Commit | Line | Data |
---|---|---|
3bd1e081 | 1 | /* |
21cf9b6b | 2 | * Copyright (C) 2011 EfficiOS Inc. |
ab5be9fa MJ |
3 | * Copyright (C) 2011 Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
4 | * Copyright (C) 2012 David Goulet <dgoulet@efficios.com> | |
3bd1e081 | 5 | * |
ab5be9fa | 6 | * SPDX-License-Identifier: GPL-2.0-only |
3bd1e081 | 7 | * |
3bd1e081 MD |
8 | */ |
9 | ||
6c1c0768 | 10 | #define _LGPL_SOURCE |
c9e313bc SM |
11 | #include <common/align.hpp> |
12 | #include <common/common.hpp> | |
13 | #include <common/compat/endian.hpp> | |
14 | #include <common/compat/poll.hpp> | |
15 | #include <common/consumer/consumer-metadata-cache.hpp> | |
16 | #include <common/consumer/consumer-stream.hpp> | |
17 | #include <common/consumer/consumer-testpoint.hpp> | |
18 | #include <common/consumer/consumer-timer.hpp> | |
19 | #include <common/consumer/consumer.hpp> | |
20 | #include <common/dynamic-array.hpp> | |
21 | #include <common/index/ctf-index.hpp> | |
22 | #include <common/index/index.hpp> | |
671e39d7 | 23 | #include <common/io-hint.hpp> |
c9e313bc SM |
24 | #include <common/kernel-consumer/kernel-consumer.hpp> |
25 | #include <common/kernel-ctl/kernel-ctl.hpp> | |
3d46ea1a | 26 | #include <common/pthread-lock.hpp> |
c9e313bc SM |
27 | #include <common/relayd/relayd.hpp> |
28 | #include <common/sessiond-comm/relayd.hpp> | |
29 | #include <common/sessiond-comm/sessiond-comm.hpp> | |
30 | #include <common/string-utils/format.hpp> | |
31 | #include <common/time.hpp> | |
32 | #include <common/trace-chunk-registry.hpp> | |
33 | #include <common/trace-chunk.hpp> | |
56047f5a | 34 | #include <common/urcu.hpp> |
c9e313bc SM |
35 | #include <common/ust-consumer/ust-consumer.hpp> |
36 | #include <common/utils.hpp> | |
3bd1e081 | 37 | |
28ab034a | 38 | #include <bin/lttng-consumerd/health-consumerd.hpp> |
671e39d7 | 39 | #include <fcntl.h> |
28ab034a JG |
40 | #include <inttypes.h> |
41 | #include <poll.h> | |
42 | #include <pthread.h> | |
43 | #include <signal.h> | |
44 | #include <stdlib.h> | |
45 | #include <string.h> | |
46 | #include <sys/mman.h> | |
47 | #include <sys/socket.h> | |
48 | #include <sys/types.h> | |
fbd566c2 | 49 | #include <type_traits> |
28ab034a JG |
50 | #include <unistd.h> |
51 | ||
97535efa | 52 | lttng_consumer_global_data the_consumer_data; |
3bd1e081 | 53 | |
d8ef542d MD |
54 | enum consumer_channel_action { |
55 | CONSUMER_CHANNEL_ADD, | |
a0cbdd2e | 56 | CONSUMER_CHANNEL_DEL, |
d8ef542d MD |
57 | CONSUMER_CHANNEL_QUIT, |
58 | }; | |
59 | ||
f1494934 | 60 | namespace { |
d8ef542d MD |
61 | struct consumer_channel_msg { |
62 | enum consumer_channel_action action; | |
28ab034a JG |
63 | struct lttng_consumer_channel *chan; /* add */ |
64 | uint64_t key; /* del */ | |
d8ef542d MD |
65 | }; |
66 | ||
f1494934 JG |
67 | /* |
68 | * Global hash table containing respectively metadata and data streams. The | |
69 | * stream element in this ht should only be updated by the metadata poll thread | |
70 | * for the metadata and the data poll thread for the data. | |
71 | */ | |
72 | struct lttng_ht *metadata_ht; | |
73 | struct lttng_ht *data_ht; | |
74 | } /* namespace */ | |
75 | ||
80957876 | 76 | /* Flag used to temporarily pause data consumption from testpoints. */ |
cf0bcb51 JG |
77 | int data_consumption_paused; |
78 | ||
3bd1e081 MD |
79 | /* |
80 | * Flag to inform the polling thread to quit when all fd hung up. Updated by | |
81 | * the consumer_thread_receive_fds when it notices that all fds has hung up. | |
82 | * Also updated by the signal handler (consumer_should_exit()). Read by the | |
83 | * polling threads. | |
84 | */ | |
10211f5c | 85 | int consumer_quit; |
3bd1e081 | 86 | |
cd9adb8b | 87 | static const char *get_consumer_domain() |
5da88b0f | 88 | { |
fa29bfbf | 89 | switch (the_consumer_data.type) { |
5da88b0f MD |
90 | case LTTNG_CONSUMER_KERNEL: |
91 | return DEFAULT_KERNEL_TRACE_DIR; | |
92 | case LTTNG_CONSUMER64_UST: | |
93 | /* Fall-through. */ | |
94 | case LTTNG_CONSUMER32_UST: | |
95 | return DEFAULT_UST_TRACE_DIR; | |
96 | default: | |
97 | abort(); | |
98 | } | |
99 | } | |
100 | ||
acdb9057 DG |
101 | /* |
102 | * Notify a thread lttng pipe to poll back again. This usually means that some | |
103 | * global state has changed so we just send back the thread in a poll wait | |
104 | * call. | |
105 | */ | |
106 | static void notify_thread_lttng_pipe(struct lttng_pipe *pipe) | |
107 | { | |
cd9adb8b | 108 | struct lttng_consumer_stream *null_stream = nullptr; |
acdb9057 | 109 | |
a0377dfe | 110 | LTTNG_ASSERT(pipe); |
acdb9057 | 111 | |
5c7248cd JG |
112 | (void) lttng_pipe_write(pipe, &null_stream, sizeof(null_stream)); /* NOLINT sizeof used on a |
113 | pointer. */ | |
acdb9057 DG |
114 | } |
115 | ||
5c635c72 MD |
116 | static void notify_health_quit_pipe(int *pipe) |
117 | { | |
6cd525e8 | 118 | ssize_t ret; |
5c635c72 | 119 | |
6cd525e8 MD |
120 | ret = lttng_write(pipe[1], "4", 1); |
121 | if (ret < 1) { | |
5c635c72 MD |
122 | PERROR("write consumer health quit"); |
123 | } | |
124 | } | |
125 | ||
d8ef542d | 126 | static void notify_channel_pipe(struct lttng_consumer_local_data *ctx, |
28ab034a JG |
127 | struct lttng_consumer_channel *chan, |
128 | uint64_t key, | |
129 | enum consumer_channel_action action) | |
d8ef542d MD |
130 | { |
131 | struct consumer_channel_msg msg; | |
6cd525e8 | 132 | ssize_t ret; |
d8ef542d | 133 | |
e56251fc DG |
134 | memset(&msg, 0, sizeof(msg)); |
135 | ||
d8ef542d MD |
136 | msg.action = action; |
137 | msg.chan = chan; | |
f21dae48 | 138 | msg.key = key; |
6cd525e8 MD |
139 | ret = lttng_write(ctx->consumer_channel_pipe[1], &msg, sizeof(msg)); |
140 | if (ret < sizeof(msg)) { | |
141 | PERROR("notify_channel_pipe write error"); | |
142 | } | |
d8ef542d MD |
143 | } |
144 | ||
28ab034a | 145 | void notify_thread_del_channel(struct lttng_consumer_local_data *ctx, uint64_t key) |
a0cbdd2e | 146 | { |
cd9adb8b | 147 | notify_channel_pipe(ctx, nullptr, key, CONSUMER_CHANNEL_DEL); |
a0cbdd2e MD |
148 | } |
149 | ||
d8ef542d | 150 | static int read_channel_pipe(struct lttng_consumer_local_data *ctx, |
28ab034a JG |
151 | struct lttng_consumer_channel **chan, |
152 | uint64_t *key, | |
153 | enum consumer_channel_action *action) | |
d8ef542d MD |
154 | { |
155 | struct consumer_channel_msg msg; | |
6cd525e8 | 156 | ssize_t ret; |
d8ef542d | 157 | |
6cd525e8 MD |
158 | ret = lttng_read(ctx->consumer_channel_pipe[0], &msg, sizeof(msg)); |
159 | if (ret < sizeof(msg)) { | |
160 | ret = -1; | |
161 | goto error; | |
d8ef542d | 162 | } |
6cd525e8 MD |
163 | *action = msg.action; |
164 | *chan = msg.chan; | |
165 | *key = msg.key; | |
166 | error: | |
167 | return (int) ret; | |
d8ef542d MD |
168 | } |
169 | ||
212d67a2 DG |
170 | /* |
171 | * Cleanup the stream list of a channel. Those streams are not yet globally | |
172 | * visible | |
173 | */ | |
174 | static void clean_channel_stream_list(struct lttng_consumer_channel *channel) | |
175 | { | |
a0377dfe | 176 | LTTNG_ASSERT(channel); |
212d67a2 DG |
177 | |
178 | /* Delete streams that might have been left in the stream list. */ | |
a1a1df65 JG |
179 | for (auto stream : lttng::urcu::list_iteration_adapter<lttng_consumer_stream, |
180 | <tng_consumer_stream::send_node>( | |
181 | channel->streams.head)) { | |
cd9adb8b | 182 | consumer_stream_destroy(stream, nullptr); |
212d67a2 DG |
183 | } |
184 | } | |
185 | ||
3bd1e081 MD |
186 | /* |
187 | * Find a stream. The consumer_data.lock must be locked during this | |
188 | * call. | |
189 | */ | |
28ab034a | 190 | static struct lttng_consumer_stream *find_stream(uint64_t key, struct lttng_ht *ht) |
3bd1e081 | 191 | { |
e4421fec | 192 | struct lttng_ht_iter iter; |
d88aee68 | 193 | struct lttng_ht_node_u64 *node; |
cd9adb8b | 194 | struct lttng_consumer_stream *stream = nullptr; |
3bd1e081 | 195 | |
a0377dfe | 196 | LTTNG_ASSERT(ht); |
8389e4f8 | 197 | |
d88aee68 DG |
198 | /* -1ULL keys are lookup failures */ |
199 | if (key == (uint64_t) -1ULL) { | |
cd9adb8b | 200 | return nullptr; |
7a57cf92 | 201 | } |
e4421fec | 202 | |
07c4863f | 203 | const lttng::urcu::read_lock_guard read_lock; |
6065ceec | 204 | |
d88aee68 | 205 | lttng_ht_lookup(ht, &key, &iter); |
00d7d903 | 206 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
cd9adb8b | 207 | if (node != nullptr) { |
0114db0e | 208 | stream = lttng::utils::container_of(node, <tng_consumer_stream::node); |
3bd1e081 | 209 | } |
e4421fec DG |
210 | |
211 | return stream; | |
3bd1e081 MD |
212 | } |
213 | ||
da009f2c | 214 | static void steal_stream_key(uint64_t key, struct lttng_ht *ht) |
7ad0a0cb MD |
215 | { |
216 | struct lttng_consumer_stream *stream; | |
217 | ||
07c4863f | 218 | const lttng::urcu::read_lock_guard read_lock; |
ffe60014 | 219 | stream = find_stream(key, ht); |
04253271 | 220 | if (stream) { |
da009f2c | 221 | stream->key = (uint64_t) -1ULL; |
04253271 MD |
222 | /* |
223 | * We don't want the lookup to match, but we still need | |
224 | * to iterate on this stream when iterating over the hash table. Just | |
225 | * change the node key. | |
226 | */ | |
da009f2c | 227 | stream->node.key = (uint64_t) -1ULL; |
04253271 | 228 | } |
7ad0a0cb MD |
229 | } |
230 | ||
d56db448 DG |
231 | /* |
232 | * Return a channel object for the given key. | |
233 | * | |
234 | * RCU read side lock MUST be acquired before calling this function and | |
235 | * protects the channel ptr. | |
236 | */ | |
d88aee68 | 237 | struct lttng_consumer_channel *consumer_find_channel(uint64_t key) |
3bd1e081 | 238 | { |
e4421fec | 239 | struct lttng_ht_iter iter; |
d88aee68 | 240 | struct lttng_ht_node_u64 *node; |
cd9adb8b | 241 | struct lttng_consumer_channel *channel = nullptr; |
3bd1e081 | 242 | |
48b7cdc2 FD |
243 | ASSERT_RCU_READ_LOCKED(); |
244 | ||
d88aee68 DG |
245 | /* -1ULL keys are lookup failures */ |
246 | if (key == (uint64_t) -1ULL) { | |
cd9adb8b | 247 | return nullptr; |
7a57cf92 | 248 | } |
e4421fec | 249 | |
fa29bfbf | 250 | lttng_ht_lookup(the_consumer_data.channel_ht, &key, &iter); |
00d7d903 | 251 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
cd9adb8b | 252 | if (node != nullptr) { |
0114db0e | 253 | channel = lttng::utils::container_of(node, <tng_consumer_channel::node); |
3bd1e081 | 254 | } |
e4421fec DG |
255 | |
256 | return channel; | |
3bd1e081 MD |
257 | } |
258 | ||
b5a6470f DG |
259 | /* |
260 | * There is a possibility that the consumer does not have enough time between | |
261 | * the close of the channel on the session daemon and the cleanup in here thus | |
262 | * once we have a channel add with an existing key, we know for sure that this | |
263 | * channel will eventually get cleaned up by all streams being closed. | |
264 | * | |
265 | * This function just nullifies the already existing channel key. | |
266 | */ | |
267 | static void steal_channel_key(uint64_t key) | |
268 | { | |
269 | struct lttng_consumer_channel *channel; | |
270 | ||
07c4863f | 271 | const lttng::urcu::read_lock_guard read_lock; |
b5a6470f DG |
272 | channel = consumer_find_channel(key); |
273 | if (channel) { | |
274 | channel->key = (uint64_t) -1ULL; | |
275 | /* | |
276 | * We don't want the lookup to match, but we still need to iterate on | |
277 | * this channel when iterating over the hash table. Just change the | |
278 | * node key. | |
279 | */ | |
280 | channel->node.key = (uint64_t) -1ULL; | |
281 | } | |
b5a6470f DG |
282 | } |
283 | ||
ffe60014 | 284 | static void free_channel_rcu(struct rcu_head *head) |
702b1ea4 | 285 | { |
28ab034a | 286 | struct lttng_ht_node_u64 *node = lttng::utils::container_of(head, <tng_ht_node_u64::head); |
ffe60014 | 287 | struct lttng_consumer_channel *channel = |
0114db0e | 288 | lttng::utils::container_of(node, <tng_consumer_channel::node); |
702b1ea4 | 289 | |
fa29bfbf | 290 | switch (the_consumer_data.type) { |
b83e03c4 MD |
291 | case LTTNG_CONSUMER_KERNEL: |
292 | break; | |
293 | case LTTNG_CONSUMER32_UST: | |
294 | case LTTNG_CONSUMER64_UST: | |
295 | lttng_ustconsumer_free_channel(channel); | |
296 | break; | |
297 | default: | |
298 | ERR("Unknown consumer_data type"); | |
299 | abort(); | |
300 | } | |
32670d71 JG |
301 | |
302 | delete channel; | |
702b1ea4 MD |
303 | } |
304 | ||
00e2e675 DG |
305 | /* |
306 | * RCU protected relayd socket pair free. | |
307 | */ | |
ffe60014 | 308 | static void free_relayd_rcu(struct rcu_head *head) |
00e2e675 | 309 | { |
28ab034a | 310 | struct lttng_ht_node_u64 *node = lttng::utils::container_of(head, <tng_ht_node_u64::head); |
00e2e675 | 311 | struct consumer_relayd_sock_pair *relayd = |
0114db0e | 312 | lttng::utils::container_of(node, &consumer_relayd_sock_pair::node); |
00e2e675 | 313 | |
8994307f DG |
314 | /* |
315 | * Close all sockets. This is done in the call RCU since we don't want the | |
316 | * socket fds to be reassigned thus potentially creating bad state of the | |
317 | * relayd object. | |
318 | * | |
319 | * We do not have to lock the control socket mutex here since at this stage | |
320 | * there is no one referencing to this relayd object. | |
321 | */ | |
322 | (void) relayd_close(&relayd->control_sock); | |
323 | (void) relayd_close(&relayd->data_sock); | |
324 | ||
3a84e2f3 | 325 | pthread_mutex_destroy(&relayd->ctrl_sock_mutex); |
00e2e675 DG |
326 | free(relayd); |
327 | } | |
328 | ||
329 | /* | |
330 | * Destroy and free relayd socket pair object. | |
00e2e675 | 331 | */ |
51230d70 | 332 | void consumer_destroy_relayd(struct consumer_relayd_sock_pair *relayd) |
00e2e675 DG |
333 | { |
334 | int ret; | |
335 | struct lttng_ht_iter iter; | |
336 | ||
cd9adb8b | 337 | if (relayd == nullptr) { |
173af62f DG |
338 | return; |
339 | } | |
340 | ||
00e2e675 DG |
341 | DBG("Consumer destroy and close relayd socket pair"); |
342 | ||
343 | iter.iter.node = &relayd->node.node; | |
fa29bfbf | 344 | ret = lttng_ht_del(the_consumer_data.relayd_ht, &iter); |
173af62f | 345 | if (ret != 0) { |
8994307f | 346 | /* We assume the relayd is being or is destroyed */ |
173af62f DG |
347 | return; |
348 | } | |
00e2e675 | 349 | |
00e2e675 | 350 | /* RCU free() call */ |
ffe60014 DG |
351 | call_rcu(&relayd->node.head, free_relayd_rcu); |
352 | } | |
353 | ||
354 | /* | |
355 | * Remove a channel from the global list protected by a mutex. This function is | |
356 | * also responsible for freeing its data structures. | |
357 | */ | |
358 | void consumer_del_channel(struct lttng_consumer_channel *channel) | |
359 | { | |
ffe60014 DG |
360 | struct lttng_ht_iter iter; |
361 | ||
d88aee68 | 362 | DBG("Consumer delete channel key %" PRIu64, channel->key); |
ffe60014 | 363 | |
fa29bfbf | 364 | pthread_mutex_lock(&the_consumer_data.lock); |
a9838785 | 365 | pthread_mutex_lock(&channel->lock); |
ffe60014 | 366 | |
212d67a2 DG |
367 | /* Destroy streams that might have been left in the stream list. */ |
368 | clean_channel_stream_list(channel); | |
51e762e5 | 369 | |
d3e2ba59 JD |
370 | if (channel->live_timer_enabled == 1) { |
371 | consumer_timer_live_stop(channel); | |
372 | } | |
e9404c27 JG |
373 | if (channel->monitor_timer_enabled == 1) { |
374 | consumer_timer_monitor_stop(channel); | |
375 | } | |
d3e2ba59 | 376 | |
319dcddc JG |
377 | /* |
378 | * Send a last buffer statistics sample to the session daemon | |
379 | * to ensure it tracks the amount of data consumed by this channel. | |
380 | */ | |
381 | sample_and_send_channel_buffer_stats(channel); | |
382 | ||
fa29bfbf | 383 | switch (the_consumer_data.type) { |
ffe60014 DG |
384 | case LTTNG_CONSUMER_KERNEL: |
385 | break; | |
386 | case LTTNG_CONSUMER32_UST: | |
387 | case LTTNG_CONSUMER64_UST: | |
388 | lttng_ustconsumer_del_channel(channel); | |
389 | break; | |
390 | default: | |
391 | ERR("Unknown consumer_data type"); | |
a0377dfe | 392 | abort(); |
ffe60014 DG |
393 | goto end; |
394 | } | |
395 | ||
d2956687 | 396 | lttng_trace_chunk_put(channel->trace_chunk); |
cd9adb8b | 397 | channel->trace_chunk = nullptr; |
5c3892a6 | 398 | |
d2956687 JG |
399 | if (channel->is_published) { |
400 | int ret; | |
401 | ||
07c4863f | 402 | const lttng::urcu::read_lock_guard read_lock; |
d2956687 | 403 | iter.iter.node = &channel->node.node; |
fa29bfbf | 404 | ret = lttng_ht_del(the_consumer_data.channel_ht, &iter); |
a0377dfe | 405 | LTTNG_ASSERT(!ret); |
ffe60014 | 406 | |
d2956687 | 407 | iter.iter.node = &channel->channels_by_session_id_ht_node.node; |
28ab034a | 408 | ret = lttng_ht_del(the_consumer_data.channels_by_session_id_ht, &iter); |
a0377dfe | 409 | LTTNG_ASSERT(!ret); |
d2956687 JG |
410 | } |
411 | ||
b6921a17 JG |
412 | channel->is_deleted = true; |
413 | call_rcu(&channel->node.head, free_channel_rcu); | |
ffe60014 | 414 | end: |
a9838785 | 415 | pthread_mutex_unlock(&channel->lock); |
fa29bfbf | 416 | pthread_mutex_unlock(&the_consumer_data.lock); |
00e2e675 DG |
417 | } |
418 | ||
228b5bf7 DG |
419 | /* |
420 | * Iterate over the relayd hash table and destroy each element. Finally, | |
421 | * destroy the whole hash table. | |
422 | */ | |
cd9adb8b | 423 | static void cleanup_relayd_ht() |
228b5bf7 | 424 | { |
c3ade133 JG |
425 | for (auto *relayd : |
426 | lttng::urcu::lfht_iteration_adapter<consumer_relayd_sock_pair, | |
427 | decltype(consumer_relayd_sock_pair::node), | |
428 | &consumer_relayd_sock_pair::node>( | |
429 | *the_consumer_data.relayd_ht->ht)) { | |
430 | consumer_destroy_relayd(relayd); | |
228b5bf7 DG |
431 | } |
432 | ||
fa29bfbf | 433 | lttng_ht_destroy(the_consumer_data.relayd_ht); |
228b5bf7 DG |
434 | } |
435 | ||
8994307f DG |
436 | /* |
437 | * Update the end point status of all streams having the given network sequence | |
438 | * index (relayd index). | |
439 | * | |
440 | * It's atomically set without having the stream mutex locked which is fine | |
441 | * because we handle the write/read race with a pipe wakeup for each thread. | |
442 | */ | |
da009f2c | 443 | static void update_endpoint_status_by_netidx(uint64_t net_seq_idx, |
28ab034a | 444 | enum consumer_endpoint_status status) |
8994307f | 445 | { |
da009f2c | 446 | DBG("Consumer set delete flag on stream by idx %" PRIu64, net_seq_idx); |
8994307f | 447 | |
8994307f | 448 | /* Let's begin with metadata */ |
c3ade133 JG |
449 | for (auto *stream : |
450 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
451 | decltype(lttng_consumer_stream::node), | |
452 | <tng_consumer_stream::node>(*metadata_ht->ht)) { | |
8994307f DG |
453 | if (stream->net_seq_idx == net_seq_idx) { |
454 | uatomic_set(&stream->endpoint_status, status); | |
32670d71 | 455 | stream->chan->metadata_pushed_wait_queue.wake_all(); |
f40b76ae | 456 | |
8994307f DG |
457 | DBG("Delete flag set to metadata stream %d", stream->wait_fd); |
458 | } | |
459 | } | |
460 | ||
461 | /* Follow up by the data streams */ | |
c3ade133 JG |
462 | for (auto *stream : |
463 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
464 | decltype(lttng_consumer_stream::node), | |
465 | <tng_consumer_stream::node>(*data_ht->ht)) { | |
8994307f DG |
466 | if (stream->net_seq_idx == net_seq_idx) { |
467 | uatomic_set(&stream->endpoint_status, status); | |
468 | DBG("Delete flag set to data stream %d", stream->wait_fd); | |
469 | } | |
470 | } | |
8994307f DG |
471 | } |
472 | ||
473 | /* | |
474 | * Cleanup a relayd object by flagging every associated streams for deletion, | |
475 | * destroying the object meaning removing it from the relayd hash table, | |
476 | * closing the sockets and freeing the memory in a RCU call. | |
477 | * | |
478 | * If a local data context is available, notify the threads that the streams' | |
479 | * state have changed. | |
480 | */ | |
9276e5c8 | 481 | void lttng_consumer_cleanup_relayd(struct consumer_relayd_sock_pair *relayd) |
8994307f | 482 | { |
da009f2c | 483 | uint64_t netidx; |
8994307f | 484 | |
a0377dfe | 485 | LTTNG_ASSERT(relayd); |
8994307f | 486 | |
97535efa | 487 | DBG("Cleaning up relayd object ID %" PRIu64, relayd->net_seq_idx); |
9617607b | 488 | |
8994307f DG |
489 | /* Save the net sequence index before destroying the object */ |
490 | netidx = relayd->net_seq_idx; | |
491 | ||
492 | /* | |
493 | * Delete the relayd from the relayd hash table, close the sockets and free | |
494 | * the object in a RCU call. | |
495 | */ | |
51230d70 | 496 | consumer_destroy_relayd(relayd); |
8994307f DG |
497 | |
498 | /* Set inactive endpoint to all streams */ | |
499 | update_endpoint_status_by_netidx(netidx, CONSUMER_ENDPOINT_INACTIVE); | |
500 | ||
501 | /* | |
502 | * With a local data context, notify the threads that the streams' state | |
503 | * have changed. The write() action on the pipe acts as an "implicit" | |
504 | * memory barrier ordering the updates of the end point status from the | |
505 | * read of this status which happens AFTER receiving this notify. | |
506 | */ | |
9276e5c8 JR |
507 | notify_thread_lttng_pipe(relayd->ctx->consumer_data_pipe); |
508 | notify_thread_lttng_pipe(relayd->ctx->consumer_metadata_pipe); | |
8994307f DG |
509 | } |
510 | ||
a6ba4fe1 DG |
511 | /* |
512 | * Flag a relayd socket pair for destruction. Destroy it if the refcount | |
513 | * reaches zero. | |
514 | * | |
515 | * RCU read side lock MUST be aquired before calling this function. | |
516 | */ | |
517 | void consumer_flag_relayd_for_destroy(struct consumer_relayd_sock_pair *relayd) | |
518 | { | |
a0377dfe | 519 | LTTNG_ASSERT(relayd); |
48b7cdc2 | 520 | ASSERT_RCU_READ_LOCKED(); |
a6ba4fe1 DG |
521 | |
522 | /* Set destroy flag for this object */ | |
523 | uatomic_set(&relayd->destroy_flag, 1); | |
524 | ||
525 | /* Destroy the relayd if refcount is 0 */ | |
526 | if (uatomic_read(&relayd->refcount) == 0) { | |
51230d70 | 527 | consumer_destroy_relayd(relayd); |
a6ba4fe1 DG |
528 | } |
529 | } | |
530 | ||
3bd1e081 | 531 | /* |
1d1a276c DG |
532 | * Completly destroy stream from every visiable data structure and the given |
533 | * hash table if one. | |
534 | * | |
535 | * One this call returns, the stream object is not longer usable nor visible. | |
3bd1e081 | 536 | */ |
28ab034a | 537 | void consumer_del_stream(struct lttng_consumer_stream *stream, struct lttng_ht *ht) |
3bd1e081 | 538 | { |
1d1a276c | 539 | consumer_stream_destroy(stream, ht); |
3bd1e081 MD |
540 | } |
541 | ||
5ab66908 MD |
542 | /* |
543 | * XXX naming of del vs destroy is all mixed up. | |
544 | */ | |
545 | void consumer_del_stream_for_data(struct lttng_consumer_stream *stream) | |
546 | { | |
547 | consumer_stream_destroy(stream, data_ht); | |
548 | } | |
549 | ||
550 | void consumer_del_stream_for_metadata(struct lttng_consumer_stream *stream) | |
551 | { | |
552 | consumer_stream_destroy(stream, metadata_ht); | |
553 | } | |
554 | ||
28ab034a JG |
555 | void consumer_stream_update_channel_attributes(struct lttng_consumer_stream *stream, |
556 | struct lttng_consumer_channel *channel) | |
d9a2e16e | 557 | { |
28ab034a | 558 | stream->channel_read_only_attributes.tracefile_size = channel->tracefile_size; |
d9a2e16e JD |
559 | } |
560 | ||
3bd1e081 MD |
561 | /* |
562 | * Add a stream to the global list protected by a mutex. | |
563 | */ | |
66d583dc | 564 | void consumer_add_data_stream(struct lttng_consumer_stream *stream) |
3bd1e081 | 565 | { |
5ab66908 | 566 | struct lttng_ht *ht = data_ht; |
3bd1e081 | 567 | |
a0377dfe FD |
568 | LTTNG_ASSERT(stream); |
569 | LTTNG_ASSERT(ht); | |
c77fc10a | 570 | |
d88aee68 | 571 | DBG3("Adding consumer stream %" PRIu64, stream->key); |
e316aad5 | 572 | |
fa29bfbf | 573 | pthread_mutex_lock(&the_consumer_data.lock); |
a9838785 | 574 | pthread_mutex_lock(&stream->chan->lock); |
ec6ea7d0 | 575 | pthread_mutex_lock(&stream->chan->timer_lock); |
2e818a6a | 576 | pthread_mutex_lock(&stream->lock); |
07c4863f | 577 | const lttng::urcu::read_lock_guard read_lock; |
e316aad5 | 578 | |
43c34bc3 | 579 | /* Steal stream identifier to avoid having streams with the same key */ |
ffe60014 | 580 | steal_stream_key(stream->key, ht); |
43c34bc3 | 581 | |
d88aee68 | 582 | lttng_ht_add_unique_u64(ht, &stream->node); |
00e2e675 | 583 | |
28ab034a | 584 | lttng_ht_add_u64(the_consumer_data.stream_per_chan_id_ht, &stream->node_channel_id); |
d8ef542d | 585 | |
ca22feea DG |
586 | /* |
587 | * Add stream to the stream_list_ht of the consumer data. No need to steal | |
588 | * the key since the HT does not use it and we allow to add redundant keys | |
589 | * into this table. | |
590 | */ | |
28ab034a | 591 | lttng_ht_add_u64(the_consumer_data.stream_list_ht, &stream->node_session_id); |
ca22feea | 592 | |
e316aad5 | 593 | /* |
ffe60014 DG |
594 | * When nb_init_stream_left reaches 0, we don't need to trigger any action |
595 | * in terms of destroying the associated channel, because the action that | |
e316aad5 DG |
596 | * causes the count to become 0 also causes a stream to be added. The |
597 | * channel deletion will thus be triggered by the following removal of this | |
598 | * stream. | |
599 | */ | |
ffe60014 | 600 | if (uatomic_read(&stream->chan->nb_init_stream_left) > 0) { |
f2ad556d MD |
601 | /* Increment refcount before decrementing nb_init_stream_left */ |
602 | cmm_smp_wmb(); | |
ffe60014 | 603 | uatomic_dec(&stream->chan->nb_init_stream_left); |
e316aad5 DG |
604 | } |
605 | ||
606 | /* Update consumer data once the node is inserted. */ | |
fa29bfbf SM |
607 | the_consumer_data.stream_count++; |
608 | the_consumer_data.need_update = 1; | |
3bd1e081 | 609 | |
2e818a6a | 610 | pthread_mutex_unlock(&stream->lock); |
ec6ea7d0 | 611 | pthread_mutex_unlock(&stream->chan->timer_lock); |
a9838785 | 612 | pthread_mutex_unlock(&stream->chan->lock); |
fa29bfbf | 613 | pthread_mutex_unlock(&the_consumer_data.lock); |
3bd1e081 MD |
614 | } |
615 | ||
00e2e675 | 616 | /* |
3f8e211f DG |
617 | * Add relayd socket to global consumer data hashtable. RCU read side lock MUST |
618 | * be acquired before calling this. | |
00e2e675 | 619 | */ |
d09e1200 | 620 | static int add_relayd(struct consumer_relayd_sock_pair *relayd) |
00e2e675 | 621 | { |
07c4863f | 622 | const int ret = 0; |
d88aee68 | 623 | struct lttng_ht_node_u64 *node; |
00e2e675 DG |
624 | struct lttng_ht_iter iter; |
625 | ||
a0377dfe | 626 | LTTNG_ASSERT(relayd); |
48b7cdc2 | 627 | ASSERT_RCU_READ_LOCKED(); |
00e2e675 | 628 | |
28ab034a | 629 | lttng_ht_lookup(the_consumer_data.relayd_ht, &relayd->net_seq_idx, &iter); |
00d7d903 | 630 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
cd9adb8b | 631 | if (node != nullptr) { |
00e2e675 DG |
632 | goto end; |
633 | } | |
