Commit | Line | Data |
---|---|---|
852c2936 MD |
1 | /* |
2 | * ring_buffer_iterator.c | |
3 | * | |
4 | * (C) Copyright 2010 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> | |
5 | * | |
6 | * Ring buffer and channel iterators. Get each event of a channel in order. Uses | |
7 | * a prio heap for per-cpu buffers, giving a O(log(NR_CPUS)) algorithmic | |
8 | * complexity for the "get next event" operation. | |
9 | * | |
10 | * Author: | |
11 | * Mathieu Desnoyers <mathieu.desnoyers@efficios.com> | |
12 | * | |
13 | * Dual LGPL v2.1/GPL v2 license. | |
14 | */ | |
15 | ||
4931a13e | 16 | #include "iterator.h" |
852c2936 MD |
17 | |
18 | /* | |
19 | * Safety factor taking into account internal kernel interrupt latency. | |
20 | * Assuming 250ms worse-case latency. | |
21 | */ | |
22 | #define MAX_SYSTEM_LATENCY 250 | |
23 | ||
24 | /* | |
25 | * Maximum delta expected between trace clocks. At most 1 jiffy delta. | |
26 | */ | |
27 | #define MAX_CLOCK_DELTA (jiffies_to_usecs(1) * 1000) | |
28 | ||
29 | /** | |
30 | * lib_ring_buffer_get_next_record - Get the next record in a buffer. | |
31 | * @chan: channel | |
32 | * @buf: buffer | |
33 | * | |
34 | * Returns the size of the event read, -EAGAIN if buffer is empty, -ENODATA if | |
35 | * buffer is empty and finalized. The buffer must already be opened for reading. | |
36 | */ | |
37 | ssize_t lib_ring_buffer_get_next_record(struct channel *chan, | |
38 | struct lib_ring_buffer *buf) | |
39 | { | |
40 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
41 | struct lib_ring_buffer_iter *iter = &buf->iter; | |
42 | int ret; | |
43 | ||
44 | restart: | |
45 | switch (iter->state) { | |
46 | case ITER_GET_SUBBUF: | |
47 | ret = lib_ring_buffer_get_next_subbuf(buf); | |
48 | if (ret && !ACCESS_ONCE(buf->finalized) | |
49 | && config->alloc == RING_BUFFER_ALLOC_GLOBAL) { | |
50 | /* | |
51 | * Use "pull" scheme for global buffers. The reader | |
52 | * itself flushes the buffer to "pull" data not visible | |
53 | * to readers yet. Flush current subbuffer and re-try. | |
54 | * | |
55 | * Per-CPU buffers rather use a "push" scheme because | |
56 | * the IPI needed to flush all CPU's buffers is too | |
57 | * costly. In the "push" scheme, the reader waits for | |
58 | * the writer periodic deferrable timer to flush the | |
59 | * buffers (keeping track of a quiescent state | |
60 | * timestamp). Therefore, the writer "pushes" data out | |
61 | * of the buffers rather than letting the reader "pull" | |
62 | * data from the buffer. | |
63 | */ | |
64 | lib_ring_buffer_switch_slow(buf, SWITCH_ACTIVE); | |
65 | ret = lib_ring_buffer_get_next_subbuf(buf); | |
66 | } | |
67 | if (ret) | |
68 | return ret; | |
69 | iter->consumed = buf->cons_snapshot; | |
70 | iter->data_size = lib_ring_buffer_get_read_data_size(config, buf); | |
71 | iter->read_offset = iter->consumed; | |
72 | /* skip header */ | |
73 | iter->read_offset += config->cb.subbuffer_header_size(); | |
74 | iter->state = ITER_TEST_RECORD; | |
75 | goto restart; | |
76 | case ITER_TEST_RECORD: | |
77 | if (iter->read_offset - iter->consumed >= iter->data_size) { | |
78 | iter->state = ITER_PUT_SUBBUF; | |
79 | } else { | |
80 | CHAN_WARN_ON(chan, !config->cb.record_get); | |
81 | config->cb.record_get(config, chan, buf, | |
82 | iter->read_offset, | |
83 | &iter->header_len, | |
84 | &iter->payload_len, | |
85 | &iter->timestamp); | |
86 | iter->read_offset += iter->header_len; | |
87 | subbuffer_consume_record(config, &buf->backend); | |
