1 FW |
2413 | /* time delta must be non zero */ |
2414 | if (!event->time_delta) | |
2415 | event->time_delta = 1; | |
2416 | } | |
2417 | ||
a1863c21 SR |
2418 | /* |
2419 | * Decrement the entries to the page that an event is on. | |
2420 | * The event does not even need to exist, only the pointer | |
2421 | * to the page it is on. This may only be called before the commit | |
2422 | * takes place. | |
2423 | */ | |
2424 | static inline void | |
2425 | rb_decrement_entry(struct ring_buffer_per_cpu *cpu_buffer, | |
2426 | struct ring_buffer_event *event) | |
2427 | { | |
2428 | unsigned long addr = (unsigned long)event; | |
2429 | struct buffer_page *bpage = cpu_buffer->commit_page; | |
2430 | struct buffer_page *start; | |
2431 | ||
2432 | addr &= PAGE_MASK; | |
2433 | ||
2434 | /* Do the likely case first */ | |
2435 | if (likely(bpage->page == (void *)addr)) { | |
2436 | local_dec(&bpage->entries); | |
2437 | return; | |
2438 | } | |
2439 | ||
2440 | /* | |
2441 | * Because the commit page may be on the reader page we | |
2442 | * start with the next page and check the end loop there. | |
2443 | */ | |
2444 | rb_inc_page(cpu_buffer, &bpage); | |
2445 | start = bpage; | |
2446 | do { | |
2447 | if (bpage->page == (void *)addr) { | |
2448 | local_dec(&bpage->entries); | |
2449 | return; | |
2450 | } | |
2451 | rb_inc_page(cpu_buffer, &bpage); | |
2452 | } while (bpage != start); | |
2453 | ||
2454 | /* commit not part of this buffer?? */ | |
2455 | RB_WARN_ON(cpu_buffer, 1); | |
2456 | } | |
2457 | ||
fa1b47dd SR |
2458 | /** |
2459 | * ring_buffer_commit_discard - discard an event that has not been committed | |
2460 | * @buffer: the ring buffer | |
2461 | * @event: non committed event to discard | |
2462 | * | |
dc892f73 SR |
2463 | * Sometimes an event that is in the ring buffer needs to be ignored. |
2464 | * This function lets the user discard an event in the ring buffer | |
2465 | * and then that event will not be read later. | |
2466 | * | |
2467 | * This function only works if it is called before the the item has been | |
2468 | * committed. It will try to free the event from the ring buffer | |
fa1b47dd SR |
2469 | * if another event has not been added behind it. |
2470 | * | |
2471 | * If another event has been added behind it, it will set the event | |
2472 | * up as discarded, and perform the commit. | |
2473 | * | |
2474 | * If this function is called, do not call ring_buffer_unlock_commit on | |
2475 | * the event. | |
2476 | */ | |
2477 | void ring_buffer_discard_commit(struct ring_buffer *buffer, | |
2478 | struct ring_buffer_event *event) | |
2479 | { | |
2480 | struct ring_buffer_per_cpu *cpu_buffer; | |
fa1b47dd SR |
2481 | int cpu; |
2482 | ||
2483 | /* The event is discarded regardless */ | |
f3b9aae1 | 2484 | rb_event_discard(event); |
fa1b47dd | 2485 | |
fa743953 SR |
2486 | cpu = smp_processor_id(); |
2487 | cpu_buffer = buffer->buffers[cpu]; | |
2488 | ||
fa1b47dd SR |
2489 | /* |
2490 | * This must only be called if the event has not been | |
2491 | * committed yet. Thus we can assume that preemption | |
2492 | * is still disabled. | |
2493 | */ | |
fa743953 | 2494 | RB_WARN_ON(buffer, !local_read(&cpu_buffer->committing)); |
fa1b47dd | 2495 | |
a1863c21 | 2496 | rb_decrement_entry(cpu_buffer, event); |
0f2541d2 | 2497 | if (rb_try_to_discard(cpu_buffer, event)) |
edd813bf | 2498 | goto out; |
fa1b47dd SR |
2499 | |
2500 | /* | |
2501 | * The commit is still visible by the reader, so we | |
a1863c21 | 2502 | * must still update the timestamp. |
fa1b47dd | 2503 | */ |
a1863c21 | 2504 | rb_update_write_stamp(cpu_buffer, event); |
fa1b47dd | 2505 | out: |
fa743953 | 2506 | rb_end_commit(cpu_buffer); |
fa1b47dd | 2507 | |
f3b9aae1 FW |
2508 | trace_recursive_unlock(); |
2509 | ||
5168ae50 | 2510 | preempt_enable_notrace(); |
fa1b47dd SR |
2511 | |
2512 | } | |
2513 | EXPORT_SYMBOL_GPL(ring_buffer_discard_commit); | |
2514 | ||
7a8e76a3 SR |
2515 | /** |
2516 | * ring_buffer_write - write data to the buffer without reserving | |
2517 | * @buffer: The ring buffer to write to. | |
2518 | * @length: The length of the data being written (excluding the event header) | |
2519 | * @data: The data to write to the buffer. | |
2520 | * | |
2521 | * This is like ring_buffer_lock_reserve and ring_buffer_unlock_commit as | |
2522 | * one function. If you already have the data to write to the buffer, it | |
2523 | * may be easier to simply call this function. | |
2524 | * | |
2525 | * Note, like ring_buffer_lock_reserve, the length is the length of the data | |
2526 | * and not the length of the event which would hold the header. | |
2527 | */ | |
2528 | int ring_buffer_write(struct ring_buffer *buffer, | |
2529 | unsigned long length, | |
2530 | void *data) | |
2531 | { | |
2532 | struct ring_buffer_per_cpu *cpu_buffer; | |
2533 | struct ring_buffer_event *event; | |
7a8e76a3 SR |
2534 | void *body; |
2535 | int ret = -EBUSY; | |
5168ae50 | 2536 | int cpu; |
7a8e76a3 | 2537 | |
033601a3 | 2538 | if (ring_buffer_flags != RB_BUFFERS_ON) |
a3583244 SR |
2539 | return -EBUSY; |
2540 | ||
5168ae50 | 2541 | preempt_disable_notrace(); |
bf41a158 | 2542 | |
52fbe9cd LJ |
2543 | if (atomic_read(&buffer->record_disabled)) |
2544 | goto out; | |
2545 | ||
7a8e76a3 SR |
2546 | cpu = raw_smp_processor_id(); |
2547 | ||
9e01c1b7 | 2548 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
d769041f | 2549 | goto out; |
7a8e76a3 SR |
2550 | |
2551 | cpu_buffer = buffer->buffers[cpu]; | |
7a8e76a3 SR |
2552 | |
2553 | if (atomic_read(&cpu_buffer->record_disabled)) | |
2554 | goto out; | |
2555 | ||
be957c44 SR |
2556 | if (length > BUF_MAX_DATA_SIZE) |
2557 | goto out; | |
2558 | ||
62f0b3eb | 2559 | event = rb_reserve_next_event(buffer, cpu_buffer, length); |
7a8e76a3 SR |
2560 | if (!event) |
2561 | goto out; | |
2562 | ||
2563 | body = rb_event_data(event); | |
2564 | ||
2565 | memcpy(body, data, length); | |
2566 | ||
2567 | rb_commit(cpu_buffer, event); | |
2568 | ||
2569 | ret = 0; | |
2570 | out: | |
5168ae50 | 2571 | preempt_enable_notrace(); |
7a8e76a3 SR |
2572 | |
2573 | return ret; | |
2574 | } | |
c4f50183 | 2575 | EXPORT_SYMBOL_GPL(ring_buffer_write); |
7a8e76a3 | 2576 | |
34a148bf | 2577 | static int rb_per_cpu_empty(struct ring_buffer_per_cpu *cpu_buffer) |
bf41a158 SR |
2578 | { |
2579 | struct buffer_page *reader = cpu_buffer->reader_page; | |
77ae365e | 2580 | struct buffer_page *head = rb_set_head_page(cpu_buffer); |
bf41a158 SR |
2581 | struct buffer_page *commit = cpu_buffer->commit_page; |
2582 | ||
77ae365e SR |
2583 | /* In case of error, head will be NULL */ |
2584 | if (unlikely(!head)) | |
2585 | return 1; | |
2586 | ||
bf41a158 SR |
2587 | return reader->read == rb_page_commit(reader) && |
2588 | (commit == reader || | |
2589 | (commit == head && | |
2590 | head->read == rb_page_commit(commit))); | |
2591 | } | |
2592 | ||
7a8e76a3 SR |
2593 | /** |
2594 | * ring_buffer_record_disable - stop all writes into the buffer | |
2595 | * @buffer: The ring buffer to stop writes to. | |
2596 | * | |
2597 | * This prevents all writes to the buffer. Any attempt to write | |
2598 | * to the buffer after this will fail and return NULL. | |
2599 | * | |
2600 | * The caller should call synchronize_sched() after this. | |
2601 | */ | |
2602 | void ring_buffer_record_disable(struct ring_buffer *buffer) | |
2603 | { | |
2604 | atomic_inc(&buffer->record_disabled); | |
2605 | } | |
c4f50183 | 2606 | EXPORT_SYMBOL_GPL(ring_buffer_record_disable); |
7a8e76a3 SR |
2607 | |
2608 | /** | |
2609 | * ring_buffer_record_enable - enable writes to the buffer | |
2610 | * @buffer: The ring buffer to enable writes | |
2611 | * | |
2612 | * Note, multiple disables will need the same number of enables | |
c41b20e7 | 2613 | * to truly enable the writing (much like preempt_disable). |
7a8e76a3 SR |
2614 | */ |
2615 | void ring_buffer_record_enable(struct ring_buffer *buffer) | |
2616 | { | |
2617 | atomic_dec(&buffer->record_disabled); | |
2618 | } | |
c4f50183 | 2619 | EXPORT_SYMBOL_GPL(ring_buffer_record_enable); |
7a8e76a3 SR |
2620 | |
2621 | /** | |
2622 | * ring_buffer_record_disable_cpu - stop all writes into the cpu_buffer | |
2623 | * @buffer: The ring buffer to stop writes to. | |
2624 | * @cpu: The CPU buffer to stop | |
2625 | * | |
2626 | * This prevents all writes to the buffer. Any attempt to write | |
2627 | * to the buffer after this will fail and return NULL. | |
2628 | * | |