fa29bfbf | 634 | lttng_ht_add_unique_u64(the_consumer_data.relayd_ht, &relayd->node); |
00e2e675 | 635 | |
00e2e675 DG |
636 | end: |
637 | return ret; | |
638 | } | |
639 | ||
640 | /* | |
641 | * Allocate and return a consumer relayd socket. | |
642 | */ | |
28ab034a | 643 | static struct consumer_relayd_sock_pair *consumer_allocate_relayd_sock_pair(uint64_t net_seq_idx) |
00e2e675 | 644 | { |
cd9adb8b | 645 | struct consumer_relayd_sock_pair *obj = nullptr; |
00e2e675 | 646 | |
da009f2c MD |
647 | /* net sequence index of -1 is a failure */ |
648 | if (net_seq_idx == (uint64_t) -1ULL) { | |
00e2e675 DG |
649 | goto error; |
650 | } | |
651 | ||
64803277 | 652 | obj = zmalloc<consumer_relayd_sock_pair>(); |
cd9adb8b | 653 | if (obj == nullptr) { |
00e2e675 DG |
654 | PERROR("zmalloc relayd sock"); |
655 | goto error; | |
656 | } | |
657 | ||
658 | obj->net_seq_idx = net_seq_idx; | |
659 | obj->refcount = 0; | |
173af62f | 660 | obj->destroy_flag = 0; |
f96e4545 MD |
661 | obj->control_sock.sock.fd = -1; |
662 | obj->data_sock.sock.fd = -1; | |
d88aee68 | 663 | lttng_ht_node_init_u64(&obj->node, obj->net_seq_idx); |
cd9adb8b | 664 | pthread_mutex_init(&obj->ctrl_sock_mutex, nullptr); |
00e2e675 DG |
665 | |
666 | error: | |
667 | return obj; | |
668 | } | |
669 | ||
670 | /* | |
671 | * Find a relayd socket pair in the global consumer data. | |
672 | * | |
673 | * Return the object if found else NULL. | |
b0b335c8 MD |
674 | * RCU read-side lock must be held across this call and while using the |
675 | * returned object. | |
00e2e675 | 676 | */ |
d88aee68 | 677 | struct consumer_relayd_sock_pair *consumer_find_relayd(uint64_t key) |
00e2e675 DG |
678 | { |
679 | struct lttng_ht_iter iter; | |
d88aee68 | 680 | struct lttng_ht_node_u64 *node; |
cd9adb8b | 681 | struct consumer_relayd_sock_pair *relayd = nullptr; |
00e2e675 | 682 | |
48b7cdc2 FD |
683 | ASSERT_RCU_READ_LOCKED(); |
684 | ||
00e2e675 | 685 | /* Negative keys are lookup failures */ |
d88aee68 | 686 | if (key == (uint64_t) -1ULL) { |
00e2e675 DG |
687 | goto error; |
688 | } | |
689 | ||
fa29bfbf | 690 | lttng_ht_lookup(the_consumer_data.relayd_ht, &key, &iter); |
00d7d903 | 691 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
cd9adb8b | 692 | if (node != nullptr) { |
0114db0e | 693 | relayd = lttng::utils::container_of(node, &consumer_relayd_sock_pair::node); |
00e2e675 DG |
694 | } |
695 | ||
00e2e675 DG |
696 | error: |
697 | return relayd; | |
698 | } | |
699 | ||
10a50311 JD |
700 | /* |
701 | * Find a relayd and send the stream | |
702 | * | |
703 | * Returns 0 on success, < 0 on error | |
704 | */ | |
28ab034a | 705 | int consumer_send_relayd_stream(struct lttng_consumer_stream *stream, char *path) |
10a50311 JD |
706 | { |
707 | int ret = 0; | |
708 | struct consumer_relayd_sock_pair *relayd; | |
709 | ||
a0377dfe FD |
710 | LTTNG_ASSERT(stream); |
711 | LTTNG_ASSERT(stream->net_seq_idx != -1ULL); | |
712 | LTTNG_ASSERT(path); | |
10a50311 JD |
713 | |
714 | /* The stream is not metadata. Get relayd reference if exists. */ | |
07c4863f | 715 | const lttng::urcu::read_lock_guard read_lock; |
10a50311 | 716 | relayd = consumer_find_relayd(stream->net_seq_idx); |
cd9adb8b | 717 | if (relayd != nullptr) { |
10a50311 JD |
718 | /* Add stream on the relayd */ |
719 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
28ab034a JG |
720 | ret = relayd_add_stream(&relayd->control_sock, |
721 | stream->name, | |
722 | get_consumer_domain(), | |
723 | path, | |
724 | &stream->relayd_stream_id, | |
725 | stream->chan->tracefile_size, | |
726 | stream->chan->tracefile_count, | |
727 | stream->trace_chunk); | |
10a50311 JD |
728 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); |
729 | if (ret < 0) { | |
28ab034a JG |
730 | ERR("Relayd add stream failed. Cleaning up relayd %" PRIu64 ".", |
731 | relayd->net_seq_idx); | |
9276e5c8 | 732 | lttng_consumer_cleanup_relayd(relayd); |
10a50311 JD |
733 | goto end; |
734 | } | |
1c20f0e2 | 735 | |
10a50311 | 736 | uatomic_inc(&relayd->refcount); |
d01178b6 | 737 | stream->sent_to_relayd = 1; |
10a50311 JD |
738 | } else { |
739 | ERR("Stream %" PRIu64 " relayd ID %" PRIu64 " unknown. Can't send it.", | |
28ab034a JG |
740 | stream->key, |
741 | stream->net_seq_idx); | |
10a50311 JD |
742 | ret = -1; |
743 | goto end; | |
744 | } | |
745 | ||
746 | DBG("Stream %s with key %" PRIu64 " sent to relayd id %" PRIu64, | |
28ab034a JG |
747 | stream->name, |
748 | stream->key, | |
749 | stream->net_seq_idx); | |
10a50311 JD |
750 | |
751 | end: | |
10a50311 JD |
752 | return ret; |
753 | } | |
754 | ||
a4baae1b JD |
755 | /* |
756 | * Find a relayd and send the streams sent message | |
757 | * | |
758 | * Returns 0 on success, < 0 on error | |
759 | */ | |
760 | int consumer_send_relayd_streams_sent(uint64_t net_seq_idx) | |
761 | { | |
762 | int ret = 0; | |
763 | struct consumer_relayd_sock_pair *relayd; | |
764 | ||
a0377dfe | 765 | LTTNG_ASSERT(net_seq_idx != -1ULL); |
a4baae1b JD |
766 | |
767 | /* The stream is not metadata. Get relayd reference if exists. */ | |
07c4863f | 768 | const lttng::urcu::read_lock_guard read_lock; |
a4baae1b | 769 | relayd = consumer_find_relayd(net_seq_idx); |
cd9adb8b | 770 | if (relayd != nullptr) { |
a4baae1b JD |
771 | /* Add stream on the relayd */ |
772 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
773 | ret = relayd_streams_sent(&relayd->control_sock); | |
774 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); | |
775 | if (ret < 0) { | |
28ab034a JG |
776 | ERR("Relayd streams sent failed. Cleaning up relayd %" PRIu64 ".", |
777 | relayd->net_seq_idx); | |
9276e5c8 | 778 | lttng_consumer_cleanup_relayd(relayd); |
a4baae1b JD |
779 | goto end; |
780 | } | |
781 | } else { | |
28ab034a | 782 | ERR("Relayd ID %" PRIu64 " unknown. Can't send streams_sent.", net_seq_idx); |
a4baae1b JD |
783 | ret = -1; |
784 | goto end; | |
785 | } | |
786 | ||
787 | ret = 0; | |
788 | DBG("All streams sent relayd id %" PRIu64, net_seq_idx); | |
789 | ||
790 | end: | |
a4baae1b JD |
791 | return ret; |
792 | } | |
793 | ||
10a50311 JD |
794 | /* |
795 | * Find a relayd and close the stream | |
796 | */ | |
797 | void close_relayd_stream(struct lttng_consumer_stream *stream) | |
798 | { | |
799 | struct consumer_relayd_sock_pair *relayd; | |
800 | ||
801 | /* The stream is not metadata. Get relayd reference if exists. */ | |
07c4863f | 802 | const lttng::urcu::read_lock_guard read_lock; |
10a50311 JD |
803 | relayd = consumer_find_relayd(stream->net_seq_idx); |
804 | if (relayd) { | |
805 | consumer_stream_relayd_close(stream, relayd); | |
806 | } | |
10a50311 JD |
807 | } |
808 | ||
00e2e675 DG |
809 | /* |
810 | * Handle stream for relayd transmission if the stream applies for network | |
811 | * streaming where the net sequence index is set. | |
812 | * | |
813 | * Return destination file descriptor or negative value on error. | |
814 | */ | |
6197aea7 | 815 | static int write_relayd_stream_header(struct lttng_consumer_stream *stream, |
28ab034a JG |
816 | size_t data_size, |
817 | unsigned long padding, | |
818 | struct consumer_relayd_sock_pair *relayd) | |
00e2e675 DG |
819 | { |
820 | int outfd = -1, ret; | |
00e2e675 DG |
821 | struct lttcomm_relayd_data_hdr data_hdr; |
822 | ||
823 | /* Safety net */ | |
a0377dfe FD |
824 | LTTNG_ASSERT(stream); |
825 | LTTNG_ASSERT(relayd); | |
00e2e675 DG |
826 | |
827 | /* Reset data header */ | |
828 | memset(&data_hdr, 0, sizeof(data_hdr)); | |
829 | ||
00e2e675 DG |
830 | if (stream->metadata_flag) { |
831 | /* Caller MUST acquire the relayd control socket lock */ | |
832 | ret = relayd_send_metadata(&relayd->control_sock, data_size); | |
833 | if (ret < 0) { | |
834 | goto error; | |
835 | } | |
836 | ||
837 | /* Metadata are always sent on the control socket. */ | |
6151a90f | 838 | outfd = relayd->control_sock.sock.fd; |
00e2e675 DG |
839 | } else { |
840 | /* Set header with stream information */ | |
841 | data_hdr.stream_id = htobe64(stream->relayd_stream_id); | |
842 | data_hdr.data_size = htobe32(data_size); | |
1d4dfdef | 843 | data_hdr.padding_size = htobe32(padding); |
c35f9726 | 844 | |
39df6d9f DG |
845 | /* |
846 | * Note that net_seq_num below is assigned with the *current* value of | |
847 | * next_net_seq_num and only after that the next_net_seq_num will be | |
848 | * increment. This is why when issuing a command on the relayd using | |
849 | * this next value, 1 should always be substracted in order to compare | |
850 | * the last seen sequence number on the relayd side to the last sent. | |
851 | */ | |
3604f373 | 852 | data_hdr.net_seq_num = htobe64(stream->next_net_seq_num); |
00e2e675 DG |
853 | /* Other fields are zeroed previously */ |
854 | ||
28ab034a | 855 | ret = relayd_send_data_hdr(&relayd->data_sock, &data_hdr, sizeof(data_hdr)); |
00e2e675 DG |
856 | if (ret < 0) { |
857 | goto error; | |
858 | } | |
859 | ||
3604f373 DG |
860 | ++stream->next_net_seq_num; |
861 | ||
00e2e675 | 862 | /* Set to go on data socket */ |
6151a90f | 863 | outfd = relayd->data_sock.sock.fd; |
00e2e675 DG |
864 | } |
865 | ||
866 | error: | |
867 | return outfd; | |
868 | } | |
869 | ||
b1316da1 JG |
870 | /* |
871 | * Write a character on the metadata poll pipe to wake the metadata thread. | |
872 | * Returns 0 on success, -1 on error. | |
873 | */ | |
874 | int consumer_metadata_wakeup_pipe(const struct lttng_consumer_channel *channel) | |
875 | { | |
876 | int ret = 0; | |
877 | ||
28ab034a | 878 | DBG("Waking up metadata poll thread (writing to pipe): channel name = '%s'", channel->name); |
b1316da1 JG |
879 | if (channel->monitor && channel->metadata_stream) { |
880 | const char dummy = 'c'; | |
28ab034a JG |
881 | const ssize_t write_ret = |
882 | lttng_write(channel->metadata_stream->ust_metadata_poll_pipe[1], &dummy, 1); | |
b1316da1 JG |
883 | |
884 | if (write_ret < 1) { | |
885 | if (errno == EWOULDBLOCK) { | |
886 | /* | |
887 | * This is fine, the metadata poll thread | |
888 | * is having a hard time keeping-up, but | |
889 | * it will eventually wake-up and consume | |
890 | * the available data. | |
891 | */ | |
892 | ret = 0; | |
893 | } else { | |
894 | PERROR("Failed to write to UST metadata pipe while attempting to wake-up the metadata poll thread"); | |
895 | ret = -1; | |
896 | goto end; | |
897 | } | |
898 | } | |
899 | } | |
900 | ||
901 | end: | |
902 | return ret; | |
903 | } | |
904 | ||
d2956687 JG |
905 | /* |
906 | * Trigger a dump of the metadata content. Following/during the succesful | |
907 | * completion of this call, the metadata poll thread will start receiving | |
908 | * metadata packets to consume. | |
909 | * | |
910 | * The caller must hold the channel and stream locks. | |
911 | */ | |
28ab034a | 912 | static int consumer_metadata_stream_dump(struct lttng_consumer_stream *stream) |
d2956687 JG |
913 | { |
914 | int ret; | |
915 | ||
916 | ASSERT_LOCKED(stream->chan->lock); | |
917 | ASSERT_LOCKED(stream->lock); | |
a0377dfe FD |
918 | LTTNG_ASSERT(stream->metadata_flag); |
919 | LTTNG_ASSERT(stream->chan->trace_chunk); | |
d2956687 | 920 | |
fa29bfbf | 921 | switch (the_consumer_data.type) { |
d2956687 JG |
922 | case LTTNG_CONSUMER_KERNEL: |
923 | /* | |
924 | * Reset the position of what has been read from the | |
925 | * metadata cache to 0 so we can dump it again. | |
926 | */ | |
927 | ret = kernctl_metadata_cache_dump(stream->wait_fd); | |
928 | break; | |
929 | case LTTNG_CONSUMER32_UST: | |
930 | case LTTNG_CONSUMER64_UST: | |
931 | /* | |
932 | * Reset the position pushed from the metadata cache so it | |
933 | * will write from the beginning on the next push. | |
934 | */ | |
935 | stream->ust_metadata_pushed = 0; | |
936 | ret = consumer_metadata_wakeup_pipe(stream->chan); | |
937 | break; | |
938 | default: | |
939 | ERR("Unknown consumer_data type"); | |
940 | abort(); | |
941 | } | |
942 | if (ret < 0) { | |
943 | ERR("Failed to dump the metadata cache"); | |
944 | } | |
945 | return ret; | |
946 | } | |
947 | ||
28ab034a JG |
948 | static int lttng_consumer_channel_set_trace_chunk(struct lttng_consumer_channel *channel, |
949 | struct lttng_trace_chunk *new_trace_chunk) | |
d2956687 | 950 | { |
d2956687 | 951 | pthread_mutex_lock(&channel->lock); |
b6921a17 JG |
952 | if (channel->is_deleted) { |
953 | /* | |
954 | * The channel has been logically deleted and should no longer | |
955 | * be used. It has released its reference to its current trace | |
956 | * chunk and should not acquire a new one. | |
957 | * | |
958 | * Return success as there is nothing for the caller to do. | |
959 | */ | |
960 | goto end; | |
961 | } | |
d2956687 JG |
962 | |
963 | /* | |
964 | * The acquisition of the reference cannot fail (barring | |
965 | * a severe internal error) since a reference to the published | |
966 | * chunk is already held by the caller. | |
967 | */ | |
968 | if (new_trace_chunk) { | |
28ab034a | 969 | const bool acquired_reference = lttng_trace_chunk_get(new_trace_chunk); |
d2956687 | 970 | |
a0377dfe | 971 | LTTNG_ASSERT(acquired_reference); |
d2956687 JG |
972 | } |
973 | ||
974 | lttng_trace_chunk_put(channel->trace_chunk); | |
975 | channel->trace_chunk = new_trace_chunk; | |
d2956687 JG |
976 | end: |
977 | pthread_mutex_unlock(&channel->lock); | |
ce1aa6fe | 978 | return 0; |
d2956687 JG |
979 | } |
980 | ||
3bd1e081 | 981 | /* |
ffe60014 DG |
982 | * Allocate and return a new lttng_consumer_channel object using the given key |
983 | * to initialize the hash table node. | |
984 | * | |
985 | * On error, return NULL. | |
3bd1e081 | 986 | */ |
886224ff | 987 | struct lttng_consumer_channel *consumer_allocate_channel(uint64_t key, |
28ab034a JG |
988 | uint64_t session_id, |
989 | const uint64_t *chunk_id, | |
990 | const char *pathname, | |
991 | const char *name, | |
992 | uint64_t relayd_id, | |
993 | enum lttng_event_output output, | |
994 | uint64_t tracefile_size, | |
995 | uint64_t tracefile_count, | |
996 | uint64_t session_id_per_pid, | |
997 | unsigned int monitor, | |
998 | unsigned int live_timer_interval, | |
999 | bool is_in_live_session, | |
1000 | const char *root_shm_path, | |
1001 | const char *shm_path) | |
3bd1e081 | 1002 | { |
cd9adb8b JG |
1003 | struct lttng_consumer_channel *channel = nullptr; |
1004 | struct lttng_trace_chunk *trace_chunk = nullptr; | |
d2956687 JG |
1005 | |
1006 | if (chunk_id) { | |
1007 | trace_chunk = lttng_trace_chunk_registry_find_chunk( | |
28ab034a | 1008 | the_consumer_data.chunk_registry, session_id, *chunk_id); |
d2956687 JG |
1009 | if (!trace_chunk) { |
1010 | ERR("Failed to find trace chunk reference during creation of channel"); | |
1011 | goto end; | |
1012 | } | |
1013 | } | |
3bd1e081 | 1014 | |
32670d71 JG |
1015 | try { |
1016 | channel = new lttng_consumer_channel; | |
1017 | } catch (const std::bad_alloc& e) { | |
1018 | ERR("Failed to allocate lttng_consumer_channel: %s", e.what()); | |
1019 | channel = nullptr; | |
3bd1e081 MD |
1020 | goto end; |
1021 | } | |
ffe60014 DG |
1022 | |
1023 | channel->key = key; | |
3bd1e081 | 1024 | channel->refcount = 0; |
ffe60014 | 1025 | channel->session_id = session_id; |
1950109e | 1026 | channel->session_id_per_pid = session_id_per_pid; |
ffe60014 | 1027 | channel->relayd_id = relayd_id; |
1624d5b7 JD |
1028 | channel->tracefile_size = tracefile_size; |
1029 | channel->tracefile_count = tracefile_count; | |
2bba9e53 | 1030 | channel->monitor = monitor; |
ecc48a90 | 1031 | channel->live_timer_interval = live_timer_interval; |
a2814ea7 | 1032 | channel->is_live = is_in_live_session; |
07c4863f JG |
1033 | pthread_mutex_init(&channel->lock, nullptr); |
1034 | pthread_mutex_init(&channel->timer_lock, nullptr); | |
ffe60014 | 1035 | |
0c759fc9 DG |
1036 | switch (output) { |
1037 | case LTTNG_EVENT_SPLICE: | |
1038 | channel->output = CONSUMER_CHANNEL_SPLICE; | |
1039 | break; | |
1040 | case LTTNG_EVENT_MMAP: | |
1041 | channel->output = CONSUMER_CHANNEL_MMAP; | |
1042 | break; | |
1043 | default: | |
a0377dfe | 1044 | abort(); |
32670d71 | 1045 | delete channel; |
cd9adb8b | 1046 | channel = nullptr; |
0c759fc9 DG |
1047 | goto end; |
1048 | } | |
1049 | ||
07b86b52 JD |
1050 | /* |
1051 | * In monitor mode, the streams associated with the channel will be put in | |
1052 | * a special list ONLY owned by this channel. So, the refcount is set to 1 | |
1053 | * here meaning that the channel itself has streams that are referenced. | |
1054 | * | |
1055 | * On a channel deletion, once the channel is no longer visible, the | |
1056 | * refcount is decremented and checked for a zero value to delete it. With | |
1057 | * streams in no monitor mode, it will now be safe to destroy the channel. | |
1058 | */ | |
1059 | if (!channel->monitor) { | |
1060 | channel->refcount = 1; | |
1061 | } | |
1062 | ||
ffe60014 DG |
1063 | strncpy(channel->pathname, pathname, sizeof(channel->pathname)); |
1064 | channel->pathname[sizeof(channel->pathname) - 1] = '\0'; | |
1065 | ||
1066 | strncpy(channel->name, name, sizeof(channel->name)); | |
1067 | channel->name[sizeof(channel->name) - 1] = '\0'; | |
1068 | ||
3d071855 MD |
1069 | if (root_shm_path) { |
1070 | strncpy(channel->root_shm_path, root_shm_path, sizeof(channel->root_shm_path)); | |
1071 | channel->root_shm_path[sizeof(channel->root_shm_path) - 1] = '\0'; | |
1072 | } | |
d7ba1388 MD |
1073 | if (shm_path) { |
1074 | strncpy(channel->shm_path, shm_path, sizeof(channel->shm_path)); | |
1075 | channel->shm_path[sizeof(channel->shm_path) - 1] = '\0'; | |
1076 | } | |
1077 | ||
d88aee68 | 1078 | lttng_ht_node_init_u64(&channel->node, channel->key); |
28ab034a | 1079 | lttng_ht_node_init_u64(&channel->channels_by_session_id_ht_node, channel->session_id); |
d8ef542d MD |
1080 | |
1081 | channel->wait_fd = -1; | |
ffe60014 DG |
1082 | CDS_INIT_LIST_HEAD(&channel->streams.head); |
1083 | ||
d2956687 | 1084 | if (trace_chunk) { |
07c4863f | 1085 | const int ret = lttng_consumer_channel_set_trace_chunk(channel, trace_chunk); |
d2956687 JG |
1086 | if (ret) { |
1087 | goto error; | |
1088 | } | |
1089 | } | |
1090 | ||
62a7b8ed | 1091 | DBG("Allocated channel (key %" PRIu64 ")", channel->key); |
3bd1e081 | 1092 | |
3bd1e081 | 1093 | end: |
d2956687 | 1094 | lttng_trace_chunk_put(trace_chunk); |
3bd1e081 | 1095 | return channel; |
d2956687 JG |
1096 | error: |
1097 | consumer_del_channel(channel); | |
cd9adb8b | 1098 | channel = nullptr; |
d2956687 | 1099 | goto end; |
3bd1e081 MD |
1100 | } |
1101 | ||
1102 | /* | |
1103 | * Add a channel to the global list protected by a mutex. | |
821fffb2 | 1104 | * |
b5a6470f | 1105 | * Always return 0 indicating success. |
3bd1e081 | 1106 | */ |
d8ef542d | 1107 | int consumer_add_channel(struct lttng_consumer_channel *channel, |
28ab034a | 1108 | struct lttng_consumer_local_data *ctx) |
3bd1e081 | 1109 | { |
fa29bfbf | 1110 | pthread_mutex_lock(&the_consumer_data.lock); |
a9838785 | 1111 | pthread_mutex_lock(&channel->lock); |
ec6ea7d0 | 1112 | pthread_mutex_lock(&channel->timer_lock); |
c77fc10a | 1113 | |
b5a6470f DG |
1114 | /* |
1115 | * This gives us a guarantee that the channel we are about to add to the | |
1116 | * channel hash table will be unique. See this function comment on the why | |
1117 | * we need to steel the channel key at this stage. | |
1118 | */ | |
1119 | steal_channel_key(channel->key); | |
c77fc10a | 1120 | |
07c4863f | 1121 | const lttng::urcu::read_lock_guard read_lock; |
fa29bfbf SM |
1122 | lttng_ht_add_unique_u64(the_consumer_data.channel_ht, &channel->node); |
1123 | lttng_ht_add_u64(the_consumer_data.channels_by_session_id_ht, | |
28ab034a | 1124 | &channel->channels_by_session_id_ht_node); |
d2956687 | 1125 | channel->is_published = true; |
b5a6470f | 1126 | |
ec6ea7d0 | 1127 | pthread_mutex_unlock(&channel->timer_lock); |
a9838785 | 1128 | pthread_mutex_unlock(&channel->lock); |
fa29bfbf | 1129 | pthread_mutex_unlock(&the_consumer_data.lock); |
702b1ea4 | 1130 | |
b5a6470f | 1131 | if (channel->wait_fd != -1 && channel->type == CONSUMER_CHANNEL_TYPE_DATA) { |
a0cbdd2e | 1132 | notify_channel_pipe(ctx, channel, -1, CONSUMER_CHANNEL_ADD); |
d8ef542d | 1133 | } |
b5a6470f DG |
1134 | |
1135 | return 0; | |
3bd1e081 MD |
1136 | } |
1137 | ||
1138 | /* | |
1139 | * Allocate the pollfd structure and the local view of the out fds to avoid | |
1140 | * doing a lookup in the linked list and concurrency issues when writing is | |
1141 | * needed. Called with consumer_data.lock held. | |
1142 | * | |
1143 | * Returns the number of fds in the structures. | |
1144 | */ | |
ffe60014 | 1145 | static int update_poll_array(struct lttng_consumer_local_data *ctx, |
28ab034a JG |
1146 | struct pollfd **pollfd, |
1147 | struct lttng_consumer_stream **local_stream, | |
1148 | struct lttng_ht *ht, | |
1149 | int *nb_inactive_fd) | |
3bd1e081 | 1150 | { |
3bd1e081 MD |
1151 | int i = 0; |
1152 | ||
a0377dfe FD |
1153 | LTTNG_ASSERT(ctx); |
1154 | LTTNG_ASSERT(ht); | |
1155 | LTTNG_ASSERT(pollfd); | |
1156 | LTTNG_ASSERT(local_stream); | |
ffe60014 | 1157 | |
3bd1e081 | 1158 | DBG("Updating poll fd array"); |
9a2fcf78 | 1159 | *nb_inactive_fd = 0; |
56047f5a | 1160 | |
c3ade133 JG |
1161 | for (auto *stream : |
1162 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
1163 | decltype(lttng_consumer_stream::node), | |
1164 | <tng_consumer_stream::node>(*ht->ht)) { | |
1165 | /* | |
1166 | * Only active streams with an active end point can be added to the | |
1167 | * poll set and local stream storage of the thread. | |
1168 | * | |
1169 | * There is a potential race here for endpoint_status to be updated | |
1170 | * just after the check. However, this is OK since the stream(s) will | |
1171 | * be deleted once the thread is notified that the end point state has | |
1172 | * changed where this function will be called back again. | |
1173 | * | |
1174 | * We track the number of inactive FDs because they still need to be | |
1175 | * closed by the polling thread after a wakeup on the data_pipe or | |
1176 | * metadata_pipe. | |
1177 | */ | |
1178 | if (stream->endpoint_status == CONSUMER_ENDPOINT_INACTIVE) { | |
1179 | (*nb_inactive_fd)++; | |
1180 | continue; | |
3bd1e081 | 1181 | } |
c3ade133 JG |
1182 | |
1183 | (*pollfd)[i].fd = stream->wait_fd; | |
1184 | (*pollfd)[i].events = POLLIN | POLLPRI; | |
1185 | local_stream[i] = stream; | |
1186 | i++; | |
3bd1e081 MD |
1187 | } |
1188 | ||
1189 | /* | |
50f8ae69 | 1190 | * Insert the consumer_data_pipe at the end of the array and don't |
3bd1e081 MD |
1191 | * increment i so nb_fd is the number of real FD. |
1192 | */ | |
acdb9057 | 1193 | (*pollfd)[i].fd = lttng_pipe_get_readfd(ctx->consumer_data_pipe); |
509bb1cf | 1194 | (*pollfd)[i].events = POLLIN | POLLPRI; |
02b3d176 DG |
1195 | |
1196 | (*pollfd)[i + 1].fd = lttng_pipe_get_readfd(ctx->consumer_wakeup_pipe); | |
1197 | (*pollfd)[i + 1].events = POLLIN | POLLPRI; | |
3bd1e081 MD |
1198 | return i; |
1199 | } | |
1200 | ||
1201 | /* | |
84382d49 MD |
1202 | * Poll on the should_quit pipe and the command socket return -1 on |
1203 | * error, 1 if should exit, 0 if data is available on the command socket | |
3bd1e081 MD |
1204 | */ |
1205 | int lttng_consumer_poll_socket(struct pollfd *consumer_sockpoll) | |
1206 | { | |
1207 | int num_rdy; | |
1208 | ||
88f2b785 | 1209 | restart: |
3bd1e081 MD |
1210 | num_rdy = poll(consumer_sockpoll, 2, -1); |
1211 | if (num_rdy == -1) { | |
88f2b785 MD |
1212 | /* |
1213 | * Restart interrupted system call. | |
1214 | */ | |
1215 | if (errno == EINTR) { | |
1216 | goto restart; | |
1217 | } | |
7a57cf92 | 1218 | PERROR("Poll error"); |
84382d49 | 1219 | return -1; |
3bd1e081 | 1220 | } |
509bb1cf | 1221 | if (consumer_sockpoll[0].revents & (POLLIN | POLLPRI)) { |
3bd1e081 | 1222 | DBG("consumer_should_quit wake up"); |
84382d49 | 1223 | return 1; |
3bd1e081 MD |
1224 | } |
1225 | return 0; | |
3bd1e081 MD |
1226 | } |
1227 | ||
1228 | /* | |
1229 | * Set the error socket. | |
1230 | */ | |
28ab034a | 1231 | void lttng_consumer_set_error_sock(struct lttng_consumer_local_data *ctx, int sock) |
3bd1e081 MD |
1232 | { |
1233 | ctx->consumer_error_socket = sock; | |
1234 | } | |
1235 | ||
1236 | /* | |
1237 | * Set the command socket path. | |
1238 | */ | |
28ab034a | 1239 | void lttng_consumer_set_command_sock_path(struct lttng_consumer_local_data *ctx, char *sock) |
3bd1e081 MD |
1240 | { |
1241 | ctx->consumer_command_sock_path = sock; | |
1242 | } | |
1243 | ||
1244 | /* | |
1245 | * Send return code to the session daemon. | |
1246 | * If the socket is not defined, we return 0, it is not a fatal error | |
1247 | */ | |
fbd566c2 JG |
1248 | int lttng_consumer_send_error(struct lttng_consumer_local_data *ctx, |
1249 | enum lttcomm_return_code error_code) | |
3bd1e081 MD |
1250 | { |
1251 | if (ctx->consumer_error_socket > 0) { | |
fbd566c2 JG |
1252 | const std::int32_t comm_code = std::int32_t(error_code); |
1253 | ||
1254 | static_assert( | |
1255 | sizeof(comm_code) >= sizeof(std::underlying_type<lttcomm_return_code>), | |
1256 | "Fixed-size communication type too small to accomodate lttcomm_return_code"); | |
28ab034a | 1257 | return lttcomm_send_unix_sock( |
fbd566c2 | 1258 | ctx->consumer_error_socket, &comm_code, sizeof(comm_code)); |
3bd1e081 MD |
1259 | } |
1260 | ||
1261 | return 0; | |
1262 | } | |
1263 | ||
1264 | /* | |
228b5bf7 DG |
1265 | * Close all the tracefiles and stream fds and MUST be called when all |
1266 | * instances are destroyed i.e. when all threads were joined and are ended. | |
3bd1e081 | 1267 | */ |
cd9adb8b | 1268 | void lttng_consumer_cleanup() |
3bd1e081 | 1269 | { |
e10aec8f | 1270 | unsigned int trace_chunks_left; |
6065ceec | 1271 | |
c3ade133 JG |
1272 | for (auto *channel : |
1273 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_channel, | |
1274 | decltype(lttng_consumer_channel::node), | |
1275 | <tng_consumer_channel::node>( | |
1276 | *the_consumer_data.channel_ht->ht)) { | |
1277 | consumer_del_channel(channel); | |
3bd1e081 | 1278 | } |
6065ceec | 1279 | |
fa29bfbf SM |
1280 | lttng_ht_destroy(the_consumer_data.channel_ht); |
1281 | lttng_ht_destroy(the_consumer_data.channels_by_session_id_ht); | |
228b5bf7 DG |
1282 | |
1283 | cleanup_relayd_ht(); | |
1284 | ||
fa29bfbf | 1285 | lttng_ht_destroy(the_consumer_data.stream_per_chan_id_ht); |
d8ef542d | 1286 | |
228b5bf7 DG |
1287 | /* |
1288 | * This HT contains streams that are freed by either the metadata thread or | |
1289 | * the data thread so we do *nothing* on the hash table and simply destroy | |
