Commit | Line | Data |
---|---|---|
b8aa1682 JD |
1 | /* |
2 | * Copyright (C) 2012 - Julien Desfossez <jdesfossez@efficios.com> | |
3 | * David Goulet <dgoulet@efficios.com> | |
cd60b05a | 4 | * 2013 - Jérémie Galarneau <jeremie.galarneau@efficios.com> |
b8aa1682 JD |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License, version 2 only, | |
8 | * as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along | |
16 | * with this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
20 | #define _GNU_SOURCE | |
21 | #include <getopt.h> | |
22 | #include <grp.h> | |
23 | #include <limits.h> | |
24 | #include <pthread.h> | |
25 | #include <signal.h> | |
26 | #include <stdio.h> | |
27 | #include <stdlib.h> | |
28 | #include <string.h> | |
29 | #include <sys/mman.h> | |
30 | #include <sys/mount.h> | |
31 | #include <sys/resource.h> | |
32 | #include <sys/socket.h> | |
33 | #include <sys/stat.h> | |
34 | #include <sys/types.h> | |
35 | #include <sys/wait.h> | |
173af62f | 36 | #include <inttypes.h> |
b8aa1682 JD |
37 | #include <urcu/futex.h> |
38 | #include <urcu/uatomic.h> | |
39 | #include <unistd.h> | |
40 | #include <fcntl.h> | |
41 | #include <config.h> | |
42 | ||
43 | #include <lttng/lttng.h> | |
44 | #include <common/common.h> | |
45 | #include <common/compat/poll.h> | |
46 | #include <common/compat/socket.h> | |
47 | #include <common/defaults.h> | |
48 | #include <common/futex.h> | |
49 | #include <common/sessiond-comm/sessiond-comm.h> | |
50 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
51 | #include <common/sessiond-comm/relayd.h> |
52 | #include <common/uri.h> | |
a02de639 | 53 | #include <common/utils.h> |
cd60b05a | 54 | #include <common/config/config.h> |
b8aa1682 | 55 | |
0f907de1 | 56 | #include "cmd.h" |
d3e2ba59 | 57 | #include "ctf-trace.h" |
1c20f0e2 | 58 | #include "index.h" |
0f907de1 | 59 | #include "utils.h" |
b8aa1682 | 60 | #include "lttng-relayd.h" |
d3e2ba59 | 61 | #include "live.h" |
55706a7d | 62 | #include "health-relayd.h" |
b8aa1682 JD |
63 | |
64 | /* command line options */ | |
0f907de1 | 65 | char *opt_output_path; |
b8aa1682 | 66 | static int opt_daemon; |
095a4ae5 MD |
67 | static struct lttng_uri *control_uri; |
68 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 69 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
70 | |
71 | const char *progname; | |
b8aa1682 | 72 | |
65931c8b | 73 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
74 | static int tracing_group_name_override; |
75 | ||
76 | const char * const config_section_name = "relayd"; | |
65931c8b | 77 | |
b8aa1682 JD |
78 | /* |
79 | * Quit pipe for all threads. This permits a single cancellation point | |
80 | * for all threads when receiving an event on the pipe. | |
81 | */ | |
82 | static int thread_quit_pipe[2] = { -1, -1 }; | |
83 | ||
84 | /* | |
85 | * This pipe is used to inform the worker thread that a command is queued and | |
86 | * ready to be processed. | |
87 | */ | |
88 | static int relay_cmd_pipe[2] = { -1, -1 }; | |
89 | ||
26c9d55e | 90 | /* Shared between threads */ |
b8aa1682 JD |
91 | static int dispatch_thread_exit; |
92 | ||
93 | static pthread_t listener_thread; | |
94 | static pthread_t dispatcher_thread; | |
95 | static pthread_t worker_thread; | |
65931c8b | 96 | static pthread_t health_thread; |
b8aa1682 | 97 | |
095a4ae5 MD |
98 | static uint64_t last_relay_stream_id; |
99 | static uint64_t last_relay_session_id; | |
b8aa1682 JD |
100 | |
101 | /* | |
102 | * Relay command queue. | |
103 | * | |
104 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
105 | * queue. | |
106 | */ | |
107 | static struct relay_cmd_queue relay_cmd_queue; | |
108 | ||
109 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
110 | static char *data_buffer; |
111 | static unsigned int data_buffer_size; | |
b8aa1682 | 112 | |
1c20f0e2 JD |
113 | /* We need those values for the file/dir creation. */ |