2629 | * The caller should call synchronize_sched() after this. | |
2630 | */ | |
2631 | void ring_buffer_record_disable_cpu(struct ring_buffer *buffer, int cpu) | |
2632 | { | |
2633 | struct ring_buffer_per_cpu *cpu_buffer; | |
2634 | ||
9e01c1b7 | 2635 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 2636 | return; |
7a8e76a3 SR |
2637 | |
2638 | cpu_buffer = buffer->buffers[cpu]; | |
2639 | atomic_inc(&cpu_buffer->record_disabled); | |
2640 | } | |
c4f50183 | 2641 | EXPORT_SYMBOL_GPL(ring_buffer_record_disable_cpu); |
7a8e76a3 SR |
2642 | |
2643 | /** | |
2644 | * ring_buffer_record_enable_cpu - enable writes to the buffer | |
2645 | * @buffer: The ring buffer to enable writes | |
2646 | * @cpu: The CPU to enable. | |
2647 | * | |
2648 | * Note, multiple disables will need the same number of enables | |
c41b20e7 | 2649 | * to truly enable the writing (much like preempt_disable). |
7a8e76a3 SR |
2650 | */ |
2651 | void ring_buffer_record_enable_cpu(struct ring_buffer *buffer, int cpu) | |
2652 | { | |
2653 | struct ring_buffer_per_cpu *cpu_buffer; | |
2654 | ||
9e01c1b7 | 2655 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 2656 | return; |
7a8e76a3 SR |
2657 | |
2658 | cpu_buffer = buffer->buffers[cpu]; | |
2659 | atomic_dec(&cpu_buffer->record_disabled); | |
2660 | } | |
c4f50183 | 2661 | EXPORT_SYMBOL_GPL(ring_buffer_record_enable_cpu); |
7a8e76a3 | 2662 | |
f6195aa0 SR |
2663 | /* |
2664 | * The total entries in the ring buffer is the running counter | |
2665 | * of entries entered into the ring buffer, minus the sum of | |
2666 | * the entries read from the ring buffer and the number of | |
2667 | * entries that were overwritten. | |
2668 | */ | |
2669 | static inline unsigned long | |
2670 | rb_num_of_entries(struct ring_buffer_per_cpu *cpu_buffer) | |
2671 | { | |
2672 | return local_read(&cpu_buffer->entries) - | |
2673 | (local_read(&cpu_buffer->overrun) + cpu_buffer->read); | |
2674 | } | |
2675 | ||
c64e148a VN |
2676 | /** |
2677 | * ring_buffer_oldest_event_ts - get the oldest event timestamp from the buffer | |
2678 | * @buffer: The ring buffer | |
2679 | * @cpu: The per CPU buffer to read from. | |
2680 | */ | |
2681 | unsigned long ring_buffer_oldest_event_ts(struct ring_buffer *buffer, int cpu) | |
2682 | { | |
2683 | unsigned long flags; | |
2684 | struct ring_buffer_per_cpu *cpu_buffer; | |
2685 | struct buffer_page *bpage; | |
2686 | unsigned long ret; | |
2687 | ||
2688 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) | |
2689 | return 0; | |
2690 | ||
2691 | cpu_buffer = buffer->buffers[cpu]; | |
7115e3fc | 2692 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
c64e148a VN |
2693 | /* |
2694 | * if the tail is on reader_page, oldest time stamp is on the reader | |
2695 | * page | |
2696 | */ | |
2697 | if (cpu_buffer->tail_page == cpu_buffer->reader_page) | |
2698 | bpage = cpu_buffer->reader_page; | |
2699 | else | |
2700 | bpage = rb_set_head_page(cpu_buffer); | |
2701 | ret = bpage->page->time_stamp; | |
7115e3fc | 2702 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
c64e148a VN |
2703 | |
2704 | return ret; | |
2705 | } | |
2706 | EXPORT_SYMBOL_GPL(ring_buffer_oldest_event_ts); | |
2707 | ||
2708 | /** | |
2709 | * ring_buffer_bytes_cpu - get the number of bytes consumed in a cpu buffer | |
2710 | * @buffer: The ring buffer | |
2711 | * @cpu: The per CPU buffer to read from. | |
2712 | */ | |
2713 | unsigned long ring_buffer_bytes_cpu(struct ring_buffer *buffer, int cpu) | |
2714 | { | |
2715 | struct ring_buffer_per_cpu *cpu_buffer; | |
2716 | unsigned long ret; | |
2717 | ||
2718 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) | |
2719 | return 0; | |
2720 | ||
2721 | cpu_buffer = buffer->buffers[cpu]; | |
2722 | ret = local_read(&cpu_buffer->entries_bytes) - cpu_buffer->read_bytes; | |
2723 | ||
2724 | return ret; | |
2725 | } | |
2726 | EXPORT_SYMBOL_GPL(ring_buffer_bytes_cpu); | |
2727 | ||
7a8e76a3 SR |
2728 | /** |
2729 | * ring_buffer_entries_cpu - get the number of entries in a cpu buffer | |
2730 | * @buffer: The ring buffer | |
2731 | * @cpu: The per CPU buffer to get the entries from. | |
2732 | */ | |
2733 | unsigned long ring_buffer_entries_cpu(struct ring_buffer *buffer, int cpu) | |
2734 | { | |
2735 | struct ring_buffer_per_cpu *cpu_buffer; | |
2736 | ||
9e01c1b7 | 2737 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 2738 | return 0; |
7a8e76a3 SR |
2739 | |
2740 | cpu_buffer = buffer->buffers[cpu]; | |
554f786e | 2741 | |
f6195aa0 | 2742 | return rb_num_of_entries(cpu_buffer); |
7a8e76a3 | 2743 | } |
c4f50183 | 2744 | EXPORT_SYMBOL_GPL(ring_buffer_entries_cpu); |
7a8e76a3 SR |
2745 | |
2746 | /** | |
2747 | * ring_buffer_overrun_cpu - get the number of overruns in a cpu_buffer | |
2748 | * @buffer: The ring buffer | |
2749 | * @cpu: The per CPU buffer to get the number of overruns from | |
2750 | */ | |
2751 | unsigned long ring_buffer_overrun_cpu(struct ring_buffer *buffer, int cpu) | |
2752 | { | |
2753 | struct ring_buffer_per_cpu *cpu_buffer; | |
8aabee57 | 2754 | unsigned long ret; |
7a8e76a3 | 2755 | |
9e01c1b7 | 2756 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 2757 | return 0; |
7a8e76a3 SR |
2758 | |
2759 | cpu_buffer = buffer->buffers[cpu]; | |
77ae365e | 2760 | ret = local_read(&cpu_buffer->overrun); |
554f786e SR |
2761 | |
2762 | return ret; | |
7a8e76a3 | 2763 | } |
c4f50183 | 2764 | EXPORT_SYMBOL_GPL(ring_buffer_overrun_cpu); |
7a8e76a3 | 2765 | |
f0d2c681 SR |
2766 | /** |
2767 | * ring_buffer_commit_overrun_cpu - get the number of overruns caused by commits | |
2768 | * @buffer: The ring buffer | |
2769 | * @cpu: The per CPU buffer to get the number of overruns from | |
2770 | */ | |
2771 | unsigned long | |
2772 | ring_buffer_commit_overrun_cpu(struct ring_buffer *buffer, int cpu) | |
2773 | { | |
2774 | struct ring_buffer_per_cpu *cpu_buffer; | |
2775 | unsigned long ret; | |
2776 | ||
2777 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) | |
2778 | return 0; | |
2779 | ||
2780 | cpu_buffer = buffer->buffers[cpu]; | |
77ae365e | 2781 | ret = local_read(&cpu_buffer->commit_overrun); |
f0d2c681 SR |
2782 | |
2783 | return ret; | |
2784 | } | |
2785 | EXPORT_SYMBOL_GPL(ring_buffer_commit_overrun_cpu); | |
2786 | ||
7a8e76a3 SR |
2787 | /** |
2788 | * ring_buffer_entries - get the number of entries in a buffer | |
2789 | * @buffer: The ring buffer | |
2790 | * | |
2791 | * Returns the total number of entries in the ring buffer | |
2792 | * (all CPU entries) | |
2793 | */ | |
2794 | unsigned long ring_buffer_entries(struct ring_buffer *buffer) | |
2795 | { | |
2796 | struct ring_buffer_per_cpu *cpu_buffer; | |
2797 | unsigned long entries = 0; | |
2798 | int cpu; | |
2799 | ||
2800 | /* if you care about this being correct, lock the buffer */ | |
2801 | for_each_buffer_cpu(buffer, cpu) { | |
2802 | cpu_buffer = buffer->buffers[cpu]; | |
f6195aa0 | 2803 | entries += rb_num_of_entries(cpu_buffer); |
7a8e76a3 SR |
2804 | } |
2805 | ||
2806 | return entries; | |
2807 | } | |
c4f50183 | 2808 | EXPORT_SYMBOL_GPL(ring_buffer_entries); |
7a8e76a3 SR |
2809 | |
2810 | /** | |
67b394f7 | 2811 | * ring_buffer_overruns - get the number of overruns in buffer |
7a8e76a3 SR |
2812 | * @buffer: The ring buffer |
2813 | * | |
2814 | * Returns the total number of overruns in the ring buffer | |
2815 | * (all CPU entries) | |
2816 | */ | |
2817 | unsigned long ring_buffer_overruns(struct ring_buffer *buffer) | |
2818 | { | |
2819 | struct ring_buffer_per_cpu *cpu_buffer; | |
2820 | unsigned long overruns = 0; | |
2821 | int cpu; | |
2822 | ||
2823 | /* if you care about this being correct, lock the buffer */ | |
2824 | for_each_buffer_cpu(buffer, cpu) { | |
2825 | cpu_buffer = buffer->buffers[cpu]; | |
77ae365e | 2826 | overruns += local_read(&cpu_buffer->overrun); |
7a8e76a3 SR |
2827 | } |
2828 | ||
2829 | return overruns; | |
2830 | } | |
c4f50183 | 2831 | EXPORT_SYMBOL_GPL(ring_buffer_overruns); |
7a8e76a3 | 2832 | |
642edba5 | 2833 | static void rb_iter_reset(struct ring_buffer_iter *iter) |
7a8e76a3 SR |
2834 | { |
2835 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
2836 | ||
d769041f SR |
2837 | /* Iterator usage is expected to have record disabled */ |
2838 | if (list_empty(&cpu_buffer->reader_page->list)) { | |
77ae365e SR |
2839 | iter->head_page = rb_set_head_page(cpu_buffer); |
2840 | if (unlikely(!iter->head_page)) | |
2841 | return; | |
2842 | iter->head = iter->head_page->read; | |