1290 | * it. | |
1291 | */ | |
fa29bfbf | 1292 | lttng_ht_destroy(the_consumer_data.stream_list_ht); |
28cc88f3 | 1293 | |
e10aec8f MD |
1294 | /* |
1295 | * Trace chunks in the registry may still exist if the session | |
1296 | * daemon has encountered an internal error and could not | |
1297 | * tear down its sessions and/or trace chunks properly. | |
1298 | * | |
1299 | * Release the session daemon's implicit reference to any remaining | |
1300 | * trace chunk and print an error if any trace chunk was found. Note | |
1301 | * that there are _no_ legitimate cases for trace chunks to be left, | |
1302 | * it is a leak. However, it can happen following a crash of the | |
1303 | * session daemon and not emptying the registry would cause an assertion | |
1304 | * to hit. | |
1305 | */ | |
28ab034a JG |
1306 | trace_chunks_left = |
1307 | lttng_trace_chunk_registry_put_each_chunk(the_consumer_data.chunk_registry); | |
e10aec8f MD |
1308 | if (trace_chunks_left) { |
1309 | ERR("%u trace chunks are leaked by lttng-consumerd. " | |
28ab034a JG |
1310 | "This can be caused by an internal error of the session daemon.", |
1311 | trace_chunks_left); | |
e10aec8f MD |
1312 | } |
1313 | /* Run all callbacks freeing each chunk. */ | |
1314 | rcu_barrier(); | |
fa29bfbf | 1315 | lttng_trace_chunk_registry_destroy(the_consumer_data.chunk_registry); |
3bd1e081 MD |
1316 | } |
1317 | ||
1318 | /* | |
1319 | * Called from signal handler. | |
1320 | */ | |
1321 | void lttng_consumer_should_exit(struct lttng_consumer_local_data *ctx) | |
1322 | { | |
6cd525e8 MD |
1323 | ssize_t ret; |
1324 | ||
10211f5c | 1325 | CMM_STORE_SHARED(consumer_quit, 1); |
6cd525e8 MD |
1326 | ret = lttng_write(ctx->consumer_should_quit[1], "4", 1); |
1327 | if (ret < 1) { | |
7a57cf92 | 1328 | PERROR("write consumer quit"); |
3bd1e081 | 1329 | } |
ab1027f4 DG |
1330 | |
1331 | DBG("Consumer flag that it should quit"); | |
3bd1e081 MD |
1332 | } |
1333 | ||
5199ffc4 JG |
1334 | /* |
1335 | * Flush pending writes to trace output disk file. | |
1336 | */ | |
28ab034a | 1337 | static void lttng_consumer_sync_trace_file(struct lttng_consumer_stream *stream, off_t orig_offset) |
3bd1e081 | 1338 | { |
07c4863f | 1339 | const int outfd = stream->out_fd; |
3bd1e081 MD |
1340 | |
1341 | /* | |
1342 | * This does a blocking write-and-wait on any page that belongs to the | |
1343 | * subbuffer prior to the one we just wrote. | |
1344 | * Don't care about error values, as these are just hints and ways to | |
1345 | * limit the amount of page cache used. | |
1346 | */ | |
ffe60014 | 1347 | if (orig_offset < stream->max_sb_size) { |
3bd1e081 MD |
1348 | return; |
1349 | } | |
671e39d7 MJ |
1350 | lttng::io::hint_flush_range_dont_need_sync( |
1351 | outfd, orig_offset - stream->max_sb_size, stream->max_sb_size); | |
3bd1e081 MD |
1352 | } |
1353 | ||
1354 | /* | |
1355 | * Initialise the necessary environnement : | |
1356 | * - create a new context | |
1357 | * - create the poll_pipe | |
1358 | * - create the should_quit pipe (for signal handler) | |
1359 | * - create the thread pipe (for splice) | |
1360 | * | |
1361 | * Takes a function pointer as argument, this function is called when data is | |
1362 | * available on a buffer. This function is responsible to do the | |
1363 | * kernctl_get_next_subbuf, read the data with mmap or splice depending on the | |
1364 | * buffer configuration and then kernctl_put_next_subbuf at the end. | |
1365 | * | |
1366 | * Returns a pointer to the new context or NULL on error. | |
1367 | */ | |
28ab034a JG |
1368 | struct lttng_consumer_local_data * |
1369 | lttng_consumer_create(enum lttng_consumer_type type, | |
1370 | ssize_t (*buffer_ready)(struct lttng_consumer_stream *stream, | |
1371 | struct lttng_consumer_local_data *ctx, | |
1372 | bool locked_by_caller), | |
1373 | int (*recv_channel)(struct lttng_consumer_channel *channel), | |
1374 | int (*recv_stream)(struct lttng_consumer_stream *stream), | |
1375 | int (*update_stream)(uint64_t stream_key, uint32_t state)) | |
3bd1e081 | 1376 | { |
d8ef542d | 1377 | int ret; |
3bd1e081 MD |
1378 | struct lttng_consumer_local_data *ctx; |
1379 | ||
a0377dfe | 1380 | LTTNG_ASSERT(the_consumer_data.type == LTTNG_CONSUMER_UNKNOWN || |
28ab034a | 1381 | the_consumer_data.type == type); |
fa29bfbf | 1382 | the_consumer_data.type = type; |
3bd1e081 | 1383 | |
64803277 | 1384 | ctx = zmalloc<lttng_consumer_local_data>(); |
cd9adb8b | 1385 | if (ctx == nullptr) { |
7a57cf92 | 1386 | PERROR("allocating context"); |
3bd1e081 MD |
1387 | goto error; |
1388 | } | |
1389 | ||
1390 | ctx->consumer_error_socket = -1; | |
331744e3 | 1391 | ctx->consumer_metadata_socket = -1; |
cd9adb8b | 1392 | pthread_mutex_init(&ctx->metadata_socket_lock, nullptr); |
3bd1e081 MD |
1393 | /* assign the callbacks */ |
1394 | ctx->on_buffer_ready = buffer_ready; | |
1395 | ctx->on_recv_channel = recv_channel; | |
1396 | ctx->on_recv_stream = recv_stream; | |
1397 | ctx->on_update_stream = update_stream; | |
1398 | ||
acdb9057 DG |
1399 | ctx->consumer_data_pipe = lttng_pipe_open(0); |
1400 | if (!ctx->consumer_data_pipe) { | |
3bd1e081 MD |
1401 | goto error_poll_pipe; |
1402 | } | |
1403 | ||
02b3d176 DG |
1404 | ctx->consumer_wakeup_pipe = lttng_pipe_open(0); |
1405 | if (!ctx->consumer_wakeup_pipe) { | |
1406 | goto error_wakeup_pipe; | |
1407 | } | |
1408 | ||
3bd1e081 MD |
1409 | ret = pipe(ctx->consumer_should_quit); |
1410 | if (ret < 0) { | |
7a57cf92 | 1411 | PERROR("Error creating recv pipe"); |
3bd1e081 MD |
1412 | goto error_quit_pipe; |
1413 | } | |
1414 | ||
d8ef542d MD |
1415 | ret = pipe(ctx->consumer_channel_pipe); |
1416 | if (ret < 0) { | |
1417 | PERROR("Error creating channel pipe"); | |
1418 | goto error_channel_pipe; | |
1419 | } | |
1420 | ||
13886d2d DG |
1421 | ctx->consumer_metadata_pipe = lttng_pipe_open(0); |
1422 | if (!ctx->consumer_metadata_pipe) { | |
fb3a43a9 DG |
1423 | goto error_metadata_pipe; |
1424 | } | |
3bd1e081 | 1425 | |
e9404c27 JG |
1426 | ctx->channel_monitor_pipe = -1; |
1427 | ||
fb3a43a9 | 1428 | return ctx; |
3bd1e081 | 1429 | |
fb3a43a9 | 1430 | error_metadata_pipe: |
d8ef542d MD |
1431 | utils_close_pipe(ctx->consumer_channel_pipe); |
1432 | error_channel_pipe: | |
d8ef542d | 1433 | utils_close_pipe(ctx->consumer_should_quit); |
3bd1e081 | 1434 | error_quit_pipe: |
02b3d176 DG |
1435 | lttng_pipe_destroy(ctx->consumer_wakeup_pipe); |
1436 | error_wakeup_pipe: | |
acdb9057 | 1437 | lttng_pipe_destroy(ctx->consumer_data_pipe); |
3bd1e081 MD |
1438 | error_poll_pipe: |
1439 | free(ctx); | |
1440 | error: | |
cd9adb8b | 1441 | return nullptr; |
3bd1e081 MD |
1442 | } |
1443 | ||
282dadbc MD |
1444 | /* |
1445 | * Iterate over all streams of the hashtable and free them properly. | |
1446 | */ | |
1447 | static void destroy_data_stream_ht(struct lttng_ht *ht) | |
1448 | { | |
cd9adb8b | 1449 | if (ht == nullptr) { |
282dadbc MD |
1450 | return; |
1451 | } | |
1452 | ||
c3ade133 JG |
1453 | for (auto *stream : |
1454 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
1455 | decltype(lttng_consumer_stream::node), | |
1456 | <tng_consumer_stream::node>(*ht->ht)) { | |
1457 | /* | |
1458 | * Ignore return value since we are currently cleaning up so any error | |
1459 | * can't be handled. | |
1460 | */ | |
1461 | (void) consumer_del_stream(stream, ht); | |
282dadbc | 1462 | } |
282dadbc MD |
1463 | |
1464 | lttng_ht_destroy(ht); | |
1465 | } | |
1466 | ||
1467 | /* | |
1468 | * Iterate over all streams of the metadata hashtable and free them | |
1469 | * properly. | |
1470 | */ | |
1471 | static void destroy_metadata_stream_ht(struct lttng_ht *ht) | |
1472 | { | |
cd9adb8b | 1473 | if (ht == nullptr) { |
282dadbc MD |
1474 | return; |
1475 | } | |
1476 | ||
c3ade133 JG |
1477 | for (auto *stream : |
1478 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
1479 | decltype(lttng_consumer_stream::node), | |
1480 | <tng_consumer_stream::node>(*ht->ht)) { | |
1481 | /* | |
1482 | * Ignore return value since we are currently cleaning up so any error | |
1483 | * can't be handled. | |
1484 | */ | |
1485 | (void) consumer_del_metadata_stream(stream, ht); | |
282dadbc | 1486 | } |
282dadbc MD |
1487 | |
1488 | lttng_ht_destroy(ht); | |
1489 | } | |
1490 | ||
3bd1e081 MD |
1491 | /* |
1492 | * Close all fds associated with the instance and free the context. | |
1493 | */ | |
1494 | void lttng_consumer_destroy(struct lttng_consumer_local_data *ctx) | |
1495 | { | |
4c462e79 MD |
1496 | int ret; |
1497 | ||
ab1027f4 DG |
1498 | DBG("Consumer destroying it. Closing everything."); |
1499 | ||
4f2e75b9 DG |
1500 | if (!ctx) { |
1501 | return; | |
1502 | } | |
1503 | ||
282dadbc MD |
1504 | destroy_data_stream_ht(data_ht); |
1505 | destroy_metadata_stream_ht(metadata_ht); | |
1506 | ||
4c462e79 MD |
1507 | ret = close(ctx->consumer_error_socket); |
1508 | if (ret) { | |
1509 | PERROR("close"); | |
1510 | } | |
331744e3 JD |
1511 | ret = close(ctx->consumer_metadata_socket); |
1512 | if (ret) { | |
1513 | PERROR("close"); | |
1514 | } | |
d8ef542d | 1515 | utils_close_pipe(ctx->consumer_channel_pipe); |
acdb9057 | 1516 | lttng_pipe_destroy(ctx->consumer_data_pipe); |
13886d2d | 1517 | lttng_pipe_destroy(ctx->consumer_metadata_pipe); |
02b3d176 | 1518 | lttng_pipe_destroy(ctx->consumer_wakeup_pipe); |
d8ef542d | 1519 | utils_close_pipe(ctx->consumer_should_quit); |
fb3a43a9 | 1520 | |
3bd1e081 MD |
1521 | unlink(ctx->consumer_command_sock_path); |
1522 | free(ctx); | |
1523 | } | |
1524 | ||
6197aea7 DG |
1525 | /* |
1526 | * Write the metadata stream id on the specified file descriptor. | |
1527 | */ | |
28ab034a JG |
1528 | static int |
1529 | write_relayd_metadata_id(int fd, struct lttng_consumer_stream *stream, unsigned long padding) | |
6197aea7 | 1530 | { |
6cd525e8 | 1531 | ssize_t ret; |
1d4dfdef | 1532 | struct lttcomm_relayd_metadata_payload hdr; |
6197aea7 | 1533 | |
1d4dfdef DG |
1534 | hdr.stream_id = htobe64(stream->relayd_stream_id); |
1535 | hdr.padding_size = htobe32(padding); | |
6cd525e8 MD |
1536 | ret = lttng_write(fd, (void *) &hdr, sizeof(hdr)); |
1537 | if (ret < sizeof(hdr)) { | |
d7b75ec8 | 1538 | /* |
6f04ed72 | 1539 | * This error means that the fd's end is closed so ignore the PERROR |
d7b75ec8 DG |
1540 | * not to clubber the error output since this can happen in a normal |
1541 | * code path. | |
1542 | */ | |
1543 | if (errno != EPIPE) { | |
1544 | PERROR("write metadata stream id"); | |
1545 | } | |
1546 | DBG3("Consumer failed to write relayd metadata id (errno: %d)", errno); | |
534d2592 DG |
1547 | /* |
1548 | * Set ret to a negative value because if ret != sizeof(hdr), we don't | |
1549 | * handle writting the missing part so report that as an error and | |
1550 | * don't lie to the caller. | |
1551 | */ | |
1552 | ret = -1; | |
6197aea7 DG |
1553 | goto end; |
1554 | } | |
1d4dfdef | 1555 | DBG("Metadata stream id %" PRIu64 " with padding %lu written before data", |
28ab034a JG |
1556 | stream->relayd_stream_id, |
1557 | padding); | |
6197aea7 DG |
1558 | |
1559 | end: | |
6cd525e8 | 1560 | return (int) ret; |
6197aea7 DG |
1561 | } |
1562 | ||
3bd1e081 | 1563 | /* |
09e26845 DG |
1564 | * Mmap the ring buffer, read it and write the data to the tracefile. This is a |
1565 | * core function for writing trace buffers to either the local filesystem or | |
1566 | * the network. | |
1567 | * | |
d2956687 | 1568 | * It must be called with the stream and the channel lock held. |
79d4ffb7 | 1569 | * |
09e26845 | 1570 | * Careful review MUST be put if any changes occur! |
3bd1e081 MD |
1571 | * |
1572 | * Returns the number of bytes written | |
1573 | */ | |
28ab034a JG |
1574 | ssize_t lttng_consumer_on_read_subbuffer_mmap(struct lttng_consumer_stream *stream, |
1575 | const struct lttng_buffer_view *buffer, | |
1576 | unsigned long padding) | |
3bd1e081 | 1577 | { |
994ab360 | 1578 | ssize_t ret = 0; |
f02e1e8a DG |
1579 | off_t orig_offset = stream->out_fd_offset; |
1580 | /* Default is on the disk */ | |
1581 | int outfd = stream->out_fd; | |
cd9adb8b | 1582 | struct consumer_relayd_sock_pair *relayd = nullptr; |
8994307f | 1583 | unsigned int relayd_hang_up = 0; |
fd424d99 JG |
1584 | const size_t subbuf_content_size = buffer->size - padding; |
1585 | size_t write_len; | |
f02e1e8a DG |
1586 | |
1587 | /* RCU lock for the relayd pointer */ | |
07c4863f | 1588 | const lttng::urcu::read_lock_guard read_lock; |
28ab034a | 1589 | LTTNG_ASSERT(stream->net_seq_idx != (uint64_t) -1ULL || stream->trace_chunk); |
d2956687 | 1590 | |
f02e1e8a | 1591 | /* Flag that the current stream if set for network streaming. */ |
da009f2c | 1592 | if (stream->net_seq_idx != (uint64_t) -1ULL) { |
f02e1e8a | 1593 | relayd = consumer_find_relayd(stream->net_seq_idx); |
cd9adb8b | 1594 | if (relayd == nullptr) { |
56591bac | 1595 | ret = -EPIPE; |
f02e1e8a DG |
1596 | goto end; |
1597 | } | |
1598 | } | |
1599 | ||
f02e1e8a DG |
1600 | /* Handle stream on the relayd if the output is on the network */ |
1601 | if (relayd) { | |
fd424d99 | 1602 | unsigned long netlen = subbuf_content_size; |
f02e1e8a DG |
1603 | |
1604 | /* | |
1605 | * Lock the control socket for the complete duration of the function | |
1606 | * since from this point on we will use the socket. | |
1607 | */ | |
1608 | if (stream->metadata_flag) { | |
1609 | /* Metadata requires the control socket. */ | |
1610 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
93ec662e JD |
1611 | if (stream->reset_metadata_flag) { |
1612 | ret = relayd_reset_metadata(&relayd->control_sock, | |
28ab034a JG |
1613 | stream->relayd_stream_id, |
1614 | stream->metadata_version); | |
93ec662e JD |
1615 | if (ret < 0) { |
1616 | relayd_hang_up = 1; | |
1617 | goto write_error; | |
1618 | } | |
1619 | stream->reset_metadata_flag = 0; | |
1620 | } | |
1d4dfdef | 1621 | netlen += sizeof(struct lttcomm_relayd_metadata_payload); |
f02e1e8a DG |
1622 | } |
1623 | ||
1d4dfdef | 1624 | ret = write_relayd_stream_header(stream, netlen, padding, relayd); |
994ab360 DG |
1625 | if (ret < 0) { |
1626 | relayd_hang_up = 1; | |
1627 | goto write_error; | |
1628 | } | |
1629 | /* Use the returned socket. */ | |
1630 | outfd = ret; | |
f02e1e8a | 1631 | |
994ab360 DG |
1632 | /* Write metadata stream id before payload */ |
1633 | if (stream->metadata_flag) { | |
239f61af | 1634 | ret = write_relayd_metadata_id(outfd, stream, padding); |
994ab360 | 1635 | if (ret < 0) { |
8994307f DG |
1636 | relayd_hang_up = 1; |
1637 | goto write_error; | |
1638 | } | |
f02e1e8a | 1639 | } |
1624d5b7 | 1640 | |
fd424d99 JG |
1641 | write_len = subbuf_content_size; |
1642 | } else { | |
1643 | /* No streaming; we have to write the full padding. */ | |
93ec662e JD |
1644 | if (stream->metadata_flag && stream->reset_metadata_flag) { |
1645 | ret = utils_truncate_stream_file(stream->out_fd, 0); | |
1646 | if (ret < 0) { | |
1647 | ERR("Reset metadata file"); | |
1648 | goto end; | |
1649 | } | |
1650 | stream->reset_metadata_flag = 0; | |
1651 | } | |
1652 | ||
1624d5b7 JD |
1653 | /* |
1654 | * Check if we need to change the tracefile before writing the packet. | |
1655 | */ | |
1656 | if (stream->chan->tracefile_size > 0 && | |
28ab034a JG |
1657 | (stream->tracefile_size_current + buffer->size) > |
1658 | stream->chan->tracefile_size) { | |
d2956687 JG |
1659 | ret = consumer_stream_rotate_output_files(stream); |
1660 | if (ret) { | |
1624d5b7 JD |
1661 | goto end; |
1662 | } | |
309167d2 | 1663 | outfd = stream->out_fd; |
a1ae300f | 1664 | orig_offset = 0; |
1624d5b7 | 1665 | } |
fd424d99 | 1666 | stream->tracefile_size_current += buffer->size; |
fd424d99 | 1667 | write_len = buffer->size; |
f02e1e8a DG |
1668 | } |
1669 | ||
d02b8372 DG |
1670 | /* |
1671 | * This call guarantee that len or less is returned. It's impossible to | |
1672 | * receive a ret value that is bigger than len. | |
1673 | */ | |
fd424d99 | 1674 | ret = lttng_write(outfd, buffer->data, write_len); |
e2d1190b | 1675 | DBG("Consumer mmap write() ret %zd (len %zu)", ret, write_len); |
fd424d99 | 1676 | if (ret < 0 || ((size_t) ret != write_len)) { |
d02b8372 DG |
1677 | /* |
1678 | * Report error to caller if nothing was written else at least send the | |
1679 | * amount written. | |
1680 | */ | |
1681 | if (ret < 0) { | |
994ab360 | 1682 | ret = -errno; |
f02e1e8a | 1683 | } |
994ab360 | 1684 | relayd_hang_up = 1; |
f02e1e8a | 1685 | |
d02b8372 | 1686 | /* Socket operation failed. We consider the relayd dead */ |
fcf0f774 | 1687 | if (errno == EPIPE) { |
d02b8372 DG |
1688 | /* |
1689 | * This is possible if the fd is closed on the other side | |
1690 | * (outfd) or any write problem. It can be verbose a bit for a | |
1691 | * normal execution if for instance the relayd is stopped | |
1692 | * abruptly. This can happen so set this to a DBG statement. | |
1693 | */ | |
1694 | DBG("Consumer mmap write detected relayd hang up"); | |
994ab360 DG |
1695 | } else { |
1696 | /* Unhandled error, print it and stop function right now. */ | |
28ab034a | 1697 | PERROR("Error in write mmap (ret %zd != write_len %zu)", ret, write_len); |
f02e1e8a | 1698 | } |
994ab360 | 1699 | goto write_error; |
d02b8372 DG |
1700 | } |
1701 | stream->output_written += ret; | |
d02b8372 DG |
1702 | |
1703 | /* This call is useless on a socket so better save a syscall. */ | |
1704 | if (!relayd) { | |
1705 | /* This won't block, but will start writeout asynchronously */ | |
671e39d7 | 1706 | lttng::io::hint_flush_range_async(outfd, stream->out_fd_offset, write_len); |
fd424d99 | 1707 | stream->out_fd_offset += write_len; |
f5dbe415 | 1708 | lttng_consumer_sync_trace_file(stream, orig_offset); |
f02e1e8a | 1709 | } |
f02e1e8a | 1710 | |
8994307f DG |
1711 | write_error: |
1712 | /* | |
1713 | * This is a special case that the relayd has closed its socket. Let's | |
1714 | * cleanup the relayd object and all associated streams. | |
1715 | */ | |
1716 | if (relayd && relayd_hang_up) { | |
28ab034a | 1717 | ERR("Relayd hangup. Cleaning up relayd %" PRIu64 ".", relayd->net_seq_idx); |
9276e5c8 | 1718 | lttng_consumer_cleanup_relayd(relayd); |
8994307f DG |
1719 | } |
1720 | ||
f02e1e8a DG |
1721 | end: |
1722 | /* Unlock only if ctrl socket used */ | |
1723 | if (relayd && stream->metadata_flag) { | |
1724 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); | |
1725 | } | |
1726 | ||
994ab360 | 1727 | return ret; |
3bd1e081 MD |
1728 | } |
1729 | ||
1730 | /* | |
1731 | * Splice the data from the ring buffer to the tracefile. | |
1732 | * | |
79d4ffb7 DG |
1733 | * It must be called with the stream lock held. |
1734 | * | |
3bd1e081 MD |
1735 | * Returns the number of bytes spliced. |
1736 | */ | |
28ab034a JG |
1737 | ssize_t lttng_consumer_on_read_subbuffer_splice(struct lttng_consumer_local_data *ctx, |
1738 | struct lttng_consumer_stream *stream, | |
1739 | unsigned long len, | |
1740 | unsigned long padding) | |
3bd1e081 | 1741 | { |
f02e1e8a DG |
1742 | ssize_t ret = 0, written = 0, ret_splice = 0; |
1743 | loff_t offset = 0; | |
1744 | off_t orig_offset = stream->out_fd_offset; | |
07c4863f | 1745 | const int fd = stream->wait_fd; |
f02e1e8a DG |
1746 | /* Default is on the disk */ |
1747 | int outfd = stream->out_fd; | |
cd9adb8b | 1748 | struct consumer_relayd_sock_pair *relayd = nullptr; |
fb3a43a9 | 1749 | int *splice_pipe; |
8994307f | 1750 | unsigned int relayd_hang_up = 0; |
f02e1e8a | 1751 | |
fa29bfbf | 1752 | switch (the_consumer_data.type) { |
3bd1e081 | 1753 | case LTTNG_CONSUMER_KERNEL: |
f02e1e8a | 1754 | break; |
7753dea8 MD |
1755 | case LTTNG_CONSUMER32_UST: |
1756 | case LTTNG_CONSUMER64_UST: | |
f02e1e8a | 1757 | /* Not supported for user space tracing */ |
3bd1e081 MD |
1758 | return -ENOSYS; |
1759 | default: | |
1760 | ERR("Unknown consumer_data type"); | |
a0377dfe | 1761 | abort(); |
3bd1e081 MD |
1762 | } |
1763 | ||
f02e1e8a | 1764 | /* RCU lock for the relayd pointer */ |
07c4863f | 1765 | const lttng::urcu::read_lock_guard read_lock; |
f02e1e8a DG |
1766 | |
1767 | /* Flag that the current stream if set for network streaming. */ | |
da009f2c | 1768 | if (stream->net_seq_idx != (uint64_t) -1ULL) { |
f02e1e8a | 1769 | relayd = consumer_find_relayd(stream->net_seq_idx); |
cd9adb8b | 1770 | if (relayd == nullptr) { |
ad0b0d23 | 1771 | written = -ret; |
f02e1e8a DG |
1772 | goto end; |
1773 | } | |
1774 | } | |
a2361a61 | 1775 | splice_pipe = stream->splice_pipe; |
fb3a43a9 | 1776 | |
f02e1e8a | 1777 | /* Write metadata stream id before payload */ |
1d4dfdef | 1778 | if (relayd) { |
ad0b0d23 | 1779 | unsigned long total_len = len; |
f02e1e8a | 1780 | |
1d4dfdef DG |
1781 | if (stream->metadata_flag) { |
1782 | /* | |
1783 | * Lock the control socket for the complete duration of the function | |
1784 | * since from this point on we will use the socket. | |
1785 | */ | |
1786 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
1787 | ||
93ec662e JD |
1788 | if (stream->reset_metadata_flag) { |
1789 | ret = relayd_reset_metadata(&relayd->control_sock, | |
28ab034a JG |
1790 | stream->relayd_stream_id, |
1791 | stream->metadata_version); | |
93ec662e JD |
1792 | if (ret < 0) { |
1793 | relayd_hang_up = 1; | |
1794 | goto write_error; | |
1795 | } | |
1796 | stream->reset_metadata_flag = 0; | |
1797 | } | |
28ab034a | 1798 | ret = write_relayd_metadata_id(splice_pipe[1], stream, padding); |
1d4dfdef DG |
1799 | if (ret < 0) { |
1800 | written = ret; | |
ad0b0d23 DG |
1801 | relayd_hang_up = 1; |
1802 | goto write_error; | |
1d4dfdef DG |
1803 | } |
1804 | ||
1805 | total_len += sizeof(struct lttcomm_relayd_metadata_payload); | |
1806 | } | |
1807 | ||
1808 | ret = write_relayd_stream_header(stream, total_len, padding, relayd); | |
ad0b0d23 DG |
1809 | if (ret < 0) { |
1810 | written = ret; | |
1811 | relayd_hang_up = 1; | |
1812 | goto write_error; | |
f02e1e8a | 1813 | } |
ad0b0d23 DG |
1814 | /* Use the returned socket. */ |
1815 | outfd = ret; | |
1d4dfdef DG |
1816 | } else { |
1817 | /* No streaming, we have to set the len with the full padding */ | |
1818 | len += padding; | |
1624d5b7 | 1819 | |
93ec662e JD |
1820 | if (stream->metadata_flag && stream->reset_metadata_flag) { |
1821 | ret = utils_truncate_stream_file(stream->out_fd, 0); | |
1822 | if (ret < 0) { | |
1823 | ERR("Reset metadata file"); | |
1824 | goto end; | |
1825 | } | |
1826 | stream->reset_metadata_flag = 0; | |
1827 | } | |
1624d5b7 JD |
1828 | /* |
1829 | * Check if we need to change the tracefile before writing the packet. | |
1830 | */ | |
1831 | if (stream->chan->tracefile_size > 0 && | |
28ab034a | 1832 | (stream->tracefile_size_current + len) > stream->chan->tracefile_size) { |
d2956687 | 1833 | ret = consumer_stream_rotate_output_files(stream); |
1624d5b7 | 1834 | if (ret < 0) { |
ad0b0d23 | 1835 | written = ret; |
1624d5b7 JD |
1836 | goto end; |
1837 | } | |
309167d2 | 1838 | outfd = stream->out_fd; |
a1ae300f | 1839 | orig_offset = 0; |
1624d5b7 JD |
1840 | } |
1841 | stream->tracefile_size_current += len; | |
f02e1e8a DG |
1842 | } |
1843 | ||
1844 | while (len > 0) { | |
1d4dfdef | 1845 | DBG("splice chan to pipe offset %lu of len %lu (fd : %d, pipe: %d)", |
28ab034a JG |
1846 | (unsigned long) offset, |
1847 | len, | |
1848 | fd, | |
1849 | splice_pipe[1]); | |
1850 | ret_splice = splice( | |
cd9adb8b | 1851 | fd, &offset, splice_pipe[1], nullptr, len, SPLICE_F_MOVE | SPLICE_F_MORE); |
f02e1e8a DG |
1852 | DBG("splice chan to pipe, ret %zd", ret_splice); |
1853 | if (ret_splice < 0) { | |
d02b8372 | 1854 | ret = errno; |
ad0b0d23 | 1855 | written = -ret; |
d02b8372 | 1856 | PERROR("Error in relay splice"); |
f02e1e8a DG |
1857 | goto splice_error; |
1858 | } | |
1859 | ||
1860 | /* Handle stream on the relayd if the output is on the network */ | |
ad0b0d23 | 1861 | if (relayd && stream->metadata_flag) { |
07c4863f | 1862 | const size_t metadata_payload_size = |
ad0b0d23 DG |
1863 | sizeof(struct lttcomm_relayd_metadata_payload); |
1864 | ||
1865 | /* Update counter to fit the spliced data */ | |
1866 | ret_splice += metadata_payload_size; | |
1867 | len += metadata_payload_size; | |
1868 | /* | |
1869 | * We do this so the return value can match the len passed as | |
1870 | * argument to this function. | |
1871 | */ | |
1872 | written -= metadata_payload_size; | |
f02e1e8a DG |
1873 | } |
1874 | ||
1875 | /* Splice data out */ | |
28ab034a | 1876 | ret_splice = splice(splice_pipe[0], |
cd9adb8b | 1877 | nullptr, |
28ab034a | 1878 | outfd, |
cd9adb8b | 1879 | nullptr, |
28ab034a JG |
1880 | ret_splice, |
1881 | SPLICE_F_MOVE | SPLICE_F_MORE); | |
1882 | DBG("Consumer splice pipe to file (out_fd: %d), ret %zd", outfd, ret_splice); | |
f02e1e8a | 1883 | if (ret_splice < 0) { |
d02b8372 | 1884 | ret = errno; |
ad0b0d23 DG |
1885 | written = -ret; |
1886 | relayd_hang_up = 1; | |
1887 | goto write_error; | |
f02e1e8a | 1888 | } else if (ret_splice > len) { |
d02b8372 DG |
1889 | /* |
1890 | * We don't expect this code path to be executed but you never know | |
1891 | * so this is an extra protection agains a buggy splice(). | |
1892 | */ | |
f02e1e8a | 1893 | ret = errno; |
ad0b0d23 | 1894 | written += ret_splice; |
28ab034a | 1895 | PERROR("Wrote more data than requested %zd (len: %lu)", ret_splice, len); |
f02e1e8a | 1896 | goto splice_error; |
d02b8372 DG |
1897 | } else { |
1898 | /* All good, update current len and continue. */ | |
1899 | len -= ret_splice; | |
f02e1e8a | 1900 | } |
f02e1e8a DG |
1901 | |
1902 | /* This call is useless on a socket so better save a syscall. */ | |
1903 | if (!relayd) { | |
1904 | /* This won't block, but will start writeout asynchronously */ | |
671e39d7 | 1905 | lttng::io::hint_flush_range_async(outfd, stream->out_fd_offset, ret_splice); |
f02e1e8a DG |
1906 | stream->out_fd_offset += ret_splice; |
1907 | } | |
e5d1a9b3 | 1908 | stream->output_written += ret_splice; |
f02e1e8a DG |
1909 | written += ret_splice; |
1910 | } | |
f5dbe415 JG |
1911 | if (!relayd) { |
1912 | lttng_consumer_sync_trace_file(stream, orig_offset); | |
1913 | } | |
f02e1e8a DG |
1914 | goto end; |
1915 | ||
8994307f DG |
1916 | write_error: |
1917 | /* | |
1918 | * This is a special case that the relayd has closed its socket. Let's | |
1919 | * cleanup the relayd object and all associated streams. | |
1920 | */ | |
1921 | if (relayd && relayd_hang_up) { | |
28ab034a | 1922 | ERR("Relayd hangup. Cleaning up relayd %" PRIu64 ".", relayd->net_seq_idx); |
9276e5c8 | 1923 | lttng_consumer_cleanup_relayd(relayd); |
