Commit | Line | Data |
---|---|---|
b7cdc182 | 1 | /* SPDX-License-Identifier: (GPL-2.0-only OR LGPL-2.1-only) |
9f36eaed | 2 | * |
f3bc08c5 MD |
3 | * ring_buffer_iterator.c |
4 | * | |
f3bc08c5 MD |
5 | * Ring buffer and channel iterators. Get each event of a channel in order. Uses |
6 | * a prio heap for per-cpu buffers, giving a O(log(NR_CPUS)) algorithmic | |
7 | * complexity for the "get next event" operation. | |
8 | * | |
886d51a3 | 9 | * Copyright (C) 2010-2012 Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
f3bc08c5 MD |
10 | */ |
11 | ||
24591303 | 12 | #include <ringbuffer/iterator.h> |
c075712b | 13 | #include <wrapper/file.h> |
80bb2600 | 14 | #include <wrapper/uaccess.h> |
f3bc08c5 MD |
15 | #include <linux/jiffies.h> |
16 | #include <linux/delay.h> | |
17 | #include <linux/module.h> | |
18 | ||
19 | /* | |
20 | * Safety factor taking into account internal kernel interrupt latency. | |
21 | * Assuming 250ms worse-case latency. | |
22 | */ | |
23 | #define MAX_SYSTEM_LATENCY 250 | |
24 | ||
25 | /* | |
26 | * Maximum delta expected between trace clocks. At most 1 jiffy delta. | |
27 | */ | |
28 | #define MAX_CLOCK_DELTA (jiffies_to_usecs(1) * 1000) | |
29 | ||
30 | /** | |
31 | * lib_ring_buffer_get_next_record - Get the next record in a buffer. | |
32 | * @chan: channel | |
33 | * @buf: buffer | |
34 | * | |
35 | * Returns the size of the event read, -EAGAIN if buffer is empty, -ENODATA if | |
36 | * buffer is empty and finalized. The buffer must already be opened for reading. | |
37 | */ | |
38 | ssize_t lib_ring_buffer_get_next_record(struct channel *chan, | |
39 | struct lib_ring_buffer *buf) | |
40 | { | |
5a8fd222 | 41 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
42 | struct lib_ring_buffer_iter *iter = &buf->iter; |
43 | int ret; | |
44 | ||
45 | restart: | |
46 | switch (iter->state) { | |
47 | case ITER_GET_SUBBUF: | |
48 | ret = lib_ring_buffer_get_next_subbuf(buf); | |
585e5dcc | 49 | if (ret && !LTTNG_READ_ONCE(buf->finalized) |
f3bc08c5 MD |
50 | && config->alloc == RING_BUFFER_ALLOC_GLOBAL) { |
51 | /* | |
52 | * Use "pull" scheme for global buffers. The reader | |
53 | * itself flushes the buffer to "pull" data not visible | |
54 | * to readers yet. Flush current subbuffer and re-try. | |
55 | * | |
56 | * Per-CPU buffers rather use a "push" scheme because | |
57 | * the IPI needed to flush all CPU's buffers is too | |
58 | * costly. In the "push" scheme, the reader waits for | |
da9f3fb7 | 59 | * the writer periodic timer to flush the |
f3bc08c5 MD |
60 | * buffers (keeping track of a quiescent state |
61 | * timestamp). Therefore, the writer "pushes" data out | |
62 | * of the buffers rather than letting the reader "pull" | |
63 | * data from the buffer. | |
64 | */ | |
65 | lib_ring_buffer_switch_slow(buf, SWITCH_ACTIVE); | |
66 | ret = lib_ring_buffer_get_next_subbuf(buf); | |
67 | } | |
68 | if (ret) | |
69 | return ret; | |
70 | iter->consumed = buf->cons_snapshot; | |
71 | iter->data_size = lib_ring_buffer_get_read_data_size(config, buf); | |
72 | iter->read_offset = iter->consumed; | |
73 | /* skip header */ | |
74 | iter->read_offset += config->cb.subbuffer_header_size(); | |
75 | iter->state = ITER_TEST_RECORD; | |
76 | goto restart; | |
77 | case ITER_TEST_RECORD: | |
78 | if (iter->read_offset - iter->consumed >= iter->data_size) { | |
79 | iter->state = ITER_PUT_SUBBUF; | |
80 | } else { | |
81 | CHAN_WARN_ON(chan, !config->cb.record_get); | |
82 | config->cb.record_get(config, chan, buf, | |
83 | iter->read_offset, | |
84 | &iter->header_len, | |
85 | &iter->payload_len, | |
86 | &iter->timestamp); | |
87 | iter->read_offset += iter->header_len; | |
88 | subbuffer_consume_record(config, &buf->backend); | |
89 | iter->state = ITER_NEXT_RECORD; | |
90 | return iter->payload_len; | |
91 | } | |
92 | goto restart; | |
93 | case ITER_NEXT_RECORD: | |
94 | iter->read_offset += iter->payload_len; | |
95 | iter->state = ITER_TEST_RECORD; | |
96 | goto restart; | |