114 | static uid_t relayd_uid; | |
115 | static gid_t relayd_gid; | |
116 | ||
d3e2ba59 JD |
117 | /* Global relay stream hash table. */ |
118 | struct lttng_ht *relay_streams_ht; | |
119 | ||
92c6ca54 DG |
120 | /* Global relay viewer stream hash table. */ |
121 | struct lttng_ht *viewer_streams_ht; | |
122 | ||
0a6518b0 DG |
123 | /* Global hash table that stores relay index object. */ |
124 | struct lttng_ht *indexes_ht; | |
125 | ||
55706a7d | 126 | /* Relayd health monitoring */ |
eea7556c | 127 | struct health_app *health_relayd; |
55706a7d | 128 | |
cd60b05a JG |
129 | static struct option long_options[] = { |
130 | { "control-port", 1, 0, 'C', }, | |
131 | { "data-port", 1, 0, 'D', }, | |
132 | { "daemonize", 0, 0, 'd', }, | |
133 | { "group", 1, 0, 'g', }, | |
134 | { "help", 0, 0, 'h', }, | |
135 | { "output", 1, 0, 'o', }, | |
136 | { "verbose", 0, 0, 'v', }, | |
137 | { "config", 1, 0, 'f' }, | |
138 | { NULL, 0, 0, 0, }, | |
139 | }; | |
140 | ||
141 | static const char *config_ignore_options[] = { "help", "config" }; | |
142 | ||
b8aa1682 JD |
143 | /* |
144 | * usage function on stderr | |
145 | */ | |
146 | static | |
147 | void usage(void) | |
148 | { | |
149 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
150 | fprintf(stderr, " -h, --help Display this usage.\n"); |
151 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
152 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); | |
153 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
154 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); | |
155 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 156 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 157 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
158 | } |
159 | ||
cd60b05a JG |
160 | /* |
161 | * Take an option from the getopt output and set it in the right variable to be | |
162 | * used later. | |
163 | * | |
164 | * Return 0 on success else a negative value. | |
165 | */ | |
b8aa1682 | 166 | static |
cd60b05a | 167 | int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 168 | { |
cd60b05a JG |
169 | int ret; |
170 | ||
171 | switch (opt) { | |
172 | case 0: | |
173 | fprintf(stderr, "option %s", optname); | |
174 | if (arg) { | |
175 | fprintf(stderr, " with arg %s\n", arg); | |
176 | } | |
177 | break; | |
178 | case 'C': | |
179 | ret = uri_parse(arg, &control_uri); | |
180 | if (ret < 0) { | |
181 | ERR("Invalid control URI specified"); | |
182 | goto end; | |
183 | } | |
184 | if (control_uri->port == 0) { | |
185 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
186 | } | |
187 | break; | |
188 | case 'D': | |
189 | ret = uri_parse(arg, &data_uri); | |
190 | if (ret < 0) { | |
191 | ERR("Invalid data URI specified"); | |
192 | goto end; | |
193 | } | |
194 | if (data_uri->port == 0) { | |
195 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
196 | } | |
197 | break; | |
198 | case 'd': | |
199 | opt_daemon = 1; | |
200 | break; | |
201 | case 'g': | |
202 | tracing_group_name = strdup(arg); | |
203 | tracing_group_name_override = 1; | |
204 | break; | |
205 | case 'h': | |
206 | usage(); | |
207 | exit(EXIT_FAILURE); | |
208 | case 'o': | |
209 | ret = asprintf(&opt_output_path, "%s", arg); | |
210 | if (ret < 0) { | |
211 | ret = -errno; | |
212 | PERROR("asprintf opt_output_path"); | |
213 | goto end; | |
214 | } | |
215 | break; | |
216 | case 'v': | |
217 | /* Verbose level can increase using multiple -v */ | |
218 | if (arg) { | |
219 | lttng_opt_verbose = config_parse_value(arg); | |
220 | } else { | |
221 | lttng_opt_verbose += 1; | |
222 | } | |
223 | break; | |
224 | default: | |
225 | /* Unknown option or other error. | |
226 | * Error is printed by getopt, just return */ | |
227 | ret = -1; | |
228 | goto end; | |
229 | } | |
230 | ||
231 | /* All good. */ | |
232 | ret = 0; | |
233 | ||