8994307f DG |
1924 | /* Skip splice error so the consumer does not fail */ |
1925 | goto end; | |
1926 | } | |
1927 | ||
f02e1e8a DG |
1928 | splice_error: |
1929 | /* send the appropriate error description to sessiond */ | |
1930 | switch (ret) { | |
f02e1e8a | 1931 | case EINVAL: |
f73fabfd | 1932 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_SPLICE_EINVAL); |
f02e1e8a DG |
1933 | break; |
1934 | case ENOMEM: | |
f73fabfd | 1935 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_SPLICE_ENOMEM); |
f02e1e8a DG |
1936 | break; |
1937 | case ESPIPE: | |
f73fabfd | 1938 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_SPLICE_ESPIPE); |
f02e1e8a DG |
1939 | break; |
1940 | } | |
1941 | ||
1942 | end: | |
1943 | if (relayd && stream->metadata_flag) { | |
1944 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); | |
1945 | } | |
1946 | ||
f02e1e8a | 1947 | return written; |
3bd1e081 MD |
1948 | } |
1949 | ||
15055ce5 JD |
1950 | /* |
1951 | * Sample the snapshot positions for a specific fd | |
1952 | * | |
1953 | * Returns 0 on success, < 0 on error | |
1954 | */ | |
1955 | int lttng_consumer_sample_snapshot_positions(struct lttng_consumer_stream *stream) | |
1956 | { | |
fa29bfbf | 1957 | switch (the_consumer_data.type) { |
15055ce5 JD |
1958 | case LTTNG_CONSUMER_KERNEL: |
1959 | return lttng_kconsumer_sample_snapshot_positions(stream); | |
1960 | case LTTNG_CONSUMER32_UST: | |
1961 | case LTTNG_CONSUMER64_UST: | |
1962 | return lttng_ustconsumer_sample_snapshot_positions(stream); | |
1963 | default: | |
1964 | ERR("Unknown consumer_data type"); | |
a0377dfe | 1965 | abort(); |
15055ce5 JD |
1966 | return -ENOSYS; |
1967 | } | |
1968 | } | |
3bd1e081 MD |
1969 | /* |
1970 | * Take a snapshot for a specific fd | |
1971 | * | |
1972 | * Returns 0 on success, < 0 on error | |
1973 | */ | |
ffe60014 | 1974 | int lttng_consumer_take_snapshot(struct lttng_consumer_stream *stream) |
3bd1e081 | 1975 | { |
fa29bfbf | 1976 | switch (the_consumer_data.type) { |
3bd1e081 | 1977 | case LTTNG_CONSUMER_KERNEL: |
ffe60014 | 1978 | return lttng_kconsumer_take_snapshot(stream); |
7753dea8 MD |
1979 | case LTTNG_CONSUMER32_UST: |
1980 | case LTTNG_CONSUMER64_UST: | |
ffe60014 | 1981 | return lttng_ustconsumer_take_snapshot(stream); |
3bd1e081 MD |
1982 | default: |
1983 | ERR("Unknown consumer_data type"); | |
a0377dfe | 1984 | abort(); |
3bd1e081 MD |
1985 | return -ENOSYS; |
1986 | } | |
3bd1e081 MD |
1987 | } |
1988 | ||
1989 | /* | |
1990 | * Get the produced position | |
1991 | * | |
1992 | * Returns 0 on success, < 0 on error | |
1993 | */ | |
28ab034a | 1994 | int lttng_consumer_get_produced_snapshot(struct lttng_consumer_stream *stream, unsigned long *pos) |
3bd1e081 | 1995 | { |
fa29bfbf | 1996 | switch (the_consumer_data.type) { |
3bd1e081 | 1997 | case LTTNG_CONSUMER_KERNEL: |
ffe60014 | 1998 | return lttng_kconsumer_get_produced_snapshot(stream, pos); |
7753dea8 MD |
1999 | case LTTNG_CONSUMER32_UST: |
2000 | case LTTNG_CONSUMER64_UST: | |
ffe60014 | 2001 | return lttng_ustconsumer_get_produced_snapshot(stream, pos); |
3bd1e081 MD |
2002 | default: |
2003 | ERR("Unknown consumer_data type"); | |
a0377dfe | 2004 | abort(); |
3bd1e081 MD |
2005 | return -ENOSYS; |
2006 | } | |
2007 | } | |
2008 | ||
15055ce5 JD |
2009 | /* |
2010 | * Get the consumed position (free-running counter position in bytes). | |
2011 | * | |
2012 | * Returns 0 on success, < 0 on error | |
2013 | */ | |
28ab034a | 2014 | int lttng_consumer_get_consumed_snapshot(struct lttng_consumer_stream *stream, unsigned long *pos) |
15055ce5 | 2015 | { |
fa29bfbf | 2016 | switch (the_consumer_data.type) { |
15055ce5 JD |
2017 | case LTTNG_CONSUMER_KERNEL: |
2018 | return lttng_kconsumer_get_consumed_snapshot(stream, pos); | |
2019 | case LTTNG_CONSUMER32_UST: | |
2020 | case LTTNG_CONSUMER64_UST: | |
2021 | return lttng_ustconsumer_get_consumed_snapshot(stream, pos); | |
2022 | default: | |
2023 | ERR("Unknown consumer_data type"); | |
a0377dfe | 2024 | abort(); |
15055ce5 JD |
2025 | return -ENOSYS; |
2026 | } | |
2027 | } | |
2028 | ||
3bd1e081 | 2029 | int lttng_consumer_recv_cmd(struct lttng_consumer_local_data *ctx, |
28ab034a JG |
2030 | int sock, |
2031 | struct pollfd *consumer_sockpoll) | |
3bd1e081 | 2032 | { |
fa29bfbf | 2033 | switch (the_consumer_data.type) { |
3bd1e081 MD |
2034 | case LTTNG_CONSUMER_KERNEL: |
2035 | return lttng_kconsumer_recv_cmd(ctx, sock, consumer_sockpoll); | |
7753dea8 MD |
2036 | case LTTNG_CONSUMER32_UST: |
2037 | case LTTNG_CONSUMER64_UST: | |
3bd1e081 MD |
2038 | return lttng_ustconsumer_recv_cmd(ctx, sock, consumer_sockpoll); |
2039 | default: | |
2040 | ERR("Unknown consumer_data type"); | |
a0377dfe | 2041 | abort(); |
3bd1e081 MD |
2042 | return -ENOSYS; |
2043 | } | |
2044 | } | |
2045 | ||
cd9adb8b | 2046 | static void lttng_consumer_close_all_metadata() |
d88aee68 | 2047 | { |
fa29bfbf | 2048 | switch (the_consumer_data.type) { |
d88aee68 DG |
2049 | case LTTNG_CONSUMER_KERNEL: |
2050 | /* | |
2051 | * The Kernel consumer has a different metadata scheme so we don't | |
2052 | * close anything because the stream will be closed by the session | |
2053 | * daemon. | |
2054 | */ | |
2055 | break; | |
2056 | case LTTNG_CONSUMER32_UST: | |
2057 | case LTTNG_CONSUMER64_UST: | |
2058 | /* | |
2059 | * Close all metadata streams. The metadata hash table is passed and | |
2060 | * this call iterates over it by closing all wakeup fd. This is safe | |
2061 | * because at this point we are sure that the metadata producer is | |
2062 | * either dead or blocked. | |
2063 | */ | |
6d574024 | 2064 | lttng_ustconsumer_close_all_metadata(metadata_ht); |
d88aee68 DG |
2065 | break; |
2066 | default: | |
2067 | ERR("Unknown consumer_data type"); | |
a0377dfe | 2068 | abort(); |
d88aee68 DG |
2069 | } |
2070 | } | |
2071 | ||
fb3a43a9 DG |
2072 | /* |
2073 | * Clean up a metadata stream and free its memory. | |
2074 | */ | |
28ab034a | 2075 | void consumer_del_metadata_stream(struct lttng_consumer_stream *stream, struct lttng_ht *ht) |
fb3a43a9 | 2076 | { |
cd9adb8b | 2077 | struct lttng_consumer_channel *channel = nullptr; |
a6ef8ee6 | 2078 | bool free_channel = false; |
fb3a43a9 | 2079 | |
a0377dfe | 2080 | LTTNG_ASSERT(stream); |
fb3a43a9 DG |
2081 | /* |
2082 | * This call should NEVER receive regular stream. It must always be | |
2083 | * metadata stream and this is crucial for data structure synchronization. | |
2084 | */ | |
a0377dfe | 2085 | LTTNG_ASSERT(stream->metadata_flag); |
fb3a43a9 | 2086 | |
e316aad5 DG |
2087 | DBG3("Consumer delete metadata stream %d", stream->wait_fd); |
2088 | ||
fa29bfbf | 2089 | pthread_mutex_lock(&the_consumer_data.lock); |
a6ef8ee6 JG |
2090 | /* |
2091 | * Note that this assumes that a stream's channel is never changed and | |
2092 | * that the stream's lock doesn't need to be taken to sample its | |
2093 | * channel. | |
2094 | */ | |
2095 | channel = stream->chan; | |
2096 | pthread_mutex_lock(&channel->lock); | |
3dad2c0f | 2097 | pthread_mutex_lock(&stream->lock); |
a6ef8ee6 | 2098 | if (channel->metadata_cache) { |
081424af | 2099 | /* Only applicable to userspace consumers. */ |
a6ef8ee6 | 2100 | pthread_mutex_lock(&channel->metadata_cache->lock); |
081424af | 2101 | } |
8994307f | 2102 | |
6d574024 DG |
2103 | /* Remove any reference to that stream. */ |
2104 | consumer_stream_delete(stream, ht); | |
ca22feea | 2105 | |
6d574024 | 2106 | /* Close down everything including the relayd if one. */ |
d119bd01 | 2107 | consumer_stream_close_output(stream); |
6d574024 DG |
2108 | /* Destroy tracer buffers of the stream. */ |
2109 | consumer_stream_destroy_buffers(stream); | |
fb3a43a9 DG |
2110 | |
2111 | /* Atomically decrement channel refcount since other threads can use it. */ | |
28ab034a JG |
2112 | if (!uatomic_sub_return(&channel->refcount, 1) && |
2113 | !uatomic_read(&channel->nb_init_stream_left)) { | |
c30aaa51 | 2114 | /* Go for channel deletion! */ |
a6ef8ee6 | 2115 | free_channel = true; |
fb3a43a9 DG |
2116 | } |
2117 | ||
73811ecc DG |
2118 | /* |
2119 | * Nullify the stream reference so it is not used after deletion. The | |
6d574024 DG |
2120 | * channel lock MUST be acquired before being able to check for a NULL |
2121 | * pointer value. | |
73811ecc | 2122 | */ |
cd9adb8b | 2123 | channel->metadata_stream = nullptr; |
32670d71 | 2124 | channel->metadata_pushed_wait_queue.wake_all(); |
73811ecc | 2125 | |
a6ef8ee6 JG |
2126 | if (channel->metadata_cache) { |
2127 | pthread_mutex_unlock(&channel->metadata_cache->lock); | |
081424af | 2128 | } |
3dad2c0f | 2129 | pthread_mutex_unlock(&stream->lock); |
a6ef8ee6 | 2130 | pthread_mutex_unlock(&channel->lock); |
fa29bfbf | 2131 | pthread_mutex_unlock(&the_consumer_data.lock); |
e316aad5 | 2132 | |
a6ef8ee6 JG |
2133 | if (free_channel) { |
2134 | consumer_del_channel(channel); | |
e316aad5 DG |
2135 | } |
2136 | ||
d2956687 | 2137 | lttng_trace_chunk_put(stream->trace_chunk); |
cd9adb8b | 2138 | stream->trace_chunk = nullptr; |
6d574024 | 2139 | consumer_stream_free(stream); |
fb3a43a9 DG |
2140 | } |
2141 | ||
2142 | /* | |
2143 | * Action done with the metadata stream when adding it to the consumer internal | |
2144 | * data structures to handle it. | |
2145 | */ | |
66d583dc | 2146 | void consumer_add_metadata_stream(struct lttng_consumer_stream *stream) |
fb3a43a9 | 2147 | { |
5ab66908 | 2148 | struct lttng_ht *ht = metadata_ht; |
76082088 | 2149 | struct lttng_ht_iter iter; |
d88aee68 | 2150 | struct lttng_ht_node_u64 *node; |
fb3a43a9 | 2151 | |
a0377dfe FD |
2152 | LTTNG_ASSERT(stream); |
2153 | LTTNG_ASSERT(ht); | |
e316aad5 | 2154 | |
d88aee68 | 2155 | DBG3("Adding metadata stream %" PRIu64 " to hash table", stream->key); |
e316aad5 | 2156 | |
fa29bfbf | 2157 | pthread_mutex_lock(&the_consumer_data.lock); |
a9838785 | 2158 | pthread_mutex_lock(&stream->chan->lock); |
ec6ea7d0 | 2159 | pthread_mutex_lock(&stream->chan->timer_lock); |
2e818a6a | 2160 | pthread_mutex_lock(&stream->lock); |
e316aad5 | 2161 | |
e316aad5 DG |
2162 | /* |
2163 | * From here, refcounts are updated so be _careful_ when returning an error | |
2164 | * after this point. | |
2165 | */ | |
2166 | ||
07c4863f | 2167 | const lttng::urcu::read_lock_guard read_lock; |
76082088 DG |
2168 | |
2169 | /* | |
2170 | * Lookup the stream just to make sure it does not exist in our internal | |
2171 | * state. This should NEVER happen. | |
2172 | */ | |
d88aee68 | 2173 | lttng_ht_lookup(ht, &stream->key, &iter); |
00d7d903 | 2174 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
a0377dfe | 2175 | LTTNG_ASSERT(!node); |
76082088 | 2176 | |
e316aad5 | 2177 | /* |
ffe60014 DG |
2178 | * When nb_init_stream_left reaches 0, we don't need to trigger any action |
2179 | * in terms of destroying the associated channel, because the action that | |
e316aad5 DG |
2180 | * causes the count to become 0 also causes a stream to be added. The |
2181 | * channel deletion will thus be triggered by the following removal of this | |
2182 | * stream. | |
2183 | */ | |
ffe60014 | 2184 | if (uatomic_read(&stream->chan->nb_init_stream_left) > 0) { |
f2ad556d MD |
2185 | /* Increment refcount before decrementing nb_init_stream_left */ |
2186 | cmm_smp_wmb(); | |
ffe60014 | 2187 | uatomic_dec(&stream->chan->nb_init_stream_left); |
e316aad5 DG |
2188 | } |
2189 | ||
d88aee68 | 2190 | lttng_ht_add_unique_u64(ht, &stream->node); |
ca22feea | 2191 | |
28ab034a | 2192 | lttng_ht_add_u64(the_consumer_data.stream_per_chan_id_ht, &stream->node_channel_id); |
d8ef542d | 2193 | |
ca22feea DG |
2194 | /* |
2195 | * Add stream to the stream_list_ht of the consumer data. No need to steal | |
2196 | * the key since the HT does not use it and we allow to add redundant keys | |
2197 | * into this table. | |
2198 | */ | |
28ab034a | 2199 | lttng_ht_add_u64(the_consumer_data.stream_list_ht, &stream->node_session_id); |
ca22feea | 2200 | |
2e818a6a | 2201 | pthread_mutex_unlock(&stream->lock); |
a9838785 | 2202 | pthread_mutex_unlock(&stream->chan->lock); |
ec6ea7d0 | 2203 | pthread_mutex_unlock(&stream->chan->timer_lock); |
fa29bfbf | 2204 | pthread_mutex_unlock(&the_consumer_data.lock); |
fb3a43a9 DG |
2205 | } |
2206 | ||
8994307f DG |
2207 | /* |
2208 | * Delete data stream that are flagged for deletion (endpoint_status). | |
2209 | */ | |
cd9adb8b | 2210 | static void validate_endpoint_status_data_stream() |
8994307f | 2211 | { |
8994307f DG |
2212 | DBG("Consumer delete flagged data stream"); |
2213 | ||
c3ade133 JG |
2214 | for (auto *stream : |
2215 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
2216 | decltype(lttng_consumer_stream::node), | |
2217 | <tng_consumer_stream::node>(*data_ht->ht)) { | |
2218 | /* Validate delete flag of the stream */ | |
2219 | if (stream->endpoint_status == CONSUMER_ENDPOINT_ACTIVE) { | |
2220 | continue; | |
8994307f | 2221 | } |
c3ade133 JG |
2222 | /* Delete it right now */ |
2223 | consumer_del_stream(stream, data_ht); | |
8994307f | 2224 | } |
8994307f DG |
2225 | } |
2226 | ||
2227 | /* | |
2228 | * Delete metadata stream that are flagged for deletion (endpoint_status). | |
2229 | */ | |
28ab034a | 2230 | static void validate_endpoint_status_metadata_stream(struct lttng_poll_event *pollset) |
8994307f | 2231 | { |
8994307f DG |
2232 | DBG("Consumer delete flagged metadata stream"); |
2233 | ||
a0377dfe | 2234 | LTTNG_ASSERT(pollset); |
8994307f | 2235 | |
c3ade133 JG |
2236 | for (auto *stream : |
2237 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_stream, | |
2238 | decltype(lttng_consumer_stream::node), | |
2239 | <tng_consumer_stream::node>(*metadata_ht->ht)) { | |
2240 | /* Validate delete flag of the stream */ | |
2241 | if (stream->endpoint_status == CONSUMER_ENDPOINT_ACTIVE) { | |
2242 | continue; | |
56047f5a | 2243 | } |
c3ade133 JG |
2244 | /* |
2245 | * Remove from pollset so the metadata thread can continue without | |
2246 | * blocking on a deleted stream. | |
2247 | */ | |
2248 | lttng_poll_del(pollset, stream->wait_fd); | |
2249 | ||
2250 | /* Delete it right now */ | |
2251 | consumer_del_metadata_stream(stream, metadata_ht); | |
8994307f | 2252 | } |
8994307f DG |
2253 | } |
2254 | ||
fb3a43a9 DG |
2255 | /* |
2256 | * Thread polls on metadata file descriptor and write them on disk or on the | |
2257 | * network. | |
2258 | */ | |
7d980def | 2259 | void *consumer_thread_metadata_poll(void *data) |
fb3a43a9 | 2260 | { |
1fc79fb4 | 2261 | int ret, i, pollfd, err = -1; |
fb3a43a9 | 2262 | uint32_t revents, nb_fd; |
cd9adb8b | 2263 | struct lttng_consumer_stream *stream = nullptr; |
fb3a43a9 | 2264 | struct lttng_ht_iter iter; |
d88aee68 | 2265 | struct lttng_ht_node_u64 *node; |
fb3a43a9 | 2266 | struct lttng_poll_event events; |
97535efa | 2267 | struct lttng_consumer_local_data *ctx = (lttng_consumer_local_data *) data; |
fb3a43a9 DG |
2268 | ssize_t len; |
2269 | ||
2270 | rcu_register_thread(); | |
2271 | ||
1fc79fb4 MD |
2272 | health_register(health_consumerd, HEALTH_CONSUMERD_TYPE_METADATA); |
2273 | ||
2d57de81 MD |
2274 | if (testpoint(consumerd_thread_metadata)) { |
2275 | goto error_testpoint; | |
2276 | } | |
2277 | ||
9ce5646a MD |
2278 | health_code_update(); |
2279 | ||
fb3a43a9 DG |
2280 | DBG("Thread metadata poll started"); |
2281 | ||
fb3a43a9 DG |
2282 | /* Size is set to 1 for the consumer_metadata pipe */ |
2283 | ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC); | |
2284 | if (ret < 0) { | |
2285 | ERR("Poll set creation failed"); | |
d8ef542d | 2286 | goto end_poll; |
fb3a43a9 DG |
2287 | } |
2288 | ||
28ab034a | 2289 | ret = lttng_poll_add(&events, lttng_pipe_get_readfd(ctx->consumer_metadata_pipe), LPOLLIN); |
fb3a43a9 DG |
2290 | if (ret < 0) { |
2291 | goto end; | |
2292 | } | |
2293 | ||
2294 | /* Main loop */ | |
2295 | DBG("Metadata main loop started"); | |
2296 | ||
cd9adb8b | 2297 | while (true) { |
28ab034a | 2298 | restart: |
7fa2082e | 2299 | health_code_update(); |
9ce5646a | 2300 | health_poll_entry(); |
7fa2082e | 2301 | DBG("Metadata poll wait"); |
fb3a43a9 | 2302 | ret = lttng_poll_wait(&events, -1); |
28ab034a | 2303 | DBG("Metadata poll return from wait with %d fd(s)", LTTNG_POLL_GETNB(&events)); |
9ce5646a | 2304 | health_poll_exit(); |
40063ead | 2305 | DBG("Metadata event caught in thread"); |
fb3a43a9 DG |
2306 | if (ret < 0) { |
2307 | if (errno == EINTR) { | |
40063ead | 2308 | ERR("Poll EINTR caught"); |
fb3a43a9 DG |
2309 | goto restart; |
2310 | } | |
d9607cd7 | 2311 | if (LTTNG_POLL_GETNB(&events) == 0) { |
28ab034a | 2312 | err = 0; /* All is OK */ |
d9607cd7 MD |
2313 | } |
2314 | goto end; | |
fb3a43a9 DG |
2315 | } |
2316 | ||
0d9c5d77 DG |
2317 | nb_fd = ret; |
2318 | ||
e316aad5 | 2319 | /* From here, the event is a metadata wait fd */ |
fb3a43a9 | 2320 | for (i = 0; i < nb_fd; i++) { |
9ce5646a MD |
2321 | health_code_update(); |
2322 | ||
fb3a43a9 DG |
2323 | revents = LTTNG_POLL_GETEV(&events, i); |
2324 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
2325 | ||
13886d2d | 2326 | if (pollfd == lttng_pipe_get_readfd(ctx->consumer_metadata_pipe)) { |
03e43155 | 2327 | if (revents & LPOLLIN) { |
13886d2d DG |
2328 | ssize_t pipe_len; |
2329 | ||
2330 | pipe_len = lttng_pipe_read(ctx->consumer_metadata_pipe, | |
28ab034a | 2331 | &stream, |
5c7248cd JG |
2332 | sizeof(stream)); /* NOLINT sizeof |
2333 | used on a | |
2334 | pointer. */ | |
2335 | if (pipe_len < sizeof(stream)) { /* NOLINT sizeof used on a | |
2336 | pointer. */ | |
03e43155 MD |
2337 | if (pipe_len < 0) { |
2338 | PERROR("read metadata stream"); | |
2339 | } | |
fb3a43a9 | 2340 | /* |
28ab034a JG |
2341 | * Remove the pipe from the poll set and continue |
2342 | * the loop since their might be data to consume. | |
fb3a43a9 | 2343 | */ |
28ab034a JG |
2344 | lttng_poll_del( |
2345 | &events, | |
2346 | lttng_pipe_get_readfd( | |
2347 | ctx->consumer_metadata_pipe)); | |
03e43155 | 2348 | lttng_pipe_read_close(ctx->consumer_metadata_pipe); |
fb3a43a9 DG |
2349 | continue; |
2350 | } | |
2351 | ||
8994307f | 2352 | /* A NULL stream means that the state has changed. */ |
cd9adb8b | 2353 | if (stream == nullptr) { |
8994307f DG |
2354 | /* Check for deleted streams. */ |
2355 | validate_endpoint_status_metadata_stream(&events); | |
3714380f | 2356 | goto restart; |
8994307f DG |
2357 | } |
2358 | ||
fb3a43a9 | 2359 | DBG("Adding metadata stream %d to poll set", |
28ab034a | 2360 | stream->wait_fd); |
fb3a43a9 | 2361 | |
fb3a43a9 | 2362 | /* Add metadata stream to the global poll events list */ |
28ab034a JG |
2363 | lttng_poll_add( |
2364 | &events, stream->wait_fd, LPOLLIN | LPOLLPRI); | |
2365 | } else if (revents & (LPOLLERR | LPOLLHUP)) { | |
03e43155 MD |
2366 | DBG("Metadata thread pipe hung up"); |
2367 | /* | |
2368 | * Remove the pipe from the poll set and continue the loop | |
2369 | * since their might be data to consume. | |
2370 | */ | |
28ab034a JG |
2371 | lttng_poll_del( |
2372 | &events, | |
2373 | lttng_pipe_get_readfd(ctx->consumer_metadata_pipe)); | |
03e43155 MD |
2374 | lttng_pipe_read_close(ctx->consumer_metadata_pipe); |
2375 | continue; | |
2376 | } else { | |
28ab034a JG |
2377 | ERR("Unexpected poll events %u for sock %d", |
2378 | revents, | |
2379 | pollfd); | |
03e43155 | 2380 | goto end; |
fb3a43a9 DG |
2381 | } |
2382 | ||
e316aad5 | 2383 | /* Handle other stream */ |
fb3a43a9 DG |
2384 | continue; |
2385 | } | |
2386 | ||
07c4863f | 2387 | const lttng::urcu::read_lock_guard read_lock; |
d88aee68 DG |
2388 | { |
2389 | uint64_t tmp_id = (uint64_t) pollfd; | |
2390 | ||
2391 | lttng_ht_lookup(metadata_ht, &tmp_id, &iter); | |
2392 | } | |
00d7d903 | 2393 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
a0377dfe | 2394 | LTTNG_ASSERT(node); |
fb3a43a9 | 2395 | |
ac2c50cd | 2396 | stream = lttng::utils::container_of(node, <tng_consumer_stream::node); |
fb3a43a9 | 2397 | |
03e43155 MD |
2398 | if (revents & (LPOLLIN | LPOLLPRI)) { |
2399 | /* Get the data out of the metadata file descriptor */ | |
2400 | DBG("Metadata available on fd %d", pollfd); | |
a0377dfe | 2401 | LTTNG_ASSERT(stream->wait_fd == pollfd); |
03e43155 MD |
2402 | |
2403 | do { | |
2404 | health_code_update(); | |
2405 | ||
6f9449c2 | 2406 | len = ctx->on_buffer_ready(stream, ctx, false); |
03e43155 MD |
2407 | /* |
2408 | * We don't check the return value here since if we get | |
83f4233d | 2409 | * a negative len, it means an error occurred thus we |
03e43155 MD |
2410 | * simply remove it from the poll set and free the |
2411 | * stream. | |
2412 | */ | |
2413 | } while (len > 0); | |
2414 | ||
2415 | /* It's ok to have an unavailable sub-buffer */ | |
2416 | if (len < 0 && len != -EAGAIN && len != -ENODATA) { | |
2417 | /* Clean up stream from consumer and free it. */ | |
2418 | lttng_poll_del(&events, stream->wait_fd); | |
2419 | consumer_del_metadata_stream(stream, metadata_ht); | |
2420 | } | |
2421 | } else if (revents & (LPOLLERR | LPOLLHUP)) { | |
e316aad5 | 2422 | DBG("Metadata fd %d is hup|err.", pollfd); |
fa29bfbf | 2423 | if (!stream->hangup_flush_done && |
28ab034a JG |
2424 | (the_consumer_data.type == LTTNG_CONSUMER32_UST || |
2425 | the_consumer_data.type == LTTNG_CONSUMER64_UST)) { | |
fb3a43a9 DG |
2426 | DBG("Attempting to flush and consume the UST buffers"); |
2427 | lttng_ustconsumer_on_stream_hangup(stream); | |
2428 | ||
2429 | /* We just flushed the stream now read it. */ | |
4bb94b75 | 2430 | do { |
9ce5646a MD |
2431 | health_code_update(); |
2432 | ||
6f9449c2 | 2433 | len = ctx->on_buffer_ready(stream, ctx, false); |
4bb94b75 | 2434 | /* |
28ab034a JG |
2435 | * We don't check the return value here since if we |
2436 | * get a negative len, it means an error occurred | |
2437 | * thus we simply remove it from the poll set and | |
2438 | * free the stream. | |
4bb94b75 DG |
2439 | */ |
2440 | } while (len > 0); | |
fb3a43a9 DG |
2441 | } |
2442 | ||
fb3a43a9 | 2443 | lttng_poll_del(&events, stream->wait_fd); |
e316aad5 DG |
2444 | /* |
2445 | * This call update the channel states, closes file descriptors | |
2446 | * and securely free the stream. | |
2447 | */ | |
2448 | consumer_del_metadata_stream(stream, metadata_ht); | |
03e43155 MD |
2449 | } else { |
2450 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
03e43155 | 2451 | goto end; |
fb3a43a9 | 2452 | } |
e316aad5 | 2453 | /* Release RCU lock for the stream looked up */ |
fb3a43a9 DG |
2454 | } |
2455 | } | |
2456 | ||
1fc79fb4 MD |
2457 | /* All is OK */ |
2458 | err = 0; | |
fb3a43a9 DG |
2459 | end: |
2460 | DBG("Metadata poll thread exiting"); | |
fb3a43a9 | 2461 | |
d8ef542d MD |
2462 | lttng_poll_clean(&events); |
2463 | end_poll: | |
2d57de81 | 2464 | error_testpoint: |
1fc79fb4 MD |
2465 | if (err) { |
2466 | health_error(); | |
2467 | ERR("Health error occurred in %s", __func__); | |
2468 | } | |
2469 | health_unregister(health_consumerd); | |
fb3a43a9 | 2470 | rcu_unregister_thread(); |
cd9adb8b | 2471 | return nullptr; |
fb3a43a9 DG |
2472 | } |
2473 | ||
3bd1e081 | 2474 | /* |
e4421fec | 2475 | * This thread polls the fds in the set to consume the data and write |
3bd1e081 MD |
2476 | * it to tracefile if necessary. |
2477 | */ | |
7d980def | 2478 | void *consumer_thread_data_poll(void *data) |
3bd1e081 | 2479 | { |
ff930959 | 2480 | int num_rdy, high_prio, ret, i, err = -1; |
cd9adb8b | 2481 | struct pollfd *pollfd = nullptr; |
3bd1e081 | 2482 | /* local view of the streams */ |
cd9adb8b | 2483 | struct lttng_consumer_stream **local_stream = nullptr, *new_stream = nullptr; |
3bd1e081 | 2484 | /* local view of consumer_data.fds_count */ |
8bdcc002 JG |
2485 | int nb_fd = 0; |
2486 | /* 2 for the consumer_data_pipe and wake up pipe */ | |
2487 | const int nb_pipes_fd = 2; | |
9a2fcf78 JD |
2488 | /* Number of FDs with CONSUMER_ENDPOINT_INACTIVE but still open. */ |
2489 | int nb_inactive_fd = 0; | |
97535efa | 2490 | struct lttng_consumer_local_data *ctx = (lttng_consumer_local_data *) data; |
00e2e675 | 2491 | ssize_t len; |
3bd1e081 | 2492 | |
e7b994a3 DG |
2493 | rcu_register_thread(); |
2494 | ||
1fc79fb4 MD |
2495 | health_register(health_consumerd, HEALTH_CONSUMERD_TYPE_DATA); |
2496 | ||
2d57de81 MD |
2497 | if (testpoint(consumerd_thread_data)) { |
2498 | goto error_testpoint; | |
2499 | } | |
2500 | ||
9ce5646a MD |
2501 | health_code_update(); |
2502 | ||
64803277 | 2503 | local_stream = zmalloc<lttng_consumer_stream *>(); |
cd9adb8b | 2504 | if (local_stream == nullptr) { |
4df6c8cb MD |
2505 | PERROR("local_stream malloc"); |
2506 | goto end; | |
2507 | } | |
3bd1e081 | 2508 | |
cd9adb8b | 2509 | while (true) { |
9ce5646a MD |
2510 | health_code_update(); |
2511 | ||
3bd1e081 | 2512 | high_prio = 0; |
3bd1e081 MD |
2513 | |
2514 | /* | |
e4421fec | 2515 | * the fds set has been updated, we need to update our |
3bd1e081 MD |
2516 | * local array as well |
2517 | */ | |
fa29bfbf SM |
2518 | pthread_mutex_lock(&the_consumer_data.lock); |
2519 | if (the_consumer_data.need_update) { | |
0e428499 | 2520 | free(pollfd); |
cd9adb8b | 2521 | pollfd = nullptr; |
0e428499 DG |
2522 | |
2523 | free(local_stream); | |
cd9adb8b | 2524 | local_stream = nullptr; |
3bd1e081 | 2525 | |
8bdcc002 | 2526 | /* Allocate for all fds */ |
28ab034a JG |
2527 | pollfd = |
2528 | calloc<struct pollfd>(the_consumer_data.stream_count + nb_pipes_fd); | |
cd9adb8b | 2529 | if (pollfd == nullptr) { |
7a57cf92 | 2530 | PERROR("pollfd malloc"); |
fa29bfbf | 2531 | pthread_mutex_unlock(&the_consumer_data.lock); |
3bd1e081 MD |
2532 | goto end; |
2533 | } | |
2534 | ||
28ab034a JG |
2535 | local_stream = calloc<lttng_consumer_stream *>( |
2536 | the_consumer_data.stream_count + nb_pipes_fd); | |
cd9adb8b | 2537 | if (local_stream == nullptr) { |
7a57cf92 | 2538 | PERROR("local_stream malloc"); |
fa29bfbf | 2539 | pthread_mutex_unlock(&the_consumer_data.lock); |
3bd1e081 MD |
2540 | goto end; |
2541 | } | |
28ab034a JG |
2542 | ret = update_poll_array( |
2543 | ctx, &pollfd, local_stream, data_ht, &nb_inactive_fd); | |
3bd1e081 MD |
2544 | if (ret < 0) { |
2545 | ERR("Error in allocating pollfd or local_outfds"); | |
f73fabfd | 2546 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_POLL_ERROR); |
fa29bfbf | 2547 | pthread_mutex_unlock(&the_consumer_data.lock); |
3bd1e081 MD |
2548 | goto end; |
2549 | } | |
2550 | nb_fd = ret; | |
fa29bfbf | 2551 | the_consumer_data.need_update = 0; |
3bd1e081 | 2552 | } |
fa29bfbf | 2553 | pthread_mutex_unlock(&the_consumer_data.lock); |
3bd1e081 | 2554 | |