d769041f SR |
2843 | } else { |
2844 | iter->head_page = cpu_buffer->reader_page; | |
6f807acd | 2845 | iter->head = cpu_buffer->reader_page->read; |
d769041f SR |
2846 | } |
2847 | if (iter->head) | |
2848 | iter->read_stamp = cpu_buffer->read_stamp; | |
2849 | else | |
abc9b56d | 2850 | iter->read_stamp = iter->head_page->page->time_stamp; |
492a74f4 SR |
2851 | iter->cache_reader_page = cpu_buffer->reader_page; |
2852 | iter->cache_read = cpu_buffer->read; | |
642edba5 | 2853 | } |
f83c9d0f | 2854 | |
642edba5 SR |
2855 | /** |
2856 | * ring_buffer_iter_reset - reset an iterator | |
2857 | * @iter: The iterator to reset | |
2858 | * | |
2859 | * Resets the iterator, so that it will start from the beginning | |
2860 | * again. | |
2861 | */ | |
2862 | void ring_buffer_iter_reset(struct ring_buffer_iter *iter) | |
2863 | { | |
554f786e | 2864 | struct ring_buffer_per_cpu *cpu_buffer; |
642edba5 SR |
2865 | unsigned long flags; |
2866 | ||
554f786e SR |
2867 | if (!iter) |
2868 | return; | |
2869 | ||
2870 | cpu_buffer = iter->cpu_buffer; | |
2871 | ||
5389f6fa | 2872 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
642edba5 | 2873 | rb_iter_reset(iter); |
5389f6fa | 2874 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
7a8e76a3 | 2875 | } |
c4f50183 | 2876 | EXPORT_SYMBOL_GPL(ring_buffer_iter_reset); |
7a8e76a3 SR |
2877 | |
2878 | /** | |
2879 | * ring_buffer_iter_empty - check if an iterator has no more to read | |
2880 | * @iter: The iterator to check | |
2881 | */ | |
2882 | int ring_buffer_iter_empty(struct ring_buffer_iter *iter) | |
2883 | { | |
2884 | struct ring_buffer_per_cpu *cpu_buffer; | |
2885 | ||
2886 | cpu_buffer = iter->cpu_buffer; | |
2887 | ||
bf41a158 SR |
2888 | return iter->head_page == cpu_buffer->commit_page && |
2889 | iter->head == rb_commit_index(cpu_buffer); | |
7a8e76a3 | 2890 | } |
c4f50183 | 2891 | EXPORT_SYMBOL_GPL(ring_buffer_iter_empty); |
7a8e76a3 SR |
2892 | |
2893 | static void | |
2894 | rb_update_read_stamp(struct ring_buffer_per_cpu *cpu_buffer, | |
2895 | struct ring_buffer_event *event) | |
2896 | { | |
2897 | u64 delta; | |
2898 | ||
334d4169 | 2899 | switch (event->type_len) { |
7a8e76a3 SR |
2900 | case RINGBUF_TYPE_PADDING: |
2901 | return; | |
2902 | ||
2903 | case RINGBUF_TYPE_TIME_EXTEND: | |
2904 | delta = event->array[0]; | |
2905 | delta <<= TS_SHIFT; | |
2906 | delta += event->time_delta; | |
2907 | cpu_buffer->read_stamp += delta; | |
2908 | return; | |
2909 | ||
2910 | case RINGBUF_TYPE_TIME_STAMP: | |
2911 | /* FIXME: not implemented */ | |
2912 | return; | |
2913 | ||
2914 | case RINGBUF_TYPE_DATA: | |
2915 | cpu_buffer->read_stamp += event->time_delta; | |
2916 | return; | |
2917 | ||
2918 | default: | |
2919 | BUG(); | |
2920 | } | |
2921 | return; | |
2922 | } | |
2923 | ||
2924 | static void | |
2925 | rb_update_iter_read_stamp(struct ring_buffer_iter *iter, | |
2926 | struct ring_buffer_event *event) | |
2927 | { | |
2928 | u64 delta; | |
2929 | ||
334d4169 | 2930 | switch (event->type_len) { |
7a8e76a3 SR |
2931 | case RINGBUF_TYPE_PADDING: |
2932 | return; | |
2933 | ||
2934 | case RINGBUF_TYPE_TIME_EXTEND: | |
2935 | delta = event->array[0]; | |
2936 | delta <<= TS_SHIFT; | |
2937 | delta += event->time_delta; | |
2938 | iter->read_stamp += delta; | |
2939 | return; | |
2940 | ||
2941 | case RINGBUF_TYPE_TIME_STAMP: | |
2942 | /* FIXME: not implemented */ | |
2943 | return; | |
2944 | ||
2945 | case RINGBUF_TYPE_DATA: | |
2946 | iter->read_stamp += event->time_delta; | |
2947 | return; | |
2948 | ||
2949 | default: | |
2950 | BUG(); | |
2951 | } | |
2952 | return; | |
2953 | } | |
2954 | ||
d769041f SR |
2955 | static struct buffer_page * |
2956 | rb_get_reader_page(struct ring_buffer_per_cpu *cpu_buffer) | |
7a8e76a3 | 2957 | { |
d769041f | 2958 | struct buffer_page *reader = NULL; |
66a8cb95 | 2959 | unsigned long overwrite; |
d769041f | 2960 | unsigned long flags; |
818e3dd3 | 2961 | int nr_loops = 0; |
77ae365e | 2962 | int ret; |
d769041f | 2963 | |
3e03fb7f | 2964 | local_irq_save(flags); |
0199c4e6 | 2965 | arch_spin_lock(&cpu_buffer->lock); |
d769041f SR |
2966 | |
2967 | again: | |
818e3dd3 SR |
2968 | /* |
2969 | * This should normally only loop twice. But because the | |
2970 | * start of the reader inserts an empty page, it causes | |
2971 | * a case where we will loop three times. There should be no | |
2972 | * reason to loop four times (that I know of). | |
2973 | */ | |
3e89c7bb | 2974 | if (RB_WARN_ON(cpu_buffer, ++nr_loops > 3)) { |
818e3dd3 SR |
2975 | reader = NULL; |
2976 | goto out; | |
2977 | } | |
2978 | ||
d769041f SR |
2979 | reader = cpu_buffer->reader_page; |
2980 | ||
2981 | /* If there's more to read, return this page */ | |
bf41a158 | 2982 | if (cpu_buffer->reader_page->read < rb_page_size(reader)) |
d769041f SR |
2983 | goto out; |
2984 | ||
2985 | /* Never should we have an index greater than the size */ | |
3e89c7bb SR |
2986 | if (RB_WARN_ON(cpu_buffer, |
2987 | cpu_buffer->reader_page->read > rb_page_size(reader))) | |
2988 | goto out; | |
d769041f SR |
2989 | |
2990 | /* check if we caught up to the tail */ | |
2991 | reader = NULL; | |
bf41a158 | 2992 | if (cpu_buffer->commit_page == cpu_buffer->reader_page) |
d769041f | 2993 | goto out; |
7a8e76a3 SR |
2994 | |
2995 | /* | |
d769041f | 2996 | * Reset the reader page to size zero. |
7a8e76a3 | 2997 | */ |
77ae365e SR |
2998 | local_set(&cpu_buffer->reader_page->write, 0); |
2999 | local_set(&cpu_buffer->reader_page->entries, 0); | |
3000 | local_set(&cpu_buffer->reader_page->page->commit, 0); | |
ff0ff84a | 3001 | cpu_buffer->reader_page->real_end = 0; |
7a8e76a3 | 3002 | |
77ae365e SR |
3003 | spin: |
3004 | /* | |
3005 | * Splice the empty reader page into the list around the head. | |
3006 | */ | |
3007 | reader = rb_set_head_page(cpu_buffer); | |
0e1ff5d7 | 3008 | cpu_buffer->reader_page->list.next = rb_list_head(reader->list.next); |
d769041f | 3009 | cpu_buffer->reader_page->list.prev = reader->list.prev; |
bf41a158 | 3010 | |
3adc54fa SR |
3011 | /* |
3012 | * cpu_buffer->pages just needs to point to the buffer, it | |
3013 | * has no specific buffer page to point to. Lets move it out | |
25985edc | 3014 | * of our way so we don't accidentally swap it. |
3adc54fa SR |
3015 | */ |
3016 | cpu_buffer->pages = reader->list.prev; | |
3017 | ||
77ae365e SR |
3018 | /* The reader page will be pointing to the new head */ |
3019 | rb_set_list_to_head(cpu_buffer, &cpu_buffer->reader_page->list); | |
7a8e76a3 | 3020 | |
66a8cb95 SR |
3021 | /* |
3022 | * We want to make sure we read the overruns after we set up our | |
3023 | * pointers to the next object. The writer side does a | |
3024 | * cmpxchg to cross pages which acts as the mb on the writer | |
3025 | * side. Note, the reader will constantly fail the swap | |
3026 | * while the writer is updating the pointers, so this | |
3027 | * guarantees that the overwrite recorded here is the one we | |
3028 | * want to compare with the last_overrun. | |
3029 | */ | |
3030 | smp_mb(); | |
3031 | overwrite = local_read(&(cpu_buffer->overrun)); | |
3032 | ||
77ae365e SR |
3033 | /* |
3034 | * Here's the tricky part. | |
3035 | * | |
3036 | * We need to move the pointer past the header page. | |
3037 | * But we can only do that if a writer is not currently | |
3038 | * moving it. The page before the header page has the | |
3039 | * flag bit '1' set if it is pointing to the page we want. | |
3040 | * but if the writer is in the process of moving it | |
3041 | * than it will be '2' or already moved '0'. | |
3042 | */ | |
3043 | ||
3044 | ret = rb_head_page_replace(reader, cpu_buffer->reader_page); | |
7a8e76a3 SR |
3045 | |
3046 | /* | |
77ae365e | 3047 | * If we did not convert it, then we must try again. |
7a8e76a3 | 3048 | */ |
77ae365e SR |
3049 | if (!ret) |
3050 | goto spin; | |
7a8e76a3 | 3051 | |
77ae365e SR |
3052 | /* |
3053 | * Yeah! We succeeded in replacing the page. | |
3054 | * | |
3055 | * Now make the new head point back to the reader page. | |
3056 | */ | |
5ded3dc6 | 3057 | rb_list_head(reader->list.next)->prev = &cpu_buffer->reader_page->list; |
77ae365e | 3058 | rb_inc_page(cpu_buffer, &cpu_buffer->head_page); |
d769041f SR |
3059 | |
3060 | /* Finally update the reader page to the new head */ | |
3061 | cpu_buffer->reader_page = reader; | |
3062 | rb_reset_reader_page(cpu_buffer); | |
3063 | ||
66a8cb95 SR |
3064 | if (overwrite != cpu_buffer->last_overrun) { |