97 | case ITER_PUT_SUBBUF: | |
98 | lib_ring_buffer_put_next_subbuf(buf); | |
99 | iter->state = ITER_GET_SUBBUF; | |
100 | goto restart; | |
101 | default: | |
102 | CHAN_WARN_ON(chan, 1); /* Should not happen */ | |
103 | return -EPERM; | |
104 | } | |
105 | } | |
106 | EXPORT_SYMBOL_GPL(lib_ring_buffer_get_next_record); | |
107 | ||
47d4b31d MD |
108 | void lib_ring_buffer_put_current_record(struct lib_ring_buffer *buf) |
109 | { | |
110 | struct lib_ring_buffer_iter *iter; | |
111 | ||
112 | if (!buf) | |
113 | return; | |
114 | iter = &buf->iter; | |
115 | if (iter->state != ITER_NEXT_RECORD) | |
116 | return; | |
117 | iter->read_offset += iter->payload_len; | |
118 | iter->state = ITER_TEST_RECORD; | |
119 | if (iter->read_offset - iter->consumed >= iter->data_size) { | |
120 | lib_ring_buffer_put_next_subbuf(buf); | |
121 | iter->state = ITER_GET_SUBBUF; | |
122 | } | |
123 | } | |
124 | EXPORT_SYMBOL_GPL(lib_ring_buffer_put_current_record); | |
125 | ||
f3bc08c5 MD |
126 | static int buf_is_higher(void *a, void *b) |
127 | { | |
128 | struct lib_ring_buffer *bufa = a; | |
129 | struct lib_ring_buffer *bufb = b; | |
130 | ||
131 | /* Consider lowest timestamps to be at the top of the heap */ | |
132 | return (bufa->iter.timestamp < bufb->iter.timestamp); | |
133 | } | |
134 | ||
135 | static | |
136 | void lib_ring_buffer_get_empty_buf_records(const struct lib_ring_buffer_config *config, | |
137 | struct channel *chan) | |
138 | { | |
a88db018 | 139 | struct lttng_ptr_heap *heap = &chan->iter.heap; |
f3bc08c5 MD |
140 | struct lib_ring_buffer *buf, *tmp; |
141 | ssize_t len; | |
142 | ||
143 | list_for_each_entry_safe(buf, tmp, &chan->iter.empty_head, | |
144 | iter.empty_node) { | |
145 | len = lib_ring_buffer_get_next_record(chan, buf); | |
146 | ||
147 | /* | |
148 | * Deal with -EAGAIN and -ENODATA. | |
149 | * len >= 0 means record contains data. | |
150 | * -EBUSY should never happen, because we support only one | |
151 | * reader. | |
152 | */ | |
153 | switch (len) { | |
154 | case -EAGAIN: | |
155 | /* Keep node in empty list */ | |
156 | break; | |
157 | case -ENODATA: | |
158 | /* | |
159 | * Buffer is finalized. Don't add to list of empty | |
160 | * buffer, because it has no more data to provide, ever. | |
161 | */ | |
162 | list_del(&buf->iter.empty_node); | |
163 | break; | |
164 | case -EBUSY: | |
165 | CHAN_WARN_ON(chan, 1); | |
166 | break; | |
167 | default: | |
168 | /* | |
169 | * Insert buffer into the heap, remove from empty buffer | |
ab2277d6 | 170 | * list. |
f3bc08c5 MD |
171 | */ |
172 | CHAN_WARN_ON(chan, len < 0); | |
173 | list_del(&buf->iter.empty_node); | |
a88db018 | 174 | CHAN_WARN_ON(chan, lttng_heap_insert(heap, buf)); |
f3bc08c5 MD |
175 | } |
176 | } | |
177 | } | |
178 | ||
179 | static | |
180 | void lib_ring_buffer_wait_for_qs(const struct lib_ring_buffer_config *config, | |
181 | struct channel *chan) | |
182 | { | |
183 | u64 timestamp_qs; | |
184 | unsigned long wait_msecs; | |
185 | ||
186 | /* | |
187 | * No need to wait if no empty buffers are present. | |
188 | */ | |
189 | if (list_empty(&chan->iter.empty_head)) | |
190 | return; | |
191 | ||
192 | timestamp_qs = config->cb.ring_buffer_clock_read(chan); | |
193 | /* | |
194 | * We need to consider previously empty buffers. | |
195 | * Do a get next buf record on each of them. Add them to | |
196 | * the heap if they have data. If at least one of them | |
197 | * don't have data, we need to wait for | |
198 | * switch_timer_interval + MAX_SYSTEM_LATENCY (so we are sure the | |
199 | * buffers have been switched either by the timer or idle entry) and | |
200 | * check them again, adding them if they have data. | |
201 | */ | |
202 | lib_ring_buffer_get_empty_buf_records(config, chan); | |
203 | ||
204 | /* | |
205 | * No need to wait if no empty buffers are present. | |
206 | */ | |
207 | if (list_empty(&chan->iter.empty_head)) | |
208 | return; | |
209 | ||
210 | /* | |
211 | * We need to wait for the buffer switch timer to run. If the | |