4078b776 | 2555 | /* No FDs and consumer_quit, consumer_cleanup the thread */ |
28ab034a JG |
2556 | if (nb_fd == 0 && nb_inactive_fd == 0 && CMM_LOAD_SHARED(consumer_quit) == 1) { |
2557 | err = 0; /* All is OK */ | |
4078b776 MD |
2558 | goto end; |
2559 | } | |
3bd1e081 | 2560 | /* poll on the array of fds */ |
88f2b785 | 2561 | restart: |
261de637 | 2562 | DBG("polling on %d fd", nb_fd + nb_pipes_fd); |
cf0bcb51 JG |
2563 | if (testpoint(consumerd_thread_data_poll)) { |
2564 | goto end; | |
2565 | } | |
9ce5646a | 2566 | health_poll_entry(); |
261de637 | 2567 | num_rdy = poll(pollfd, nb_fd + nb_pipes_fd, -1); |
9ce5646a | 2568 | health_poll_exit(); |
3bd1e081 MD |
2569 | DBG("poll num_rdy : %d", num_rdy); |
2570 | if (num_rdy == -1) { | |
88f2b785 MD |
2571 | /* |
2572 | * Restart interrupted system call. | |
2573 | */ | |
2574 | if (errno == EINTR) { | |
2575 | goto restart; | |
2576 | } | |
7a57cf92 | 2577 | PERROR("Poll error"); |
f73fabfd | 2578 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_POLL_ERROR); |
3bd1e081 MD |
2579 | goto end; |
2580 | } else if (num_rdy == 0) { | |
2581 | DBG("Polling thread timed out"); | |
2582 | goto end; | |
2583 | } | |
2584 | ||
80957876 JG |
2585 | if (caa_unlikely(data_consumption_paused)) { |
2586 | DBG("Data consumption paused, sleeping..."); | |
2587 | sleep(1); | |
2588 | goto restart; | |
2589 | } | |
2590 | ||
3bd1e081 | 2591 | /* |
50f8ae69 | 2592 | * If the consumer_data_pipe triggered poll go directly to the |
00e2e675 DG |
2593 | * beginning of the loop to update the array. We want to prioritize |
2594 | * array update over low-priority reads. | |
3bd1e081 | 2595 | */ |
509bb1cf | 2596 | if (pollfd[nb_fd].revents & (POLLIN | POLLPRI)) { |
ab30f567 | 2597 | ssize_t pipe_readlen; |
04fdd819 | 2598 | |
50f8ae69 | 2599 | DBG("consumer_data_pipe wake up"); |
5c7248cd JG |
2600 | pipe_readlen = lttng_pipe_read(ctx->consumer_data_pipe, |
2601 | &new_stream, | |
2602 | sizeof(new_stream)); /* NOLINT sizeof used on | |
2603 | a pointer. */ | |
2604 | if (pipe_readlen < sizeof(new_stream)) { /* NOLINT sizeof used on a pointer. | |
2605 | */ | |
6cd525e8 | 2606 | PERROR("Consumer data pipe"); |
23f5f35d DG |
2607 | /* Continue so we can at least handle the current stream(s). */ |
2608 | continue; | |
2609 | } | |
c869f647 DG |
2610 | |
2611 | /* | |
2612 | * If the stream is NULL, just ignore it. It's also possible that | |
2613 | * the sessiond poll thread changed the consumer_quit state and is | |
2614 | * waking us up to test it. | |
2615 | */ | |
cd9adb8b | 2616 | if (new_stream == nullptr) { |
8994307f | 2617 | validate_endpoint_status_data_stream(); |
c869f647 DG |
2618 | continue; |
2619 | } | |
2620 | ||
c869f647 | 2621 | /* Continue to update the local streams and handle prio ones */ |
3bd1e081 MD |
2622 | continue; |
2623 | } | |
2624 | ||
02b3d176 DG |
2625 | /* Handle wakeup pipe. */ |
2626 | if (pollfd[nb_fd + 1].revents & (POLLIN | POLLPRI)) { | |
2627 | char dummy; | |
2628 | ssize_t pipe_readlen; | |
2629 | ||
28ab034a JG |
2630 | pipe_readlen = |
2631 | lttng_pipe_read(ctx->consumer_wakeup_pipe, &dummy, sizeof(dummy)); | |
02b3d176 DG |
2632 | if (pipe_readlen < 0) { |
2633 | PERROR("Consumer data wakeup pipe"); | |
2634 | } | |
2635 | /* We've been awakened to handle stream(s). */ | |
2636 | ctx->has_wakeup = 0; | |
2637 | } | |
2638 | ||
3bd1e081 MD |
2639 | /* Take care of high priority channels first. */ |
2640 | for (i = 0; i < nb_fd; i++) { | |
9ce5646a MD |
2641 | health_code_update(); |
2642 | ||
cd9adb8b | 2643 | if (local_stream[i] == nullptr) { |
9617607b DG |
2644 | continue; |
2645 | } | |
fb3a43a9 | 2646 | if (pollfd[i].revents & POLLPRI) { |
d41f73b7 MD |
2647 | DBG("Urgent read on fd %d", pollfd[i].fd); |
2648 | high_prio = 1; | |
6f9449c2 | 2649 | len = ctx->on_buffer_ready(local_stream[i], ctx, false); |
d41f73b7 | 2650 | /* it's ok to have an unavailable sub-buffer */ |
b64403e3 | 2651 | if (len < 0 && len != -EAGAIN && len != -ENODATA) { |
ab1027f4 DG |
2652 | /* Clean the stream and free it. */ |
2653 | consumer_del_stream(local_stream[i], data_ht); | |
cd9adb8b | 2654 | local_stream[i] = nullptr; |
4078b776 | 2655 | } else if (len > 0) { |
28ab034a JG |
2656 | local_stream[i]->has_data_left_to_be_read_before_teardown = |
2657 | 1; | |
d41f73b7 | 2658 | } |
3bd1e081 MD |
2659 | } |
2660 | } | |
2661 | ||
4078b776 MD |
2662 | /* |
2663 | * If we read high prio channel in this loop, try again | |
2664 | * for more high prio data. | |
2665 | */ | |
2666 | if (high_prio) { | |
3bd1e081 MD |
2667 | continue; |
2668 | } | |
2669 | ||
2670 | /* Take care of low priority channels. */ | |
4078b776 | 2671 | for (i = 0; i < nb_fd; i++) { |
9ce5646a MD |
2672 | health_code_update(); |
2673 | ||
cd9adb8b | 2674 | if (local_stream[i] == nullptr) { |
9617607b DG |
2675 | continue; |
2676 | } | |
28ab034a JG |
2677 | if ((pollfd[i].revents & POLLIN) || local_stream[i]->hangup_flush_done || |
2678 | local_stream[i]->has_data) { | |
4078b776 | 2679 | DBG("Normal read on fd %d", pollfd[i].fd); |
6f9449c2 | 2680 | len = ctx->on_buffer_ready(local_stream[i], ctx, false); |
4078b776 | 2681 | /* it's ok to have an unavailable sub-buffer */ |
b64403e3 | 2682 | if (len < 0 && len != -EAGAIN && len != -ENODATA) { |
ab1027f4 DG |
2683 | /* Clean the stream and free it. */ |
2684 | consumer_del_stream(local_stream[i], data_ht); | |
cd9adb8b | 2685 | local_stream[i] = nullptr; |
4078b776 | 2686 | } else if (len > 0) { |
28ab034a JG |
2687 | local_stream[i]->has_data_left_to_be_read_before_teardown = |
2688 | 1; | |
4078b776 MD |
2689 | } |
2690 | } | |
2691 | } | |
2692 | ||
2693 | /* Handle hangup and errors */ | |
2694 | for (i = 0; i < nb_fd; i++) { | |
9ce5646a MD |
2695 | health_code_update(); |
2696 | ||
cd9adb8b | 2697 | if (local_stream[i] == nullptr) { |
9617607b DG |
2698 | continue; |
2699 | } | |
28ab034a JG |
2700 | if (!local_stream[i]->hangup_flush_done && |
2701 | (pollfd[i].revents & (POLLHUP | POLLERR | POLLNVAL)) && | |
2702 | (the_consumer_data.type == LTTNG_CONSUMER32_UST || | |
2703 | the_consumer_data.type == LTTNG_CONSUMER64_UST)) { | |
4078b776 | 2704 | DBG("fd %d is hup|err|nval. Attempting flush and read.", |
28ab034a | 2705 | pollfd[i].fd); |
4078b776 MD |
2706 | lttng_ustconsumer_on_stream_hangup(local_stream[i]); |
2707 | /* Attempt read again, for the data we just flushed. */ | |
c715ddc9 | 2708 | local_stream[i]->has_data_left_to_be_read_before_teardown = 1; |
4078b776 MD |
2709 | } |
2710 | /* | |
c715ddc9 JG |
2711 | * When a stream's pipe dies (hup/err/nval), an "inactive producer" flush is |
2712 | * performed. This type of flush ensures that a new packet is produced no | |
2713 | * matter the consumed/produced positions are. | |
2714 | * | |
2715 | * This, in turn, causes the next pass to see that data available for the | |
2716 | * stream. When we come back here, we can be assured that all available | |
2717 | * data has been consumed and we can finally destroy the stream. | |
2718 | * | |
4078b776 MD |
2719 | * If the poll flag is HUP/ERR/NVAL and we have |
2720 | * read no data in this pass, we can remove the | |
2721 | * stream from its hash table. | |
2722 | */ | |
2723 | if ((pollfd[i].revents & POLLHUP)) { | |
2724 | DBG("Polling fd %d tells it has hung up.", pollfd[i].fd); | |
c715ddc9 | 2725 | if (!local_stream[i]->has_data_left_to_be_read_before_teardown) { |
43c34bc3 | 2726 | consumer_del_stream(local_stream[i], data_ht); |
cd9adb8b | 2727 | local_stream[i] = nullptr; |
4078b776 MD |
2728 | } |
2729 | } else if (pollfd[i].revents & POLLERR) { | |
2730 | ERR("Error returned in polling fd %d.", pollfd[i].fd); | |
c715ddc9 | 2731 | if (!local_stream[i]->has_data_left_to_be_read_before_teardown) { |
43c34bc3 | 2732 | consumer_del_stream(local_stream[i], data_ht); |
cd9adb8b | 2733 | local_stream[i] = nullptr; |
4078b776 MD |
2734 | } |
2735 | } else if (pollfd[i].revents & POLLNVAL) { | |
2736 | ERR("Polling fd %d tells fd is not open.", pollfd[i].fd); | |
c715ddc9 | 2737 | if (!local_stream[i]->has_data_left_to_be_read_before_teardown) { |
43c34bc3 | 2738 | consumer_del_stream(local_stream[i], data_ht); |
cd9adb8b | 2739 | local_stream[i] = nullptr; |
3bd1e081 MD |
2740 | } |
2741 | } | |
cd9adb8b | 2742 | if (local_stream[i] != nullptr) { |
c715ddc9 | 2743 | local_stream[i]->has_data_left_to_be_read_before_teardown = 0; |
9617607b | 2744 | } |
3bd1e081 MD |
2745 | } |
2746 | } | |
1fc79fb4 MD |
2747 | /* All is OK */ |
2748 | err = 0; | |
3bd1e081 MD |
2749 | end: |
2750 | DBG("polling thread exiting"); | |
0e428499 DG |
2751 | free(pollfd); |
2752 | free(local_stream); | |
fb3a43a9 DG |
2753 | |
2754 | /* | |
2755 | * Close the write side of the pipe so epoll_wait() in | |
7d980def DG |
2756 | * consumer_thread_metadata_poll can catch it. The thread is monitoring the |
2757 | * read side of the pipe. If we close them both, epoll_wait strangely does | |
2758 | * not return and could create a endless wait period if the pipe is the | |
2759 | * only tracked fd in the poll set. The thread will take care of closing | |
2760 | * the read side. | |
fb3a43a9 | 2761 | */ |
13886d2d | 2762 | (void) lttng_pipe_write_close(ctx->consumer_metadata_pipe); |
fb3a43a9 | 2763 | |
2d57de81 | 2764 | error_testpoint: |
1fc79fb4 MD |
2765 | if (err) { |
2766 | health_error(); | |
2767 | ERR("Health error occurred in %s", __func__); | |
2768 | } | |
2769 | health_unregister(health_consumerd); | |
2770 | ||
e7b994a3 | 2771 | rcu_unregister_thread(); |
cd9adb8b | 2772 | return nullptr; |
3bd1e081 MD |
2773 | } |
2774 | ||
d8ef542d MD |
2775 | /* |
2776 | * Close wake-up end of each stream belonging to the channel. This will | |
2777 | * allow the poll() on the stream read-side to detect when the | |
2778 | * write-side (application) finally closes them. | |
2779 | */ | |
28ab034a | 2780 | static void consumer_close_channel_streams(struct lttng_consumer_channel *channel) |
d8ef542d | 2781 | { |
3d46ea1a JG |
2782 | const auto ht = the_consumer_data.stream_per_chan_id_ht; |
2783 | ||
2784 | for (auto *stream : lttng::urcu::lfht_filtered_iteration_adapter< | |
2785 | lttng_consumer_stream, | |
2786 | decltype(lttng_consumer_stream::node_channel_id), | |
2787 | <tng_consumer_stream::node_channel_id, | |
2788 | std::uint64_t>(*ht->ht, | |
2789 | &channel->key, | |
2790 | ht->hash_fct(&channel->key, lttng_ht_seed), | |
2791 | ht->match_fct)) { | |
f2ad556d MD |
2792 | /* |
2793 | * Protect against teardown with mutex. | |
2794 | */ | |
2795 | pthread_mutex_lock(&stream->lock); | |
2796 | if (cds_lfht_is_node_deleted(&stream->node.node)) { | |
2797 | goto next; | |
2798 | } | |
fa29bfbf | 2799 | switch (the_consumer_data.type) { |
d8ef542d MD |
2800 | case LTTNG_CONSUMER_KERNEL: |
2801 | break; | |
2802 | case LTTNG_CONSUMER32_UST: | |
2803 | case LTTNG_CONSUMER64_UST: | |
b4a650f3 DG |
2804 | if (stream->metadata_flag) { |
2805 | /* Safe and protected by the stream lock. */ | |
2806 | lttng_ustconsumer_close_metadata(stream->chan); | |
2807 | } else { | |
2808 | /* | |
2809 | * Note: a mutex is taken internally within | |
2810 | * liblttng-ust-ctl to protect timer wakeup_fd | |
2811 | * use from concurrent close. | |
2812 | */ | |
2813 | lttng_ustconsumer_close_stream_wakeup(stream); | |
2814 | } | |
d8ef542d MD |
2815 | break; |
2816 | default: | |
2817 | ERR("Unknown consumer_data type"); | |
a0377dfe | 2818 | abort(); |
d8ef542d | 2819 | } |
f2ad556d MD |
2820 | next: |
2821 | pthread_mutex_unlock(&stream->lock); | |
d8ef542d | 2822 | } |
d8ef542d MD |
2823 | } |
2824 | ||
2825 | static void destroy_channel_ht(struct lttng_ht *ht) | |
2826 | { | |
cd9adb8b | 2827 | if (ht == nullptr) { |
d8ef542d MD |
2828 | return; |
2829 | } | |
2830 | ||
c3ade133 JG |
2831 | for (auto *channel : |
2832 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_channel, | |
2833 | decltype(lttng_consumer_channel::wait_fd_node), | |
2834 | <tng_consumer_channel::wait_fd_node>(*ht->ht)) { | |
2835 | const auto ret = cds_lfht_del(ht->ht, &channel->node.node); | |
2836 | LTTNG_ASSERT(ret != 0); | |
d8ef542d | 2837 | } |
d8ef542d MD |
2838 | |
2839 | lttng_ht_destroy(ht); | |
2840 | } | |
2841 | ||
2842 | /* | |
2843 | * This thread polls the channel fds to detect when they are being | |
2844 | * closed. It closes all related streams if the channel is detected as | |
2845 | * closed. It is currently only used as a shim layer for UST because the | |
2846 | * consumerd needs to keep the per-stream wakeup end of pipes open for | |
2847 | * periodical flush. | |
2848 | */ | |
2849 | void *consumer_thread_channel_poll(void *data) | |
2850 | { | |
1fc79fb4 | 2851 | int ret, i, pollfd, err = -1; |
d8ef542d | 2852 | uint32_t revents, nb_fd; |
cd9adb8b | 2853 | struct lttng_consumer_channel *chan = nullptr; |
d8ef542d MD |
2854 | struct lttng_ht_iter iter; |
2855 | struct lttng_ht_node_u64 *node; | |
2856 | struct lttng_poll_event events; | |
97535efa | 2857 | struct lttng_consumer_local_data *ctx = (lttng_consumer_local_data *) data; |
d8ef542d MD |
2858 | struct lttng_ht *channel_ht; |
2859 | ||
2860 | rcu_register_thread(); | |
2861 | ||
1fc79fb4 MD |
2862 | health_register(health_consumerd, HEALTH_CONSUMERD_TYPE_CHANNEL); |
2863 | ||
2d57de81 MD |
2864 | if (testpoint(consumerd_thread_channel)) { |
2865 | goto error_testpoint; | |
2866 | } | |
2867 | ||
9ce5646a MD |
2868 | health_code_update(); |
2869 | ||
d8ef542d MD |
2870 | channel_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2871 | if (!channel_ht) { | |
2872 | /* ENOMEM at this point. Better to bail out. */ | |
2873 | goto end_ht; | |
2874 | } | |
2875 | ||
2876 | DBG("Thread channel poll started"); | |
2877 | ||
2878 | /* Size is set to 1 for the consumer_channel pipe */ | |
2879 | ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC); | |
2880 | if (ret < 0) { | |
2881 | ERR("Poll set creation failed"); | |
2882 | goto end_poll; | |
2883 | } | |
2884 | ||
2885 | ret = lttng_poll_add(&events, ctx->consumer_channel_pipe[0], LPOLLIN); | |
2886 | if (ret < 0) { | |
2887 | goto end; | |
2888 | } | |
2889 | ||
2890 | /* Main loop */ | |
2891 | DBG("Channel main loop started"); | |
2892 | ||
cd9adb8b | 2893 | while (true) { |
28ab034a | 2894 | restart: |
7fa2082e MD |
2895 | health_code_update(); |
2896 | DBG("Channel poll wait"); | |
9ce5646a | 2897 | health_poll_entry(); |
d8ef542d | 2898 | ret = lttng_poll_wait(&events, -1); |
28ab034a | 2899 | DBG("Channel poll return from wait with %d fd(s)", LTTNG_POLL_GETNB(&events)); |
9ce5646a | 2900 | health_poll_exit(); |
40063ead | 2901 | DBG("Channel event caught in thread"); |
d8ef542d MD |
2902 | if (ret < 0) { |
2903 | if (errno == EINTR) { | |
40063ead | 2904 | ERR("Poll EINTR caught"); |
d8ef542d MD |
2905 | goto restart; |
2906 | } | |
d9607cd7 | 2907 | if (LTTNG_POLL_GETNB(&events) == 0) { |
28ab034a | 2908 | err = 0; /* All is OK */ |
d9607cd7 | 2909 | } |
d8ef542d MD |
2910 | goto end; |
2911 | } | |
2912 | ||
2913 | nb_fd = ret; | |
2914 | ||
2915 | /* From here, the event is a channel wait fd */ | |
2916 | for (i = 0; i < nb_fd; i++) { | |
9ce5646a MD |
2917 | health_code_update(); |
2918 | ||
d8ef542d MD |
2919 | revents = LTTNG_POLL_GETEV(&events, i); |
2920 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
2921 | ||
d8ef542d | 2922 | if (pollfd == ctx->consumer_channel_pipe[0]) { |
03e43155 | 2923 | if (revents & LPOLLIN) { |
d8ef542d | 2924 | enum consumer_channel_action action; |
a0cbdd2e | 2925 | uint64_t key; |
d8ef542d | 2926 | |
a0cbdd2e | 2927 | ret = read_channel_pipe(ctx, &chan, &key, &action); |
d8ef542d | 2928 | if (ret <= 0) { |
03e43155 MD |
2929 | if (ret < 0) { |
2930 | ERR("Error reading channel pipe"); | |
2931 | } | |
28ab034a JG |
2932 | lttng_poll_del(&events, |
2933 | ctx->consumer_channel_pipe[0]); | |
d8ef542d MD |
2934 | continue; |
2935 | } | |
2936 | ||
2937 | switch (action) { | |
2938 | case CONSUMER_CHANNEL_ADD: | |
56047f5a | 2939 | { |
28ab034a | 2940 | DBG("Adding channel %d to poll set", chan->wait_fd); |
d8ef542d MD |
2941 | |
2942 | lttng_ht_node_init_u64(&chan->wait_fd_node, | |
28ab034a | 2943 | chan->wait_fd); |
07c4863f | 2944 | const lttng::urcu::read_lock_guard read_lock; |
d8ef542d | 2945 | lttng_ht_add_unique_u64(channel_ht, |
28ab034a | 2946 | &chan->wait_fd_node); |
d8ef542d | 2947 | /* Add channel to the global poll events list */ |
28ab034a JG |
2948 | // FIXME: Empty flag on a pipe pollset, this might |
2949 | // hang on FreeBSD. | |
1524f98c | 2950 | lttng_poll_add(&events, chan->wait_fd, 0); |
d8ef542d | 2951 | break; |
56047f5a | 2952 | } |
a0cbdd2e MD |
2953 | case CONSUMER_CHANNEL_DEL: |
2954 | { | |
b4a650f3 | 2955 | /* |
28ab034a JG |
2956 | * This command should never be called if the |
2957 | * channel has streams monitored by either the data | |
2958 | * or metadata thread. The consumer only notify this | |
2959 | * thread with a channel del. command if it receives | |
2960 | * a destroy channel command from the session daemon | |
2961 | * that send it if a command prior to the | |
2962 | * GET_CHANNEL failed. | |
b4a650f3 DG |
2963 | */ |
2964 | ||
07c4863f | 2965 | const lttng::urcu::read_lock_guard read_lock; |
a0cbdd2e MD |
2966 | chan = consumer_find_channel(key); |
2967 | if (!chan) { | |
28ab034a JG |
2968 | ERR("UST consumer get channel key %" PRIu64 |
2969 | " not found for del channel", | |
2970 | key); | |
a0cbdd2e MD |
2971 | break; |
2972 | } | |
2973 | lttng_poll_del(&events, chan->wait_fd); | |
f623cc0b | 2974 | iter.iter.node = &chan->wait_fd_node.node; |
a0cbdd2e | 2975 | ret = lttng_ht_del(channel_ht, &iter); |
a0377dfe | 2976 | LTTNG_ASSERT(ret == 0); |
a0cbdd2e | 2977 | |
fa29bfbf | 2978 | switch (the_consumer_data.type) { |
f2a444f1 DG |
2979 | case LTTNG_CONSUMER_KERNEL: |
2980 | break; | |
2981 | case LTTNG_CONSUMER32_UST: | |
2982 | case LTTNG_CONSUMER64_UST: | |
212d67a2 | 2983 | health_code_update(); |
28ab034a JG |
2984 | /* Destroy streams that might have been left |
2985 | * in the stream list. */ | |
212d67a2 | 2986 | clean_channel_stream_list(chan); |
f2a444f1 DG |
2987 | break; |
2988 | default: | |
2989 | ERR("Unknown consumer_data type"); | |
a0377dfe | 2990 | abort(); |
f2a444f1 DG |
2991 | } |
2992 | ||
a0cbdd2e | 2993 | /* |
28ab034a JG |
2994 | * Release our own refcount. Force channel deletion |
2995 | * even if streams were not initialized. | |
a0cbdd2e MD |
2996 | */ |
2997 | if (!uatomic_sub_return(&chan->refcount, 1)) { | |
2998 | consumer_del_channel(chan); | |
2999 | } | |
3000 | goto restart; | |
3001 | } | |
d8ef542d MD |
3002 | case CONSUMER_CHANNEL_QUIT: |
3003 | /* | |
28ab034a JG |
3004 | * Remove the pipe from the poll set and continue |
3005 | * the loop since their might be data to consume. | |
d8ef542d | 3006 | */ |
28ab034a JG |
3007 | lttng_poll_del(&events, |
3008 | ctx->consumer_channel_pipe[0]); | |
d8ef542d MD |
3009 | continue; |
3010 | default: | |
3011 | ERR("Unknown action"); | |
3012 | break; | |
3013 | } | |
03e43155 MD |
3014 | } else if (revents & (LPOLLERR | LPOLLHUP)) { |
3015 | DBG("Channel thread pipe hung up"); | |
3016 | /* | |
3017 | * Remove the pipe from the poll set and continue the loop | |
3018 | * since their might be data to consume. | |
3019 | */ | |
3020 | lttng_poll_del(&events, ctx->consumer_channel_pipe[0]); | |
3021 | continue; | |
3022 | } else { | |
28ab034a JG |
3023 | ERR("Unexpected poll events %u for sock %d", |
3024 | revents, | |
3025 | pollfd); | |
03e43155 | 3026 | goto end; |
d8ef542d MD |
3027 | } |
3028 | ||
3029 | /* Handle other stream */ | |
3030 | continue; | |
3031 | } | |
3032 | ||
07c4863f | 3033 | const lttng::urcu::read_lock_guard read_lock; |
d8ef542d MD |
3034 | { |
3035 | uint64_t tmp_id = (uint64_t) pollfd; | |
3036 | ||
3037 | lttng_ht_lookup(channel_ht, &tmp_id, &iter); | |
3038 | } | |
00d7d903 | 3039 | node = lttng_ht_iter_get_node<lttng_ht_node_u64>(&iter); |
a0377dfe | 3040 | LTTNG_ASSERT(node); |
d8ef542d | 3041 | |
c3ade133 JG |
3042 | chan = lttng::utils::container_of(node, |
3043 | <tng_consumer_channel::wait_fd_node); | |
d8ef542d MD |
3044 | |
3045 | /* Check for error event */ | |
3046 | if (revents & (LPOLLERR | LPOLLHUP)) { | |
3047 | DBG("Channel fd %d is hup|err.", pollfd); | |
3048 | ||
3049 | lttng_poll_del(&events, chan->wait_fd); | |
3050 | ret = lttng_ht_del(channel_ht, &iter); | |
a0377dfe | 3051 | LTTNG_ASSERT(ret == 0); |
b4a650f3 DG |
3052 | |
3053 | /* | |
3054 | * This will close the wait fd for each stream associated to | |
3055 | * this channel AND monitored by the data/metadata thread thus | |
3056 | * will be clean by the right thread. | |
3057 | */ | |
d8ef542d | 3058 | consumer_close_channel_streams(chan); |
f2ad556d MD |
3059 | |
3060 | /* Release our own refcount */ | |
28ab034a JG |
3061 | if (!uatomic_sub_return(&chan->refcount, 1) && |
3062 | !uatomic_read(&chan->nb_init_stream_left)) { | |
f2ad556d MD |
3063 | consumer_del_channel(chan); |
3064 | } | |
03e43155 MD |
3065 | } else { |
3066 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
03e43155 | 3067 | goto end; |
d8ef542d MD |
3068 | } |
3069 | ||
3070 | /* Release RCU lock for the channel looked up */ | |
d8ef542d MD |
3071 | } |
3072 | } | |
3073 | ||
1fc79fb4 MD |
3074 | /* All is OK */ |
3075 | err = 0; | |
d8ef542d MD |
3076 | end: |
3077 | lttng_poll_clean(&events); | |
3078 | end_poll: | |
3079 | destroy_channel_ht(channel_ht); | |
3080 | end_ht: | |
2d57de81 | 3081 | error_testpoint: |
d8ef542d | 3082 | DBG("Channel poll thread exiting"); |
1fc79fb4 MD |
3083 | if (err) { |
3084 | health_error(); | |
3085 | ERR("Health error occurred in %s", __func__); | |
3086 | } | |
3087 | health_unregister(health_consumerd); | |
d8ef542d | 3088 | rcu_unregister_thread(); |
cd9adb8b | 3089 | return nullptr; |
d8ef542d MD |
3090 | } |
3091 | ||
331744e3 | 3092 | static int set_metadata_socket(struct lttng_consumer_local_data *ctx, |
28ab034a JG |
3093 | struct pollfd *sockpoll, |
3094 | int client_socket) | |
331744e3 JD |
3095 | { |
3096 | int ret; | |
3097 | ||
a0377dfe FD |
3098 | LTTNG_ASSERT(ctx); |
3099 | LTTNG_ASSERT(sockpoll); | |
331744e3 | 3100 | |
84382d49 MD |
3101 | ret = lttng_consumer_poll_socket(sockpoll); |
3102 | if (ret) { | |
331744e3 JD |
3103 | goto error; |
3104 | } | |
3105 | DBG("Metadata connection on client_socket"); | |
3106 | ||
3107 | /* Blocking call, waiting for transmission */ | |
3108 | ctx->consumer_metadata_socket = lttcomm_accept_unix_sock(client_socket); | |
3109 | if (ctx->consumer_metadata_socket < 0) { | |
3110 | WARN("On accept metadata"); | |
3111 | ret = -1; | |
3112 | goto error; | |
3113 | } | |
3114 | ret = 0; | |
3115 | ||
3116 | error: | |
3117 | return ret; | |
3118 | } | |
3119 | ||
3bd1e081 MD |
3120 | /* |
3121 | * This thread listens on the consumerd socket and receives the file | |
3122 | * descriptors from the session daemon. | |
3123 | */ | |
7d980def | 3124 | void *consumer_thread_sessiond_poll(void *data) |
3bd1e081 | 3125 | { |
1fc79fb4 | 3126 | int sock = -1, client_socket, ret, err = -1; |
3bd1e081 MD |
3127 | /* |
3128 | * structure to poll for incoming data on communication socket avoids | |
3129 | * making blocking sockets. | |
3130 | */ | |
3131 | struct pollfd consumer_sockpoll[2]; | |
97535efa | 3132 | struct lttng_consumer_local_data *ctx = (lttng_consumer_local_data *) data; |
3bd1e081 | 3133 | |
e7b994a3 DG |
3134 | rcu_register_thread(); |
3135 | ||
1fc79fb4 MD |
3136 | health_register(health_consumerd, HEALTH_CONSUMERD_TYPE_SESSIOND); |
3137 | ||
2d57de81 MD |
3138 | if (testpoint(consumerd_thread_sessiond)) { |
3139 | goto error_testpoint; | |
3140 | } | |
3141 | ||
9ce5646a MD |
3142 | health_code_update(); |
3143 | ||
3bd1e081 MD |
3144 | DBG("Creating command socket %s", ctx->consumer_command_sock_path); |
3145 | unlink(ctx->consumer_command_sock_path); | |
3146 | client_socket = lttcomm_create_unix_sock(ctx->consumer_command_sock_path); | |
3147 | if (client_socket < 0) { | |
3148 | ERR("Cannot create command socket"); | |
3149 | goto end; | |
3150 | } | |
3151 | ||
3152 | ret = lttcomm_listen_unix_sock(client_socket); | |
3153 | if (ret < 0) { | |
3154 | goto end; | |
3155 | } | |
3156 | ||
32258573 | 3157 | DBG("Sending ready command to lttng-sessiond"); |
f73fabfd | 3158 | ret = lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_COMMAND_SOCK_READY); |
3bd1e081 MD |
3159 | /* return < 0 on error, but == 0 is not fatal */ |
3160 | if (ret < 0) { | |
32258573 | 3161 | ERR("Error sending ready command to lttng-sessiond"); |
3bd1e081 MD |
3162 | goto end; |
3163 | } | |
3164 | ||
3bd1e081 MD |
3165 | /* prepare the FDs to poll : to client socket and the should_quit pipe */ |
3166 | consumer_sockpoll[0].fd = ctx->consumer_should_quit[0]; | |
3167 | consumer_sockpoll[0].events = POLLIN | POLLPRI; | |
3168 | consumer_sockpoll[1].fd = client_socket; | |
3169 | consumer_sockpoll[1].events = POLLIN | POLLPRI; | |
3170 | ||
84382d49 MD |
3171 | ret = lttng_consumer_poll_socket(consumer_sockpoll); |
3172 | if (ret) { | |
3173 | if (ret > 0) { | |
3174 | /* should exit */ | |
3175 | err = 0; | |
3176 | } | |
3bd1e081 MD |
3177 | goto end; |
3178 | } | |
3179 | DBG("Connection on client_socket"); | |
3180 | ||
3181 | /* Blocking call, waiting for transmission */ | |
3182 | sock = lttcomm_accept_unix_sock(client_socket); | |
534d2592 | 3183 | if (sock < 0) { |
3bd1e081 MD |
3184 | WARN("On accept"); |
3185 | goto end; | |
3186 | } | |
3bd1e081 | 3187 | |
331744e3 JD |
3188 | /* |
3189 | * Setup metadata socket which is the second socket connection on the | |
3190 | * command unix socket. | |
3191 | */ | |
3192 | ret = set_metadata_socket(ctx, consumer_sockpoll, client_socket); | |
84382d49 MD |
3193 | if (ret) { |
3194 | if (ret > 0) { | |
3195 | /* should exit */ | |
3196 | err = 0; | |
3197 | } | |
331744e3 JD |
3198 | goto end; |
3199 | } | |
3200 | ||
d96f09c6 DG |
3201 | /* This socket is not useful anymore. */ |
3202 | ret = close(client_socket); | |
3203 | if (ret < 0) { | |
3204 | PERROR("close client_socket"); | |
3205 | } | |
3206 | client_socket = -1; | |
3207 | ||
3bd1e081 MD |
3208 | /* update the polling structure to poll on the established socket */ |
3209 | consumer_sockpoll[1].fd = sock; | |
3210 | consumer_sockpoll[1].events = POLLIN | POLLPRI; | |
3211 | ||
cd9adb8b | 3212 | while (true) { |
9ce5646a MD |
3213 | health_code_update(); |
3214 | ||
3215 | health_poll_entry(); | |