234 | end: | |
235 | return ret; | |
236 | } | |
237 | ||
238 | /* | |
239 | * config_entry_handler_cb used to handle options read from a config file. | |
240 | * See config_entry_handler_cb comment in common/config/config.h for the | |
241 | * return value conventions. | |
242 | */ | |
243 | static | |
244 | int config_entry_handler(const struct config_entry *entry, void *unused) | |
245 | { | |
246 | int ret = 0, i; | |
247 | ||
248 | if (!entry || !entry->name || !entry->value) { | |
249 | ret = -EINVAL; | |
250 | goto end; | |
251 | } | |
252 | ||
253 | /* Check if the option is to be ignored */ | |
254 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
255 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
256 | goto end; | |
257 | } | |
258 | } | |
259 | ||
260 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
261 | /* Ignore if entry name is not fully matched. */ | |
262 | if (strcmp(entry->name, long_options[i].name)) { | |
263 | continue; | |
264 | } | |
265 | ||
266 | /* | |
267 | * If the option takes no argument on the command line, we have to | |
268 | * check if the value is "true". We support non-zero numeric values, | |
269 | * true, on and yes. | |
270 | */ | |
271 | if (!long_options[i].has_arg) { | |
272 | ret = config_parse_value(entry->value); | |
273 | if (ret <= 0) { | |
274 | if (ret) { | |
275 | WARN("Invalid configuration value \"%s\" for option %s", | |
276 | entry->value, entry->name); | |
277 | } | |
278 | /* False, skip boolean config option. */ | |
279 | goto end; | |
280 | } | |
281 | } | |
282 | ||
283 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
284 | goto end; | |
285 | } | |
286 | ||
287 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
288 | entry->name); | |
289 | ||
290 | end: | |
291 | return ret; | |
292 | } | |
293 | ||
294 | static | |
295 | int set_options(int argc, char **argv) | |
296 | { | |
297 | int c, ret = 0, option_index = 0; | |
298 | int orig_optopt = optopt, orig_optind = optind; | |
299 | char *default_address, *optstring; | |
300 | const char *config_path = NULL; | |
301 | ||
302 | optstring = utils_generate_optstring(long_options, | |
303 | sizeof(long_options) / sizeof(struct option)); | |
304 | if (!optstring) { | |
305 | ret = -ENOMEM; | |
306 | goto exit; | |
307 | } | |
308 | ||
309 | /* Check for the --config option */ | |
310 | ||
311 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
312 | &option_index)) != -1) { | |
313 | if (c == '?') { | |
314 | ret = -EINVAL; | |
315 | goto exit; | |
316 | } else if (c != 'f') { | |
317 | continue; | |
318 | } | |
319 | ||
320 | config_path = utils_expand_path(optarg); | |
321 | if (!config_path) { | |
322 | ERR("Failed to resolve path: %s", optarg); | |
323 | } | |
324 | } | |
325 | ||
326 | ret = config_get_section_entries(config_path, config_section_name, | |
327 | config_entry_handler, NULL); | |
328 | if (ret) { | |
329 | if (ret > 0) { | |
330 | ERR("Invalid configuration option at line %i", ret); | |
331 | ret = -1; | |
332 | } | |
333 | goto exit; | |
334 | } | |
b8aa1682 | 335 | |
cd60b05a JG |
336 | /* Reset getopt's global state */ |
337 | optopt = orig_optopt; | |
338 | optind = orig_optind; | |
b8aa1682 | 339 | while (1) { |
cd60b05a | 340 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
341 | if (c == -1) { |
342 | break; | |
343 | } | |
344 | ||
cd60b05a JG |
345 | ret = set_option(c, optarg, long_options[option_index].name); |
346 | if (ret < 0) { | |
b8aa1682 JD |
347 | goto exit; |
348 | } | |
349 | } | |
350 | ||
351 | /* assign default values */ | |
352 | if (control_uri == NULL) { | |
353 | ret = asprintf(&default_address, "tcp://0.0.0.0:%d", | |
354 | DEFAULT_NETWORK_CONTROL_PORT); | |
355 | if (ret < 0) { | |
356 | PERROR("asprintf default data address"); | |
357 | goto exit; | |
358 | } | |
359 | ||
360 | ret = uri_parse(default_address, &control_uri); | |