212 | * CPU is idle, idle entry performed the switch. | |
213 | * TODO: we could optimize further by skipping the sleep if all | |
214 | * empty buffers belong to idle or offline cpus. | |
215 | */ | |
216 | wait_msecs = jiffies_to_msecs(chan->switch_timer_interval); | |
217 | wait_msecs += MAX_SYSTEM_LATENCY; | |
218 | msleep(wait_msecs); | |
219 | lib_ring_buffer_get_empty_buf_records(config, chan); | |
220 | /* | |
221 | * Any buffer still in the empty list here cannot possibly | |
222 | * contain an event with a timestamp prior to "timestamp_qs". | |
223 | * The new quiescent state timestamp is the one we grabbed | |
224 | * before waiting for buffer data. It is therefore safe to | |
225 | * ignore empty buffers up to last_qs timestamp for fusion | |
226 | * merge. | |
227 | */ | |
228 | chan->iter.last_qs = timestamp_qs; | |
229 | } | |
230 | ||
231 | /** | |
232 | * channel_get_next_record - Get the next record in a channel. | |
233 | * @chan: channel | |
234 | * @ret_buf: the buffer in which the event is located (output) | |
235 | * | |
236 | * Returns the size of new current event, -EAGAIN if all buffers are empty, | |
237 | * -ENODATA if all buffers are empty and finalized. The channel must already be | |
238 | * opened for reading. | |
239 | */ | |
240 | ||
241 | ssize_t channel_get_next_record(struct channel *chan, | |
242 | struct lib_ring_buffer **ret_buf) | |
243 | { | |
5a8fd222 | 244 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 | 245 | struct lib_ring_buffer *buf; |
a88db018 | 246 | struct lttng_ptr_heap *heap; |
f3bc08c5 MD |
247 | ssize_t len; |
248 | ||
249 | if (config->alloc == RING_BUFFER_ALLOC_GLOBAL) { | |
250 | *ret_buf = channel_get_ring_buffer(config, chan, 0); | |
251 | return lib_ring_buffer_get_next_record(chan, *ret_buf); | |
252 | } | |
253 | ||
254 | heap = &chan->iter.heap; | |
255 | ||
256 | /* | |
257 | * get next record for topmost buffer. | |
258 | */ | |
a88db018 | 259 | buf = lttng_heap_maximum(heap); |
f3bc08c5 MD |
260 | if (buf) { |
261 | len = lib_ring_buffer_get_next_record(chan, buf); | |
262 | /* | |
263 | * Deal with -EAGAIN and -ENODATA. | |
264 | * len >= 0 means record contains data. | |
265 | */ | |
266 | switch (len) { | |
267 | case -EAGAIN: | |
268 | buf->iter.timestamp = 0; | |
269 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); | |
270 | /* Remove topmost buffer from the heap */ | |
a88db018 | 271 | CHAN_WARN_ON(chan, lttng_heap_remove(heap) != buf); |
f3bc08c5 MD |
272 | break; |
273 | case -ENODATA: | |
274 | /* | |
275 | * Buffer is finalized. Remove buffer from heap and | |
276 | * don't add to list of empty buffer, because it has no | |
277 | * more data to provide, ever. | |
278 | */ | |
a88db018 | 279 | CHAN_WARN_ON(chan, lttng_heap_remove(heap) != buf); |
f3bc08c5 MD |
280 | break; |
281 | case -EBUSY: | |
282 | CHAN_WARN_ON(chan, 1); | |
283 | break; | |
284 | default: | |
285 | /* | |
286 | * Reinsert buffer into the heap. Note that heap can be | |
287 | * partially empty, so we need to use | |
a88db018 | 288 | * lttng_heap_replace_max(). |
f3bc08c5 MD |
289 | */ |
290 | CHAN_WARN_ON(chan, len < 0); | |
a88db018 | 291 | CHAN_WARN_ON(chan, lttng_heap_replace_max(heap, buf) != buf); |
f3bc08c5 MD |
292 | break; |
293 | } | |
294 | } | |
295 | ||
a88db018 | 296 | buf = lttng_heap_maximum(heap); |
f3bc08c5 MD |
297 | if (!buf || buf->iter.timestamp > chan->iter.last_qs) { |
298 | /* | |
299 | * Deal with buffers previously showing no data. | |
300 | * Add buffers containing data to the heap, update | |
301 | * last_qs. | |
302 | */ | |
303 | lib_ring_buffer_wait_for_qs(config, chan); | |
304 | } | |
305 | ||
a88db018 | 306 | *ret_buf = buf = lttng_heap_maximum(heap); |
f3bc08c5 MD |
307 | if (buf) { |
308 | /* | |
309 | * If this warning triggers, you probably need to check your | |
310 | * system interrupt latency. Typical causes: too many printk() | |
311 | * output going to a serial console with interrupts off. | |