3216 | ret = lttng_consumer_poll_socket(consumer_sockpoll); | |
3217 | health_poll_exit(); | |
84382d49 MD |
3218 | if (ret) { |
3219 | if (ret > 0) { | |
3220 | /* should exit */ | |
3221 | err = 0; | |
3222 | } | |
3bd1e081 MD |
3223 | goto end; |
3224 | } | |
3225 | DBG("Incoming command on sock"); | |
3226 | ret = lttng_consumer_recv_cmd(ctx, sock, consumer_sockpoll); | |
4cbc1a04 DG |
3227 | if (ret <= 0) { |
3228 | /* | |
3229 | * This could simply be a session daemon quitting. Don't output | |
3230 | * ERR() here. | |
3231 | */ | |
3232 | DBG("Communication interrupted on command socket"); | |
41ba6035 | 3233 | err = 0; |
3bd1e081 MD |
3234 | goto end; |
3235 | } | |
10211f5c | 3236 | if (CMM_LOAD_SHARED(consumer_quit)) { |
3bd1e081 | 3237 | DBG("consumer_thread_receive_fds received quit from signal"); |
28ab034a | 3238 | err = 0; /* All is OK */ |
3bd1e081 MD |
3239 | goto end; |
3240 | } | |
a1ed855a | 3241 | DBG("Received command on sock"); |
3bd1e081 | 3242 | } |
1fc79fb4 MD |
3243 | /* All is OK */ |
3244 | err = 0; | |
3245 | ||
3bd1e081 | 3246 | end: |
ffe60014 | 3247 | DBG("Consumer thread sessiond poll exiting"); |
3bd1e081 | 3248 | |
d88aee68 DG |
3249 | /* |
3250 | * Close metadata streams since the producer is the session daemon which | |
3251 | * just died. | |
3252 | * | |
3253 | * NOTE: for now, this only applies to the UST tracer. | |
3254 | */ | |
6d574024 | 3255 | lttng_consumer_close_all_metadata(); |
d88aee68 | 3256 | |
3bd1e081 MD |
3257 | /* |
3258 | * when all fds have hung up, the polling thread | |
3259 | * can exit cleanly | |
3260 | */ | |
10211f5c | 3261 | CMM_STORE_SHARED(consumer_quit, 1); |
3bd1e081 | 3262 | |
04fdd819 | 3263 | /* |
c869f647 | 3264 | * Notify the data poll thread to poll back again and test the |
8994307f | 3265 | * consumer_quit state that we just set so to quit gracefully. |
04fdd819 | 3266 | */ |
acdb9057 | 3267 | notify_thread_lttng_pipe(ctx->consumer_data_pipe); |
c869f647 | 3268 | |
cd9adb8b | 3269 | notify_channel_pipe(ctx, nullptr, -1, CONSUMER_CHANNEL_QUIT); |
d8ef542d | 3270 | |
5c635c72 MD |
3271 | notify_health_quit_pipe(health_quit_pipe); |
3272 | ||
d96f09c6 DG |
3273 | /* Cleaning up possibly open sockets. */ |
3274 | if (sock >= 0) { | |
3275 | ret = close(sock); | |
3276 | if (ret < 0) { | |
3277 | PERROR("close sock sessiond poll"); | |
3278 | } | |
3279 | } | |
3280 | if (client_socket >= 0) { | |
38476d24 | 3281 | ret = close(client_socket); |
d96f09c6 DG |
3282 | if (ret < 0) { |
3283 | PERROR("close client_socket sessiond poll"); | |
3284 | } | |
3285 | } | |
3286 | ||
2d57de81 | 3287 | error_testpoint: |
1fc79fb4 MD |
3288 | if (err) { |
3289 | health_error(); | |
3290 | ERR("Health error occurred in %s", __func__); | |
3291 | } | |
3292 | health_unregister(health_consumerd); | |
3293 | ||
e7b994a3 | 3294 | rcu_unregister_thread(); |
cd9adb8b | 3295 | return nullptr; |
3bd1e081 | 3296 | } |
d41f73b7 | 3297 | |
503fefca | 3298 | static int post_consume(struct lttng_consumer_stream *stream, |
28ab034a JG |
3299 | const struct stream_subbuffer *subbuffer, |
3300 | struct lttng_consumer_local_data *ctx) | |
f96af312 | 3301 | { |
503fefca | 3302 | size_t i; |
f96af312 | 3303 | int ret = 0; |
28ab034a JG |
3304 | const size_t count = |
3305 | lttng_dynamic_array_get_count(&stream->read_subbuffer_ops.post_consume_cbs); | |
f96af312 | 3306 | |
503fefca JG |
3307 | for (i = 0; i < count; i++) { |
3308 | const post_consume_cb op = *(post_consume_cb *) lttng_dynamic_array_get_element( | |
28ab034a | 3309 | &stream->read_subbuffer_ops.post_consume_cbs, i); |
503fefca JG |
3310 | |
3311 | ret = op(stream, subbuffer, ctx); | |
3312 | if (ret) { | |
3313 | goto end; | |
f96af312 | 3314 | } |
f96af312 | 3315 | } |
f96af312 JG |
3316 | end: |
3317 | return ret; | |
3318 | } | |
3319 | ||
4078b776 | 3320 | ssize_t lttng_consumer_read_subbuffer(struct lttng_consumer_stream *stream, |
28ab034a JG |
3321 | struct lttng_consumer_local_data *ctx, |
3322 | bool locked_by_caller) | |
d41f73b7 | 3323 | { |
12bddd1d | 3324 | ssize_t ret, written_bytes = 0; |
23d56598 | 3325 | int rotation_ret; |
6f9449c2 | 3326 | struct stream_subbuffer subbuffer = {}; |
b6797c8e | 3327 | enum get_next_subbuffer_status get_next_status; |
74251bb8 | 3328 | |
6f9449c2 JG |
3329 | if (!locked_by_caller) { |
3330 | stream->read_subbuffer_ops.lock(stream); | |
947bd097 JR |
3331 | } else { |
3332 | stream->read_subbuffer_ops.assert_locked(stream); | |
6f9449c2 JG |
3333 | } |
3334 | ||
3335 | if (stream->read_subbuffer_ops.on_wake_up) { | |
3336 | ret = stream->read_subbuffer_ops.on_wake_up(stream); | |
3337 | if (ret) { | |
3338 | goto end; | |
3339 | } | |
94d49140 | 3340 | } |
74251bb8 | 3341 | |
23d56598 JG |
3342 | /* |
3343 | * If the stream was flagged to be ready for rotation before we extract | |
3344 | * the next packet, rotate it now. | |
3345 | */ | |
3346 | if (stream->rotate_ready) { | |
3347 | DBG("Rotate stream before consuming data"); | |
f46376a1 | 3348 | ret = lttng_consumer_rotate_stream(stream); |
23d56598 JG |
3349 | if (ret < 0) { |
3350 | ERR("Stream rotation error before consuming data"); | |
3351 | goto end; | |
3352 | } | |
3353 | } | |
3354 | ||
28ab034a | 3355 | get_next_status = stream->read_subbuffer_ops.get_next_subbuffer(stream, &subbuffer); |
b6797c8e JG |
3356 | switch (get_next_status) { |
3357 | case GET_NEXT_SUBBUFFER_STATUS_OK: | |
3358 | break; | |
3359 | case GET_NEXT_SUBBUFFER_STATUS_NO_DATA: | |
3360 | /* Not an error. */ | |
3361 | ret = 0; | |
3362 | goto sleep_stream; | |
3363 | case GET_NEXT_SUBBUFFER_STATUS_ERROR: | |
3364 | ret = -1; | |
6f9449c2 | 3365 | goto end; |
b6797c8e JG |
3366 | default: |
3367 | abort(); | |
d41f73b7 | 3368 | } |
74251bb8 | 3369 | |
28ab034a | 3370 | ret = stream->read_subbuffer_ops.pre_consume_subbuffer(stream, &subbuffer); |
6f9449c2 JG |
3371 | if (ret) { |
3372 | goto error_put_subbuf; | |
3373 | } | |
3374 | ||
28ab034a | 3375 | written_bytes = stream->read_subbuffer_ops.consume_subbuffer(ctx, stream, &subbuffer); |
514775d9 FD |
3376 | if (written_bytes <= 0) { |
3377 | ERR("Error consuming subbuffer: (%zd)", written_bytes); | |
3378 | ret = (int) written_bytes; | |
3379 | goto error_put_subbuf; | |
6f9449c2 JG |
3380 | } |
3381 | ||
3382 | ret = stream->read_subbuffer_ops.put_next_subbuffer(stream, &subbuffer); | |
3383 | if (ret) { | |
23d56598 JG |
3384 | goto end; |
3385 | } | |
3386 | ||
503fefca JG |
3387 | ret = post_consume(stream, &subbuffer, ctx); |
3388 | if (ret) { | |
3389 | goto end; | |
6f9449c2 JG |
3390 | } |
3391 | ||
23d56598 JG |
3392 | /* |
3393 | * After extracting the packet, we check if the stream is now ready to | |
3394 | * be rotated and perform the action immediately. | |
3395 | * | |
3396 | * Don't overwrite `ret` as callers expect the number of bytes | |
3397 | * consumed to be returned on success. | |
3398 | */ | |
3399 | rotation_ret = lttng_consumer_stream_is_rotate_ready(stream); | |
3400 | if (rotation_ret == 1) { | |
f46376a1 | 3401 | rotation_ret = lttng_consumer_rotate_stream(stream); |
23d56598 JG |
3402 | if (rotation_ret < 0) { |
3403 | ret = rotation_ret; | |
3404 | ERR("Stream rotation error after consuming data"); | |
3405 | goto end; | |
3406 | } | |
503fefca | 3407 | |
23d56598 JG |
3408 | } else if (rotation_ret < 0) { |
3409 | ret = rotation_ret; | |
3410 | ERR("Failed to check if stream was ready to rotate after consuming data"); | |
3411 | goto end; | |
3412 | } | |
3413 | ||
82e72193 | 3414 | sleep_stream: |
6f9449c2 JG |
3415 | if (stream->read_subbuffer_ops.on_sleep) { |
3416 | stream->read_subbuffer_ops.on_sleep(stream, ctx); | |
3417 | } | |
3418 | ||
3419 | ret = written_bytes; | |
23d56598 | 3420 | end: |
6f9449c2 JG |
3421 | if (!locked_by_caller) { |
3422 | stream->read_subbuffer_ops.unlock(stream); | |
94d49140 | 3423 | } |
6f9449c2 | 3424 | |
74251bb8 | 3425 | return ret; |
6f9449c2 JG |
3426 | error_put_subbuf: |
3427 | (void) stream->read_subbuffer_ops.put_next_subbuffer(stream, &subbuffer); | |
3428 | goto end; | |
d41f73b7 MD |
3429 | } |
3430 | ||
3431 | int lttng_consumer_on_recv_stream(struct lttng_consumer_stream *stream) | |
3432 | { | |
fa29bfbf | 3433 | switch (the_consumer_data.type) { |
d41f73b7 MD |
3434 | case LTTNG_CONSUMER_KERNEL: |
3435 | return lttng_kconsumer_on_recv_stream(stream); | |
7753dea8 MD |
3436 | case LTTNG_CONSUMER32_UST: |
3437 | case LTTNG_CONSUMER64_UST: | |
d41f73b7 MD |
3438 | return lttng_ustconsumer_on_recv_stream(stream); |
3439 | default: | |
3440 | ERR("Unknown consumer_data type"); | |
a0377dfe | 3441 | abort(); |
d41f73b7 MD |
3442 | return -ENOSYS; |
3443 | } | |
3444 | } | |
e4421fec DG |
3445 | |
3446 | /* | |
3447 | * Allocate and set consumer data hash tables. | |
3448 | */ | |
cd9adb8b | 3449 | int lttng_consumer_init() |
e4421fec | 3450 | { |
fa29bfbf SM |
3451 | the_consumer_data.channel_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3452 | if (!the_consumer_data.channel_ht) { | |
282dadbc MD |
3453 | goto error; |
3454 | } | |
3455 | ||
28ab034a | 3456 | the_consumer_data.channels_by_session_id_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
fa29bfbf | 3457 | if (!the_consumer_data.channels_by_session_id_ht) { |
5c3892a6 JG |
3458 | goto error; |
3459 | } | |
3460 | ||
fa29bfbf SM |
3461 | the_consumer_data.relayd_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3462 | if (!the_consumer_data.relayd_ht) { | |
282dadbc MD |
3463 | goto error; |
3464 | } | |
3465 | ||
fa29bfbf SM |
3466 | the_consumer_data.stream_list_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3467 | if (!the_consumer_data.stream_list_ht) { | |
282dadbc MD |
3468 | goto error; |
3469 | } | |
3470 | ||
28ab034a | 3471 | the_consumer_data.stream_per_chan_id_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
fa29bfbf | 3472 | if (!the_consumer_data.stream_per_chan_id_ht) { |
282dadbc MD |
3473 | goto error; |
3474 | } | |
3475 | ||
3476 | data_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); | |
3477 | if (!data_ht) { | |
3478 | goto error; | |
3479 | } | |
3480 | ||
3481 | metadata_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); | |
3482 | if (!metadata_ht) { | |
3483 | goto error; | |
3484 | } | |
3485 | ||
fa29bfbf SM |
3486 | the_consumer_data.chunk_registry = lttng_trace_chunk_registry_create(); |
3487 | if (!the_consumer_data.chunk_registry) { | |
28cc88f3 JG |
3488 | goto error; |
3489 | } | |
3490 | ||
282dadbc MD |
3491 | return 0; |
3492 | ||
3493 | error: | |
3494 | return -1; | |
e4421fec | 3495 | } |
7735ef9e DG |
3496 | |
3497 | /* | |
3498 | * Process the ADD_RELAYD command receive by a consumer. | |
3499 | * | |
3500 | * This will create a relayd socket pair and add it to the relayd hash table. | |
3501 | * The caller MUST acquire a RCU read side lock before calling it. | |
3502 | */ | |
4222116f | 3503 | void consumer_add_relayd_socket(uint64_t net_seq_idx, |
28ab034a JG |
3504 | int sock_type, |
3505 | struct lttng_consumer_local_data *ctx, | |
3506 | int sock, | |
3507 | struct pollfd *consumer_sockpoll, | |
3508 | uint64_t sessiond_id, | |
3509 | uint64_t relayd_session_id, | |
3510 | uint32_t relayd_version_major, | |
3511 | uint32_t relayd_version_minor, | |
3512 | enum lttcomm_sock_proto relayd_socket_protocol) | |
7735ef9e | 3513 | { |
cd2b09ed | 3514 | int fd = -1, ret = -1, relayd_created = 0; |
0c759fc9 | 3515 | enum lttcomm_return_code ret_code = LTTCOMM_CONSUMERD_SUCCESS; |
cd9adb8b | 3516 | struct consumer_relayd_sock_pair *relayd = nullptr; |
7735ef9e | 3517 | |
a0377dfe | 3518 | LTTNG_ASSERT(ctx); |
4222116f | 3519 | LTTNG_ASSERT(sock >= 0); |
48b7cdc2 | 3520 | ASSERT_RCU_READ_LOCKED(); |
6151a90f | 3521 | |
da009f2c | 3522 | DBG("Consumer adding relayd socket (idx: %" PRIu64 ")", net_seq_idx); |
7735ef9e DG |
3523 | |
3524 | /* Get relayd reference if exists. */ | |
3525 | relayd = consumer_find_relayd(net_seq_idx); | |
cd9adb8b | 3526 | if (relayd == nullptr) { |
a0377dfe | 3527 | LTTNG_ASSERT(sock_type == LTTNG_STREAM_CONTROL); |
7735ef9e DG |
3528 | /* Not found. Allocate one. */ |
3529 | relayd = consumer_allocate_relayd_sock_pair(net_seq_idx); | |
cd9adb8b | 3530 | if (relayd == nullptr) { |
618a6a28 MD |
3531 | ret_code = LTTCOMM_CONSUMERD_ENOMEM; |
3532 | goto error; | |
0d08d75e | 3533 | } else { |
30319bcb | 3534 | relayd->sessiond_session_id = sessiond_id; |
0d08d75e | 3535 | relayd_created = 1; |
7735ef9e | 3536 | } |
0d08d75e DG |
3537 | |
3538 | /* | |
3539 | * This code path MUST continue to the consumer send status message to | |
3540 | * we can notify the session daemon and continue our work without | |
3541 | * killing everything. | |
3542 | */ | |
da009f2c MD |
3543 | } else { |
3544 | /* | |
3545 | * relayd key should never be found for control socket. | |
3546 | */ | |
a0377dfe | 3547 | LTTNG_ASSERT(sock_type != LTTNG_STREAM_CONTROL); |
0d08d75e DG |
3548 | } |
3549 | ||
3550 | /* First send a status message before receiving the fds. */ | |
0c759fc9 | 3551 | ret = consumer_send_status_msg(sock, LTTCOMM_CONSUMERD_SUCCESS); |
618a6a28 | 3552 | if (ret < 0) { |
0d08d75e | 3553 | /* Somehow, the session daemon is not responding anymore. */ |
618a6a28 MD |
3554 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_FATAL); |
3555 | goto error_nosignal; | |
7735ef9e DG |
3556 | } |
3557 | ||
3558 | /* Poll on consumer socket. */ | |
84382d49 MD |
3559 | ret = lttng_consumer_poll_socket(consumer_sockpoll); |
3560 | if (ret) { | |
3561 | /* Needing to exit in the middle of a command: error. */ | |
0d08d75e | 3562 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_POLL_ERROR); |
618a6a28 | 3563 | goto error_nosignal; |
7735ef9e DG |
3564 | } |
3565 | ||
3566 | /* Get relayd socket from session daemon */ | |
3567 | ret = lttcomm_recv_fds_unix_sock(sock, &fd, 1); | |
3568 | if (ret != sizeof(fd)) { | |
28ab034a | 3569 | fd = -1; /* Just in case it gets set with an invalid value. */ |
0d08d75e DG |
3570 | |
3571 | /* | |
3572 | * Failing to receive FDs might indicate a major problem such as | |
3573 | * reaching a fd limit during the receive where the kernel returns a | |
3574 | * MSG_CTRUNC and fails to cleanup the fd in the queue. Any case, we | |
3575 | * don't take any chances and stop everything. | |
3576 | * | |
3577 | * XXX: Feature request #558 will fix that and avoid this possible | |
3578 | * issue when reaching the fd limit. | |
3579 | */ | |
3580 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_ERROR_RECV_FD); | |
618a6a28 | 3581 | ret_code = LTTCOMM_CONSUMERD_ERROR_RECV_FD; |
f50f23d9 DG |
3582 | goto error; |
3583 | } | |
3584 | ||
7735ef9e DG |
3585 | /* Copy socket information and received FD */ |
3586 | switch (sock_type) { | |
3587 | case LTTNG_STREAM_CONTROL: | |
3588 | /* Copy received lttcomm socket */ | |
4222116f | 3589 | ret = lttcomm_populate_sock_from_open_socket( |
28ab034a | 3590 | &relayd->control_sock.sock, fd, relayd_socket_protocol); |
7735ef9e | 3591 | |
6151a90f | 3592 | /* Assign version values. */ |
4222116f JR |
3593 | relayd->control_sock.major = relayd_version_major; |
3594 | relayd->control_sock.minor = relayd_version_minor; | |
c5b6f4f0 | 3595 | |
d3e2ba59 | 3596 | relayd->relayd_session_id = relayd_session_id; |
c5b6f4f0 | 3597 | |
7735ef9e DG |
3598 | break; |
3599 | case LTTNG_STREAM_DATA: | |
3600 | /* Copy received lttcomm socket */ | |
4222116f | 3601 | ret = lttcomm_populate_sock_from_open_socket( |
28ab034a | 3602 | &relayd->data_sock.sock, fd, relayd_socket_protocol); |
6151a90f | 3603 | /* Assign version values. */ |
4222116f JR |
3604 | relayd->data_sock.major = relayd_version_major; |
3605 | relayd->data_sock.minor = relayd_version_minor; | |
7735ef9e DG |
3606 | break; |
3607 | default: | |
3608 | ERR("Unknown relayd socket type (%d)", sock_type); | |
618a6a28 | 3609 | ret_code = LTTCOMM_CONSUMERD_FATAL; |
7735ef9e DG |
3610 | goto error; |
3611 | } | |
3612 | ||
4222116f JR |
3613 | if (ret < 0) { |
3614 | ret_code = LTTCOMM_CONSUMERD_FATAL; | |
3615 | goto error; | |
3616 | } | |
3617 | ||
d88aee68 | 3618 | DBG("Consumer %s socket created successfully with net idx %" PRIu64 " (fd: %d)", |
28ab034a JG |
3619 | sock_type == LTTNG_STREAM_CONTROL ? "control" : "data", |
3620 | relayd->net_seq_idx, | |
3621 | fd); | |
39d9954c FD |
3622 | /* |
3623 | * We gave the ownership of the fd to the relayd structure. Set the | |
3624 | * fd to -1 so we don't call close() on it in the error path below. | |
3625 | */ | |
3626 | fd = -1; | |
7735ef9e | 3627 | |
618a6a28 MD |
3628 | /* We successfully added the socket. Send status back. */ |
3629 | ret = consumer_send_status_msg(sock, ret_code); | |
3630 | if (ret < 0) { | |
3631 | /* Somehow, the session daemon is not responding anymore. */ | |
3632 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_FATAL); | |
3633 | goto error_nosignal; | |
3634 | } | |
3635 | ||
7735ef9e DG |
3636 | /* |
3637 | * Add relayd socket pair to consumer data hashtable. If object already | |
3638 | * exists or on error, the function gracefully returns. | |
3639 | */ | |
9276e5c8 | 3640 | relayd->ctx = ctx; |
d09e1200 | 3641 | add_relayd(relayd); |
7735ef9e DG |
3642 | |
3643 | /* All good! */ | |
2527bf85 | 3644 | return; |
7735ef9e DG |
3645 | |
3646 | error: | |
618a6a28 MD |
3647 | if (consumer_send_status_msg(sock, ret_code) < 0) { |
3648 | lttng_consumer_send_error(ctx, LTTCOMM_CONSUMERD_FATAL); | |
3649 | } | |
3650 | ||
3651 | error_nosignal: | |
4028eeb9 DG |
3652 | /* Close received socket if valid. */ |
3653 | if (fd >= 0) { | |
3654 | if (close(fd)) { | |
3655 | PERROR("close received socket"); | |
3656 | } | |
3657 | } | |
cd2b09ed DG |
3658 | |
3659 | if (relayd_created) { | |
cd2b09ed DG |
3660 | free(relayd); |
3661 | } | |
7735ef9e | 3662 | } |
ca22feea | 3663 | |
f7079f67 DG |
3664 | /* |
3665 | * Search for a relayd associated to the session id and return the reference. | |
3666 | * | |
3667 | * A rcu read side lock MUST be acquire before calling this function and locked | |
3668 | * until the relayd object is no longer necessary. | |
3669 | */ | |
3670 | static struct consumer_relayd_sock_pair *find_relayd_by_session_id(uint64_t id) | |
3671 | { | |
f7079f67 | 3672 | /* Iterate over all relayd since they are indexed by net_seq_idx. */ |
c3ade133 JG |
3673 | for (auto *relayd : |
3674 | lttng::urcu::lfht_iteration_adapter<consumer_relayd_sock_pair, | |
3675 | decltype(consumer_relayd_sock_pair::node), | |
3676 | &consumer_relayd_sock_pair::node>( | |
3677 | *the_consumer_data.relayd_ht->ht)) { | |
18261bd1 DG |
3678 | /* |
3679 | * Check by sessiond id which is unique here where the relayd session | |
3680 | * id might not be when having multiple relayd. | |
3681 | */ | |
3682 | if (relayd->sessiond_session_id == id) { | |
f7079f67 | 3683 | /* Found the relayd. There can be only one per id. */ |
c3ade133 | 3684 | return relayd; |
f7079f67 DG |
3685 | } |
3686 | } | |
3687 | ||
cd9adb8b | 3688 | return nullptr; |
f7079f67 DG |
3689 | } |
3690 | ||
ca22feea DG |
3691 | /* |
3692 | * Check if for a given session id there is still data needed to be extract | |
3693 | * from the buffers. | |
3694 | * | |
6d805429 | 3695 | * Return 1 if data is pending or else 0 meaning ready to be read. |
ca22feea | 3696 | */ |
6d805429 | 3697 | int consumer_data_pending(uint64_t id) |
ca22feea DG |
3698 | { |
3699 | int ret; | |
3d46ea1a | 3700 | const auto ht = the_consumer_data.stream_list_ht; |
cd9adb8b | 3701 | struct consumer_relayd_sock_pair *relayd = nullptr; |
6d805429 | 3702 | int (*data_pending)(struct lttng_consumer_stream *); |
ca22feea | 3703 | |
6d805429 | 3704 | DBG("Consumer data pending command on session id %" PRIu64, id); |
ca22feea | 3705 | |
3d46ea1a | 3706 | const lttng::pthread::lock_guard consumer_data_lock(the_consumer_data.lock); |
ca22feea | 3707 | |
fa29bfbf | 3708 | switch (the_consumer_data.type) { |
ca22feea | 3709 | case LTTNG_CONSUMER_KERNEL: |
6d805429 | 3710 | data_pending = lttng_kconsumer_data_pending; |
ca22feea DG |
3711 | break; |
3712 | case LTTNG_CONSUMER32_UST: | |
3713 | case LTTNG_CONSUMER64_UST: | |
6d805429 | 3714 | data_pending = lttng_ustconsumer_data_pending; |
ca22feea DG |
3715 | break; |
3716 | default: | |
3717 | ERR("Unknown consumer data type"); | |
a0377dfe | 3718 | abort(); |
ca22feea DG |
3719 | } |
3720 | ||
3d46ea1a JG |
3721 | for (auto *stream : lttng::urcu::lfht_filtered_iteration_adapter< |
3722 | lttng_consumer_stream, | |
3723 | decltype(lttng_consumer_stream::node_session_id), | |
3724 | <tng_consumer_stream::node_session_id, | |
3725 | std::uint64_t>(*ht->ht, &id, ht->hash_fct(&id, lttng_ht_seed), ht->match_fct)) { | |
3726 | const lttng::pthread::lock_guard stream_lock(stream->lock); | |
ca22feea | 3727 | |
4e9a4686 DG |
3728 | /* |
3729 | * A removed node from the hash table indicates that the stream has | |
3730 | * been deleted thus having a guarantee that the buffers are closed | |
3731 | * on the consumer side. However, data can still be transmitted | |
3732 | * over the network so don't skip the relayd check. | |
3733 | */ | |
3734 | ret = cds_lfht_is_node_deleted(&stream->node.node); | |
3735 | if (!ret) { | |
3736 | /* Check the stream if there is data in the buffers. */ | |
6d805429 DG |
3737 | ret = data_pending(stream); |
3738 | if (ret == 1) { | |
f7079f67 | 3739 | goto data_pending; |
4e9a4686 DG |
3740 | } |
3741 | } | |
d9f0c7c7 JR |
3742 | } |
3743 | ||
3744 | relayd = find_relayd_by_session_id(id); | |
3745 | if (relayd) { | |
3746 | unsigned int is_data_inflight = 0; | |
3747 | ||
3d46ea1a JG |
3748 | const lttng::pthread::lock_guard ctrl_sock_lock(relayd->ctrl_sock_mutex); |
3749 | ||
d9f0c7c7 | 3750 | /* Send init command for data pending. */ |
28ab034a | 3751 | ret = relayd_begin_data_pending(&relayd->control_sock, relayd->relayd_session_id); |
d9f0c7c7 | 3752 | if (ret < 0) { |
d9f0c7c7 JR |
3753 | /* Communication error thus the relayd so no data pending. */ |
3754 | goto data_not_pending; | |
3755 | } | |
3756 | ||
3d46ea1a JG |
3757 | for (auto *stream : lttng::urcu::lfht_filtered_iteration_adapter< |
3758 | lttng_consumer_stream, | |
3759 | decltype(lttng_consumer_stream::node_session_id), | |
3760 | <tng_consumer_stream::node_session_id, | |
3761 | std::uint64_t>( | |
3762 | *ht->ht, &id, ht->hash_fct(&id, lttng_ht_seed), ht->match_fct)) { | |
c8f59ee5 | 3763 | if (stream->metadata_flag) { |
ad7051c0 | 3764 | ret = relayd_quiescent_control(&relayd->control_sock, |
28ab034a | 3765 | stream->relayd_stream_id); |
c8f59ee5 | 3766 | } else { |
6d805429 | 3767 | ret = relayd_data_pending(&relayd->control_sock, |
28ab034a JG |
3768 | stream->relayd_stream_id, |
3769 | stream->next_net_seq_num - 1); | |
c8f59ee5 | 3770 | } |
d9f0c7c7 JR |
3771 | |
3772 | if (ret == 1) { | |
d9f0c7c7 JR |
3773 | goto data_pending; |
3774 | } else if (ret < 0) { | |
28ab034a JG |
3775 | ERR("Relayd data pending failed. Cleaning up relayd %" PRIu64 ".", |
3776 | relayd->net_seq_idx); | |
9276e5c8 | 3777 | lttng_consumer_cleanup_relayd(relayd); |
9276e5c8 JR |
3778 | goto data_not_pending; |
3779 | } | |
c8f59ee5 | 3780 | } |
f7079f67 | 3781 | |
d9f0c7c7 | 3782 | /* Send end command for data pending. */ |
28ab034a JG |
3783 | ret = relayd_end_data_pending( |
3784 | &relayd->control_sock, relayd->relayd_session_id, &is_data_inflight); | |
bdd88757 | 3785 | if (ret < 0) { |
28ab034a JG |
3786 | ERR("Relayd end data pending failed. Cleaning up relayd %" PRIu64 ".", |
3787 | relayd->net_seq_idx); | |
9276e5c8 | 3788 | lttng_consumer_cleanup_relayd(relayd); |
f7079f67 DG |
3789 | goto data_not_pending; |
3790 | } | |
bdd88757 DG |
3791 | if (is_data_inflight) { |
3792 | goto data_pending; | |
3793 | } | |
f7079f67 DG |
3794 | } |
3795 | ||
ca22feea | 3796 | /* |
f7079f67 DG |
3797 | * Finding _no_ node in the hash table and no inflight data means that the |
3798 | * stream(s) have been removed thus data is guaranteed to be available for | |
3799 | * analysis from the trace files. | |
ca22feea DG |
3800 | */ |
3801 | ||
f7079f67 | 3802 | data_not_pending: |
ca22feea | 3803 | /* Data is available to be read by a viewer. */ |
6d805429 | 3804 | return 0; |
ca22feea | 3805 | |
f7079f67 | 3806 | data_pending: |
ca22feea | 3807 | /* Data is still being extracted from buffers. */ |
6d805429 | 3808 | return 1; |
ca22feea | 3809 | } |
f50f23d9 DG |
3810 | |
3811 | /* | |
3812 | * Send a ret code status message to the sessiond daemon. | |
3813 | * | |
3814 | * Return the sendmsg() return value. | |
3815 | */ | |
3816 | int consumer_send_status_msg(int sock, int ret_code) | |
3817 | { | |
3818 | struct lttcomm_consumer_status_msg msg; | |
3819 | ||
53efb85a | 3820 | memset(&msg, 0, sizeof(msg)); |
97535efa | 3821 | msg.ret_code = (lttcomm_return_code) ret_code; |
f50f23d9 DG |
3822 | |
3823 | return lttcomm_send_unix_sock(sock, &msg, sizeof(msg)); | |
3824 | } | |
ffe60014 DG |
3825 | |
3826 | /* | |
3827 | * Send a channel status message to the sessiond daemon. | |
3828 | * | |
3829 | * Return the sendmsg() return value. | |
3830 | */ | |
28ab034a | 3831 | int consumer_send_status_channel(int sock, struct lttng_consumer_channel *channel) |
ffe60014 DG |
3832 | { |
3833 | struct lttcomm_consumer_status_channel msg; | |
3834 | ||
a0377dfe | 3835 | LTTNG_ASSERT(sock >= 0); |
ffe60014 | 3836 | |
53efb85a | 3837 | memset(&msg, 0, sizeof(msg)); |
ffe60014 | 3838 | if (!channel) { |
0c759fc9 | 3839 | msg.ret_code = LTTCOMM_CONSUMERD_CHANNEL_FAIL; |
ffe60014 | 3840 | } else { |
0c759fc9 | 3841 | msg.ret_code = LTTCOMM_CONSUMERD_SUCCESS; |
ffe60014 DG |
3842 | msg.key = channel->key; |
3843 | msg.stream_count = channel->streams.count; | |
3844 | } | |
3845 | ||
3846 | return lttcomm_send_unix_sock(sock, &msg, sizeof(msg)); | |
3847 | } | |
5c786ded | 3848 | |
d07ceecd | 3849 | unsigned long consumer_get_consume_start_pos(unsigned long consumed_pos, |
28ab034a JG |
3850 | unsigned long produced_pos, |
3851 | uint64_t nb_packets_per_stream, | |
3852 | uint64_t max_sb_size) | |
5c786ded | 3853 | { |
d07ceecd | 3854 | unsigned long start_pos; |
5c786ded | 3855 | |
d07ceecd | 3856 | if (!nb_packets_per_stream) { |
28ab034a | 3857 | return consumed_pos; /* Grab everything */ |
d07ceecd | 3858 | } |
1cbd136b | 3859 | start_pos = produced_pos - lttng_offset_align_floor(produced_pos, max_sb_size); |
d07ceecd MD |
3860 | start_pos -= max_sb_size * nb_packets_per_stream; |
3861 | if ((long) (start_pos - consumed_pos) < 0) { | |
28ab034a | 3862 | return consumed_pos; /* Grab everything */ |