361 | free(default_address); | |
362 | if (ret < 0) { | |
363 | ERR("Invalid control URI specified"); | |
364 | goto exit; | |
365 | } | |
366 | } | |
367 | if (data_uri == NULL) { | |
368 | ret = asprintf(&default_address, "tcp://0.0.0.0:%d", | |
369 | DEFAULT_NETWORK_DATA_PORT); | |
370 | if (ret < 0) { | |
371 | PERROR("asprintf default data address"); | |
372 | goto exit; | |
373 | } | |
374 | ||
375 | ret = uri_parse(default_address, &data_uri); | |
376 | free(default_address); | |
377 | if (ret < 0) { | |
378 | ERR("Invalid data URI specified"); | |
379 | goto exit; | |
380 | } | |
381 | } | |
d3e2ba59 JD |
382 | if (live_uri == NULL) { |
383 | ret = asprintf(&default_address, "tcp://0.0.0.0:%d", | |
384 | DEFAULT_NETWORK_VIEWER_PORT); | |
385 | if (ret < 0) { | |
386 | PERROR("asprintf default viewer control address"); | |
387 | goto exit; | |
388 | } | |
389 | ||
390 | ret = uri_parse(default_address, &live_uri); | |
391 | free(default_address); | |
392 | if (ret < 0) { | |
393 | ERR("Invalid viewer control URI specified"); | |
394 | goto exit; | |
395 | } | |
396 | } | |
b8aa1682 JD |
397 | |
398 | exit: | |
cd60b05a | 399 | free(optstring); |
b8aa1682 JD |
400 | return ret; |
401 | } | |
402 | ||
403 | /* | |
404 | * Cleanup the daemon | |
405 | */ | |
406 | static | |
407 | void cleanup(void) | |
408 | { | |
b8aa1682 JD |
409 | DBG("Cleaning up"); |
410 | ||
095a4ae5 MD |
411 | /* free the dynamically allocated opt_output_path */ |
412 | free(opt_output_path); | |
413 | ||
a02de639 CB |
414 | /* Close thread quit pipes */ |
415 | utils_close_pipe(thread_quit_pipe); | |
416 | ||
710c1f73 DG |
417 | uri_free(control_uri); |
418 | uri_free(data_uri); | |
cd60b05a JG |
419 | |
420 | if (tracing_group_name_override) { | |
421 | free((void *) tracing_group_name); | |
422 | } | |
b8aa1682 JD |
423 | } |
424 | ||
425 | /* | |
426 | * Write to writable pipe used to notify a thread. | |
427 | */ | |
428 | static | |
429 | int notify_thread_pipe(int wpipe) | |
430 | { | |
6cd525e8 | 431 | ssize_t ret; |
b8aa1682 | 432 | |
6cd525e8 MD |
433 | ret = lttng_write(wpipe, "!", 1); |
434 | if (ret < 1) { | |
b8aa1682 JD |
435 | PERROR("write poll pipe"); |
436 | } | |
437 | ||
438 | return ret; | |
439 | } | |
440 | ||
65931c8b MD |
441 | static void notify_health_quit_pipe(int *pipe) |
442 | { | |
6cd525e8 | 443 | ssize_t ret; |
65931c8b | 444 | |
6cd525e8 MD |
445 | ret = lttng_write(pipe[1], "4", 1); |
446 | if (ret < 1) { | |
65931c8b MD |
447 | PERROR("write relay health quit"); |
448 | } | |
449 | } | |
450 | ||
b8aa1682 JD |
451 | /* |
452 | * Stop all threads by closing the thread quit pipe. | |
453 | */ | |
454 | static | |
455 | void stop_threads(void) | |
456 | { | |
457 | int ret; | |
458 | ||
459 | /* Stopping all threads */ | |
460 | DBG("Terminating all threads"); | |
461 | ret = notify_thread_pipe(thread_quit_pipe[1]); | |
462 | if (ret < 0) { | |
463 | ERR("write error on thread quit pipe"); | |
464 | } | |
465 | ||
65931c8b MD |
466 | notify_health_quit_pipe(health_quit_pipe); |
467 | ||
b8aa1682 | 468 | /* Dispatch thread */ |
26c9d55e | 469 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
b8aa1682 JD |
470 | futex_nto1_wake(&relay_cmd_queue.futex); |
471 | } | |
472 | ||
473 | /* | |
474 | * Signal handler for the daemon | |
475 | * | |
476 | * Simply stop all worker threads, leaving main() return gracefully after | |
477 | * joining all threads and calling cleanup(). | |
478 | */ | |
479 | static | |
480 | void sighandler(int sig) | |
481 | { | |
482 | switch (sig) { | |
483 | case SIGPIPE: | |
484 | DBG("SIGPIPE caught"); | |
485 | return; | |
486 | case SIGINT: | |
487 | DBG("SIGINT caught"); | |
488 | stop_threads(); | |
489 | break; | |
490 | case SIGTERM: | |
491 | DBG("SIGTERM caught"); | |
492 | stop_threads(); | |
493 | break; | |
494 | default: | |
495 | break; | |