312 | * Allow for MAX_CLOCK_DELTA ns timestamp delta going backward. | |
313 | * Observed on SMP KVM setups with trace_clock(). | |
314 | */ | |
315 | if (chan->iter.last_timestamp | |
316 | > (buf->iter.timestamp + MAX_CLOCK_DELTA)) { | |
5a15f70c | 317 | printk(KERN_WARNING "LTTng: ring_buffer: timestamps going " |
f3bc08c5 MD |
318 | "backward. Last time %llu ns, cpu %d, " |
319 | "current time %llu ns, cpu %d, " | |
320 | "delta %llu ns.\n", | |
321 | chan->iter.last_timestamp, chan->iter.last_cpu, | |
322 | buf->iter.timestamp, buf->backend.cpu, | |
323 | chan->iter.last_timestamp - buf->iter.timestamp); | |
324 | CHAN_WARN_ON(chan, 1); | |
325 | } | |
326 | chan->iter.last_timestamp = buf->iter.timestamp; | |
327 | chan->iter.last_cpu = buf->backend.cpu; | |
328 | return buf->iter.payload_len; | |
329 | } else { | |
330 | /* Heap is empty */ | |
331 | if (list_empty(&chan->iter.empty_head)) | |
332 | return -ENODATA; /* All buffers finalized */ | |
333 | else | |
334 | return -EAGAIN; /* Temporarily empty */ | |
335 | } | |
336 | } | |
337 | EXPORT_SYMBOL_GPL(channel_get_next_record); | |
338 | ||
339 | static | |
340 | void lib_ring_buffer_iterator_init(struct channel *chan, struct lib_ring_buffer *buf) | |
341 | { | |
342 | if (buf->iter.allocated) | |
343 | return; | |
344 | ||
345 | buf->iter.allocated = 1; | |
346 | if (chan->iter.read_open && !buf->iter.read_open) { | |
347 | CHAN_WARN_ON(chan, lib_ring_buffer_open_read(buf) != 0); | |
348 | buf->iter.read_open = 1; | |
349 | } | |
350 | ||
351 | /* Add to list of buffers without any current record */ | |
5a8fd222 | 352 | if (chan->backend.config.alloc == RING_BUFFER_ALLOC_PER_CPU) |
f3bc08c5 MD |
353 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); |
354 | } | |
355 | ||
5f4c791e | 356 | #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) |
1e367326 MD |
357 | |
358 | int lttng_cpuhp_rb_iter_online(unsigned int cpu, | |
359 | struct lttng_cpuhp_node *node) | |
360 | { | |
361 | struct channel *chan = container_of(node, struct channel, | |
362 | cpuhp_iter_online); | |
363 | struct lib_ring_buffer *buf = per_cpu_ptr(chan->backend.buf, cpu); | |
364 | const struct lib_ring_buffer_config *config = &chan->backend.config; | |
365 | ||
366 | CHAN_WARN_ON(chan, config->alloc == RING_BUFFER_ALLOC_GLOBAL); | |
367 | ||
368 | lib_ring_buffer_iterator_init(chan, buf); | |
369 | return 0; | |
370 | } | |
371 | EXPORT_SYMBOL_GPL(lttng_cpuhp_rb_iter_online); | |
372 | ||
5f4c791e | 373 | #else /* #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) */ |
1e367326 | 374 | |
f3bc08c5 MD |
375 | #ifdef CONFIG_HOTPLUG_CPU |
376 | static | |
e8f071d5 | 377 | int channel_iterator_cpu_hotplug(struct notifier_block *nb, |
f3bc08c5 MD |
378 | unsigned long action, |
379 | void *hcpu) | |
380 | { | |
381 | unsigned int cpu = (unsigned long)hcpu; | |
382 | struct channel *chan = container_of(nb, struct channel, | |
383 | hp_iter_notifier); | |
384 | struct lib_ring_buffer *buf = per_cpu_ptr(chan->backend.buf, cpu); | |
5a8fd222 | 385 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
386 | |
387 | if (!chan->hp_iter_enable) | |
388 | return NOTIFY_DONE; | |
389 | ||
390 | CHAN_WARN_ON(chan, config->alloc == RING_BUFFER_ALLOC_GLOBAL); | |
391 | ||
392 | switch (action) { | |
393 | case CPU_DOWN_FAILED: | |
394 | case CPU_DOWN_FAILED_FROZEN: | |
395 | case CPU_ONLINE: | |
396 | case CPU_ONLINE_FROZEN: | |
397 | lib_ring_buffer_iterator_init(chan, buf); | |
398 | return NOTIFY_OK; | |
399 | default: | |
400 | return NOTIFY_DONE; | |
401 | } | |
402 | } | |
403 | #endif | |
404 | ||
5f4c791e | 405 | #endif /* #else #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) */ |
1e367326 | 406 | |
f3bc08c5 MD |
407 | int channel_iterator_init(struct channel *chan) |
408 | { | |
5a8fd222 | 409 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
410 | struct lib_ring_buffer *buf; |
411 | ||
412 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
1e367326 | 413 | int ret; |