88 | iter->state = ITER_NEXT_RECORD; | |
89 | return iter->payload_len; | |
90 | } | |
91 | goto restart; | |
92 | case ITER_NEXT_RECORD: | |
93 | iter->read_offset += iter->payload_len; | |
94 | iter->state = ITER_TEST_RECORD; | |
95 | goto restart; | |
96 | case ITER_PUT_SUBBUF: | |
97 | lib_ring_buffer_put_next_subbuf(buf); | |
98 | iter->state = ITER_GET_SUBBUF; | |
99 | goto restart; | |
100 | default: | |
101 | CHAN_WARN_ON(chan, 1); /* Should not happen */ | |
102 | return -EPERM; | |
103 | } | |
104 | } | |
105 | EXPORT_SYMBOL_GPL(lib_ring_buffer_get_next_record); | |
106 | ||
107 | static int buf_is_higher(void *a, void *b) | |
108 | { | |
109 | struct lib_ring_buffer *bufa = a; | |
110 | struct lib_ring_buffer *bufb = b; | |
111 | ||
112 | /* Consider lowest timestamps to be at the top of the heap */ | |
113 | return (bufa->iter.timestamp < bufb->iter.timestamp); | |
114 | } | |
115 | ||
116 | static | |
117 | void lib_ring_buffer_get_empty_buf_records(const struct lib_ring_buffer_config *config, | |
118 | struct channel *chan) | |
119 | { | |
120 | struct lttng_ptr_heap *heap = &chan->iter.heap; | |
121 | struct lib_ring_buffer *buf, *tmp; | |
122 | ssize_t len; | |
123 | ||
124 | list_for_each_entry_safe(buf, tmp, &chan->iter.empty_head, | |
125 | iter.empty_node) { | |
126 | len = lib_ring_buffer_get_next_record(chan, buf); | |
127 | ||
128 | /* | |
129 | * Deal with -EAGAIN and -ENODATA. | |
130 | * len >= 0 means record contains data. | |
131 | * -EBUSY should never happen, because we support only one | |
132 | * reader. | |
133 | */ | |
134 | switch (len) { | |
135 | case -EAGAIN: | |
136 | /* Keep node in empty list */ | |
137 | break; | |
138 | case -ENODATA: | |
139 | /* | |
140 | * Buffer is finalized. Don't add to list of empty | |
141 | * buffer, because it has no more data to provide, ever. | |
142 | */ | |
143 | list_del(&buf->iter.empty_node); | |
144 | break; | |
145 | case -EBUSY: | |
146 | CHAN_WARN_ON(chan, 1); | |
147 | break; | |
148 | default: | |
149 | /* | |
150 | * Insert buffer into the heap, remove from empty buffer | |
151 | * list. | |
152 | */ | |
153 | CHAN_WARN_ON(chan, len < 0); | |
154 | list_del(&buf->iter.empty_node); | |
155 | CHAN_WARN_ON(chan, lttng_heap_insert(heap, buf)); | |
156 | } | |
157 | } | |
158 | } | |
159 | ||
160 | static | |
161 | void lib_ring_buffer_wait_for_qs(const struct lib_ring_buffer_config *config, | |
162 | struct channel *chan) | |
163 | { | |
164 | u64 timestamp_qs; | |
165 | unsigned long wait_msecs; | |
166 | ||
167 | /* | |
168 | * No need to wait if no empty buffers are present. | |
169 | */ | |
170 | if (list_empty(&chan->iter.empty_head)) | |
171 | return; | |
172 | ||
173 | timestamp_qs = config->cb.ring_buffer_clock_read(chan); | |
174 | /* | |
175 | * We need to consider previously empty buffers. | |
176 | * Do a get next buf record on each of them. Add them to | |
177 | * the heap if they have data. If at least one of them | |
178 | * don't have data, we need to wait for | |
179 | * switch_timer_interval + MAX_SYSTEM_LATENCY (so we are sure the | |
180 | * buffers have been switched either by the timer or idle entry) and | |
181 | * check them again, adding them if they have data. | |
182 | */ | |
183 | lib_ring_buffer_get_empty_buf_records(config, chan); | |
184 | ||
185 | /* | |
186 | * No need to wait if no empty buffers are present. | |
187 | */ | |
188 | if (list_empty(&chan->iter.empty_head)) | |
189 | return; | |
190 | ||
191 | /* | |