496 | } | |
497 | } | |
498 | ||
499 | /* | |
500 | * Setup signal handler for : | |
501 | * SIGINT, SIGTERM, SIGPIPE | |
502 | */ | |
503 | static | |
504 | int set_signal_handler(void) | |
505 | { | |
506 | int ret = 0; | |
507 | struct sigaction sa; | |
508 | sigset_t sigset; | |
509 | ||
510 | if ((ret = sigemptyset(&sigset)) < 0) { | |
511 | PERROR("sigemptyset"); | |
512 | return ret; | |
513 | } | |
514 | ||
515 | sa.sa_handler = sighandler; | |
516 | sa.sa_mask = sigset; | |
517 | sa.sa_flags = 0; | |
518 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
519 | PERROR("sigaction"); | |
520 | return ret; | |
521 | } | |
522 | ||
523 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
524 | PERROR("sigaction"); | |
525 | return ret; | |
526 | } | |
527 | ||
528 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
529 | PERROR("sigaction"); | |
530 | return ret; | |
531 | } | |
532 | ||
533 | DBG("Signal handler set for SIGTERM, SIGPIPE and SIGINT"); | |
534 | ||
535 | return ret; | |
536 | } | |
537 | ||
538 | /* | |
539 | * Init thread quit pipe. | |
540 | * | |
541 | * Return -1 on error or 0 if all pipes are created. | |
542 | */ | |
543 | static | |
544 | int init_thread_quit_pipe(void) | |
545 | { | |
a02de639 | 546 | int ret; |
b8aa1682 | 547 | |
a02de639 | 548 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 549 | |
b8aa1682 JD |
550 | return ret; |
551 | } | |
552 | ||
553 | /* | |
554 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
555 | */ | |
556 | static | |
557 | int create_thread_poll_set(struct lttng_poll_event *events, int size) | |
558 | { | |
559 | int ret; | |
560 | ||
561 | if (events == NULL || size == 0) { | |
562 | ret = -1; | |
563 | goto error; | |
564 | } | |
565 | ||
566 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
567 | if (ret < 0) { | |
568 | goto error; | |
569 | } | |
570 | ||
571 | /* Add quit pipe */ | |
572 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN); | |
573 | if (ret < 0) { | |
574 | goto error; | |
575 | } | |
576 | ||
577 | return 0; | |
578 | ||
579 | error: | |
580 | return ret; | |
581 | } | |
582 | ||
583 | /* | |
584 | * Check if the thread quit pipe was triggered. | |
585 | * | |
586 | * Return 1 if it was triggered else 0; | |
587 | */ | |
588 | static | |
589 | int check_thread_quit_pipe(int fd, uint32_t events) | |
590 | { | |
591 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
592 | return 1; | |
593 | } | |
594 | ||
595 | return 0; | |
596 | } | |
597 | ||
598 | /* | |
599 | * Create and init socket from uri. | |
600 | */ | |
601 | static | |
602 | struct lttcomm_sock *relay_init_sock(struct lttng_uri *uri) | |
603 | { | |
604 | int ret; | |
605 | struct lttcomm_sock *sock = NULL; | |
606 | ||
607 | sock = lttcomm_alloc_sock_from_uri(uri); | |
608 | if (sock == NULL) { | |
609 | ERR("Allocating socket"); | |
610 | goto error; | |
611 | } | |
612 | ||
613 | ret = lttcomm_create_sock(sock); | |
614 | if (ret < 0) { | |
615 | goto error; | |
616 | } | |
617 | DBG("Listening on sock %d", sock->fd); | |
618 | ||
619 | ret = sock->ops->bind(sock); | |
620 | if (ret < 0) { | |
621 | goto error; | |
622 | } | |
623 | ||
624 | ret = sock->ops->listen(sock, -1); | |
625 | if (ret < 0) { | |
626 | goto error; | |
627 | ||
628 | } | |
629 | ||
630 | return sock; | |
631 | ||
632 | error: | |
633 | if (sock) { | |
634 | lttcomm_destroy_sock(sock); | |
635 | } | |
636 | return NULL; | |
637 | } | |
638 | ||
173af62f DG |
639 | /* |
640 | * Return nonzero if stream needs to be closed. | |
641 | */ | |
642 | static | |
643 | int close_stream_check(struct relay_stream *stream) | |
644 | { | |
645 | ||
646 | if (stream->close_flag && stream->prev_seq == stream->last_net_seq_num) { | |
f7079f67 DG |
647 | /* |
648 | * We are about to close the stream so set the data pending flag to 1 | |