f3bc08c5 MD |
414 | |
415 | INIT_LIST_HEAD(&chan->iter.empty_head); | |
a88db018 | 416 | ret = lttng_heap_init(&chan->iter.heap, |
41affe31 | 417 | num_possible_cpus(), |
f3bc08c5 MD |
418 | GFP_KERNEL, buf_is_higher); |
419 | if (ret) | |
420 | return ret; | |
1e367326 | 421 | |
5f4c791e | 422 | #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) |
1e367326 MD |
423 | chan->cpuhp_iter_online.component = LTTNG_RING_BUFFER_ITER; |
424 | ret = cpuhp_state_add_instance(lttng_rb_hp_online, | |
425 | &chan->cpuhp_iter_online.node); | |
426 | if (ret) | |
427 | return ret; | |
5f4c791e | 428 | #else /* #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) */ |
1e367326 MD |
429 | { |
430 | int cpu; | |
431 | ||
432 | /* | |
433 | * In case of non-hotplug cpu, if the ring-buffer is allocated | |
434 | * in early initcall, it will not be notified of secondary cpus. | |
435 | * In that off case, we need to allocate for all possible cpus. | |
436 | */ | |
f3bc08c5 | 437 | #ifdef CONFIG_HOTPLUG_CPU |
1e367326 MD |
438 | chan->hp_iter_notifier.notifier_call = |
439 | channel_iterator_cpu_hotplug; | |
440 | chan->hp_iter_notifier.priority = 10; | |
441 | register_cpu_notifier(&chan->hp_iter_notifier); | |
442 | ||
443 | get_online_cpus(); | |
444 | for_each_online_cpu(cpu) { | |
445 | buf = per_cpu_ptr(chan->backend.buf, cpu); | |
446 | lib_ring_buffer_iterator_init(chan, buf); | |
447 | } | |
448 | chan->hp_iter_enable = 1; | |
449 | put_online_cpus(); | |
f3bc08c5 | 450 | #else |
1e367326 MD |
451 | for_each_possible_cpu(cpu) { |
452 | buf = per_cpu_ptr(chan->backend.buf, cpu); | |
453 | lib_ring_buffer_iterator_init(chan, buf); | |
454 | } | |
f3bc08c5 | 455 | #endif |
1e367326 | 456 | } |
5f4c791e | 457 | #endif /* #else #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) */ |
f3bc08c5 MD |
458 | } else { |
459 | buf = channel_get_ring_buffer(config, chan, 0); | |
460 | lib_ring_buffer_iterator_init(chan, buf); | |
461 | } | |
462 | return 0; | |
463 | } | |
464 | ||
465 | void channel_iterator_unregister_notifiers(struct channel *chan) | |
466 | { | |
5a8fd222 | 467 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
468 | |
469 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
5f4c791e | 470 | #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) |
1e367326 MD |
471 | { |
472 | int ret; | |
473 | ||
474 | ret = cpuhp_state_remove_instance(lttng_rb_hp_online, | |
475 | &chan->cpuhp_iter_online.node); | |
476 | WARN_ON(ret); | |
477 | } | |
5f4c791e | 478 | #else /* #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) */ |
f3bc08c5 MD |
479 | chan->hp_iter_enable = 0; |
480 | unregister_cpu_notifier(&chan->hp_iter_notifier); | |
5f4c791e | 481 | #endif /* #else #if (LTTNG_LINUX_VERSION_CODE >= LTTNG_KERNEL_VERSION(4,10,0)) */ |
f3bc08c5 MD |
482 | } |
483 | } | |
484 | ||
485 | void channel_iterator_free(struct channel *chan) | |
486 | { | |
5a8fd222 | 487 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
488 | |
489 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) | |
a88db018 | 490 | lttng_heap_free(&chan->iter.heap); |
f3bc08c5 MD |
491 | } |
492 | ||
493 | int lib_ring_buffer_iterator_open(struct lib_ring_buffer *buf) | |
494 | { | |
495 | struct channel *chan = buf->backend.chan; | |
5a8fd222 | 496 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
497 | CHAN_WARN_ON(chan, config->output != RING_BUFFER_ITERATOR); |
498 | return lib_ring_buffer_open_read(buf); | |
499 | } | |
500 | EXPORT_SYMBOL_GPL(lib_ring_buffer_iterator_open); | |
501 | ||
502 | /* | |
503 | * Note: Iterators must not be mixed with other types of outputs, because an | |
504 | * iterator can leave the buffer in "GET" state, which is not consistent with | |
505 | * other types of output (mmap, splice, raw data read). | |
506 | */ | |
507 | void lib_ring_buffer_iterator_release(struct lib_ring_buffer *buf) | |
508 | { | |
509 | lib_ring_buffer_release_read(buf); | |
510 | } | |
511 | EXPORT_SYMBOL_GPL(lib_ring_buffer_iterator_release); | |