3065 | cpu_buffer->lost_events = overwrite - cpu_buffer->last_overrun; | |
3066 | cpu_buffer->last_overrun = overwrite; | |
3067 | } | |
3068 | ||
d769041f SR |
3069 | goto again; |
3070 | ||
3071 | out: | |
0199c4e6 | 3072 | arch_spin_unlock(&cpu_buffer->lock); |
3e03fb7f | 3073 | local_irq_restore(flags); |
d769041f SR |
3074 | |
3075 | return reader; | |
3076 | } | |
3077 | ||
3078 | static void rb_advance_reader(struct ring_buffer_per_cpu *cpu_buffer) | |
3079 | { | |
3080 | struct ring_buffer_event *event; | |
3081 | struct buffer_page *reader; | |
3082 | unsigned length; | |
3083 | ||
3084 | reader = rb_get_reader_page(cpu_buffer); | |
7a8e76a3 | 3085 | |
d769041f | 3086 | /* This function should not be called when buffer is empty */ |
3e89c7bb SR |
3087 | if (RB_WARN_ON(cpu_buffer, !reader)) |
3088 | return; | |
7a8e76a3 | 3089 | |
d769041f SR |
3090 | event = rb_reader_event(cpu_buffer); |
3091 | ||
a1863c21 | 3092 | if (event->type_len <= RINGBUF_TYPE_DATA_TYPE_LEN_MAX) |
e4906eff | 3093 | cpu_buffer->read++; |
d769041f SR |
3094 | |
3095 | rb_update_read_stamp(cpu_buffer, event); | |
3096 | ||
3097 | length = rb_event_length(event); | |
6f807acd | 3098 | cpu_buffer->reader_page->read += length; |
7a8e76a3 SR |
3099 | } |
3100 | ||
3101 | static void rb_advance_iter(struct ring_buffer_iter *iter) | |
3102 | { | |
7a8e76a3 SR |
3103 | struct ring_buffer_per_cpu *cpu_buffer; |
3104 | struct ring_buffer_event *event; | |
3105 | unsigned length; | |
3106 | ||
3107 | cpu_buffer = iter->cpu_buffer; | |
7a8e76a3 SR |
3108 | |
3109 | /* | |
3110 | * Check if we are at the end of the buffer. | |
3111 | */ | |
bf41a158 | 3112 | if (iter->head >= rb_page_size(iter->head_page)) { |
ea05b57c SR |
3113 | /* discarded commits can make the page empty */ |
3114 | if (iter->head_page == cpu_buffer->commit_page) | |
3e89c7bb | 3115 | return; |
d769041f | 3116 | rb_inc_iter(iter); |
7a8e76a3 SR |
3117 | return; |
3118 | } | |
3119 | ||
3120 | event = rb_iter_head_event(iter); | |
3121 | ||
3122 | length = rb_event_length(event); | |
3123 | ||
3124 | /* | |
3125 | * This should not be called to advance the header if we are | |
3126 | * at the tail of the buffer. | |
3127 | */ | |
3e89c7bb | 3128 | if (RB_WARN_ON(cpu_buffer, |
f536aafc | 3129 | (iter->head_page == cpu_buffer->commit_page) && |
3e89c7bb SR |
3130 | (iter->head + length > rb_commit_index(cpu_buffer)))) |
3131 | return; | |
7a8e76a3 SR |
3132 | |
3133 | rb_update_iter_read_stamp(iter, event); | |
3134 | ||
3135 | iter->head += length; | |
3136 | ||
3137 | /* check for end of page padding */ | |
bf41a158 SR |
3138 | if ((iter->head >= rb_page_size(iter->head_page)) && |
3139 | (iter->head_page != cpu_buffer->commit_page)) | |
7a8e76a3 SR |
3140 | rb_advance_iter(iter); |
3141 | } | |
3142 | ||
66a8cb95 SR |
3143 | static int rb_lost_events(struct ring_buffer_per_cpu *cpu_buffer) |
3144 | { | |
3145 | return cpu_buffer->lost_events; | |
3146 | } | |
3147 | ||
f83c9d0f | 3148 | static struct ring_buffer_event * |
66a8cb95 SR |
3149 | rb_buffer_peek(struct ring_buffer_per_cpu *cpu_buffer, u64 *ts, |
3150 | unsigned long *lost_events) | |
7a8e76a3 | 3151 | { |
7a8e76a3 | 3152 | struct ring_buffer_event *event; |
d769041f | 3153 | struct buffer_page *reader; |
818e3dd3 | 3154 | int nr_loops = 0; |
7a8e76a3 | 3155 | |
7a8e76a3 | 3156 | again: |
818e3dd3 | 3157 | /* |
69d1b839 SR |
3158 | * We repeat when a time extend is encountered. |
3159 | * Since the time extend is always attached to a data event, | |
3160 | * we should never loop more than once. | |
3161 | * (We never hit the following condition more than twice). | |
818e3dd3 | 3162 | */ |
69d1b839 | 3163 | if (RB_WARN_ON(cpu_buffer, ++nr_loops > 2)) |
818e3dd3 | 3164 | return NULL; |
818e3dd3 | 3165 | |
d769041f SR |
3166 | reader = rb_get_reader_page(cpu_buffer); |
3167 | if (!reader) | |
7a8e76a3 SR |
3168 | return NULL; |
3169 | ||
d769041f | 3170 | event = rb_reader_event(cpu_buffer); |
7a8e76a3 | 3171 | |
334d4169 | 3172 | switch (event->type_len) { |
7a8e76a3 | 3173 | case RINGBUF_TYPE_PADDING: |
2d622719 TZ |
3174 | if (rb_null_event(event)) |
3175 | RB_WARN_ON(cpu_buffer, 1); | |
3176 | /* | |
3177 | * Because the writer could be discarding every | |
3178 | * event it creates (which would probably be bad) | |
3179 | * if we were to go back to "again" then we may never | |
3180 | * catch up, and will trigger the warn on, or lock | |
3181 | * the box. Return the padding, and we will release | |
3182 | * the current locks, and try again. | |
3183 | */ | |
2d622719 | 3184 | return event; |
7a8e76a3 SR |
3185 | |
3186 | case RINGBUF_TYPE_TIME_EXTEND: | |
3187 | /* Internal data, OK to advance */ | |
d769041f | 3188 | rb_advance_reader(cpu_buffer); |
7a8e76a3 SR |
3189 | goto again; |
3190 | ||
3191 | case RINGBUF_TYPE_TIME_STAMP: | |
3192 | /* FIXME: not implemented */ | |
d769041f | 3193 | rb_advance_reader(cpu_buffer); |
7a8e76a3 SR |
3194 | goto again; |
3195 | ||
3196 | case RINGBUF_TYPE_DATA: | |
3197 | if (ts) { | |
3198 | *ts = cpu_buffer->read_stamp + event->time_delta; | |
d8eeb2d3 | 3199 | ring_buffer_normalize_time_stamp(cpu_buffer->buffer, |
37886f6a | 3200 | cpu_buffer->cpu, ts); |
7a8e76a3 | 3201 | } |
66a8cb95 SR |
3202 | if (lost_events) |
3203 | *lost_events = rb_lost_events(cpu_buffer); | |
7a8e76a3 SR |
3204 | return event; |
3205 | ||
3206 | default: | |
3207 | BUG(); | |
3208 | } | |
3209 | ||
3210 | return NULL; | |
3211 | } | |
c4f50183 | 3212 | EXPORT_SYMBOL_GPL(ring_buffer_peek); |
7a8e76a3 | 3213 | |
f83c9d0f SR |
3214 | static struct ring_buffer_event * |
3215 | rb_iter_peek(struct ring_buffer_iter *iter, u64 *ts) | |
7a8e76a3 SR |
3216 | { |
3217 | struct ring_buffer *buffer; | |
3218 | struct ring_buffer_per_cpu *cpu_buffer; | |
3219 | struct ring_buffer_event *event; | |
818e3dd3 | 3220 | int nr_loops = 0; |
7a8e76a3 | 3221 | |
7a8e76a3 SR |
3222 | cpu_buffer = iter->cpu_buffer; |
3223 | buffer = cpu_buffer->buffer; | |
3224 | ||
492a74f4 SR |
3225 | /* |
3226 | * Check if someone performed a consuming read to | |
3227 | * the buffer. A consuming read invalidates the iterator | |
3228 | * and we need to reset the iterator in this case. | |
3229 | */ | |
3230 | if (unlikely(iter->cache_read != cpu_buffer->read || | |
3231 | iter->cache_reader_page != cpu_buffer->reader_page)) | |
3232 | rb_iter_reset(iter); | |
3233 | ||
7a8e76a3 | 3234 | again: |
3c05d748 SR |
3235 | if (ring_buffer_iter_empty(iter)) |
3236 | return NULL; | |
3237 | ||
818e3dd3 | 3238 | /* |
69d1b839 SR |
3239 | * We repeat when a time extend is encountered. |
3240 | * Since the time extend is always attached to a data event, | |
3241 | * we should never loop more than once. | |
3242 | * (We never hit the following condition more than twice). | |
818e3dd3 | 3243 | */ |
69d1b839 | 3244 | if (RB_WARN_ON(cpu_buffer, ++nr_loops > 2)) |
818e3dd3 | 3245 | return NULL; |
818e3dd3 | 3246 | |
7a8e76a3 SR |
3247 | if (rb_per_cpu_empty(cpu_buffer)) |
3248 | return NULL; | |
3249 | ||
3c05d748 SR |
3250 | if (iter->head >= local_read(&iter->head_page->page->commit)) { |
3251 | rb_inc_iter(iter); | |
3252 | goto again; | |
3253 | } | |
3254 | ||
7a8e76a3 SR |
3255 | event = rb_iter_head_event(iter); |
3256 | ||
334d4169 | 3257 | switch (event->type_len) { |
7a8e76a3 | 3258 | case RINGBUF_TYPE_PADDING: |
2d622719 TZ |
3259 | if (rb_null_event(event)) { |
3260 | rb_inc_iter(iter); | |
3261 | goto again; | |
3262 | } | |
3263 | rb_advance_iter(iter); | |
3264 | return event; | |
7a8e76a3 SR |
3265 | |
3266 | case RINGBUF_TYPE_TIME_EXTEND: | |
3267 | /* Internal data, OK to advance */ | |
3268 | rb_advance_iter(iter); | |
3269 | goto again; | |
3270 | ||
3271 | case RINGBUF_TYPE_TIME_STAMP: | |
3272 | /* FIXME: not implemented */ | |
3273 | rb_advance_iter(iter); | |
3274 | goto again; | |
3275 | ||
3276 | case RINGBUF_TYPE_DATA: | |
3277 | if (ts) { | |
3278 | *ts = iter->read_stamp + event->time_delta; | |
37886f6a SR |
3279 | ring_buffer_normalize_time_stamp(buffer, |
3280 | cpu_buffer->cpu, ts); | |
7a8e76a3 SR |
3281 | } |
3282 | return event; | |
3283 | ||
3284 | default: | |
3285 | BUG(); | |
3286 | } | |
3287 | ||
3288 | return NULL; | |
3289 | } | |
c4f50183 | 3290 | EXPORT_SYMBOL_GPL(ring_buffer_iter_peek); |
7a8e76a3 | 3291 | |
8d707e8e SR |
3292 | static inline int rb_ok_to_lock(void) |
3293 | { | |
3294 | /* | |
3295 | * If an NMI die dumps out the content of the ring buffer | |