649 | * which will make the end data pending command skip the stream which | |
650 | * is now closed and ready. Note that after proceeding to a file close, | |
651 | * the written file is ready for reading. | |
652 | */ | |
653 | stream->data_pending_check_done = 1; | |
173af62f DG |
654 | return 1; |
655 | } | |
656 | return 0; | |
657 | } | |
658 | ||
b8aa1682 JD |
659 | /* |
660 | * This thread manages the listening for new connections on the network | |
661 | */ | |
662 | static | |
663 | void *relay_thread_listener(void *data) | |
664 | { | |
095a4ae5 | 665 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
666 | int val = 1; |
667 | uint32_t revents, nb_fd; | |
668 | struct lttng_poll_event events; | |
669 | struct lttcomm_sock *control_sock, *data_sock; | |
670 | ||
b8aa1682 JD |
671 | DBG("[thread] Relay listener started"); |
672 | ||
55706a7d MD |
673 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
674 | ||
f385ae0a MD |
675 | health_code_update(); |
676 | ||
b8aa1682 JD |
677 | control_sock = relay_init_sock(control_uri); |
678 | if (!control_sock) { | |
095a4ae5 | 679 | goto error_sock_control; |
b8aa1682 JD |
680 | } |
681 | ||
682 | data_sock = relay_init_sock(data_uri); | |
683 | if (!data_sock) { | |
095a4ae5 | 684 | goto error_sock_relay; |
b8aa1682 JD |
685 | } |
686 | ||
687 | /* | |
688 | * Pass 3 as size here for the thread quit pipe, control and data socket. | |
689 | */ | |
690 | ret = create_thread_poll_set(&events, 3); | |
691 | if (ret < 0) { | |
692 | goto error_create_poll; | |
693 | } | |
694 | ||
695 | /* Add the control socket */ | |
696 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
697 | if (ret < 0) { | |
698 | goto error_poll_add; | |
699 | } | |
700 | ||
701 | /* Add the data socket */ | |
702 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
703 | if (ret < 0) { | |
704 | goto error_poll_add; | |
705 | } | |
706 | ||
707 | while (1) { | |
f385ae0a MD |
708 | health_code_update(); |
709 | ||
b8aa1682 JD |
710 | DBG("Listener accepting connections"); |
711 | ||
b8aa1682 | 712 | restart: |
f385ae0a | 713 | health_poll_entry(); |
b8aa1682 | 714 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 715 | health_poll_exit(); |
b8aa1682 JD |
716 | if (ret < 0) { |
717 | /* | |
718 | * Restart interrupted system call. | |
719 | */ | |
720 | if (errno == EINTR) { | |
721 | goto restart; | |
722 | } | |
723 | goto error; | |
724 | } | |
725 | ||
0d9c5d77 DG |
726 | nb_fd = ret; |
727 | ||
b8aa1682 JD |
728 | DBG("Relay new connection received"); |
729 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
730 | health_code_update(); |
731 | ||
b8aa1682 JD |
732 | /* Fetch once the poll data */ |
733 | revents = LTTNG_POLL_GETEV(&events, i); | |
734 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
735 | ||
736 | /* Thread quit pipe has been closed. Killing thread. */ | |
737 | ret = check_thread_quit_pipe(pollfd, revents); | |
738 | if (ret) { | |
095a4ae5 MD |
739 | err = 0; |
740 | goto exit; | |
b8aa1682 JD |
741 | } |
742 | ||
743 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
744 | ERR("socket poll error"); | |
745 | goto error; | |
746 | } else if (revents & LPOLLIN) { | |
4b7f17b2 MD |
747 | /* |
748 | * Get allocated in this thread, | |
749 | * enqueued to a global queue, dequeued | |
750 | * and freed in the worker thread. | |
751 | */ | |
752 | struct relay_command *relay_cmd; | |
753 | struct lttcomm_sock *newsock; | |
b8aa1682 JD |
754 | |
755 | relay_cmd = zmalloc(sizeof(struct relay_command)); | |
756 | if (relay_cmd == NULL) { | |
757 | PERROR("relay command zmalloc"); | |
758 | goto error; | |
759 | } | |
760 | ||
761 | if (pollfd == data_sock->fd) { | |
762 | newsock = data_sock->ops->accept(data_sock); | |
4b7f17b2 | 763 | if (!newsock) { |
b8aa1682 | 764 | PERROR("accepting data sock"); |
4b7f17b2 | 765 | free(relay_cmd); |
b8aa1682 JD |
766 | goto error; |