192 | * We need to wait for the buffer switch timer to run. If the | |
193 | * CPU is idle, idle entry performed the switch. | |
194 | * TODO: we could optimize further by skipping the sleep if all | |
195 | * empty buffers belong to idle or offline cpus. | |
196 | */ | |
197 | wait_msecs = jiffies_to_msecs(chan->switch_timer_interval); | |
198 | wait_msecs += MAX_SYSTEM_LATENCY; | |
199 | msleep(wait_msecs); | |
200 | lib_ring_buffer_get_empty_buf_records(config, chan); | |
201 | /* | |
202 | * Any buffer still in the empty list here cannot possibly | |
203 | * contain an event with a timestamp prior to "timestamp_qs". | |
204 | * The new quiescent state timestamp is the one we grabbed | |
205 | * before waiting for buffer data. It is therefore safe to | |
206 | * ignore empty buffers up to last_qs timestamp for fusion | |
207 | * merge. | |
208 | */ | |
209 | chan->iter.last_qs = timestamp_qs; | |
210 | } | |
211 | ||
212 | /** | |
213 | * channel_get_next_record - Get the next record in a channel. | |
214 | * @chan: channel | |
215 | * @ret_buf: the buffer in which the event is located (output) | |
216 | * | |
217 | * Returns the size of new current event, -EAGAIN if all buffers are empty, | |
218 | * -ENODATA if all buffers are empty and finalized. The channel must already be | |
219 | * opened for reading. | |
220 | */ | |
221 | ||
222 | ssize_t channel_get_next_record(struct channel *chan, | |
223 | struct lib_ring_buffer **ret_buf) | |
224 | { | |
225 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
226 | struct lib_ring_buffer *buf; | |
227 | struct lttng_ptr_heap *heap; | |
228 | ssize_t len; | |
229 | ||
230 | if (config->alloc == RING_BUFFER_ALLOC_GLOBAL) { | |
231 | *ret_buf = channel_get_ring_buffer(config, chan, 0); | |
232 | return lib_ring_buffer_get_next_record(chan, *ret_buf); | |
233 | } | |
234 | ||
235 | heap = &chan->iter.heap; | |
236 | ||
237 | /* | |
238 | * get next record for topmost buffer. | |
239 | */ | |
240 | buf = lttng_heap_maximum(heap); | |
241 | if (buf) { | |
242 | len = lib_ring_buffer_get_next_record(chan, buf); | |
243 | /* | |
244 | * Deal with -EAGAIN and -ENODATA. | |
245 | * len >= 0 means record contains data. | |
246 | */ | |
247 | switch (len) { | |
248 | case -EAGAIN: | |
249 | buf->iter.timestamp = 0; | |
250 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); | |
251 | /* Remove topmost buffer from the heap */ | |
252 | CHAN_WARN_ON(chan, lttng_heap_remove(heap) != buf); | |
253 | break; | |
254 | case -ENODATA: | |
255 | /* | |
256 | * Buffer is finalized. Remove buffer from heap and | |
257 | * don't add to list of empty buffer, because it has no | |
258 | * more data to provide, ever. | |
259 | */ | |
260 | CHAN_WARN_ON(chan, lttng_heap_remove(heap) != buf); | |
261 | break; | |
262 | case -EBUSY: | |
263 | CHAN_WARN_ON(chan, 1); | |
264 | break; | |
265 | default: | |
266 | /* | |
267 | * Reinsert buffer into the heap. Note that heap can be | |
268 | * partially empty, so we need to use | |
269 | * lttng_heap_replace_max(). | |
270 | */ | |
271 | CHAN_WARN_ON(chan, len < 0); | |
272 | CHAN_WARN_ON(chan, lttng_heap_replace_max(heap, buf) != buf); | |
273 | break; | |
274 | } | |
275 | } | |
276 | ||
277 | buf = lttng_heap_maximum(heap); | |
278 | if (!buf || buf->iter.timestamp > chan->iter.last_qs) { | |
279 | /* | |
280 | * Deal with buffers previously showing no data. | |
281 | * Add buffers containing data to the heap, update | |
282 | * last_qs. | |
283 | */ | |
284 | lib_ring_buffer_wait_for_qs(config, chan); | |