3296 | * do not grab locks. We also permanently disable the ring | |
3297 | * buffer too. A one time deal is all you get from reading | |
3298 | * the ring buffer from an NMI. | |
3299 | */ | |
464e85eb | 3300 | if (likely(!in_nmi())) |
8d707e8e SR |
3301 | return 1; |
3302 | ||
3303 | tracing_off_permanent(); | |
3304 | return 0; | |
3305 | } | |
3306 | ||
f83c9d0f SR |
3307 | /** |
3308 | * ring_buffer_peek - peek at the next event to be read | |
3309 | * @buffer: The ring buffer to read | |
3310 | * @cpu: The cpu to peak at | |
3311 | * @ts: The timestamp counter of this event. | |
66a8cb95 | 3312 | * @lost_events: a variable to store if events were lost (may be NULL) |
f83c9d0f SR |
3313 | * |
3314 | * This will return the event that will be read next, but does | |
3315 | * not consume the data. | |
3316 | */ | |
3317 | struct ring_buffer_event * | |
66a8cb95 SR |
3318 | ring_buffer_peek(struct ring_buffer *buffer, int cpu, u64 *ts, |
3319 | unsigned long *lost_events) | |
f83c9d0f SR |
3320 | { |
3321 | struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu]; | |
8aabee57 | 3322 | struct ring_buffer_event *event; |
f83c9d0f | 3323 | unsigned long flags; |
8d707e8e | 3324 | int dolock; |
f83c9d0f | 3325 | |
554f786e | 3326 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 3327 | return NULL; |
554f786e | 3328 | |
8d707e8e | 3329 | dolock = rb_ok_to_lock(); |
2d622719 | 3330 | again: |
8d707e8e SR |
3331 | local_irq_save(flags); |
3332 | if (dolock) | |
5389f6fa | 3333 | raw_spin_lock(&cpu_buffer->reader_lock); |
66a8cb95 | 3334 | event = rb_buffer_peek(cpu_buffer, ts, lost_events); |
469535a5 RR |
3335 | if (event && event->type_len == RINGBUF_TYPE_PADDING) |
3336 | rb_advance_reader(cpu_buffer); | |
8d707e8e | 3337 | if (dolock) |
5389f6fa | 3338 | raw_spin_unlock(&cpu_buffer->reader_lock); |
8d707e8e | 3339 | local_irq_restore(flags); |
f83c9d0f | 3340 | |
1b959e18 | 3341 | if (event && event->type_len == RINGBUF_TYPE_PADDING) |
2d622719 | 3342 | goto again; |
2d622719 | 3343 | |
f83c9d0f SR |
3344 | return event; |
3345 | } | |
3346 | ||
3347 | /** | |
3348 | * ring_buffer_iter_peek - peek at the next event to be read | |
3349 | * @iter: The ring buffer iterator | |
3350 | * @ts: The timestamp counter of this event. | |
3351 | * | |
3352 | * This will return the event that will be read next, but does | |
3353 | * not increment the iterator. | |
3354 | */ | |
3355 | struct ring_buffer_event * | |
3356 | ring_buffer_iter_peek(struct ring_buffer_iter *iter, u64 *ts) | |
3357 | { | |
3358 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
3359 | struct ring_buffer_event *event; | |
3360 | unsigned long flags; | |
3361 | ||
2d622719 | 3362 | again: |
5389f6fa | 3363 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
f83c9d0f | 3364 | event = rb_iter_peek(iter, ts); |
5389f6fa | 3365 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
f83c9d0f | 3366 | |
1b959e18 | 3367 | if (event && event->type_len == RINGBUF_TYPE_PADDING) |
2d622719 | 3368 | goto again; |
2d622719 | 3369 | |
f83c9d0f SR |
3370 | return event; |
3371 | } | |
3372 | ||
7a8e76a3 SR |
3373 | /** |
3374 | * ring_buffer_consume - return an event and consume it | |
3375 | * @buffer: The ring buffer to get the next event from | |
66a8cb95 SR |
3376 | * @cpu: the cpu to read the buffer from |
3377 | * @ts: a variable to store the timestamp (may be NULL) | |
3378 | * @lost_events: a variable to store if events were lost (may be NULL) | |
7a8e76a3 SR |
3379 | * |
3380 | * Returns the next event in the ring buffer, and that event is consumed. | |
3381 | * Meaning, that sequential reads will keep returning a different event, | |
3382 | * and eventually empty the ring buffer if the producer is slower. | |
3383 | */ | |
3384 | struct ring_buffer_event * | |
66a8cb95 SR |
3385 | ring_buffer_consume(struct ring_buffer *buffer, int cpu, u64 *ts, |
3386 | unsigned long *lost_events) | |
7a8e76a3 | 3387 | { |
554f786e SR |
3388 | struct ring_buffer_per_cpu *cpu_buffer; |
3389 | struct ring_buffer_event *event = NULL; | |
f83c9d0f | 3390 | unsigned long flags; |
8d707e8e SR |
3391 | int dolock; |
3392 | ||
3393 | dolock = rb_ok_to_lock(); | |
7a8e76a3 | 3394 | |
2d622719 | 3395 | again: |
554f786e SR |
3396 | /* might be called in atomic */ |
3397 | preempt_disable(); | |
3398 | ||
9e01c1b7 | 3399 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
554f786e | 3400 | goto out; |
7a8e76a3 | 3401 | |
554f786e | 3402 | cpu_buffer = buffer->buffers[cpu]; |
8d707e8e SR |
3403 | local_irq_save(flags); |
3404 | if (dolock) | |
5389f6fa | 3405 | raw_spin_lock(&cpu_buffer->reader_lock); |
f83c9d0f | 3406 | |
66a8cb95 SR |
3407 | event = rb_buffer_peek(cpu_buffer, ts, lost_events); |
3408 | if (event) { | |
3409 | cpu_buffer->lost_events = 0; | |
469535a5 | 3410 | rb_advance_reader(cpu_buffer); |
66a8cb95 | 3411 | } |
7a8e76a3 | 3412 | |
8d707e8e | 3413 | if (dolock) |
5389f6fa | 3414 | raw_spin_unlock(&cpu_buffer->reader_lock); |
8d707e8e | 3415 | local_irq_restore(flags); |
f83c9d0f | 3416 | |
554f786e SR |
3417 | out: |
3418 | preempt_enable(); | |
3419 | ||
1b959e18 | 3420 | if (event && event->type_len == RINGBUF_TYPE_PADDING) |
2d622719 | 3421 | goto again; |
2d622719 | 3422 | |
7a8e76a3 SR |
3423 | return event; |
3424 | } | |
c4f50183 | 3425 | EXPORT_SYMBOL_GPL(ring_buffer_consume); |
7a8e76a3 SR |
3426 | |
3427 | /** | |
72c9ddfd | 3428 | * ring_buffer_read_prepare - Prepare for a non consuming read of the buffer |
7a8e76a3 SR |
3429 | * @buffer: The ring buffer to read from |
3430 | * @cpu: The cpu buffer to iterate over | |
3431 | * | |
72c9ddfd DM |
3432 | * This performs the initial preparations necessary to iterate |
3433 | * through the buffer. Memory is allocated, buffer recording | |
3434 | * is disabled, and the iterator pointer is returned to the caller. | |
7a8e76a3 | 3435 | * |
72c9ddfd DM |
3436 | * Disabling buffer recordng prevents the reading from being |
3437 | * corrupted. This is not a consuming read, so a producer is not | |
3438 | * expected. | |
3439 | * | |
3440 | * After a sequence of ring_buffer_read_prepare calls, the user is | |
3441 | * expected to make at least one call to ring_buffer_prepare_sync. | |
3442 | * Afterwards, ring_buffer_read_start is invoked to get things going | |
3443 | * for real. | |
3444 | * | |
3445 | * This overall must be paired with ring_buffer_finish. | |
7a8e76a3 SR |
3446 | */ |
3447 | struct ring_buffer_iter * | |
72c9ddfd | 3448 | ring_buffer_read_prepare(struct ring_buffer *buffer, int cpu) |
7a8e76a3 SR |
3449 | { |
3450 | struct ring_buffer_per_cpu *cpu_buffer; | |
8aabee57 | 3451 | struct ring_buffer_iter *iter; |
7a8e76a3 | 3452 | |
9e01c1b7 | 3453 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 3454 | return NULL; |
7a8e76a3 SR |
3455 | |
3456 | iter = kmalloc(sizeof(*iter), GFP_KERNEL); | |
3457 | if (!iter) | |
8aabee57 | 3458 | return NULL; |
7a8e76a3 SR |
3459 | |
3460 | cpu_buffer = buffer->buffers[cpu]; | |
3461 | ||
3462 | iter->cpu_buffer = cpu_buffer; | |
3463 | ||
3464 | atomic_inc(&cpu_buffer->record_disabled); | |
72c9ddfd DM |
3465 | |
3466 | return iter; | |
3467 | } | |
3468 | EXPORT_SYMBOL_GPL(ring_buffer_read_prepare); | |
3469 | ||
3470 | /** | |
3471 | * ring_buffer_read_prepare_sync - Synchronize a set of prepare calls | |
3472 | * | |
3473 | * All previously invoked ring_buffer_read_prepare calls to prepare | |
3474 | * iterators will be synchronized. Afterwards, read_buffer_read_start | |
3475 | * calls on those iterators are allowed. | |
3476 | */ | |
3477 | void | |
3478 | ring_buffer_read_prepare_sync(void) | |
3479 | { | |
7a8e76a3 | 3480 | synchronize_sched(); |
72c9ddfd DM |
3481 | } |
3482 | EXPORT_SYMBOL_GPL(ring_buffer_read_prepare_sync); | |
3483 | ||
3484 | /** | |
3485 | * ring_buffer_read_start - start a non consuming read of the buffer | |
3486 | * @iter: The iterator returned by ring_buffer_read_prepare | |
3487 | * | |
3488 | * This finalizes the startup of an iteration through the buffer. | |
3489 | * The iterator comes from a call to ring_buffer_read_prepare and | |
3490 | * an intervening ring_buffer_read_prepare_sync must have been | |
3491 | * performed. | |
3492 | * | |
3493 | * Must be paired with ring_buffer_finish. | |
3494 | */ | |
3495 | void | |
3496 | ring_buffer_read_start(struct ring_buffer_iter *iter) | |
3497 | { | |
3498 | struct ring_buffer_per_cpu *cpu_buffer; | |
3499 | unsigned long flags; | |
3500 | ||
3501 | if (!iter) | |