512 | ||
513 | int channel_iterator_open(struct channel *chan) | |
514 | { | |
5a8fd222 | 515 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
516 | struct lib_ring_buffer *buf; |
517 | int ret = 0, cpu; | |
518 | ||
519 | CHAN_WARN_ON(chan, config->output != RING_BUFFER_ITERATOR); | |
520 | ||
521 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
522 | get_online_cpus(); | |
523 | /* Allow CPU hotplug to keep track of opened reader */ | |
524 | chan->iter.read_open = 1; | |
525 | for_each_channel_cpu(cpu, chan) { | |
526 | buf = channel_get_ring_buffer(config, chan, cpu); | |
527 | ret = lib_ring_buffer_iterator_open(buf); | |
528 | if (ret) | |
529 | goto error; | |
530 | buf->iter.read_open = 1; | |
531 | } | |
532 | put_online_cpus(); | |
533 | } else { | |
534 | buf = channel_get_ring_buffer(config, chan, 0); | |
535 | ret = lib_ring_buffer_iterator_open(buf); | |
536 | } | |
537 | return ret; | |
538 | error: | |
539 | /* Error should always happen on CPU 0, hence no close is required. */ | |
540 | CHAN_WARN_ON(chan, cpu != 0); | |
541 | put_online_cpus(); | |
542 | return ret; | |
543 | } | |
544 | EXPORT_SYMBOL_GPL(channel_iterator_open); | |
545 | ||
546 | void channel_iterator_release(struct channel *chan) | |
547 | { | |
5a8fd222 | 548 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
549 | struct lib_ring_buffer *buf; |
550 | int cpu; | |
551 | ||
552 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) { | |
553 | get_online_cpus(); | |
554 | for_each_channel_cpu(cpu, chan) { | |
555 | buf = channel_get_ring_buffer(config, chan, cpu); | |
556 | if (buf->iter.read_open) { | |
557 | lib_ring_buffer_iterator_release(buf); | |
558 | buf->iter.read_open = 0; | |
559 | } | |
560 | } | |
561 | chan->iter.read_open = 0; | |
562 | put_online_cpus(); | |
563 | } else { | |
564 | buf = channel_get_ring_buffer(config, chan, 0); | |
565 | lib_ring_buffer_iterator_release(buf); | |
566 | } | |
567 | } | |
568 | EXPORT_SYMBOL_GPL(channel_iterator_release); | |
569 | ||
570 | void lib_ring_buffer_iterator_reset(struct lib_ring_buffer *buf) | |
571 | { | |
572 | struct channel *chan = buf->backend.chan; | |
573 | ||
574 | if (buf->iter.state != ITER_GET_SUBBUF) | |
575 | lib_ring_buffer_put_next_subbuf(buf); | |
576 | buf->iter.state = ITER_GET_SUBBUF; | |
577 | /* Remove from heap (if present). */ | |
a88db018 | 578 | if (lttng_heap_cherrypick(&chan->iter.heap, buf)) |
f3bc08c5 MD |
579 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); |
580 | buf->iter.timestamp = 0; | |
581 | buf->iter.header_len = 0; | |
582 | buf->iter.payload_len = 0; | |
583 | buf->iter.consumed = 0; | |
584 | buf->iter.read_offset = 0; | |
585 | buf->iter.data_size = 0; | |
586 | /* Don't reset allocated and read_open */ | |
587 | } | |
588 | ||
589 | void channel_iterator_reset(struct channel *chan) | |
590 | { | |
5a8fd222 | 591 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
592 | struct lib_ring_buffer *buf; |
593 | int cpu; | |
594 | ||
595 | /* Empty heap, put into empty_head */ | |
a88db018 | 596 | while ((buf = lttng_heap_remove(&chan->iter.heap)) != NULL) |
f3bc08c5 MD |
597 | list_add(&buf->iter.empty_node, &chan->iter.empty_head); |
598 | ||
599 | for_each_channel_cpu(cpu, chan) { | |
600 | buf = channel_get_ring_buffer(config, chan, cpu); | |
601 | lib_ring_buffer_iterator_reset(buf); | |
602 | } | |
603 | /* Don't reset read_open */ | |
604 | chan->iter.last_qs = 0; | |
605 | chan->iter.last_timestamp = 0; | |
606 | chan->iter.last_cpu = 0; | |
607 | chan->iter.len_left = 0; | |
608 | } | |
609 | ||
610 | /* | |
611 | * Ring buffer payload extraction read() implementation. | |
612 | */ | |
613 | static | |
614 | ssize_t channel_ring_buffer_file_read(struct file *filp, | |
615 | char __user *user_buf, | |
616 | size_t count, | |
617 | loff_t *ppos, | |
618 | struct channel *chan, | |
619 | struct lib_ring_buffer *buf, | |
620 | int fusionmerge) | |
621 | { | |