767 | } | |
768 | relay_cmd->type = RELAY_DATA; | |
769 | DBG("Relay data connection accepted, socket %d", newsock->fd); | |
4b7f17b2 MD |
770 | } else { |
771 | assert(pollfd == control_sock->fd); | |
b8aa1682 | 772 | newsock = control_sock->ops->accept(control_sock); |
4b7f17b2 | 773 | if (!newsock) { |
b8aa1682 | 774 | PERROR("accepting control sock"); |
4b7f17b2 | 775 | free(relay_cmd); |
b8aa1682 JD |
776 | goto error; |
777 | } | |
778 | relay_cmd->type = RELAY_CONTROL; | |
779 | DBG("Relay control connection accepted, socket %d", newsock->fd); | |
780 | } | |
781 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, | |
782 | &val, sizeof(int)); | |
783 | if (ret < 0) { | |
784 | PERROR("setsockopt inet"); | |
4b7f17b2 MD |
785 | lttcomm_destroy_sock(newsock); |
786 | free(relay_cmd); | |
b8aa1682 JD |
787 | goto error; |
788 | } | |
789 | relay_cmd->sock = newsock; | |
790 | /* | |
791 | * Lock free enqueue the request. | |
792 | */ | |
793 | cds_wfq_enqueue(&relay_cmd_queue.queue, &relay_cmd->node); | |
794 | ||
795 | /* | |
796 | * Wake the dispatch queue futex. Implicit memory | |
797 | * barrier with the exchange in cds_wfq_enqueue. | |
798 | */ | |
799 | futex_nto1_wake(&relay_cmd_queue.futex); | |
800 | } | |
801 | } | |
802 | } | |
803 | ||
095a4ae5 | 804 | exit: |
b8aa1682 JD |
805 | error: |
806 | error_poll_add: | |
807 | lttng_poll_clean(&events); | |
808 | error_create_poll: | |
095a4ae5 MD |
809 | if (data_sock->fd >= 0) { |
810 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
811 | if (ret) { |
812 | PERROR("close"); | |
813 | } | |
b8aa1682 | 814 | } |
095a4ae5 MD |
815 | lttcomm_destroy_sock(data_sock); |
816 | error_sock_relay: | |
817 | if (control_sock->fd >= 0) { | |
818 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
819 | if (ret) { |
820 | PERROR("close"); | |
821 | } | |
b8aa1682 | 822 | } |
095a4ae5 MD |
823 | lttcomm_destroy_sock(control_sock); |
824 | error_sock_control: | |
825 | if (err) { | |
f385ae0a MD |
826 | health_error(); |
827 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 828 | } |
55706a7d | 829 | health_unregister(health_relayd); |
b8aa1682 JD |
830 | DBG("Relay listener thread cleanup complete"); |
831 | stop_threads(); | |
b8aa1682 JD |
832 | return NULL; |
833 | } | |
834 | ||
835 | /* | |
836 | * This thread manages the dispatching of the requests to worker threads | |
837 | */ | |
838 | static | |
839 | void *relay_thread_dispatcher(void *data) | |
840 | { | |
6cd525e8 MD |
841 | int err = -1; |
842 | ssize_t ret; | |
b8aa1682 JD |
843 | struct cds_wfq_node *node; |
844 | struct relay_command *relay_cmd = NULL; | |
845 | ||
846 | DBG("[thread] Relay dispatcher started"); | |
847 | ||
55706a7d MD |
848 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
849 | ||
f385ae0a MD |
850 | health_code_update(); |
851 | ||
26c9d55e | 852 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
853 | health_code_update(); |
854 | ||
b8aa1682 JD |
855 | /* Atomically prepare the queue futex */ |
856 | futex_nto1_prepare(&relay_cmd_queue.futex); | |
857 | ||
858 | do { | |
f385ae0a MD |
859 | health_code_update(); |
860 | ||
b8aa1682 JD |
861 | /* Dequeue commands */ |
862 | node = cds_wfq_dequeue_blocking(&relay_cmd_queue.queue); | |
863 | if (node == NULL) { | |
864 | DBG("Woken up but nothing in the relay command queue"); | |
865 | /* Continue thread execution */ | |
866 | break; | |
867 | } | |
868 | ||
869 | relay_cmd = caa_container_of(node, struct relay_command, node); | |
870 | DBG("Dispatching request waiting on sock %d", relay_cmd->sock->fd); | |
871 | ||
872 | /* | |
873 | * Inform worker thread of the new request. This | |
874 | * call is blocking so we can be assured that the data will be read | |
875 | * at some point in time or wait to the end of the world :) | |
876 | */ | |
6cd525e8 MD |
877 | ret = lttng_write(relay_cmd_pipe[1], relay_cmd, |