3502 | return; | |
3503 | ||
3504 | cpu_buffer = iter->cpu_buffer; | |
7a8e76a3 | 3505 | |
5389f6fa | 3506 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
0199c4e6 | 3507 | arch_spin_lock(&cpu_buffer->lock); |
642edba5 | 3508 | rb_iter_reset(iter); |
0199c4e6 | 3509 | arch_spin_unlock(&cpu_buffer->lock); |
5389f6fa | 3510 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
7a8e76a3 | 3511 | } |
c4f50183 | 3512 | EXPORT_SYMBOL_GPL(ring_buffer_read_start); |
7a8e76a3 SR |
3513 | |
3514 | /** | |
3515 | * ring_buffer_finish - finish reading the iterator of the buffer | |
3516 | * @iter: The iterator retrieved by ring_buffer_start | |
3517 | * | |
3518 | * This re-enables the recording to the buffer, and frees the | |
3519 | * iterator. | |
3520 | */ | |
3521 | void | |
3522 | ring_buffer_read_finish(struct ring_buffer_iter *iter) | |
3523 | { | |
3524 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; | |
3525 | ||
3526 | atomic_dec(&cpu_buffer->record_disabled); | |
3527 | kfree(iter); | |
3528 | } | |
c4f50183 | 3529 | EXPORT_SYMBOL_GPL(ring_buffer_read_finish); |
7a8e76a3 SR |
3530 | |
3531 | /** | |
3532 | * ring_buffer_read - read the next item in the ring buffer by the iterator | |
3533 | * @iter: The ring buffer iterator | |
3534 | * @ts: The time stamp of the event read. | |
3535 | * | |
3536 | * This reads the next event in the ring buffer and increments the iterator. | |
3537 | */ | |
3538 | struct ring_buffer_event * | |
3539 | ring_buffer_read(struct ring_buffer_iter *iter, u64 *ts) | |
3540 | { | |
3541 | struct ring_buffer_event *event; | |
f83c9d0f SR |
3542 | struct ring_buffer_per_cpu *cpu_buffer = iter->cpu_buffer; |
3543 | unsigned long flags; | |
7a8e76a3 | 3544 | |
5389f6fa | 3545 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
7e9391cf | 3546 | again: |
f83c9d0f | 3547 | event = rb_iter_peek(iter, ts); |
7a8e76a3 | 3548 | if (!event) |
f83c9d0f | 3549 | goto out; |
7a8e76a3 | 3550 | |
7e9391cf SR |
3551 | if (event->type_len == RINGBUF_TYPE_PADDING) |
3552 | goto again; | |
3553 | ||
7a8e76a3 | 3554 | rb_advance_iter(iter); |
f83c9d0f | 3555 | out: |
5389f6fa | 3556 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
7a8e76a3 SR |
3557 | |
3558 | return event; | |
3559 | } | |
c4f50183 | 3560 | EXPORT_SYMBOL_GPL(ring_buffer_read); |
7a8e76a3 SR |
3561 | |
3562 | /** | |
3563 | * ring_buffer_size - return the size of the ring buffer (in bytes) | |
3564 | * @buffer: The ring buffer. | |
3565 | */ | |
3566 | unsigned long ring_buffer_size(struct ring_buffer *buffer) | |
3567 | { | |
3568 | return BUF_PAGE_SIZE * buffer->pages; | |
3569 | } | |
c4f50183 | 3570 | EXPORT_SYMBOL_GPL(ring_buffer_size); |
7a8e76a3 SR |
3571 | |
3572 | static void | |
3573 | rb_reset_cpu(struct ring_buffer_per_cpu *cpu_buffer) | |
3574 | { | |
77ae365e SR |
3575 | rb_head_page_deactivate(cpu_buffer); |
3576 | ||
7a8e76a3 | 3577 | cpu_buffer->head_page |
3adc54fa | 3578 | = list_entry(cpu_buffer->pages, struct buffer_page, list); |
bf41a158 | 3579 | local_set(&cpu_buffer->head_page->write, 0); |
778c55d4 | 3580 | local_set(&cpu_buffer->head_page->entries, 0); |
abc9b56d | 3581 | local_set(&cpu_buffer->head_page->page->commit, 0); |
d769041f | 3582 | |
6f807acd | 3583 | cpu_buffer->head_page->read = 0; |
bf41a158 SR |
3584 | |
3585 | cpu_buffer->tail_page = cpu_buffer->head_page; | |
3586 | cpu_buffer->commit_page = cpu_buffer->head_page; | |
3587 | ||
3588 | INIT_LIST_HEAD(&cpu_buffer->reader_page->list); | |
3589 | local_set(&cpu_buffer->reader_page->write, 0); | |
778c55d4 | 3590 | local_set(&cpu_buffer->reader_page->entries, 0); |
abc9b56d | 3591 | local_set(&cpu_buffer->reader_page->page->commit, 0); |
6f807acd | 3592 | cpu_buffer->reader_page->read = 0; |
7a8e76a3 | 3593 | |
77ae365e | 3594 | local_set(&cpu_buffer->commit_overrun, 0); |
c64e148a | 3595 | local_set(&cpu_buffer->entries_bytes, 0); |
77ae365e | 3596 | local_set(&cpu_buffer->overrun, 0); |
e4906eff | 3597 | local_set(&cpu_buffer->entries, 0); |
fa743953 SR |
3598 | local_set(&cpu_buffer->committing, 0); |
3599 | local_set(&cpu_buffer->commits, 0); | |
77ae365e | 3600 | cpu_buffer->read = 0; |
c64e148a | 3601 | cpu_buffer->read_bytes = 0; |
69507c06 SR |
3602 | |
3603 | cpu_buffer->write_stamp = 0; | |
3604 | cpu_buffer->read_stamp = 0; | |
77ae365e | 3605 | |
66a8cb95 SR |
3606 | cpu_buffer->lost_events = 0; |
3607 | cpu_buffer->last_overrun = 0; | |
3608 | ||
77ae365e | 3609 | rb_head_page_activate(cpu_buffer); |
7a8e76a3 SR |
3610 | } |
3611 | ||
3612 | /** | |
3613 | * ring_buffer_reset_cpu - reset a ring buffer per CPU buffer | |
3614 | * @buffer: The ring buffer to reset a per cpu buffer of | |
3615 | * @cpu: The CPU buffer to be reset | |
3616 | */ | |
3617 | void ring_buffer_reset_cpu(struct ring_buffer *buffer, int cpu) | |
3618 | { | |
3619 | struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu]; | |
3620 | unsigned long flags; | |
3621 | ||
9e01c1b7 | 3622 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 3623 | return; |
7a8e76a3 | 3624 | |
41ede23e SR |
3625 | atomic_inc(&cpu_buffer->record_disabled); |
3626 | ||
5389f6fa | 3627 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
f83c9d0f | 3628 | |
41b6a95d SR |
3629 | if (RB_WARN_ON(cpu_buffer, local_read(&cpu_buffer->committing))) |
3630 | goto out; | |
3631 | ||
0199c4e6 | 3632 | arch_spin_lock(&cpu_buffer->lock); |
7a8e76a3 SR |
3633 | |
3634 | rb_reset_cpu(cpu_buffer); | |
3635 | ||
0199c4e6 | 3636 | arch_spin_unlock(&cpu_buffer->lock); |
f83c9d0f | 3637 | |
41b6a95d | 3638 | out: |
5389f6fa | 3639 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
41ede23e SR |
3640 | |
3641 | atomic_dec(&cpu_buffer->record_disabled); | |
7a8e76a3 | 3642 | } |
c4f50183 | 3643 | EXPORT_SYMBOL_GPL(ring_buffer_reset_cpu); |
7a8e76a3 SR |
3644 | |
3645 | /** | |
3646 | * ring_buffer_reset - reset a ring buffer | |
3647 | * @buffer: The ring buffer to reset all cpu buffers | |
3648 | */ | |
3649 | void ring_buffer_reset(struct ring_buffer *buffer) | |
3650 | { | |
7a8e76a3 SR |
3651 | int cpu; |
3652 | ||
7a8e76a3 | 3653 | for_each_buffer_cpu(buffer, cpu) |
d769041f | 3654 | ring_buffer_reset_cpu(buffer, cpu); |
7a8e76a3 | 3655 | } |
c4f50183 | 3656 | EXPORT_SYMBOL_GPL(ring_buffer_reset); |
7a8e76a3 SR |
3657 | |
3658 | /** | |
3659 | * rind_buffer_empty - is the ring buffer empty? | |
3660 | * @buffer: The ring buffer to test | |
3661 | */ | |
3662 | int ring_buffer_empty(struct ring_buffer *buffer) | |
3663 | { | |
3664 | struct ring_buffer_per_cpu *cpu_buffer; | |
d4788207 | 3665 | unsigned long flags; |
8d707e8e | 3666 | int dolock; |
7a8e76a3 | 3667 | int cpu; |
d4788207 | 3668 | int ret; |
7a8e76a3 | 3669 | |
8d707e8e | 3670 | dolock = rb_ok_to_lock(); |
7a8e76a3 SR |
3671 | |
3672 | /* yes this is racy, but if you don't like the race, lock the buffer */ | |
3673 | for_each_buffer_cpu(buffer, cpu) { | |
3674 | cpu_buffer = buffer->buffers[cpu]; | |
8d707e8e SR |
3675 | local_irq_save(flags); |
3676 | if (dolock) | |
5389f6fa | 3677 | raw_spin_lock(&cpu_buffer->reader_lock); |
d4788207 | 3678 | ret = rb_per_cpu_empty(cpu_buffer); |
8d707e8e | 3679 | if (dolock) |
5389f6fa | 3680 | raw_spin_unlock(&cpu_buffer->reader_lock); |
8d707e8e SR |
3681 | local_irq_restore(flags); |
3682 | ||
d4788207 | 3683 | if (!ret) |
7a8e76a3 SR |
3684 | return 0; |
3685 | } | |
554f786e | 3686 | |
7a8e76a3 SR |
3687 | return 1; |
3688 | } | |
c4f50183 | 3689 | EXPORT_SYMBOL_GPL(ring_buffer_empty); |
7a8e76a3 SR |
3690 | |
3691 | /** | |
3692 | * ring_buffer_empty_cpu - is a cpu buffer of a ring buffer empty? | |
3693 | * @buffer: The ring buffer | |
3694 | * @cpu: The CPU buffer to test | |
3695 | */ | |
3696 | int ring_buffer_empty_cpu(struct ring_buffer *buffer, int cpu) | |
3697 | { | |
3698 | struct ring_buffer_per_cpu *cpu_buffer; | |
d4788207 | 3699 | unsigned long flags; |
8d707e8e | 3700 | int dolock; |
8aabee57 | 3701 | int ret; |
7a8e76a3 | 3702 | |
9e01c1b7 | 3703 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
8aabee57 | 3704 | return 1; |
7a8e76a3 | 3705 | |
8d707e8e SR |
3706 | dolock = rb_ok_to_lock(); |
3707 | ||
7a8e76a3 | 3708 | cpu_buffer = buffer->buffers[cpu]; |
8d707e8e SR |
3709 | local_irq_save(flags); |
3710 | if (dolock) | |
5389f6fa | 3711 | raw_spin_lock(&cpu_buffer->reader_lock); |
554f786e | 3712 | ret = rb_per_cpu_empty(cpu_buffer); |
8d707e8e | 3713 | if (dolock) |
5389f6fa | 3714 | raw_spin_unlock(&cpu_buffer->reader_lock); |