285 | } | |
286 | ||
287 | *ret_buf = buf = lttng_heap_maximum(heap); | |
288 | if (buf) { | |
289 | /* | |
290 | * If this warning triggers, you probably need to check your | |
291 | * system interrupt latency. Typical causes: too many printk() | |
292 | * output going to a serial console with interrupts off. | |
293 | * Allow for MAX_CLOCK_DELTA ns timestamp delta going backward. | |
294 | * Observed on SMP KVM setups with trace_clock(). | |
295 | */ | |
296 | if (chan->iter.last_timestamp | |
297 | > (buf->iter.timestamp + MAX_CLOCK_DELTA)) { | |
298 | printk(KERN_WARNING "ring_buffer: timestamps going " | |
299 | "backward. Last time %llu ns, cpu %d, " | |
300 | "current time %llu ns, cpu %d, " | |
301 | "delta %llu ns.\n", | |
302 | chan->iter.last_timestamp, chan->iter.last_cpu, | |
303 | buf->iter.timestamp, buf->backend.cpu, | |
304 | chan->iter.last_timestamp - buf->iter.timestamp); | |
305 | CHAN_WARN_ON(chan, 1); | |
306 | } | |
307 | chan->iter.last_timestamp = buf->iter.timestamp; | |
308 | chan->iter.last_cpu = buf->backend.cpu; | |
309 | return buf->iter.payload_len; | |
310 | } else { | |
311 | /* Heap is empty */ | |
312 | if (list_empty(&chan->iter.empty_head)) | |
313 | return -ENODATA; /* All buffers finalized */ | |
314 | else | |
315 | return -EAGAIN; /* Temporarily empty */ | |
316 | } | |
317 | } | |
318 | EXPORT_SYMBOL_GPL(channel_get_next_record); | |
319 | ||
320 | static | |
321 | void lib_ring_buffer_iterator_init(struct channel *chan, struct lib_ring_buffer *buf) | |
322 | { | |
323 | if (buf->iter.allocated) | |
324 | return; | |
325 | ||
326 | buf->iter.allocated = 1; | |
327 | if (chan->iter.read_open && !buf->iter.read_open) { | |
328 | CHAN_WARN_ON(chan, lib_ring_buffer_open_read(buf) != 0); | |
329 | buf->iter.read_open = 1; | |
330 | } | |
331 | ||
332 | /* Add to list of buffers without any current record */ | |
333 | if (chan->backend.config->alloc == RING_BUFFER_ALLOC_PER_CPU) | |
334 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); | |
335 | } | |
336 | ||
337 | #ifdef CONFIG_HOTPLUG_CPU | |
338 | static | |
339 | int __cpuinit channel_iterator_cpu_hotplug(struct notifier_block *nb, | |
340 | unsigned long action, | |
341 | void *hcpu) | |
342 | { | |
343 | unsigned int cpu = (unsigned long)hcpu; | |
344 | struct channel *chan = container_of(nb, struct channel, | |
345 | hp_iter_notifier); | |
346 | struct lib_ring_buffer *buf = per_cpu_ptr(chan->backend.buf, cpu); | |
347 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
348 | ||
349 | if (!chan->hp_iter_enable) | |
350 | return NOTIFY_DONE; | |
351 | ||
352 | CHAN_WARN_ON(chan, config->alloc == RING_BUFFER_ALLOC_GLOBAL); | |
353 | ||
354 | switch (action) { | |
355 | case CPU_DOWN_FAILED: | |
356 | case CPU_DOWN_FAILED_FROZEN: | |
357 | case CPU_ONLINE: | |
358 | case CPU_ONLINE_FROZEN: | |
359 | lib_ring_buffer_iterator_init(chan, buf); | |
360 | return NOTIFY_OK; | |
361 | default: | |
362 | return NOTIFY_DONE; | |
363 | } | |
364 | } | |
365 | #endif | |
366 | ||
367 | int channel_iterator_init(struct channel *chan) | |
368 | { | |
369 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
370 | struct lib_ring_buffer *buf; | |
371 | ||
372 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
373 | int cpu, ret; | |
374 | ||
375 | INIT_LIST_HEAD(&chan->iter.empty_head); | |
376 | ret = lttng_heap_init(&chan->iter.heap, | |
377 | num_possible_cpus(), | |
378 | GFP_KERNEL, buf_is_higher); | |
379 | if (ret) | |
380 | return ret; | |