5a8fd222 | 622 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
623 | size_t read_count = 0, read_offset; |
624 | ssize_t len; | |
625 | ||
626 | might_sleep(); | |
80bb2600 | 627 | if (!lttng_access_ok(VERIFY_WRITE, user_buf, count)) |
f3bc08c5 MD |
628 | return -EFAULT; |
629 | ||
630 | /* Finish copy of previous record */ | |
631 | if (*ppos != 0) { | |
632 | if (read_count < count) { | |
633 | len = chan->iter.len_left; | |
634 | read_offset = *ppos; | |
635 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU | |
636 | && fusionmerge) | |
a88db018 | 637 | buf = lttng_heap_maximum(&chan->iter.heap); |
f3bc08c5 MD |
638 | CHAN_WARN_ON(chan, !buf); |
639 | goto skip_get_next; | |
640 | } | |
641 | } | |
642 | ||
643 | while (read_count < count) { | |
644 | size_t copy_len, space_left; | |
645 | ||
646 | if (fusionmerge) | |
647 | len = channel_get_next_record(chan, &buf); | |
648 | else | |
649 | len = lib_ring_buffer_get_next_record(chan, buf); | |
650 | len_test: | |
651 | if (len < 0) { | |
652 | /* | |
653 | * Check if buffer is finalized (end of file). | |
654 | */ | |
655 | if (len == -ENODATA) { | |
656 | /* A 0 read_count will tell about end of file */ | |
657 | goto nodata; | |
658 | } | |
659 | if (filp->f_flags & O_NONBLOCK) { | |
660 | if (!read_count) | |
661 | read_count = -EAGAIN; | |
662 | goto nodata; | |
663 | } else { | |
664 | int error; | |
665 | ||
666 | /* | |
667 | * No data available at the moment, return what | |
668 | * we got. | |
669 | */ | |
670 | if (read_count) | |
671 | goto nodata; | |
672 | ||
673 | /* | |
674 | * Wait for returned len to be >= 0 or -ENODATA. | |
675 | */ | |
676 | if (fusionmerge) | |
677 | error = wait_event_interruptible( | |
678 | chan->read_wait, | |
679 | ((len = channel_get_next_record(chan, | |
680 | &buf)), len != -EAGAIN)); | |
681 | else | |
682 | error = wait_event_interruptible( | |
683 | buf->read_wait, | |
684 | ((len = lib_ring_buffer_get_next_record( | |
685 | chan, buf)), len != -EAGAIN)); | |
686 | CHAN_WARN_ON(chan, len == -EBUSY); | |
687 | if (error) { | |
688 | read_count = error; | |
689 | goto nodata; | |
690 | } | |
691 | CHAN_WARN_ON(chan, len < 0 && len != -ENODATA); | |
692 | goto len_test; | |
693 | } | |
694 | } | |
695 | read_offset = buf->iter.read_offset; | |
696 | skip_get_next: | |
697 | space_left = count - read_count; | |
698 | if (len <= space_left) { | |
699 | copy_len = len; | |
700 | chan->iter.len_left = 0; | |
701 | *ppos = 0; | |
702 | } else { | |
703 | copy_len = space_left; | |
704 | chan->iter.len_left = len - copy_len; | |
705 | *ppos = read_offset + copy_len; | |
706 | } | |
707 | if (__lib_ring_buffer_copy_to_user(&buf->backend, read_offset, | |
708 | &user_buf[read_count], | |
709 | copy_len)) { | |
710 | /* | |
711 | * Leave the len_left and ppos values at their current | |
712 | * state, as we currently have a valid event to read. | |
713 | */ | |
714 | return -EFAULT; | |
715 | } | |
716 | read_count += copy_len; | |
47d4b31d MD |
717 | } |
718 | goto put_record; | |
f3bc08c5 MD |
719 | |
720 | nodata: | |
721 | *ppos = 0; | |
722 | chan->iter.len_left = 0; | |
47d4b31d MD |
723 | put_record: |
724 | lib_ring_buffer_put_current_record(buf); | |
f3bc08c5 MD |
725 | return read_count; |
726 | } | |
727 | ||
728 | /** | |
729 | * lib_ring_buffer_file_read - Read buffer record payload. | |
730 | * @filp: file structure pointer. | |
731 | * @buffer: user buffer to read data into. | |
732 | * @count: number of bytes to read. | |
733 | * @ppos: file read position. | |
734 | * | |
735 | * Returns a negative value on error, or the number of bytes read on success. | |
736 | * ppos is used to save the position _within the current record_ between calls | |
737 | * to read(). | |
738 | */ | |
739 | static | |
740 | ssize_t lib_ring_buffer_file_read(struct file *filp, | |
741 | char __user *user_buf, | |
742 | size_t count, | |
743 | loff_t *ppos) | |
744 | { | |
b06ed645 | 745 | struct inode *inode = filp->lttng_f_dentry->d_inode; |
f3bc08c5 MD |