8d707e8e | 3715 | local_irq_restore(flags); |
554f786e SR |
3716 | |
3717 | return ret; | |
7a8e76a3 | 3718 | } |
c4f50183 | 3719 | EXPORT_SYMBOL_GPL(ring_buffer_empty_cpu); |
7a8e76a3 | 3720 | |
85bac32c | 3721 | #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP |
7a8e76a3 SR |
3722 | /** |
3723 | * ring_buffer_swap_cpu - swap a CPU buffer between two ring buffers | |
3724 | * @buffer_a: One buffer to swap with | |
3725 | * @buffer_b: The other buffer to swap with | |
3726 | * | |
3727 | * This function is useful for tracers that want to take a "snapshot" | |
3728 | * of a CPU buffer and has another back up buffer lying around. | |
3729 | * it is expected that the tracer handles the cpu buffer not being | |
3730 | * used at the moment. | |
3731 | */ | |
3732 | int ring_buffer_swap_cpu(struct ring_buffer *buffer_a, | |
3733 | struct ring_buffer *buffer_b, int cpu) | |
3734 | { | |
3735 | struct ring_buffer_per_cpu *cpu_buffer_a; | |
3736 | struct ring_buffer_per_cpu *cpu_buffer_b; | |
554f786e SR |
3737 | int ret = -EINVAL; |
3738 | ||
9e01c1b7 RR |
3739 | if (!cpumask_test_cpu(cpu, buffer_a->cpumask) || |
3740 | !cpumask_test_cpu(cpu, buffer_b->cpumask)) | |
554f786e | 3741 | goto out; |
7a8e76a3 SR |
3742 | |
3743 | /* At least make sure the two buffers are somewhat the same */ | |
6d102bc6 | 3744 | if (buffer_a->pages != buffer_b->pages) |
554f786e SR |
3745 | goto out; |
3746 | ||
3747 | ret = -EAGAIN; | |
7a8e76a3 | 3748 | |
97b17efe | 3749 | if (ring_buffer_flags != RB_BUFFERS_ON) |
554f786e | 3750 | goto out; |
97b17efe SR |
3751 | |
3752 | if (atomic_read(&buffer_a->record_disabled)) | |
554f786e | 3753 | goto out; |
97b17efe SR |
3754 | |
3755 | if (atomic_read(&buffer_b->record_disabled)) | |
554f786e | 3756 | goto out; |
97b17efe | 3757 | |
7a8e76a3 SR |
3758 | cpu_buffer_a = buffer_a->buffers[cpu]; |
3759 | cpu_buffer_b = buffer_b->buffers[cpu]; | |
3760 | ||
97b17efe | 3761 | if (atomic_read(&cpu_buffer_a->record_disabled)) |
554f786e | 3762 | goto out; |
97b17efe SR |
3763 | |
3764 | if (atomic_read(&cpu_buffer_b->record_disabled)) | |
554f786e | 3765 | goto out; |
97b17efe | 3766 | |
7a8e76a3 SR |
3767 | /* |
3768 | * We can't do a synchronize_sched here because this | |
3769 | * function can be called in atomic context. | |
3770 | * Normally this will be called from the same CPU as cpu. | |
3771 | * If not it's up to the caller to protect this. | |
3772 | */ | |
3773 | atomic_inc(&cpu_buffer_a->record_disabled); | |
3774 | atomic_inc(&cpu_buffer_b->record_disabled); | |
3775 | ||
98277991 SR |
3776 | ret = -EBUSY; |
3777 | if (local_read(&cpu_buffer_a->committing)) | |
3778 | goto out_dec; | |
3779 | if (local_read(&cpu_buffer_b->committing)) | |
3780 | goto out_dec; | |
3781 | ||
7a8e76a3 SR |
3782 | buffer_a->buffers[cpu] = cpu_buffer_b; |
3783 | buffer_b->buffers[cpu] = cpu_buffer_a; | |
3784 | ||
3785 | cpu_buffer_b->buffer = buffer_a; | |
3786 | cpu_buffer_a->buffer = buffer_b; | |
3787 | ||
98277991 SR |
3788 | ret = 0; |
3789 | ||
3790 | out_dec: | |
7a8e76a3 SR |
3791 | atomic_dec(&cpu_buffer_a->record_disabled); |
3792 | atomic_dec(&cpu_buffer_b->record_disabled); | |
554f786e | 3793 | out: |
554f786e | 3794 | return ret; |
7a8e76a3 | 3795 | } |
c4f50183 | 3796 | EXPORT_SYMBOL_GPL(ring_buffer_swap_cpu); |
85bac32c | 3797 | #endif /* CONFIG_RING_BUFFER_ALLOW_SWAP */ |
7a8e76a3 | 3798 | |
8789a9e7 SR |
3799 | /** |
3800 | * ring_buffer_alloc_read_page - allocate a page to read from buffer | |
3801 | * @buffer: the buffer to allocate for. | |
3802 | * | |
3803 | * This function is used in conjunction with ring_buffer_read_page. | |
3804 | * When reading a full page from the ring buffer, these functions | |
3805 | * can be used to speed up the process. The calling function should | |
3806 | * allocate a few pages first with this function. Then when it | |
3807 | * needs to get pages from the ring buffer, it passes the result | |
3808 | * of this function into ring_buffer_read_page, which will swap | |
3809 | * the page that was allocated, with the read page of the buffer. | |
3810 | * | |
3811 | * Returns: | |
3812 | * The page allocated, or NULL on error. | |
3813 | */ | |
7ea59064 | 3814 | void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu) |
8789a9e7 | 3815 | { |
044fa782 | 3816 | struct buffer_data_page *bpage; |
7ea59064 | 3817 | struct page *page; |
8789a9e7 | 3818 | |
d7ec4bfe VN |
3819 | page = alloc_pages_node(cpu_to_node(cpu), |
3820 | GFP_KERNEL | __GFP_NORETRY, 0); | |
7ea59064 | 3821 | if (!page) |
8789a9e7 SR |
3822 | return NULL; |
3823 | ||
7ea59064 | 3824 | bpage = page_address(page); |
8789a9e7 | 3825 | |
ef7a4a16 SR |
3826 | rb_init_page(bpage); |
3827 | ||
044fa782 | 3828 | return bpage; |
8789a9e7 | 3829 | } |
d6ce96da | 3830 | EXPORT_SYMBOL_GPL(ring_buffer_alloc_read_page); |
8789a9e7 SR |
3831 | |
3832 | /** | |
3833 | * ring_buffer_free_read_page - free an allocated read page | |
3834 | * @buffer: the buffer the page was allocate for | |
3835 | * @data: the page to free | |
3836 | * | |
3837 | * Free a page allocated from ring_buffer_alloc_read_page. | |
3838 | */ | |
3839 | void ring_buffer_free_read_page(struct ring_buffer *buffer, void *data) | |
3840 | { | |
3841 | free_page((unsigned long)data); | |
3842 | } | |
d6ce96da | 3843 | EXPORT_SYMBOL_GPL(ring_buffer_free_read_page); |
8789a9e7 SR |
3844 | |
3845 | /** | |
3846 | * ring_buffer_read_page - extract a page from the ring buffer | |
3847 | * @buffer: buffer to extract from | |
3848 | * @data_page: the page to use allocated from ring_buffer_alloc_read_page | |
ef7a4a16 | 3849 | * @len: amount to extract |
8789a9e7 SR |
3850 | * @cpu: the cpu of the buffer to extract |
3851 | * @full: should the extraction only happen when the page is full. | |
3852 | * | |
3853 | * This function will pull out a page from the ring buffer and consume it. | |
3854 | * @data_page must be the address of the variable that was returned | |
3855 | * from ring_buffer_alloc_read_page. This is because the page might be used | |
3856 | * to swap with a page in the ring buffer. | |
3857 | * | |
3858 | * for example: | |
b85fa01e | 3859 | * rpage = ring_buffer_alloc_read_page(buffer); |
8789a9e7 SR |
3860 | * if (!rpage) |
3861 | * return error; | |
ef7a4a16 | 3862 | * ret = ring_buffer_read_page(buffer, &rpage, len, cpu, 0); |
667d2412 LJ |
3863 | * if (ret >= 0) |
3864 | * process_page(rpage, ret); | |
8789a9e7 SR |
3865 | * |
3866 | * When @full is set, the function will not return true unless | |
3867 | * the writer is off the reader page. | |
3868 | * | |
3869 | * Note: it is up to the calling functions to handle sleeps and wakeups. | |
3870 | * The ring buffer can be used anywhere in the kernel and can not | |
3871 | * blindly call wake_up. The layer that uses the ring buffer must be | |
3872 | * responsible for that. | |
3873 | * | |
3874 | * Returns: | |
667d2412 LJ |
3875 | * >=0 if data has been transferred, returns the offset of consumed data. |
3876 | * <0 if no data has been transferred. | |
8789a9e7 SR |
3877 | */ |
3878 | int ring_buffer_read_page(struct ring_buffer *buffer, | |
ef7a4a16 | 3879 | void **data_page, size_t len, int cpu, int full) |
8789a9e7 SR |
3880 | { |
3881 | struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu]; | |
3882 | struct ring_buffer_event *event; | |
044fa782 | 3883 | struct buffer_data_page *bpage; |
ef7a4a16 | 3884 | struct buffer_page *reader; |
ff0ff84a | 3885 | unsigned long missed_events; |
8789a9e7 | 3886 | unsigned long flags; |
ef7a4a16 | 3887 | unsigned int commit; |
667d2412 | 3888 | unsigned int read; |
4f3640f8 | 3889 | u64 save_timestamp; |
667d2412 | 3890 | int ret = -1; |
8789a9e7 | 3891 | |
554f786e SR |
3892 | if (!cpumask_test_cpu(cpu, buffer->cpumask)) |
3893 | goto out; | |
3894 | ||
474d32b6 SR |
3895 | /* |
3896 | * If len is not big enough to hold the page header, then | |
3897 | * we can not copy anything. | |
3898 | */ | |
3899 | if (len <= BUF_PAGE_HDR_SIZE) | |
554f786e | 3900 | goto out; |
474d32b6 SR |
3901 | |
3902 | len -= BUF_PAGE_HDR_SIZE; | |
3903 | ||
8789a9e7 | 3904 | if (!data_page) |
554f786e | 3905 | goto out; |
8789a9e7 | 3906 | |
044fa782 SR |
3907 | bpage = *data_page; |
3908 | if (!bpage) | |
554f786e | 3909 | goto out; |
8789a9e7 | 3910 | |
5389f6fa | 3911 | raw_spin_lock_irqsave(&cpu_buffer->reader_lock, flags); |
8789a9e7 | 3912 | |
ef7a4a16 SR |
3913 | reader = rb_get_reader_page(cpu_buffer); |
3914 | if (!reader) | |