381 | /* | |
382 | * In case of non-hotplug cpu, if the ring-buffer is allocated | |
383 | * in early initcall, it will not be notified of secondary cpus. | |
384 | * In that off case, we need to allocate for all possible cpus. | |
385 | */ | |
386 | #ifdef CONFIG_HOTPLUG_CPU | |
387 | chan->hp_iter_notifier.notifier_call = | |
388 | channel_iterator_cpu_hotplug; | |
389 | chan->hp_iter_notifier.priority = 10; | |
390 | register_cpu_notifier(&chan->hp_iter_notifier); | |
391 | get_online_cpus(); | |
392 | for_each_online_cpu(cpu) { | |
393 | buf = per_cpu_ptr(chan->backend.buf, cpu); | |
394 | lib_ring_buffer_iterator_init(chan, buf); | |
395 | } | |
396 | chan->hp_iter_enable = 1; | |
397 | put_online_cpus(); | |
398 | #else | |
399 | for_each_possible_cpu(cpu) { | |
400 | buf = per_cpu_ptr(chan->backend.buf, cpu); | |
401 | lib_ring_buffer_iterator_init(chan, buf); | |
402 | } | |
403 | #endif | |
404 | } else { | |
405 | buf = channel_get_ring_buffer(config, chan, 0); | |
406 | lib_ring_buffer_iterator_init(chan, buf); | |
407 | } | |
408 | return 0; | |
409 | } | |
410 | ||
411 | void channel_iterator_unregister_notifiers(struct channel *chan) | |
412 | { | |
413 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
414 | ||
415 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
416 | chan->hp_iter_enable = 0; | |
417 | unregister_cpu_notifier(&chan->hp_iter_notifier); | |
418 | } | |
419 | } | |
420 | ||
421 | void channel_iterator_free(struct channel *chan) | |
422 | { | |
423 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
424 | ||
425 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) | |
426 | lttng_heap_free(&chan->iter.heap); | |
427 | } | |
428 | ||
429 | int lib_ring_buffer_iterator_open(struct lib_ring_buffer *buf) | |
430 | { | |
431 | struct channel *chan = buf->backend.chan; | |
432 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
433 | CHAN_WARN_ON(chan, config->output != RING_BUFFER_ITERATOR); | |
434 | return lib_ring_buffer_open_read(buf); | |
435 | } | |
436 | EXPORT_SYMBOL_GPL(lib_ring_buffer_iterator_open); | |
437 | ||
438 | /* | |
439 | * Note: Iterators must not be mixed with other types of outputs, because an | |
440 | * iterator can leave the buffer in "GET" state, which is not consistent with | |
441 | * other types of output (mmap, splice, raw data read). | |
442 | */ | |
443 | void lib_ring_buffer_iterator_release(struct lib_ring_buffer *buf) | |
444 | { | |
445 | lib_ring_buffer_release_read(buf); | |
446 | } | |
447 | EXPORT_SYMBOL_GPL(lib_ring_buffer_iterator_release); | |
448 | ||
449 | int channel_iterator_open(struct channel *chan) | |
450 | { | |
451 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
452 | struct lib_ring_buffer *buf; | |
453 | int ret = 0, cpu; | |
454 | ||
455 | CHAN_WARN_ON(chan, config->output != RING_BUFFER_ITERATOR); | |
456 | ||
457 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
458 | get_online_cpus(); | |
459 | /* Allow CPU hotplug to keep track of opened reader */ | |
460 | chan->iter.read_open = 1; | |
461 | for_each_channel_cpu(cpu, chan) { | |
462 | buf = channel_get_ring_buffer(config, chan, cpu); | |
463 | ret = lib_ring_buffer_iterator_open(buf); | |
464 | if (ret) | |
465 | goto error; | |
466 | buf->iter.read_open = 1; | |
467 | } | |
468 | put_online_cpus(); | |
469 | } else { | |
470 | buf = channel_get_ring_buffer(config, chan, 0); | |
471 | ret = lib_ring_buffer_iterator_open(buf); | |
472 | } | |
473 | return ret; | |
474 | error: | |