746 | struct lib_ring_buffer *buf = inode->i_private; |
747 | struct channel *chan = buf->backend.chan; | |
748 | ||
749 | return channel_ring_buffer_file_read(filp, user_buf, count, ppos, | |
750 | chan, buf, 0); | |
751 | } | |
752 | ||
753 | /** | |
754 | * channel_file_read - Read channel record payload. | |
755 | * @filp: file structure pointer. | |
756 | * @buffer: user buffer to read data into. | |
757 | * @count: number of bytes to read. | |
758 | * @ppos: file read position. | |
759 | * | |
760 | * Returns a negative value on error, or the number of bytes read on success. | |
761 | * ppos is used to save the position _within the current record_ between calls | |
762 | * to read(). | |
763 | */ | |
764 | static | |
765 | ssize_t channel_file_read(struct file *filp, | |
766 | char __user *user_buf, | |
767 | size_t count, | |
768 | loff_t *ppos) | |
769 | { | |
b06ed645 | 770 | struct inode *inode = filp->lttng_f_dentry->d_inode; |
f3bc08c5 | 771 | struct channel *chan = inode->i_private; |
5a8fd222 | 772 | const struct lib_ring_buffer_config *config = &chan->backend.config; |
f3bc08c5 MD |
773 | |
774 | if (config->alloc == RING_BUFFER_ALLOC_PER_CPU) | |
775 | return channel_ring_buffer_file_read(filp, user_buf, count, | |
776 | ppos, chan, NULL, 1); | |
777 | else { | |
778 | struct lib_ring_buffer *buf = | |
779 | channel_get_ring_buffer(config, chan, 0); | |
780 | return channel_ring_buffer_file_read(filp, user_buf, count, | |
781 | ppos, chan, buf, 0); | |
782 | } | |
783 | } | |
784 | ||
785 | static | |
786 | int lib_ring_buffer_file_open(struct inode *inode, struct file *file) | |
787 | { | |
788 | struct lib_ring_buffer *buf = inode->i_private; | |
789 | int ret; | |
790 | ||
791 | ret = lib_ring_buffer_iterator_open(buf); | |
792 | if (ret) | |
793 | return ret; | |
794 | ||
795 | file->private_data = buf; | |
796 | ret = nonseekable_open(inode, file); | |
797 | if (ret) | |
798 | goto release_iter; | |
799 | return 0; | |
800 | ||
801 | release_iter: | |
802 | lib_ring_buffer_iterator_release(buf); | |
803 | return ret; | |
804 | } | |
805 | ||
806 | static | |
807 | int lib_ring_buffer_file_release(struct inode *inode, struct file *file) | |
808 | { | |
809 | struct lib_ring_buffer *buf = inode->i_private; | |
810 | ||
811 | lib_ring_buffer_iterator_release(buf); | |
812 | return 0; | |
813 | } | |
814 | ||
815 | static | |
816 | int channel_file_open(struct inode *inode, struct file *file) | |
817 | { | |
818 | struct channel *chan = inode->i_private; | |
819 | int ret; | |
820 | ||
821 | ret = channel_iterator_open(chan); | |
822 | if (ret) | |
823 | return ret; | |
824 | ||
825 | file->private_data = chan; | |
826 | ret = nonseekable_open(inode, file); | |
827 | if (ret) | |
828 | goto release_iter; | |
829 | return 0; | |
830 | ||
831 | release_iter: | |
832 | channel_iterator_release(chan); | |
833 | return ret; | |
834 | } | |
835 | ||
836 | static | |
837 | int channel_file_release(struct inode *inode, struct file *file) | |
838 | { | |
839 | struct channel *chan = inode->i_private; | |
840 | ||
841 | channel_iterator_release(chan); | |
842 | return 0; | |
843 | } | |
844 | ||
845 | const struct file_operations channel_payload_file_operations = { | |
a33c9927 | 846 | .owner = THIS_MODULE, |
f3bc08c5 MD |
847 | .open = channel_file_open, |
848 | .release = channel_file_release, | |
849 | .read = channel_file_read, | |
d83004aa | 850 | .llseek = vfs_lib_ring_buffer_no_llseek, |
f3bc08c5 MD |
851 | }; |
852 | EXPORT_SYMBOL_GPL(channel_payload_file_operations); | |
853 | ||
854 | const struct file_operations lib_ring_buffer_payload_file_operations = { | |
a33c9927 | 855 | .owner = THIS_MODULE, |
f3bc08c5 MD |
856 | .open = lib_ring_buffer_file_open, |
857 | .release = lib_ring_buffer_file_release, | |
858 | .read = lib_ring_buffer_file_read, | |
d83004aa | 859 | .llseek = vfs_lib_ring_buffer_no_llseek, |
f3bc08c5 MD |
860 | }; |
861 | EXPORT_SYMBOL_GPL(lib_ring_buffer_payload_file_operations); |