d07ceecd MD |
3863 | } |
3864 | return start_pos; | |
5c786ded | 3865 | } |
a1ae2ea5 | 3866 | |
c1dcb8bb JG |
3867 | /* Stream lock must be held by the caller. */ |
3868 | static int sample_stream_positions(struct lttng_consumer_stream *stream, | |
28ab034a JG |
3869 | unsigned long *produced, |
3870 | unsigned long *consumed) | |
c1dcb8bb JG |
3871 | { |
3872 | int ret; | |
3873 | ||
3874 | ASSERT_LOCKED(stream->lock); | |
3875 | ||
3876 | ret = lttng_consumer_sample_snapshot_positions(stream); | |
3877 | if (ret < 0) { | |
3878 | ERR("Failed to sample snapshot positions"); | |
3879 | goto end; | |
3880 | } | |
3881 | ||
3882 | ret = lttng_consumer_get_produced_snapshot(stream, produced); | |
3883 | if (ret < 0) { | |
3884 | ERR("Failed to sample produced position"); | |
3885 | goto end; | |
3886 | } | |
3887 | ||
3888 | ret = lttng_consumer_get_consumed_snapshot(stream, consumed); | |
3889 | if (ret < 0) { | |
3890 | ERR("Failed to sample consumed position"); | |
3891 | goto end; | |
3892 | } | |
3893 | ||
3894 | end: | |
3895 | return ret; | |
3896 | } | |
3897 | ||
b99a8d42 JD |
3898 | /* |
3899 | * Sample the rotate position for all the streams of a channel. If a stream | |
3900 | * is already at the rotate position (produced == consumed), we flag it as | |
3901 | * ready for rotation. The rotation of ready streams occurs after we have | |
3902 | * replied to the session daemon that we have finished sampling the positions. | |
92b7a7f8 | 3903 | * Must be called with RCU read-side lock held to ensure existence of channel. |
b99a8d42 JD |
3904 | * |
3905 | * Returns 0 on success, < 0 on error | |
3906 | */ | |
92b7a7f8 | 3907 | int lttng_consumer_rotate_channel(struct lttng_consumer_channel *channel, |
28ab034a JG |
3908 | uint64_t key, |
3909 | uint64_t relayd_id) | |
b99a8d42 JD |
3910 | { |
3911 | int ret; | |
3d46ea1a | 3912 | const auto ht = the_consumer_data.stream_per_chan_id_ht; |
c35f9726 JG |
3913 | struct lttng_dynamic_array stream_rotation_positions; |
3914 | uint64_t next_chunk_id, stream_count = 0; | |
3915 | enum lttng_trace_chunk_status chunk_status; | |
3916 | const bool is_local_trace = relayd_id == -1ULL; | |
cd9adb8b | 3917 | struct consumer_relayd_sock_pair *relayd = nullptr; |
c35f9726 | 3918 | bool rotating_to_new_chunk = true; |
b32703d6 JG |
3919 | /* Array of `struct lttng_consumer_stream *` */ |
3920 | struct lttng_dynamic_pointer_array streams_packet_to_open; | |
b99a8d42 | 3921 | |
48b7cdc2 FD |
3922 | ASSERT_RCU_READ_LOCKED(); |
3923 | ||
b99a8d42 JD |
3924 | DBG("Consumer sample rotate position for channel %" PRIu64, key); |
3925 | ||
cd9adb8b JG |
3926 | lttng_dynamic_array_init(&stream_rotation_positions, |
3927 | sizeof(struct relayd_stream_rotation_position), | |
3928 | nullptr); | |
3929 | lttng_dynamic_pointer_array_init(&streams_packet_to_open, nullptr); | |
c35f9726 | 3930 | |
3d46ea1a | 3931 | const lttng::pthread::lock_guard channel_lock(channel->lock); |
b99a8d42 | 3932 | |
a0377dfe | 3933 | LTTNG_ASSERT(channel->trace_chunk); |
28ab034a | 3934 | chunk_status = lttng_trace_chunk_get_id(channel->trace_chunk, &next_chunk_id); |
c35f9726 JG |
3935 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
3936 | ret = -1; | |
3d46ea1a | 3937 | goto end; |
c35f9726 | 3938 | } |
b99a8d42 | 3939 | |
3d46ea1a JG |
3940 | for (auto *stream : lttng::urcu::lfht_filtered_iteration_adapter< |
3941 | lttng_consumer_stream, | |
3942 | decltype(lttng_consumer_stream::node_channel_id), | |
3943 | <tng_consumer_stream::node_channel_id, | |
3944 | std::uint64_t>(*ht->ht, | |
3945 | &channel->key, | |
3946 | ht->hash_fct(&channel->key, lttng_ht_seed), | |
3947 | ht->match_fct)) { | |
a40a503f | 3948 | unsigned long produced_pos = 0, consumed_pos = 0; |
b99a8d42 JD |
3949 | |
3950 | health_code_update(); | |
3951 | ||
3952 | /* | |
3953 | * Lock stream because we are about to change its state. | |
3954 | */ | |
3d46ea1a | 3955 | const lttng::pthread::lock_guard stream_lock(stream->lock); |
b99a8d42 | 3956 | |
c35f9726 JG |
3957 | if (stream->trace_chunk == stream->chan->trace_chunk) { |
3958 | rotating_to_new_chunk = false; | |
3959 | } | |
3960 | ||
a40a503f | 3961 | /* |
c1dcb8bb | 3962 | * Do not flush a packet when rotating from a NULL trace |
a9dde553 | 3963 | * chunk. The stream has no means to output data, and the prior |
c1dcb8bb JG |
3964 | * rotation which rotated to NULL performed that side-effect |
3965 | * already. No new data can be produced when a stream has no | |
3966 | * associated trace chunk (e.g. a stop followed by a rotate). | |
a40a503f | 3967 | */ |
a9dde553 | 3968 | if (stream->trace_chunk) { |
c1dcb8bb JG |
3969 | bool flush_active; |
3970 | ||
3971 | if (stream->metadata_flag) { | |
3972 | /* | |
3973 | * Don't produce an empty metadata packet, | |
3974 | * simply close the current one. | |
3975 | * | |
3976 | * Metadata is regenerated on every trace chunk | |
3977 | * switch; there is no concern that no data was | |
3978 | * produced. | |
3979 | */ | |
3980 | flush_active = true; | |
3981 | } else { | |
3982 | /* | |
3983 | * Only flush an empty packet if the "packet | |
3984 | * open" could not be performed on transition | |
3985 | * to a new trace chunk and no packets were | |
3986 | * consumed within the chunk's lifetime. | |
3987 | */ | |
3988 | if (stream->opened_packet_in_current_trace_chunk) { | |
3989 | flush_active = true; | |
3990 | } else { | |
3991 | /* | |
3992 | * Stream could have been full at the | |
3993 | * time of rotation, but then have had | |
3994 | * no activity at all. | |
3995 | * | |
3996 | * It is important to flush a packet | |
3997 | * to prevent 0-length files from being | |
3998 | * produced as most viewers choke on | |
3999 | * them. | |
4000 | * | |
4001 | * Unfortunately viewers will not be | |
4002 | * able to know that tracing was active | |
4003 | * for this stream during this trace | |
4004 | * chunk's lifetime. | |
4005 | */ | |
28ab034a JG |
4006 | ret = sample_stream_positions( |
4007 | stream, &produced_pos, &consumed_pos); | |
c1dcb8bb | 4008 | if (ret) { |
3d46ea1a | 4009 | goto end; |
c1dcb8bb JG |
4010 | } |
4011 | ||
4012 | /* | |
4013 | * Don't flush an empty packet if data | |
4014 | * was produced; it will be consumed | |
4015 | * before the rotation completes. | |
4016 | */ | |
4017 | flush_active = produced_pos != consumed_pos; | |
4018 | if (!flush_active) { | |
c1dcb8bb JG |
4019 | const char *trace_chunk_name; |
4020 | uint64_t trace_chunk_id; | |
4021 | ||
4022 | chunk_status = lttng_trace_chunk_get_name( | |
28ab034a JG |
4023 | stream->trace_chunk, |
4024 | &trace_chunk_name, | |
cd9adb8b | 4025 | nullptr); |
c1dcb8bb JG |
4026 | if (chunk_status == LTTNG_TRACE_CHUNK_STATUS_NONE) { |
4027 | trace_chunk_name = "none"; | |
4028 | } | |
4029 | ||
4030 | /* | |
4031 | * Consumer trace chunks are | |
4032 | * never anonymous. | |
4033 | */ | |
4034 | chunk_status = lttng_trace_chunk_get_id( | |
28ab034a | 4035 | stream->trace_chunk, &trace_chunk_id); |
a0377dfe | 4036 | LTTNG_ASSERT(chunk_status == |
28ab034a | 4037 | LTTNG_TRACE_CHUNK_STATUS_OK); |
c1dcb8bb JG |
4038 | |
4039 | DBG("Unable to open packet for stream during trace chunk's lifetime. " | |
28ab034a JG |
4040 | "Flushing an empty packet to prevent an empty file from being created: " |
4041 | "stream id = %" PRIu64 | |
4042 | ", trace chunk name = `%s`, trace chunk id = %" PRIu64, | |
4043 | stream->key, | |
4044 | trace_chunk_name, | |
4045 | trace_chunk_id); | |
c1dcb8bb JG |
4046 | } |
4047 | } | |
4048 | } | |
4049 | ||
a9dde553 | 4050 | /* |
c1dcb8bb JG |
4051 | * Close the current packet before sampling the |
4052 | * ring buffer positions. | |
a9dde553 | 4053 | */ |
c1dcb8bb | 4054 | ret = consumer_stream_flush_buffer(stream, flush_active); |
a9dde553 MD |
4055 | if (ret < 0) { |
4056 | ERR("Failed to flush stream %" PRIu64 " during channel rotation", | |
28ab034a | 4057 | stream->key); |
3d46ea1a | 4058 | goto end; |
a9dde553 | 4059 | } |
b99a8d42 JD |
4060 | } |
4061 | ||
a40a503f MD |
4062 | ret = lttng_consumer_take_snapshot(stream); |
4063 | if (ret < 0 && ret != -ENODATA && ret != -EAGAIN) { | |
4064 | ERR("Failed to sample snapshot position during channel rotation"); | |
3d46ea1a | 4065 | goto end; |
b99a8d42 | 4066 | } |
a40a503f | 4067 | if (!ret) { |
28ab034a | 4068 | ret = lttng_consumer_get_produced_snapshot(stream, &produced_pos); |
a40a503f MD |
4069 | if (ret < 0) { |
4070 | ERR("Failed to sample produced position during channel rotation"); | |
3d46ea1a | 4071 | goto end; |
a40a503f | 4072 | } |
b99a8d42 | 4073 | |
28ab034a | 4074 | ret = lttng_consumer_get_consumed_snapshot(stream, &consumed_pos); |
a40a503f MD |
4075 | if (ret < 0) { |
4076 | ERR("Failed to sample consumed position during channel rotation"); | |
3d46ea1a | 4077 | goto end; |
a40a503f MD |
4078 | } |
4079 | } | |
4080 | /* | |
4081 | * Align produced position on the start-of-packet boundary of the first | |
4082 | * packet going into the next trace chunk. | |
4083 | */ | |
1cbd136b | 4084 | produced_pos = lttng_align_floor(produced_pos, stream->max_sb_size); |
a40a503f | 4085 | if (consumed_pos == produced_pos) { |
f8528c7a | 4086 | DBG("Set rotate ready for stream %" PRIu64 " produced = %lu consumed = %lu", |
28ab034a JG |
4087 | stream->key, |
4088 | produced_pos, | |
4089 | consumed_pos); | |
b99a8d42 | 4090 | stream->rotate_ready = true; |
f8528c7a MD |
4091 | } else { |
4092 | DBG("Different consumed and produced positions " | |
28ab034a JG |
4093 | "for stream %" PRIu64 " produced = %lu consumed = %lu", |
4094 | stream->key, | |
4095 | produced_pos, | |
4096 | consumed_pos); | |
b99a8d42 | 4097 | } |
633d0182 | 4098 | /* |
a40a503f MD |
4099 | * The rotation position is based on the packet_seq_num of the |
4100 | * packet following the last packet that was consumed for this | |
4101 | * stream, incremented by the offset between produced and | |
4102 | * consumed positions. This rotation position is a lower bound | |
4103 | * (inclusive) at which the next trace chunk starts. Since it | |
4104 | * is a lower bound, it is OK if the packet_seq_num does not | |
4105 | * correspond exactly to the same packet identified by the | |
4106 | * consumed_pos, which can happen in overwrite mode. | |
633d0182 | 4107 | */ |
a40a503f MD |
4108 | if (stream->sequence_number_unavailable) { |
4109 | /* | |
4110 | * Rotation should never be performed on a session which | |
4111 | * interacts with a pre-2.8 lttng-modules, which does | |
4112 | * not implement packet sequence number. | |
4113 | */ | |
4114 | ERR("Failure to rotate stream %" PRIu64 ": sequence number unavailable", | |
28ab034a | 4115 | stream->key); |
a40a503f | 4116 | ret = -1; |
3d46ea1a | 4117 | goto end; |
b99a8d42 | 4118 | } |
a40a503f | 4119 | stream->rotate_position = stream->last_sequence_number + 1 + |
28ab034a | 4120 | ((produced_pos - consumed_pos) / stream->max_sb_size); |
f8528c7a | 4121 | DBG("Set rotation position for stream %" PRIu64 " at position %" PRIu64, |
28ab034a JG |
4122 | stream->key, |
4123 | stream->rotate_position); | |
b99a8d42 | 4124 | |
c35f9726 | 4125 | if (!is_local_trace) { |
633d0182 JG |
4126 | /* |
4127 | * The relay daemon control protocol expects a rotation | |
4128 | * position as "the sequence number of the first packet | |
a40a503f | 4129 | * _after_ the current trace chunk". |
633d0182 | 4130 | */ |
c35f9726 JG |
4131 | const struct relayd_stream_rotation_position position = { |
4132 | .stream_id = stream->relayd_stream_id, | |
a40a503f | 4133 | .rotate_at_seq_num = stream->rotate_position, |
c35f9726 JG |
4134 | }; |
4135 | ||
28ab034a JG |
4136 | ret = lttng_dynamic_array_add_element(&stream_rotation_positions, |
4137 | &position); | |
c35f9726 JG |
4138 | if (ret) { |
4139 | ERR("Failed to allocate stream rotation position"); | |
3d46ea1a | 4140 | goto end; |
c35f9726 JG |
4141 | } |
4142 | stream_count++; | |
4143 | } | |
f96af312 JG |
4144 | |
4145 | stream->opened_packet_in_current_trace_chunk = false; | |
4146 | ||
4147 | if (rotating_to_new_chunk && !stream->metadata_flag) { | |
4148 | /* | |
4149 | * Attempt to flush an empty packet as close to the | |
4150 | * rotation point as possible. In the event where a | |
4151 | * stream remains inactive after the rotation point, | |
4152 | * this ensures that the new trace chunk has a | |
4153 | * beginning timestamp set at the begining of the | |
4154 | * trace chunk instead of only creating an empty | |
4155 | * packet when the trace chunk is stopped. | |
4156 | * | |
4157 | * This indicates to the viewers that the stream | |
4158 | * was being recorded, but more importantly it | |
4159 | * allows viewers to determine a useable trace | |
4160 | * intersection. | |
4161 | * | |
4162 | * This presents a problem in the case where the | |
4163 | * ring-buffer is completely full. | |
4164 | * | |
4165 | * Consider the following scenario: | |
4166 | * - The consumption of data is slow (slow network, | |
4167 | * for instance), | |
4168 | * - The ring buffer is full, | |
4169 | * - A rotation is initiated, | |
4170 | * - The flush below does nothing (no space left to | |
4171 | * open a new packet), | |
4172 | * - The other streams rotate very soon, and new | |
4173 | * data is produced in the new chunk, | |
4174 | * - This stream completes its rotation long after the | |
4175 | * rotation was initiated | |
4176 | * - The session is stopped before any event can be | |
4177 | * produced in this stream's buffers. | |
4178 | * | |
4179 | * The resulting trace chunk will have a single packet | |
4180 | * temporaly at the end of the trace chunk for this | |
4181 | * stream making the stream intersection more narrow | |
4182 | * than it should be. | |
4183 | * | |
4184 | * To work-around this, an empty flush is performed | |
4185 | * after the first consumption of a packet during a | |
4186 | * rotation if open_packet fails. The idea is that | |
4187 | * consuming a packet frees enough space to switch | |
4188 | * packets in this scenario and allows the tracer to | |
4189 | * "stamp" the beginning of the new trace chunk at the | |
4190 | * earliest possible point. | |
b32703d6 JG |
4191 | * |
4192 | * The packet open is performed after the channel | |
4193 | * rotation to ensure that no attempt to open a packet | |
4194 | * is performed in a stream that has no active trace | |
4195 | * chunk. | |
f96af312 | 4196 | */ |
28ab034a JG |
4197 | ret = lttng_dynamic_pointer_array_add_pointer(&streams_packet_to_open, |
4198 | stream); | |
b32703d6 JG |
4199 | if (ret) { |
4200 | PERROR("Failed to add a stream pointer to array of streams in which to open a packet"); | |
f96af312 | 4201 | ret = -1; |
3d46ea1a | 4202 | goto end; |
f96af312 JG |
4203 | } |
4204 | } | |
b99a8d42 | 4205 | } |
b99a8d42 | 4206 | |
b32703d6 JG |
4207 | if (!is_local_trace) { |
4208 | relayd = consumer_find_relayd(relayd_id); | |
4209 | if (!relayd) { | |
4210 | ERR("Failed to find relayd %" PRIu64, relayd_id); | |
4211 | ret = -1; | |
3d46ea1a | 4212 | goto end; |
b32703d6 | 4213 | } |
c35f9726 | 4214 | |
b32703d6 | 4215 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); |
28ab034a JG |
4216 | ret = relayd_rotate_streams(&relayd->control_sock, |
4217 | stream_count, | |
cd9adb8b | 4218 | rotating_to_new_chunk ? &next_chunk_id : nullptr, |
28ab034a JG |
4219 | (const struct relayd_stream_rotation_position *) |
4220 | stream_rotation_positions.buffer.data); | |
b32703d6 JG |
4221 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); |
4222 | if (ret < 0) { | |
4223 | ERR("Relayd rotate stream failed. Cleaning up relayd %" PRIu64, | |
28ab034a | 4224 | relayd->net_seq_idx); |
b32703d6 | 4225 | lttng_consumer_cleanup_relayd(relayd); |
3d46ea1a | 4226 | goto end; |
b32703d6 | 4227 | } |
c35f9726 JG |
4228 | } |
4229 | ||
3d46ea1a | 4230 | for (std::size_t stream_idx = 0; |
28ab034a JG |
4231 | stream_idx < lttng_dynamic_pointer_array_get_count(&streams_packet_to_open); |
4232 | stream_idx++) { | |
b32703d6 | 4233 | enum consumer_stream_open_packet_status status; |
3d46ea1a | 4234 | auto *stream = (lttng_consumer_stream *) lttng_dynamic_pointer_array_get_pointer( |
28ab034a | 4235 | &streams_packet_to_open, stream_idx); |
b32703d6 | 4236 | |
3d46ea1a | 4237 | const lttng::pthread::lock_guard stream_lock(stream->lock); |
b32703d6 | 4238 | status = consumer_stream_open_packet(stream); |
b32703d6 JG |
4239 | switch (status) { |
4240 | case CONSUMER_STREAM_OPEN_PACKET_STATUS_OPENED: | |
4241 | DBG("Opened a packet after a rotation: stream id = %" PRIu64 | |
4242 | ", channel name = %s, session id = %" PRIu64, | |
28ab034a JG |
4243 | stream->key, |
4244 | stream->chan->name, | |
4245 | stream->chan->session_id); | |
b32703d6 JG |
4246 | break; |
4247 | case CONSUMER_STREAM_OPEN_PACKET_STATUS_NO_SPACE: | |
4248 | /* | |
4249 | * Can't open a packet as there is no space left | |
4250 | * in the buffer. A new packet will be opened | |
4251 | * once one has been consumed. | |
4252 | */ | |
4253 | DBG("No space left to open a packet after a rotation: stream id = %" PRIu64 | |
4254 | ", channel name = %s, session id = %" PRIu64, | |
28ab034a JG |
4255 | stream->key, |
4256 | stream->chan->name, | |
4257 | stream->chan->session_id); | |
b32703d6 JG |
4258 | break; |
4259 | case CONSUMER_STREAM_OPEN_PACKET_STATUS_ERROR: | |
4260 | /* Logged by callee. */ | |
4261 | ret = -1; | |
3d46ea1a | 4262 | goto end; |
b32703d6 JG |
4263 | default: |
4264 | abort(); | |
4265 | } | |
c35f9726 JG |
4266 | } |
4267 | ||
b99a8d42 | 4268 | ret = 0; |
b99a8d42 | 4269 | end: |
c35f9726 | 4270 | lttng_dynamic_array_reset(&stream_rotation_positions); |
b32703d6 | 4271 | lttng_dynamic_pointer_array_reset(&streams_packet_to_open); |
b99a8d42 JD |
4272 | return ret; |
4273 | } | |
4274 | ||
28ab034a | 4275 | static int consumer_clear_buffer(struct lttng_consumer_stream *stream) |
5f3aff8b MD |
4276 | { |
4277 | int ret = 0; | |
4278 | unsigned long consumed_pos_before, consumed_pos_after; | |
4279 | ||
4280 | ret = lttng_consumer_sample_snapshot_positions(stream); | |
4281 | if (ret < 0) { | |
4282 | ERR("Taking snapshot positions"); | |
4283 | goto end; | |
4284 | } | |
4285 | ||
4286 | ret = lttng_consumer_get_consumed_snapshot(stream, &consumed_pos_before); | |
4287 | if (ret < 0) { | |
4288 | ERR("Consumed snapshot position"); | |
4289 | goto end; | |
4290 | } | |
4291 | ||
fa29bfbf | 4292 | switch (the_consumer_data.type) { |
5f3aff8b MD |
4293 | case LTTNG_CONSUMER_KERNEL: |
4294 | ret = kernctl_buffer_clear(stream->wait_fd); | |
4295 | if (ret < 0) { | |
96393977 | 4296 | ERR("Failed to clear kernel stream (ret = %d)", ret); |
5f3aff8b MD |
4297 | goto end; |
4298 | } | |
4299 | break; | |
4300 | case LTTNG_CONSUMER32_UST: | |
4301 | case LTTNG_CONSUMER64_UST: | |
881fc67f MD |
4302 | ret = lttng_ustconsumer_clear_buffer(stream); |
4303 | if (ret < 0) { | |
4304 | ERR("Failed to clear ust stream (ret = %d)", ret); | |
4305 | goto end; | |
4306 | } | |
5f3aff8b MD |
4307 | break; |
4308 | default: | |
4309 | ERR("Unknown consumer_data type"); | |
4310 | abort(); | |
4311 | } | |
4312 | ||
4313 | ret = lttng_consumer_sample_snapshot_positions(stream); | |
4314 | if (ret < 0) { | |
4315 | ERR("Taking snapshot positions"); | |
4316 | goto end; | |
4317 | } | |
4318 | ret = lttng_consumer_get_consumed_snapshot(stream, &consumed_pos_after); | |
4319 | if (ret < 0) { | |
4320 | ERR("Consumed snapshot position"); | |
4321 | goto end; | |
4322 | } | |
4323 | DBG("clear: before: %lu after: %lu", consumed_pos_before, consumed_pos_after); | |
4324 | end: | |
4325 | return ret; | |
4326 | } | |
4327 | ||
28ab034a | 4328 | static int consumer_clear_stream(struct lttng_consumer_stream *stream) |
5f3aff8b MD |
4329 | { |
4330 | int ret; | |
4331 | ||
cd9adb8b | 4332 | ret = consumer_stream_flush_buffer(stream, true); |
5f3aff8b | 4333 | if (ret < 0) { |
28ab034a | 4334 | ERR("Failed to flush stream %" PRIu64 " during channel clear", stream->key); |
5f3aff8b MD |
4335 | ret = LTTCOMM_CONSUMERD_FATAL; |
4336 | goto error; | |
4337 | } | |
4338 | ||
4339 | ret = consumer_clear_buffer(stream); | |
4340 | if (ret < 0) { | |
28ab034a | 4341 | ERR("Failed to clear stream %" PRIu64 " during channel clear", stream->key); |
5f3aff8b MD |
4342 | ret = LTTCOMM_CONSUMERD_FATAL; |
4343 | goto error; | |
4344 | } | |
4345 | ||
4346 | ret = LTTCOMM_CONSUMERD_SUCCESS; | |
4347 | error: | |
4348 | return ret; | |
4349 | } | |
4350 | ||
28ab034a | 4351 | static int consumer_clear_unmonitored_channel(struct lttng_consumer_channel *channel) |
5f3aff8b | 4352 | { |
07c4863f | 4353 | const lttng::urcu::read_lock_guard read_lock; |
a1a1df65 JG |
4354 | const lttng::pthread::lock_guard channel_lock(channel->lock); |
4355 | ||
4356 | for (auto stream : lttng::urcu::list_iteration_adapter<lttng_consumer_stream, | |
4357 | <tng_consumer_stream::send_node>( | |
4358 | channel->streams.head)) { | |
5f3aff8b | 4359 | health_code_update(); |
a1a1df65 JG |
4360 | |
4361 | const lttng::pthread::lock_guard stream_lock(stream->lock); | |
4362 | const auto ret = consumer_clear_stream(stream); | |
5f3aff8b | 4363 | if (ret) { |
a1a1df65 | 4364 | return ret; |
5f3aff8b | 4365 | } |
5f3aff8b | 4366 | } |
5f3aff8b | 4367 | |
a1a1df65 | 4368 | return 0; |
5f3aff8b MD |
4369 | } |
4370 | ||
02d02e31 JD |
4371 | /* |
4372 | * Check if a stream is ready to be rotated after extracting it. | |
4373 | * | |
4374 | * Return 1 if it is ready for rotation, 0 if it is not, a negative value on | |
4375 | * error. Stream lock must be held. | |
4376 | */ | |
4377 | int lttng_consumer_stream_is_rotate_ready(struct lttng_consumer_stream *stream) | |
4378 | { | |
28ab034a JG |
4379 | DBG("Check is rotate ready for stream %" PRIu64 " ready %u rotate_position %" PRIu64 |
4380 | " last_sequence_number %" PRIu64, | |
4381 | stream->key, | |
4382 | stream->rotate_ready, | |
4383 | stream->rotate_position, | |
4384 | stream->last_sequence_number); | |
02d02e31 | 4385 | if (stream->rotate_ready) { |
a40a503f | 4386 | return 1; |
02d02e31 JD |
4387 | } |
4388 | ||
4389 | /* | |
a40a503f MD |
4390 | * If packet seq num is unavailable, it means we are interacting |
4391 | * with a pre-2.8 lttng-modules which does not implement the | |
4392 | * sequence number. Rotation should never be used by sessiond in this | |
4393 | * scenario. | |
02d02e31 | 4394 | */ |
a40a503f MD |
4395 | if (stream->sequence_number_unavailable) { |
4396 | ERR("Internal error: rotation used on stream %" PRIu64 | |
28ab034a JG |
4397 | " with unavailable sequence number", |
4398 | stream->key); | |
a40a503f | 4399 | return -1; |
02d02e31 JD |
4400 | } |
4401 | ||
28ab034a | 4402 | if (stream->rotate_position == -1ULL || stream->last_sequence_number == -1ULL) { |
a40a503f | 4403 | return 0; |
02d02e31 JD |
4404 | } |
4405 | ||
a40a503f MD |
4406 | /* |
4407 | * Rotate position not reached yet. The stream rotate position is | |
4408 | * the position of the next packet belonging to the next trace chunk, | |
4409 | * but consumerd considers rotation ready when reaching the last | |
4410 | * packet of the current chunk, hence the "rotate_position - 1". | |
4411 | */ | |
f8528c7a | 4412 | |
28ab034a JG |
4413 | DBG("Check is rotate ready for stream %" PRIu64 " last_sequence_number %" PRIu64 |
4414 | " rotate_position %" PRIu64, | |
4415 | stream->key, | |
4416 | stream->last_sequence_number, | |
4417 | stream->rotate_position); | |
a40a503f MD |
4418 | if (stream->last_sequence_number >= stream->rotate_position - 1) { |
4419 | return 1; | |
02d02e31 | 4420 | } |
02d02e31 | 4421 | |
a40a503f | 4422 | return 0; |
02d02e31 JD |
4423 | } |
4424 | ||
d73bf3d7 JD |
4425 | /* |
4426 | * Reset the state for a stream after a rotation occurred. | |
4427 | */ | |
4428 | void lttng_consumer_reset_stream_rotate_state(struct lttng_consumer_stream *stream) | |
4429 | { | |
28ab034a | 4430 | DBG("lttng_consumer_reset_stream_rotate_state for stream %" PRIu64, stream->key); |
a40a503f | 4431 | stream->rotate_position = -1ULL; |
d73bf3d7 JD |
4432 | stream->rotate_ready = false; |
4433 | } | |
4434 | ||
4435 | /* | |
4436 | * Perform the rotation a local stream file. | |
4437 | */ | |
28ab034a | 4438 | static int rotate_local_stream(struct lttng_consumer_stream *stream) |
d73bf3d7 | 4439 | { |
d2956687 | 4440 | int ret = 0; |
d73bf3d7 | 4441 | |
d2956687 | 4442 | DBG("Rotate local stream: stream key %" PRIu64 ", channel key %" PRIu64, |
28ab034a JG |
4443 | stream->key, |
4444 | stream->chan->key); | |
d73bf3d7 | 4445 | stream->tracefile_size_current = 0; |
d2956687 | 4446 | stream->tracefile_count_current = 0; |
d73bf3d7 | 4447 | |
d2956687 JG |
4448 | if (stream->out_fd >= 0) { |
4449 | ret = close(stream->out_fd); | |
4450 | if (ret) { | |
4451 | PERROR("Failed to close stream out_fd of channel \"%s\"", | |
28ab034a | 4452 | stream->chan->name); |
d2956687 JG |
4453 | } |
4454 | stream->out_fd = -1; | |
4455 | } | |
d73bf3d7 | 4456 | |
d2956687 | 4457 | if (stream->index_file) { |
d73bf3d7 | 4458 | lttng_index_file_put(stream->index_file); |
cd9adb8b | 4459 | stream->index_file = nullptr; |
d73bf3d7 JD |
4460 | } |
4461 | ||
d2956687 JG |
4462 | if (!stream->trace_chunk) { |
4463 | goto end; | |
4464 | } | |
d73bf3d7 | 4465 | |
d2956687 | 4466 | ret = consumer_stream_create_output_files(stream, true); |
d73bf3d7 JD |
4467 | end: |
4468 | return ret; | |
d73bf3d7 JD |
4469 | } |
4470 | ||
d73bf3d7 JD |
4471 | /* |
4472 | * Performs the stream rotation for the rotate session feature if needed. | |
d2956687 | 4473 | * It must be called with the channel and stream locks held. |
d73bf3d7 JD |
4474 | * |
4475 | * Return 0 on success, a negative number of error. | |
4476 | */ | |
f46376a1 | 4477 | int lttng_consumer_rotate_stream(struct lttng_consumer_stream *stream) |
d73bf3d7 JD |
4478 | { |
4479 | int ret; | |
4480 | ||
4481 | DBG("Consumer rotate stream %" PRIu64, stream->key); | |
4482 | ||
d2956687 JG |
4483 | /* |
4484 | * Update the stream's 'current' chunk to the session's (channel) | |
4485 | * now-current chunk. | |
4486 | */ | |
4487 | lttng_trace_chunk_put(stream->trace_chunk); | |
4488 | if (stream->chan->trace_chunk == stream->trace_chunk) { | |
4489 | /* | |
4490 | * A channel can be rotated and not have a "next" chunk | |
4491 | * to transition to. In that case, the channel's "current chunk" | |
4492 | * has not been closed yet, but it has not been updated to | |
4493 | * a "next" trace chunk either. Hence, the stream, like its | |