554f786e | 3915 | goto out_unlock; |
8789a9e7 | 3916 | |
ef7a4a16 SR |
3917 | event = rb_reader_event(cpu_buffer); |
3918 | ||
3919 | read = reader->read; | |
3920 | commit = rb_page_commit(reader); | |
667d2412 | 3921 | |
66a8cb95 | 3922 | /* Check if any events were dropped */ |
ff0ff84a | 3923 | missed_events = cpu_buffer->lost_events; |
66a8cb95 | 3924 | |
8789a9e7 | 3925 | /* |
474d32b6 SR |
3926 | * If this page has been partially read or |
3927 | * if len is not big enough to read the rest of the page or | |
3928 | * a writer is still on the page, then | |
3929 | * we must copy the data from the page to the buffer. | |
3930 | * Otherwise, we can simply swap the page with the one passed in. | |
8789a9e7 | 3931 | */ |
474d32b6 | 3932 | if (read || (len < (commit - read)) || |
ef7a4a16 | 3933 | cpu_buffer->reader_page == cpu_buffer->commit_page) { |
667d2412 | 3934 | struct buffer_data_page *rpage = cpu_buffer->reader_page->page; |
474d32b6 SR |
3935 | unsigned int rpos = read; |
3936 | unsigned int pos = 0; | |
ef7a4a16 | 3937 | unsigned int size; |
8789a9e7 SR |
3938 | |
3939 | if (full) | |
554f786e | 3940 | goto out_unlock; |
8789a9e7 | 3941 | |
ef7a4a16 SR |
3942 | if (len > (commit - read)) |
3943 | len = (commit - read); | |
3944 | ||
69d1b839 SR |
3945 | /* Always keep the time extend and data together */ |
3946 | size = rb_event_ts_length(event); | |
ef7a4a16 SR |
3947 | |
3948 | if (len < size) | |
554f786e | 3949 | goto out_unlock; |
ef7a4a16 | 3950 | |
4f3640f8 SR |
3951 | /* save the current timestamp, since the user will need it */ |
3952 | save_timestamp = cpu_buffer->read_stamp; | |
3953 | ||
ef7a4a16 SR |
3954 | /* Need to copy one event at a time */ |
3955 | do { | |
e1e35927 DS |
3956 | /* We need the size of one event, because |
3957 | * rb_advance_reader only advances by one event, | |
3958 | * whereas rb_event_ts_length may include the size of | |
3959 | * one or two events. | |
3960 | * We have already ensured there's enough space if this | |
3961 | * is a time extend. */ | |
3962 | size = rb_event_length(event); | |
474d32b6 | 3963 | memcpy(bpage->data + pos, rpage->data + rpos, size); |
ef7a4a16 SR |
3964 | |
3965 | len -= size; | |
3966 | ||
3967 | rb_advance_reader(cpu_buffer); | |
474d32b6 SR |
3968 | rpos = reader->read; |
3969 | pos += size; | |
ef7a4a16 | 3970 | |
18fab912 HY |
3971 | if (rpos >= commit) |
3972 | break; | |
3973 | ||
ef7a4a16 | 3974 | event = rb_reader_event(cpu_buffer); |
69d1b839 SR |
3975 | /* Always keep the time extend and data together */ |
3976 | size = rb_event_ts_length(event); | |
e1e35927 | 3977 | } while (len >= size); |
667d2412 LJ |
3978 | |
3979 | /* update bpage */ | |
ef7a4a16 | 3980 | local_set(&bpage->commit, pos); |
4f3640f8 | 3981 | bpage->time_stamp = save_timestamp; |
ef7a4a16 | 3982 | |
474d32b6 SR |
3983 | /* we copied everything to the beginning */ |
3984 | read = 0; | |
8789a9e7 | 3985 | } else { |
afbab76a | 3986 | /* update the entry counter */ |
77ae365e | 3987 | cpu_buffer->read += rb_page_entries(reader); |
c64e148a | 3988 | cpu_buffer->read_bytes += BUF_PAGE_SIZE; |
afbab76a | 3989 | |
8789a9e7 | 3990 | /* swap the pages */ |
044fa782 | 3991 | rb_init_page(bpage); |
ef7a4a16 SR |
3992 | bpage = reader->page; |
3993 | reader->page = *data_page; | |
3994 | local_set(&reader->write, 0); | |
778c55d4 | 3995 | local_set(&reader->entries, 0); |
ef7a4a16 | 3996 | reader->read = 0; |
044fa782 | 3997 | *data_page = bpage; |
ff0ff84a SR |
3998 | |
3999 | /* | |
4000 | * Use the real_end for the data size, | |
4001 | * This gives us a chance to store the lost events | |
4002 | * on the page. | |
4003 | */ | |
4004 | if (reader->real_end) | |
4005 | local_set(&bpage->commit, reader->real_end); | |
8789a9e7 | 4006 | } |
667d2412 | 4007 | ret = read; |
8789a9e7 | 4008 | |
66a8cb95 | 4009 | cpu_buffer->lost_events = 0; |
2711ca23 SR |
4010 | |
4011 | commit = local_read(&bpage->commit); | |
66a8cb95 SR |
4012 | /* |
4013 | * Set a flag in the commit field if we lost events | |
4014 | */ | |
ff0ff84a | 4015 | if (missed_events) { |
ff0ff84a SR |
4016 | /* If there is room at the end of the page to save the |
4017 | * missed events, then record it there. | |
4018 | */ | |
4019 | if (BUF_PAGE_SIZE - commit >= sizeof(missed_events)) { | |
4020 | memcpy(&bpage->data[commit], &missed_events, | |
4021 | sizeof(missed_events)); | |
4022 | local_add(RB_MISSED_STORED, &bpage->commit); | |
2711ca23 | 4023 | commit += sizeof(missed_events); |
ff0ff84a | 4024 | } |
66a8cb95 | 4025 | local_add(RB_MISSED_EVENTS, &bpage->commit); |
ff0ff84a | 4026 | } |
66a8cb95 | 4027 | |
2711ca23 SR |
4028 | /* |
4029 | * This page may be off to user land. Zero it out here. | |
4030 | */ | |
4031 | if (commit < BUF_PAGE_SIZE) | |
4032 | memset(&bpage->data[commit], 0, BUF_PAGE_SIZE - commit); | |
4033 | ||
554f786e | 4034 | out_unlock: |
5389f6fa | 4035 | raw_spin_unlock_irqrestore(&cpu_buffer->reader_lock, flags); |
8789a9e7 | 4036 | |
554f786e | 4037 | out: |
8789a9e7 SR |
4038 | return ret; |
4039 | } | |
d6ce96da | 4040 | EXPORT_SYMBOL_GPL(ring_buffer_read_page); |
8789a9e7 | 4041 | |
1155de47 | 4042 | #ifdef CONFIG_TRACING |
a3583244 SR |
4043 | static ssize_t |
4044 | rb_simple_read(struct file *filp, char __user *ubuf, | |
4045 | size_t cnt, loff_t *ppos) | |
4046 | { | |
5e39841c | 4047 | unsigned long *p = filp->private_data; |
a3583244 SR |
4048 | char buf[64]; |
4049 | int r; | |
4050 | ||
033601a3 SR |
4051 | if (test_bit(RB_BUFFERS_DISABLED_BIT, p)) |
4052 | r = sprintf(buf, "permanently disabled\n"); | |
4053 | else | |
4054 | r = sprintf(buf, "%d\n", test_bit(RB_BUFFERS_ON_BIT, p)); | |
a3583244 SR |
4055 | |
4056 | return simple_read_from_buffer(ubuf, cnt, ppos, buf, r); | |
4057 | } | |
4058 | ||
4059 | static ssize_t | |
4060 | rb_simple_write(struct file *filp, const char __user *ubuf, | |
4061 | size_t cnt, loff_t *ppos) | |
4062 | { | |
5e39841c | 4063 | unsigned long *p = filp->private_data; |
5e39841c | 4064 | unsigned long val; |
a3583244 SR |
4065 | int ret; |
4066 | ||
22fe9b54 PH |
4067 | ret = kstrtoul_from_user(ubuf, cnt, 10, &val); |
4068 | if (ret) | |
a3583244 SR |
4069 | return ret; |
4070 | ||
033601a3 SR |
4071 | if (val) |
4072 | set_bit(RB_BUFFERS_ON_BIT, p); | |
4073 | else | |
4074 | clear_bit(RB_BUFFERS_ON_BIT, p); | |
a3583244 SR |
4075 | |
4076 | (*ppos)++; | |
4077 | ||
4078 | return cnt; | |
4079 | } | |
4080 | ||
5e2336a0 | 4081 | static const struct file_operations rb_simple_fops = { |
a3583244 SR |
4082 | .open = tracing_open_generic, |
4083 | .read = rb_simple_read, | |
4084 | .write = rb_simple_write, | |
6038f373 | 4085 | .llseek = default_llseek, |
a3583244 SR |
4086 | }; |
4087 | ||
4088 | ||
4089 | static __init int rb_init_debugfs(void) | |
4090 | { | |
4091 | struct dentry *d_tracer; | |
a3583244 SR |
4092 | |
4093 | d_tracer = tracing_init_dentry(); | |
4094 | ||
5452af66 FW |
4095 | trace_create_file("tracing_on", 0644, d_tracer, |
4096 | &ring_buffer_flags, &rb_simple_fops); | |
a3583244 SR |
4097 | |
4098 | return 0; | |
4099 | } | |
4100 | ||
4101 | fs_initcall(rb_init_debugfs); | |
1155de47 | 4102 | #endif |
554f786e | 4103 | |
59222efe | 4104 | #ifdef CONFIG_HOTPLUG_CPU |
09c9e84d FW |
4105 | static int rb_cpu_notify(struct notifier_block *self, |
4106 | unsigned long action, void *hcpu) | |
554f786e SR |
4107 | { |
4108 | struct ring_buffer *buffer = | |
4109 | container_of(self, struct ring_buffer, cpu_notify); | |
4110 | long cpu = (long)hcpu; | |
4111 | ||
4112 | switch (action) { | |
4113 | case CPU_UP_PREPARE: | |
4114 | case CPU_UP_PREPARE_FROZEN: | |
3f237a79 | 4115 | if (cpumask_test_cpu(cpu, buffer->cpumask)) |
554f786e SR |
4116 | return NOTIFY_OK; |
4117 | ||
4118 | buffer->buffers[cpu] = | |
4119 | rb_allocate_cpu_buffer(buffer, cpu); | |
4120 | if (!buffer->buffers[cpu]) { | |
4121 | WARN(1, "failed to allocate ring buffer on CPU %ld\n", | |
4122 | cpu); | |
4123 | return NOTIFY_OK; | |
4124 | } | |
4125 | smp_wmb(); | |
3f237a79 | 4126 | cpumask_set_cpu(cpu, buffer->cpumask); |
554f786e SR |
4127 | break; |
4128 | case CPU_DOWN_PREPARE: | |
4129 | case CPU_DOWN_PREPARE_FROZEN: | |
4130 | /* | |
4131 | * Do nothing. | |
4132 | * If we were to free the buffer, then the user would | |
4133 | * lose any trace that was in the buffer. | |
4134 | */ | |
4135 | break; | |
4136 | default: | |
4137 | break; | |
4138 | } | |
4139 | return NOTIFY_OK; | |
4140 | } | |
4141 | #endif |