475 | /* Error should always happen on CPU 0, hence no close is required. */ | |
476 | CHAN_WARN_ON(chan, cpu != 0); | |
477 | put_online_cpus(); | |
478 | return ret; | |
479 | } | |
480 | EXPORT_SYMBOL_GPL(channel_iterator_open); | |
481 | ||
482 | void channel_iterator_release(struct channel *chan) | |
483 | { | |
484 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
485 | struct lib_ring_buffer *buf; | |
486 | int cpu; | |
487 | ||
488 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
489 | get_online_cpus(); | |
490 | for_each_channel_cpu(cpu, chan) { | |
491 | buf = channel_get_ring_buffer(config, chan, cpu); | |
492 | if (buf->iter.read_open) { | |
493 | lib_ring_buffer_iterator_release(buf); | |
494 | buf->iter.read_open = 0; | |
495 | } | |
496 | } | |
497 | chan->iter.read_open = 0; | |
498 | put_online_cpus(); | |
499 | } else { | |
500 | buf = channel_get_ring_buffer(config, chan, 0); | |
501 | lib_ring_buffer_iterator_release(buf); | |
502 | } | |
503 | } | |
504 | EXPORT_SYMBOL_GPL(channel_iterator_release); | |
505 | ||
506 | void lib_ring_buffer_iterator_reset(struct lib_ring_buffer *buf) | |
507 | { | |
508 | struct channel *chan = buf->backend.chan; | |
509 | ||
510 | if (buf->iter.state != ITER_GET_SUBBUF) | |
511 | lib_ring_buffer_put_next_subbuf(buf); | |
512 | buf->iter.state = ITER_GET_SUBBUF; | |
513 | /* Remove from heap (if present). */ | |
514 | if (lttng_heap_cherrypick(&chan->iter.heap, buf)) | |
515 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); | |
516 | buf->iter.timestamp = 0; | |
517 | buf->iter.header_len = 0; | |
518 | buf->iter.payload_len = 0; | |
519 | buf->iter.consumed = 0; | |
520 | buf->iter.read_offset = 0; | |
521 | buf->iter.data_size = 0; | |
522 | /* Don't reset allocated and read_open */ | |
523 | } | |
524 | ||
525 | void channel_iterator_reset(struct channel *chan) | |
526 | { | |
527 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
528 | struct lib_ring_buffer *buf; | |
529 | int cpu; | |
530 | ||
531 | /* Empty heap, put into empty_head */ | |
532 | while ((buf = lttng_heap_remove(&chan->iter.heap)) != NULL) | |
533 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); | |
534 | ||
535 | for_each_channel_cpu(cpu, chan) { | |
536 | buf = channel_get_ring_buffer(config, chan, cpu); | |
537 | lib_ring_buffer_iterator_reset(buf); | |
538 | } | |
539 | /* Don't reset read_open */ | |
540 | chan->iter.last_qs = 0; | |
541 | chan->iter.last_timestamp = 0; | |
542 | chan->iter.last_cpu = 0; | |
543 | chan->iter.len_left = 0; | |
544 | } | |
545 | ||
546 | /* | |
547 | * Ring buffer payload extraction read() implementation. | |
548 | */ | |
549 | static | |
550 | ssize_t channel_ring_buffer_file_read(struct file *filp, | |
551 | char __user *user_buf, | |
552 | size_t count, | |
553 | loff_t *ppos, | |
554 | struct channel *chan, | |
555 | struct lib_ring_buffer *buf, | |
556 | int fusionmerge) | |
557 | { | |
558 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
559 | size_t read_count = 0, read_offset; | |
560 | ssize_t len; | |
561 | ||
562 | might_sleep(); | |
563 | if (!access_ok(VERIFY_WRITE, user_buf, count)) | |
564 | return -EFAULT; | |
565 | ||
566 | /* Finish copy of previous record */ | |
567 | if (*ppos != 0) { | |
568 | if (read_count < count) { | |
569 | len = chan->iter.len_left; | |
570 | read_offset = *ppos; | |
571 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU | |
572 | && fusionmerge) | |
573 | buf = lttng_heap_maximum(&chan->iter.heap); | |