4494 | * parent channel, becomes part of no chunk and can't output | |
4495 | * anything until a new trace chunk is created. | |
4496 | */ | |
cd9adb8b | 4497 | stream->trace_chunk = nullptr; |
28ab034a | 4498 | } else if (stream->chan->trace_chunk && !lttng_trace_chunk_get(stream->chan->trace_chunk)) { |
d2956687 JG |
4499 | ERR("Failed to acquire a reference to channel's trace chunk during stream rotation"); |
4500 | ret = -1; | |
4501 | goto error; | |
4502 | } else { | |
4503 | /* | |
4504 | * Update the stream's trace chunk to its parent channel's | |
4505 | * current trace chunk. | |
4506 | */ | |
4507 | stream->trace_chunk = stream->chan->trace_chunk; | |
4508 | } | |
4509 | ||
c35f9726 | 4510 | if (stream->net_seq_idx == (uint64_t) -1ULL) { |
f46376a1 | 4511 | ret = rotate_local_stream(stream); |
c35f9726 JG |
4512 | if (ret < 0) { |
4513 | ERR("Failed to rotate stream, ret = %i", ret); | |
4514 | goto error; | |
4515 | } | |
d73bf3d7 JD |
4516 | } |
4517 | ||
d2956687 JG |
4518 | if (stream->metadata_flag && stream->trace_chunk) { |
4519 | /* | |
4520 | * If the stream has transitioned to a new trace | |
4521 | * chunk, the metadata should be re-dumped to the | |
4522 | * newest chunk. | |
4523 | * | |
4524 | * However, it is possible for a stream to transition to | |
4525 | * a "no-chunk" state. This can happen if a rotation | |
4526 | * occurs on an inactive session. In such cases, the metadata | |
4527 | * regeneration will happen when the next trace chunk is | |
4528 | * created. | |
4529 | */ | |
4530 | ret = consumer_metadata_stream_dump(stream); | |
4531 | if (ret) { | |
4532 | goto error; | |
d73bf3d7 JD |
4533 | } |
4534 | } | |
4535 | lttng_consumer_reset_stream_rotate_state(stream); | |
4536 | ||
4537 | ret = 0; | |
4538 | ||
4539 | error: | |
4540 | return ret; | |
4541 | } | |
4542 | ||
b99a8d42 JD |
4543 | /* |
4544 | * Rotate all the ready streams now. | |
4545 | * | |
4546 | * This is especially important for low throughput streams that have already | |
4547 | * been consumed, we cannot wait for their next packet to perform the | |
4548 | * rotation. | |
92b7a7f8 MD |
4549 | * Need to be called with RCU read-side lock held to ensure existence of |
4550 | * channel. | |
b99a8d42 JD |
4551 | * |
4552 | * Returns 0 on success, < 0 on error | |
4553 | */ | |
28ab034a | 4554 | int lttng_consumer_rotate_ready_streams(struct lttng_consumer_channel *channel, uint64_t key) |
b99a8d42 JD |
4555 | { |
4556 | int ret; | |
3d46ea1a | 4557 | const auto ht = the_consumer_data.stream_per_chan_id_ht; |
b99a8d42 | 4558 | |
48b7cdc2 FD |
4559 | ASSERT_RCU_READ_LOCKED(); |
4560 | ||
b99a8d42 JD |
4561 | DBG("Consumer rotate ready streams in channel %" PRIu64, key); |
4562 | ||
3d46ea1a JG |
4563 | for (auto *stream : lttng::urcu::lfht_filtered_iteration_adapter< |
4564 | lttng_consumer_stream, | |
4565 | decltype(lttng_consumer_stream::node_channel_id), | |
4566 | <tng_consumer_stream::node_channel_id, | |
4567 | std::uint64_t>(*ht->ht, | |
4568 | &channel->key, | |
4569 | ht->hash_fct(&channel->key, lttng_ht_seed), | |
4570 | ht->match_fct)) { | |
b99a8d42 JD |
4571 | health_code_update(); |
4572 | ||
3d46ea1a JG |
4573 | const lttng::pthread::lock_guard channel_lock(stream->chan->lock); |
4574 | const lttng::pthread::lock_guard stream_lock(stream->lock); | |
b99a8d42 JD |
4575 | |
4576 | if (!stream->rotate_ready) { | |
b99a8d42 JD |
4577 | continue; |
4578 | } | |
b99a8d42 | 4579 | |
3d46ea1a | 4580 | DBG("Consumer rotate ready stream %" PRIu64, stream->key); |
f46376a1 | 4581 | ret = lttng_consumer_rotate_stream(stream); |
b99a8d42 JD |
4582 | if (ret) { |
4583 | goto end; | |
4584 | } | |
4585 | } | |
4586 | ||
4587 | ret = 0; | |
4588 | ||
4589 | end: | |
b99a8d42 JD |
4590 | return ret; |
4591 | } | |
4592 | ||
28ab034a JG |
4593 | enum lttcomm_return_code lttng_consumer_init_command(struct lttng_consumer_local_data *ctx, |
4594 | const lttng_uuid& sessiond_uuid) | |
00fb02ac | 4595 | { |
d2956687 | 4596 | enum lttcomm_return_code ret; |
c70636a7 | 4597 | char uuid_str[LTTNG_UUID_STR_LEN]; |
00fb02ac | 4598 | |
d2956687 JG |
4599 | if (ctx->sessiond_uuid.is_set) { |
4600 | ret = LTTCOMM_CONSUMERD_ALREADY_SET; | |
00fb02ac JD |
4601 | goto end; |
4602 | } | |
4603 | ||
d2956687 | 4604 | ctx->sessiond_uuid.is_set = true; |
328c2fe7 | 4605 | ctx->sessiond_uuid.value = sessiond_uuid; |
d2956687 JG |
4606 | ret = LTTCOMM_CONSUMERD_SUCCESS; |
4607 | lttng_uuid_to_str(sessiond_uuid, uuid_str); | |
4608 | DBG("Received session daemon UUID: %s", uuid_str); | |
00fb02ac JD |
4609 | end: |
4610 | return ret; | |
4611 | } | |
4612 | ||
28ab034a JG |
4613 | enum lttcomm_return_code |
4614 | lttng_consumer_create_trace_chunk(const uint64_t *relayd_id, | |
4615 | uint64_t session_id, | |
4616 | uint64_t chunk_id, | |
4617 | time_t chunk_creation_timestamp, | |
4618 | const char *chunk_override_name, | |
4619 | const struct lttng_credentials *credentials, | |
4620 | struct lttng_directory_handle *chunk_directory_handle) | |
00fb02ac JD |
4621 | { |
4622 | int ret; | |
d2956687 | 4623 | enum lttcomm_return_code ret_code = LTTCOMM_CONSUMERD_SUCCESS; |
cd9adb8b | 4624 | struct lttng_trace_chunk *created_chunk = nullptr, *published_chunk = nullptr; |
d2956687 JG |
4625 | enum lttng_trace_chunk_status chunk_status; |
4626 | char relayd_id_buffer[MAX_INT_DEC_LEN(*relayd_id)]; | |
4627 | char creation_timestamp_buffer[ISO8601_STR_LEN]; | |
4628 | const char *relayd_id_str = "(none)"; | |
4629 | const char *creation_timestamp_str; | |
92816cc3 | 4630 | |
d2956687 JG |
4631 | if (relayd_id) { |
4632 | /* Only used for logging purposes. */ | |
28ab034a | 4633 | ret = snprintf(relayd_id_buffer, sizeof(relayd_id_buffer), "%" PRIu64, *relayd_id); |
d2956687 JG |
4634 | if (ret > 0 && ret < sizeof(relayd_id_buffer)) { |
4635 | relayd_id_str = relayd_id_buffer; | |
4636 | } else { | |
4637 | relayd_id_str = "(formatting error)"; | |
4638 | } | |
d01ef216 | 4639 | } |
d2956687 | 4640 | |
d01ef216 | 4641 | /* Local protocol error. */ |
a0377dfe | 4642 | LTTNG_ASSERT(chunk_creation_timestamp); |
d2956687 | 4643 | ret = time_to_iso8601_str(chunk_creation_timestamp, |
28ab034a JG |
4644 | creation_timestamp_buffer, |
4645 | sizeof(creation_timestamp_buffer)); | |
4646 | creation_timestamp_str = !ret ? creation_timestamp_buffer : "(formatting error)"; | |
d2956687 JG |
4647 | |
4648 | DBG("Consumer create trace chunk command: relay_id = %s" | |
28ab034a JG |
4649 | ", session_id = %" PRIu64 ", chunk_id = %" PRIu64 ", chunk_override_name = %s" |
4650 | ", chunk_creation_timestamp = %s", | |
4651 | relayd_id_str, | |
4652 | session_id, | |
4653 | chunk_id, | |
4654 | chunk_override_name ?: "(none)", | |
4655 | creation_timestamp_str); | |
92816cc3 JG |
4656 | |
4657 | /* | |
d2956687 JG |
4658 | * The trace chunk registry, as used by the consumer daemon, implicitly |
4659 | * owns the trace chunks. This is only needed in the consumer since | |
4660 | * the consumer has no notion of a session beyond session IDs being | |
4661 | * used to identify other objects. | |
4662 | * | |
4663 | * The lttng_trace_chunk_registry_publish() call below provides a | |
4664 | * reference which is not released; it implicitly becomes the session | |
4665 | * daemon's reference to the chunk in the consumer daemon. | |
4666 | * | |
4667 | * The lifetime of trace chunks in the consumer daemon is managed by | |
4668 | * the session daemon through the LTTNG_CONSUMER_CREATE_TRACE_CHUNK | |
4669 | * and LTTNG_CONSUMER_DESTROY_TRACE_CHUNK commands. | |
92816cc3 | 4670 | */ |
cd9adb8b | 4671 | created_chunk = lttng_trace_chunk_create(chunk_id, chunk_creation_timestamp, nullptr); |
d2956687 JG |
4672 | if (!created_chunk) { |
4673 | ERR("Failed to create trace chunk"); | |
4674 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
7ea24db3 | 4675 | goto error; |
d2956687 | 4676 | } |
92816cc3 | 4677 | |
d2956687 | 4678 | if (chunk_override_name) { |
28ab034a | 4679 | chunk_status = lttng_trace_chunk_override_name(created_chunk, chunk_override_name); |
d2956687 JG |
4680 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
4681 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
7ea24db3 | 4682 | goto error; |
92816cc3 JG |
4683 | } |
4684 | } | |
4685 | ||
d2956687 | 4686 | if (chunk_directory_handle) { |
28ab034a | 4687 | chunk_status = lttng_trace_chunk_set_credentials(created_chunk, credentials); |
d2956687 JG |
4688 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
4689 | ERR("Failed to set trace chunk credentials"); | |
4690 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
7ea24db3 | 4691 | goto error; |
d2956687 JG |
4692 | } |
4693 | /* | |
4694 | * The consumer daemon has no ownership of the chunk output | |
4695 | * directory. | |
4696 | */ | |
28ab034a | 4697 | chunk_status = lttng_trace_chunk_set_as_user(created_chunk, chunk_directory_handle); |
cd9adb8b | 4698 | chunk_directory_handle = nullptr; |
d2956687 JG |
4699 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
4700 | ERR("Failed to set trace chunk's directory handle"); | |
4701 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
7ea24db3 | 4702 | goto error; |
92816cc3 JG |
4703 | } |
4704 | } | |
4705 | ||
d2956687 | 4706 | published_chunk = lttng_trace_chunk_registry_publish_chunk( |
28ab034a | 4707 | the_consumer_data.chunk_registry, session_id, created_chunk); |
d2956687 | 4708 | lttng_trace_chunk_put(created_chunk); |
cd9adb8b | 4709 | created_chunk = nullptr; |
d2956687 JG |
4710 | if (!published_chunk) { |
4711 | ERR("Failed to publish trace chunk"); | |
4712 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
7ea24db3 | 4713 | goto error; |
d88744a4 JD |
4714 | } |
4715 | ||
3d46ea1a JG |
4716 | for (auto *channel : lttng::urcu::lfht_filtered_iteration_adapter< |
4717 | lttng_consumer_channel, | |
4718 | decltype(lttng_consumer_channel::channels_by_session_id_ht_node), | |
4719 | <tng_consumer_channel::channels_by_session_id_ht_node, | |
4720 | std::uint64_t>(*the_consumer_data.channels_by_session_id_ht->ht, | |
4721 | &session_id, | |
4722 | the_consumer_data.channels_by_session_id_ht->hash_fct( | |
4723 | &session_id, lttng_ht_seed), | |
4724 | the_consumer_data.channels_by_session_id_ht->match_fct)) { | |
4725 | ret = lttng_consumer_channel_set_trace_chunk(channel, published_chunk); | |
4726 | if (ret) { | |
4727 | /* | |
4728 | * Roll-back the creation of this chunk. | |
4729 | * | |
4730 | * This is important since the session daemon will | |
4731 | * assume that the creation of this chunk failed and | |
4732 | * will never ask for it to be closed, resulting | |
4733 | * in a leak and an inconsistent state for some | |
4734 | * channels. | |
4735 | */ | |
4736 | enum lttcomm_return_code close_ret; | |
4737 | char path[LTTNG_PATH_MAX]; | |
d2956687 | 4738 | |
3d46ea1a JG |
4739 | DBG("Failed to set new trace chunk on existing channels, rolling back"); |
4740 | close_ret = lttng_consumer_close_trace_chunk(relayd_id, | |
4741 | session_id, | |
4742 | chunk_id, | |
4743 | chunk_creation_timestamp, | |
4744 | nullptr, | |
4745 | path); | |
4746 | if (close_ret != LTTCOMM_CONSUMERD_SUCCESS) { | |
4747 | ERR("Failed to roll-back the creation of new chunk: session_id = %" PRIu64 | |
4748 | ", chunk_id = %" PRIu64, | |
4749 | session_id, | |
4750 | chunk_id); | |
d2956687 | 4751 | } |
3d46ea1a JG |
4752 | |
4753 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
4754 | break; | |
d2956687 | 4755 | } |
a1ae2ea5 JD |
4756 | } |
4757 | ||
e5add6d0 JG |
4758 | if (relayd_id) { |
4759 | struct consumer_relayd_sock_pair *relayd; | |
4760 | ||
4761 | relayd = consumer_find_relayd(*relayd_id); | |
4762 | if (relayd) { | |
4763 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
28ab034a | 4764 | ret = relayd_create_trace_chunk(&relayd->control_sock, published_chunk); |
e5add6d0 JG |
4765 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); |
4766 | } else { | |
4767 | ERR("Failed to find relay daemon socket: relayd_id = %" PRIu64, *relayd_id); | |
4768 | } | |
4769 | ||
4770 | if (!relayd || ret) { | |
4771 | enum lttcomm_return_code close_ret; | |
ecd1a12f | 4772 | char path[LTTNG_PATH_MAX]; |
e5add6d0 JG |
4773 | |
4774 | close_ret = lttng_consumer_close_trace_chunk(relayd_id, | |
28ab034a JG |
4775 | session_id, |
4776 | chunk_id, | |
4777 | chunk_creation_timestamp, | |
cd9adb8b | 4778 | nullptr, |
28ab034a | 4779 | path); |
e5add6d0 | 4780 | if (close_ret != LTTCOMM_CONSUMERD_SUCCESS) { |
28ab034a JG |
4781 | ERR("Failed to roll-back the creation of new chunk: session_id = %" PRIu64 |
4782 | ", chunk_id = %" PRIu64, | |
4783 | session_id, | |
4784 | chunk_id); | |
e5add6d0 JG |
4785 | } |
4786 | ||
4787 | ret_code = LTTCOMM_CONSUMERD_CREATE_TRACE_CHUNK_FAILED; | |
7ea24db3 | 4788 | goto error_unlock; |
e5add6d0 JG |
4789 | } |
4790 | } | |
7ea24db3 | 4791 | error_unlock: |
7ea24db3 | 4792 | error: |
d2956687 JG |
4793 | /* Release the reference returned by the "publish" operation. */ |
4794 | lttng_trace_chunk_put(published_chunk); | |
9bb5f1f8 | 4795 | lttng_trace_chunk_put(created_chunk); |
d2956687 | 4796 | return ret_code; |
a1ae2ea5 JD |
4797 | } |
4798 | ||
28ab034a JG |
4799 | enum lttcomm_return_code |
4800 | lttng_consumer_close_trace_chunk(const uint64_t *relayd_id, | |
4801 | uint64_t session_id, | |
4802 | uint64_t chunk_id, | |
4803 | time_t chunk_close_timestamp, | |
4804 | const enum lttng_trace_chunk_command_type *close_command, | |
4805 | char *path) | |
a1ae2ea5 | 4806 | { |
d2956687 JG |
4807 | enum lttcomm_return_code ret_code = LTTCOMM_CONSUMERD_SUCCESS; |
4808 | struct lttng_trace_chunk *chunk; | |
4809 | char relayd_id_buffer[MAX_INT_DEC_LEN(*relayd_id)]; | |
4810 | const char *relayd_id_str = "(none)"; | |
bbc4768c | 4811 | const char *close_command_name = "none"; |
d2956687 | 4812 | enum lttng_trace_chunk_status chunk_status; |
a1ae2ea5 | 4813 | |
d2956687 JG |
4814 | if (relayd_id) { |
4815 | int ret; | |
4816 | ||
4817 | /* Only used for logging purposes. */ | |
28ab034a | 4818 | ret = snprintf(relayd_id_buffer, sizeof(relayd_id_buffer), "%" PRIu64, *relayd_id); |
d2956687 JG |
4819 | if (ret > 0 && ret < sizeof(relayd_id_buffer)) { |
4820 | relayd_id_str = relayd_id_buffer; | |
4821 | } else { | |
4822 | relayd_id_str = "(formatting error)"; | |
4823 | } | |
bbc4768c JG |
4824 | } |
4825 | if (close_command) { | |
28ab034a | 4826 | close_command_name = lttng_trace_chunk_command_type_get_name(*close_command); |
bbc4768c | 4827 | } |
d2956687 JG |
4828 | |
4829 | DBG("Consumer close trace chunk command: relayd_id = %s" | |
28ab034a JG |
4830 | ", session_id = %" PRIu64 ", chunk_id = %" PRIu64 ", close command = %s", |
4831 | relayd_id_str, | |
4832 | session_id, | |
4833 | chunk_id, | |
4834 | close_command_name); | |
bbc4768c | 4835 | |
d2956687 | 4836 | chunk = lttng_trace_chunk_registry_find_chunk( |
28ab034a | 4837 | the_consumer_data.chunk_registry, session_id, chunk_id); |
bbc4768c | 4838 | if (!chunk) { |
28ab034a JG |
4839 | ERR("Failed to find chunk: session_id = %" PRIu64 ", chunk_id = %" PRIu64, |
4840 | session_id, | |
4841 | chunk_id); | |
d2956687 | 4842 | ret_code = LTTCOMM_CONSUMERD_UNKNOWN_TRACE_CHUNK; |
a1ae2ea5 JD |
4843 | goto end; |
4844 | } | |
4845 | ||
28ab034a | 4846 | chunk_status = lttng_trace_chunk_set_close_timestamp(chunk, chunk_close_timestamp); |
d2956687 JG |
4847 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
4848 | ret_code = LTTCOMM_CONSUMERD_CLOSE_TRACE_CHUNK_FAILED; | |
4849 | goto end; | |
45f1d9a1 | 4850 | } |
bbc4768c JG |
4851 | |
4852 | if (close_command) { | |
28ab034a | 4853 | chunk_status = lttng_trace_chunk_set_close_command(chunk, *close_command); |
bbc4768c JG |
4854 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
4855 | ret_code = LTTCOMM_CONSUMERD_CLOSE_TRACE_CHUNK_FAILED; | |
4856 | goto end; | |
4857 | } | |
4858 | } | |
a1ae2ea5 | 4859 | |
d2956687 JG |
4860 | /* |
4861 | * chunk is now invalid to access as we no longer hold a reference to | |
4862 | * it; it is only kept around to compare it (by address) to the | |
4863 | * current chunk found in the session's channels. | |
4864 | */ | |
c3ade133 JG |
4865 | for (auto *channel : |
4866 | lttng::urcu::lfht_iteration_adapter<lttng_consumer_channel, | |
4867 | decltype(lttng_consumer_channel::node), | |
4868 | <tng_consumer_channel::node>( | |
4869 | *the_consumer_data.channel_ht->ht)) { | |
4870 | int ret; | |
a1ae2ea5 | 4871 | |
c3ade133 JG |
4872 | /* |
4873 | * Only change the channel's chunk to NULL if it still | |
4874 | * references the chunk being closed. The channel may | |
4875 | * reference a newer channel in the case of a session | |
4876 | * rotation. When a session rotation occurs, the "next" | |
4877 | * chunk is created before the "current" chunk is closed. | |
4878 | */ | |
4879 | if (channel->trace_chunk != chunk) { | |
4880 | continue; | |
4881 | } | |
4882 | ret = lttng_consumer_channel_set_trace_chunk(channel, nullptr); | |
4883 | if (ret) { | |
d2956687 | 4884 | /* |
c3ade133 JG |
4885 | * Attempt to close the chunk on as many channels as |
4886 | * possible. | |
d2956687 | 4887 | */ |
c3ade133 | 4888 | ret_code = LTTCOMM_CONSUMERD_CLOSE_TRACE_CHUNK_FAILED; |
d2956687 | 4889 | } |
a1ae2ea5 | 4890 | } |
c3ade133 | 4891 | |
bbc4768c JG |
4892 | if (relayd_id) { |
4893 | int ret; | |
4894 | struct consumer_relayd_sock_pair *relayd; | |
4895 | ||
4896 | relayd = consumer_find_relayd(*relayd_id); | |
4897 | if (relayd) { | |
4898 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
28ab034a | 4899 | ret = relayd_close_trace_chunk(&relayd->control_sock, chunk, path); |
bbc4768c JG |
4900 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); |
4901 | } else { | |
28ab034a | 4902 | ERR("Failed to find relay daemon socket: relayd_id = %" PRIu64, *relayd_id); |
bbc4768c JG |
4903 | } |
4904 | ||
4905 | if (!relayd || ret) { | |
4906 | ret_code = LTTCOMM_CONSUMERD_CLOSE_TRACE_CHUNK_FAILED; | |
4907 | goto error_unlock; | |
4908 | } | |
4909 | } | |
4910 | error_unlock: | |
d2956687 | 4911 | end: |
bbc4768c JG |
4912 | /* |
4913 | * Release the reference returned by the "find" operation and | |
4914 | * the session daemon's implicit reference to the chunk. | |
4915 | */ | |
4916 | lttng_trace_chunk_put(chunk); | |
4917 | lttng_trace_chunk_put(chunk); | |
4918 | ||
d2956687 | 4919 | return ret_code; |
a1ae2ea5 | 4920 | } |
3654ed19 | 4921 | |
28ab034a JG |
4922 | enum lttcomm_return_code |
4923 | lttng_consumer_trace_chunk_exists(const uint64_t *relayd_id, uint64_t session_id, uint64_t chunk_id) | |
3654ed19 | 4924 | { |
c35f9726 | 4925 | int ret; |
d2956687 | 4926 | enum lttcomm_return_code ret_code; |
d2956687 JG |
4927 | char relayd_id_buffer[MAX_INT_DEC_LEN(*relayd_id)]; |
4928 | const char *relayd_id_str = "(none)"; | |
c35f9726 | 4929 | const bool is_local_trace = !relayd_id; |
cd9adb8b | 4930 | struct consumer_relayd_sock_pair *relayd = nullptr; |
6b584c2e | 4931 | bool chunk_exists_local, chunk_exists_remote; |
07c4863f | 4932 | const lttng::urcu::read_lock_guard read_lock; |
d2956687 JG |
4933 | |
4934 | if (relayd_id) { | |
d2956687 | 4935 | /* Only used for logging purposes. */ |
28ab034a | 4936 | ret = snprintf(relayd_id_buffer, sizeof(relayd_id_buffer), "%" PRIu64, *relayd_id); |
d2956687 JG |
4937 | if (ret > 0 && ret < sizeof(relayd_id_buffer)) { |
4938 | relayd_id_str = relayd_id_buffer; | |
4939 | } else { | |
4940 | relayd_id_str = "(formatting error)"; | |
4941 | } | |
d01ef216 | 4942 | } |
d2956687 JG |
4943 | |
4944 | DBG("Consumer trace chunk exists command: relayd_id = %s" | |
28ab034a JG |
4945 | ", chunk_id = %" PRIu64, |
4946 | relayd_id_str, | |
4947 | chunk_id); | |
6b584c2e | 4948 | ret = lttng_trace_chunk_registry_chunk_exists( |
28ab034a | 4949 | the_consumer_data.chunk_registry, session_id, chunk_id, &chunk_exists_local); |
6b584c2e JG |
4950 | if (ret) { |
4951 | /* Internal error. */ | |
4952 | ERR("Failed to query the existence of a trace chunk"); | |
4953 | ret_code = LTTCOMM_CONSUMERD_FATAL; | |
13e3b280 | 4954 | goto end; |
6b584c2e | 4955 | } |
28ab034a | 4956 | DBG("Trace chunk %s locally", chunk_exists_local ? "exists" : "does not exist"); |
6b584c2e | 4957 | if (chunk_exists_local) { |
c35f9726 | 4958 | ret_code = LTTCOMM_CONSUMERD_TRACE_CHUNK_EXISTS_LOCAL; |
c35f9726 JG |
4959 | goto end; |
4960 | } else if (is_local_trace) { | |
4961 | ret_code = LTTCOMM_CONSUMERD_UNKNOWN_TRACE_CHUNK; | |
4962 | goto end; | |
4963 | } | |
4964 | ||
c35f9726 JG |
4965 | relayd = consumer_find_relayd(*relayd_id); |
4966 | if (!relayd) { | |
4967 | ERR("Failed to find relayd %" PRIu64, *relayd_id); | |
4968 | ret_code = LTTCOMM_CONSUMERD_INVALID_PARAMETERS; | |
4969 | goto end_rcu_unlock; | |
4970 | } | |
4971 | DBG("Looking up existence of trace chunk on relay daemon"); | |
4972 | pthread_mutex_lock(&relayd->ctrl_sock_mutex); | |
28ab034a | 4973 | ret = relayd_trace_chunk_exists(&relayd->control_sock, chunk_id, &chunk_exists_remote); |
c35f9726 JG |
4974 | pthread_mutex_unlock(&relayd->ctrl_sock_mutex); |
4975 | if (ret < 0) { | |
4976 | ERR("Failed to look-up the existence of trace chunk on relay daemon"); | |
4977 | ret_code = LTTCOMM_CONSUMERD_RELAYD_FAIL; | |
4978 | goto end_rcu_unlock; | |
4979 | } | |
4980 | ||
28ab034a JG |
4981 | ret_code = chunk_exists_remote ? LTTCOMM_CONSUMERD_TRACE_CHUNK_EXISTS_REMOTE : |
4982 | LTTCOMM_CONSUMERD_UNKNOWN_TRACE_CHUNK; | |
4983 | DBG("Trace chunk %s on relay daemon", chunk_exists_remote ? "exists" : "does not exist"); | |
d2956687 | 4984 | |
c35f9726 | 4985 | end_rcu_unlock: |
c35f9726 | 4986 | end: |
d2956687 | 4987 | return ret_code; |
3654ed19 | 4988 | } |
5f3aff8b | 4989 | |
28ab034a | 4990 | static int consumer_clear_monitored_channel(struct lttng_consumer_channel *channel) |
5f3aff8b | 4991 | { |
5f3aff8b | 4992 | int ret; |
3d46ea1a JG |
4993 | const auto ht = the_consumer_data.stream_per_chan_id_ht; |
4994 | ||
4995 | for (auto *stream : lttng::urcu::lfht_filtered_iteration_adapter< | |
4996 | lttng_consumer_stream, | |
4997 | decltype(lttng_consumer_stream::node_channel_id), | |
4998 | <tng_consumer_stream::node_channel_id, | |
4999 | std::uint64_t>(*ht->ht, | |
5000 | &channel->key, | |
5001 | ht->hash_fct(&channel->key, lttng_ht_seed), | |
5002 | ht->match_fct)) { | |
5f3aff8b MD |
5003 | /* |
5004 | * Protect against teardown with mutex. | |
5005 | */ | |
3d46ea1a | 5006 | const lttng::pthread::lock_guard stream_lock(stream->lock); |
5f3aff8b | 5007 | if (cds_lfht_is_node_deleted(&stream->node.node)) { |
3d46ea1a | 5008 | continue; |
5f3aff8b | 5009 | } |
3d46ea1a | 5010 | |
5f3aff8b MD |
5011 | ret = consumer_clear_stream(stream); |
5012 | if (ret) { | |
3d46ea1a | 5013 | return ret; |
5f3aff8b | 5014 | } |
5f3aff8b | 5015 | } |
5f3aff8b | 5016 | |
3d46ea1a | 5017 | return LTTCOMM_CONSUMERD_SUCCESS; |
5f3aff8b MD |
5018 | } |
5019 | ||
5020 | int lttng_consumer_clear_channel(struct lttng_consumer_channel *channel) | |
5021 | { | |
5022 | int ret; | |
5023 | ||
5024 | DBG("Consumer clear channel %" PRIu64, channel->key); | |
5025 | ||
5026 | if (channel->type == CONSUMER_CHANNEL_TYPE_METADATA) { | |
5027 | /* | |
5028 | * Nothing to do for the metadata channel/stream. | |
5029 | * Snapshot mechanism already take care of the metadata | |
5030 | * handling/generation, and monitored channels only need to | |
5031 | * have their data stream cleared.. | |
5032 | */ | |
5033 | ret = LTTCOMM_CONSUMERD_SUCCESS; | |
5034 | goto end; | |
5035 | } | |
5036 | ||
5037 | if (!channel->monitor) { | |
5038 | ret = consumer_clear_unmonitored_channel(channel); | |
5039 | } else { | |
5040 | ret = consumer_clear_monitored_channel(channel); | |
5041 | } | |
5042 | end: | |
5043 | return ret; | |
5044 | } | |
04ed9e10 | 5045 | |
28ab034a | 5046 | enum lttcomm_return_code lttng_consumer_open_channel_packets(struct lttng_consumer_channel *channel) |
04ed9e10 | 5047 | { |
04ed9e10 JG |
5048 | enum lttcomm_return_code ret = LTTCOMM_CONSUMERD_SUCCESS; |
5049 | ||
5050 | if (channel->metadata_stream) { | |
5051 | ERR("Open channel packets command attempted on a metadata channel"); | |
a1a1df65 | 5052 | return LTTCOMM_CONSUMERD_INVALID_PARAMETERS; |
04ed9e10 JG |
5053 | } |
5054 | ||
a1a1df65 JG |
5055 | const lttng::urcu::read_lock_guard read_lock; |
5056 | for (auto stream : lttng::urcu::list_iteration_adapter<lttng_consumer_stream, | |
5057 | <tng_consumer_stream::send_node>( | |
5058 | channel->streams.head)) { | |
5059 | enum consumer_stream_open_packet_status status; | |
04ed9e10 | 5060 | |
a1a1df65 JG |
5061 | const lttng::pthread::lock_guard stream_lock(stream->lock); |
5062 | if (cds_lfht_is_node_deleted(&stream->node.node)) { | |
5063 | continue; | |
5064 | } | |
04ed9e10 | 5065 | |
a1a1df65 JG |
5066 | status = consumer_stream_open_packet(stream); |
5067 | switch (status) { | |
5068 | case CONSUMER_STREAM_OPEN_PACKET_STATUS_OPENED: | |
5069 | DBG("Opened a packet in \"open channel packets\" command: stream id = %" PRIu64 | |
5070 | ", channel name = %s, session id = %" PRIu64, | |
5071 | stream->key, | |
5072 | stream->chan->name, | |
5073 | stream->chan->session_id); | |
5074 | stream->opened_packet_in_current_trace_chunk = true; | |
5075 | break; | |
5076 | case CONSUMER_STREAM_OPEN_PACKET_STATUS_NO_SPACE: | |
5077 | DBG("No space left to open a packet in \"open channel packets\" command: stream id = %" PRIu64 | |
5078 | ", channel name = %s, session id = %" PRIu64, | |
5079 | stream->key, | |
5080 | stream->chan->name, | |
5081 | stream->chan->session_id); | |
5082 | break; | |
5083 | case CONSUMER_STREAM_OPEN_PACKET_STATUS_ERROR: | |
5084 | /* | |
5085 | * Only unexpected internal errors can lead to this | |
5086 | * failing. Report an unknown error. | |
5087 | */ | |
5088 | ERR("Failed to flush empty buffer in \"open channel packets\" command: stream id = %" PRIu64 | |
5089 | ", channel id = %" PRIu64 ", channel name = %s" | |
5090 | ", session id = %" PRIu64, | |
5091 | stream->key, | |
5092 | channel->key, | |
5093 | channel->name, | |
5094 | channel->session_id); | |
5095 | return LTTCOMM_CONSUMERD_UNKNOWN_ERROR; | |
5096 | default: | |
5097 | abort(); | |
56047f5a | 5098 | } |
04ed9e10 | 5099 | } |
04ed9e10 | 5100 | |
a1a1df65 | 5101 | return ret; |
04ed9e10 | 5102 | } |
881fc67f MD |
5103 | |
5104 | void lttng_consumer_sigbus_handle(void *addr) | |
5105 | { | |
5106 | lttng_ustconsumer_sigbus_handle(addr); | |
5107 | } |