574 | CHAN_WARN_ON(chan, !buf); | |
575 | goto skip_get_next; | |
576 | } | |
577 | } | |
578 | ||
579 | while (read_count < count) { | |
580 | size_t copy_len, space_left; | |
581 | ||
582 | if (fusionmerge) | |
583 | len = channel_get_next_record(chan, &buf); | |
584 | else | |
585 | len = lib_ring_buffer_get_next_record(chan, buf); | |
586 | len_test: | |
587 | if (len < 0) { | |
588 | /* | |
589 | * Check if buffer is finalized (end of file). | |
590 | */ | |
591 | if (len == -ENODATA) { | |
592 | /* A 0 read_count will tell about end of file */ | |
593 | goto nodata; | |
594 | } | |
595 | if (filp->f_flags & O_NONBLOCK) { | |
596 | if (!read_count) | |
597 | read_count = -EAGAIN; | |
598 | goto nodata; | |
599 | } else { | |
600 | int error; | |
601 | ||
602 | /* | |
603 | * No data available at the moment, return what | |
604 | * we got. | |
605 | */ | |
606 | if (read_count) | |
607 | goto nodata; | |
608 | ||
609 | /* | |
610 | * Wait for returned len to be >= 0 or -ENODATA. | |
611 | */ | |
612 | if (fusionmerge) | |
613 | error = wait_event_interruptible( | |
614 | chan->read_wait, | |
615 | ((len = channel_get_next_record(chan, | |
616 | &buf)), len != -EAGAIN)); | |
617 | else | |
618 | error = wait_event_interruptible( | |
619 | buf->read_wait, | |
620 | ((len = lib_ring_buffer_get_next_record( | |
621 | chan, buf)), len != -EAGAIN)); | |
622 | CHAN_WARN_ON(chan, len == -EBUSY); | |
623 | if (error) { | |
624 | read_count = error; | |
625 | goto nodata; | |
626 | } | |
627 | CHAN_WARN_ON(chan, len < 0 && len != -ENODATA); | |
628 | goto len_test; | |
629 | } | |
630 | } | |
631 | read_offset = buf->iter.read_offset; | |
632 | skip_get_next: | |
633 | space_left = count - read_count; | |
634 | if (len <= space_left) { | |
635 | copy_len = len; | |
636 | chan->iter.len_left = 0; | |
637 | *ppos = 0; | |
638 | } else { | |
639 | copy_len = space_left; | |
640 | chan->iter.len_left = len - copy_len; | |
641 | *ppos = read_offset + copy_len; | |
642 | } | |
643 | if (__lib_ring_buffer_copy_to_user(&buf->backend, read_offset, | |
644 | &user_buf[read_count], | |
645 | copy_len)) { | |
646 | /* | |
647 | * Leave the len_left and ppos values at their current | |
648 | * state, as we currently have a valid event to read. | |
649 | */ | |
650 | return -EFAULT; | |
651 | } | |
652 | read_count += copy_len; | |
653 | }; | |
654 | return read_count; | |
655 | ||
656 | nodata: | |
657 | *ppos = 0; | |
658 | chan->iter.len_left = 0; | |
659 | return read_count; | |
660 | } | |
661 | ||
662 | /** | |
663 | * lib_ring_buffer_file_read - Read buffer record payload. | |
664 | * @filp: file structure pointer. | |
665 | * @buffer: user buffer to read data into. | |
666 | * @count: number of bytes to read. | |
667 | * @ppos: file read position. | |
668 | * | |
669 | * Returns a negative value on error, or the number of bytes read on success. | |
670 | * ppos is used to save the position _within the current record_ between calls | |
671 | * to read(). | |
672 | */ | |
673 | static | |
674 | ssize_t lib_ring_buffer_file_read(struct file *filp, | |
675 | char __user *user_buf, | |
676 | size_t count, | |
677 | loff_t *ppos) | |
678 | { | |
679 | struct inode *inode = filp->f_dentry->d_inode; | |
680 | struct lib_ring_buffer *buf = inode->i_private; | |
681 | struct channel *chan = buf->backend.chan; | |
682 | ||
683 | return channel_ring_buffer_file_read(filp, user_buf, count, ppos, | |
684 | chan, buf, 0); | |
685 | } | |
686 | ||
687 | /** | |
688 | * channel_file_read - Read channel record payload. | |
689 | * @filp: file structure pointer. | |
690 | * @buffer: user buffer to read data into. | |
691 | * @count: number of bytes to read. | |
692 | * @ppos: file read position. | |
693 | * | |
694 | * Returns a negative value on error, or the number of bytes read on success. | |
695 | * ppos is used to save the position _within the current record_ between calls | |
696 | * to read(). | |
697 | */ | |
698 | static | |
699 | ssize_t channel_file_read(struct file *filp, | |
700 | char __user *user_buf, | |
701 | size_t count, | |
702 | loff_t *ppos) | |
703 | { | |
704 | struct inode *inode = filp->f_dentry->d_inode; | |
705 | struct channel *chan = inode->i_private; | |
706 | const struct lib_ring_buffer_config *config = chan->backend.config; | |
707 | ||
708 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) | |
709 | return channel_ring_buffer_file_read(filp, user_buf, count, | |
710 | ppos, chan, NULL, 1); | |
711 | else { | |
712 | struct lib_ring_buffer *buf = | |
713 | channel_get_ring_buffer(config, chan, 0); | |
714 | return channel_ring_buffer_file_read(filp, user_buf, count, | |
715 | ppos, chan, buf, 0); | |
716 | } | |
717 | } | |
718 | ||
719 | static | |
720 | int lib_ring_buffer_file_open(struct inode *inode, struct file *file) | |
721 | { | |
722 | struct lib_ring_buffer *buf = inode->i_private; | |
723 | int ret; | |
724 | ||
725 | ret = lib_ring_buffer_iterator_open(buf); | |
726 | if (ret) | |
727 | return ret; | |
728 | ||
729 | file->private_data = buf; | |
730 | ret = nonseekable_open(inode, file); | |
731 | if (ret) | |
732 | goto release_iter; | |
733 | return 0; | |
734 | ||
735 | release_iter: | |
736 | lib_ring_buffer_iterator_release(buf); | |
737 | return ret; | |
738 | } | |
739 | ||
740 | static | |
741 | int lib_ring_buffer_file_release(struct inode *inode, struct file *file) | |
742 | { | |
743 | struct lib_ring_buffer *buf = inode->i_private; | |
744 | ||
745 | lib_ring_buffer_iterator_release(buf); | |
746 | return 0; | |
747 | } | |
748 | ||
749 | static | |
750 | int channel_file_open(struct inode *inode, struct file *file) | |
751 | { | |
752 | struct channel *chan = inode->i_private; | |
753 | int ret; | |
754 | ||
755 | ret = channel_iterator_open(chan); | |
756 | if (ret) | |
757 | return ret; | |
758 | ||
759 | file->private_data = chan; | |
760 | ret = nonseekable_open(inode, file); | |
761 | if (ret) | |
762 | goto release_iter; | |
763 | return 0; | |
764 | ||
765 | release_iter: | |
766 | channel_iterator_release(chan); | |
767 | return ret; | |
768 | } | |
769 | ||
770 | static | |
771 | int channel_file_release(struct inode *inode, struct file *file) | |
772 | { | |
773 | struct channel *chan = inode->i_private; | |
774 | ||
775 | channel_iterator_release(chan); | |
776 | return 0; | |
777 | } | |
778 | ||
779 | const struct file_operations channel_payload_file_operations = { | |
780 | .open = channel_file_open, | |
781 | .release = channel_file_release, | |
782 | .read = channel_file_read, | |
783 | .llseek = lib_ring_buffer_no_llseek, | |
784 | }; | |
785 | EXPORT_SYMBOL_GPL(channel_payload_file_operations); | |
786 | ||
787 | const struct file_operations lib_ring_buffer_payload_file_operations = { | |
788 | .open = lib_ring_buffer_file_open, | |
789 | .release = lib_ring_buffer_file_release, | |
790 | .read = lib_ring_buffer_file_read, | |
791 | .llseek = lib_ring_buffer_no_llseek, | |
792 | }; | |
793 | EXPORT_SYMBOL_GPL(lib_ring_buffer_payload_file_operations); |