878 | sizeof(struct relay_command)); | |
b8aa1682 | 879 | free(relay_cmd); |
6cd525e8 | 880 | if (ret < sizeof(struct relay_command)) { |
b8aa1682 JD |
881 | PERROR("write cmd pipe"); |
882 | goto error; | |
883 | } | |
884 | } while (node != NULL); | |
885 | ||
886 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 887 | health_poll_entry(); |
b8aa1682 | 888 | futex_nto1_wait(&relay_cmd_queue.futex); |
f385ae0a | 889 | health_poll_exit(); |
b8aa1682 JD |
890 | } |
891 | ||
f385ae0a MD |
892 | /* Normal exit, no error */ |
893 | err = 0; | |
894 | ||
b8aa1682 | 895 | error: |
f385ae0a MD |
896 | if (err) { |
897 | health_error(); | |
898 | ERR("Health error occurred in %s", __func__); | |
899 | } | |
55706a7d | 900 | health_unregister(health_relayd); |
b8aa1682 JD |
901 | DBG("Dispatch thread dying"); |
902 | stop_threads(); | |
903 | return NULL; | |
904 | } | |
905 | ||
de91f48a DG |
906 | /* |
907 | * Get stream from stream id. | |
908 | * Need to be called with RCU read-side lock held. | |
909 | */ | |
d3e2ba59 | 910 | struct relay_stream *relay_stream_find_by_id(uint64_t stream_id) |
de91f48a DG |
911 | { |
912 | struct lttng_ht_node_ulong *node; | |
913 | struct lttng_ht_iter iter; | |
914 | struct relay_stream *ret; | |
915 | ||
d3e2ba59 | 916 | lttng_ht_lookup(relay_streams_ht, |
de91f48a DG |
917 | (void *)((unsigned long) stream_id), |
918 | &iter); | |
919 | node = lttng_ht_iter_get_node_ulong(&iter); | |
920 | if (node == NULL) { | |
921 | DBG("Relay stream %" PRIu64 " not found", stream_id); | |
922 | ret = NULL; | |
923 | goto end; | |
924 | } | |
925 | ||
926 | ret = caa_container_of(node, struct relay_stream, stream_n); | |
927 | ||
928 | end: | |
929 | return ret; | |
930 | } | |
931 | ||
9d1bbf21 MD |
932 | static |
933 | void deferred_free_stream(struct rcu_head *head) | |
934 | { | |
935 | struct relay_stream *stream = | |
936 | caa_container_of(head, struct relay_stream, rcu_node); | |
d3e2ba59 | 937 | |
0f907de1 JD |
938 | free(stream->path_name); |
939 | free(stream->channel_name); | |
9d1bbf21 MD |
940 | free(stream); |
941 | } | |
942 | ||
d3e2ba59 JD |
943 | static |
944 | void deferred_free_session(struct rcu_head *head) | |
945 | { | |
946 | struct relay_session *session = | |
947 | caa_container_of(head, struct relay_session, rcu_node); | |
948 | free(session); | |
949 | } | |
950 | ||
4c80d9a2 DG |
951 | /* |
952 | * Close a given stream. The stream is freed using a call RCU. | |
953 | * | |
954 | * RCU read side lock MUST be acquired. If NO close_stream_check() was called | |
955 | * BEFORE the stream lock MUST be acquired. | |
956 | */ | |
c07d8a78 | 957 | static void destroy_stream(struct relay_stream *stream) |
94d49140 JD |
958 | { |
959 | int delret; | |
960 | struct relay_viewer_stream *vstream; | |
961 | struct lttng_ht_iter iter; | |
962 | ||
963 | assert(stream); | |
94d49140 JD |
964 | |
965 | delret = close(stream->fd); | |
966 | if (delret < 0) { | |
967 | PERROR("close stream"); | |
968 | } | |
969 | ||
970 | if (stream->index_fd >= 0) { | |
971 | delret = close(stream->index_fd); | |
972 | if (delret < 0) { | |
973 | PERROR("close stream index_fd"); | |
974 | } | |
975 | } | |
976 | ||
92c6ca54 | 977 | vstream = live_find_viewer_stream_by_id(stream->stream_handle); |
94d49140 JD |
978 | if (vstream) { |
979 | /* | |
980 | * Set the last good value into the viewer stream. This is done | |
981 | * right before the stream gets deleted from the hash table. The | |
982 | * lookup failure on the live thread side of a stream indicates | |
983 | * that the viewer stream index received value should be used. | |
984 | */ | |
cef0f7d5 | 985 | pthread_mutex_lock(&stream->viewer_stream_rotation_lock); |
94d49140 | 986 | vstream->total_index_received = stream->total_index_received; |
a020f610 | 987 | vstream->tracefile_count_last = stream->tracefile_count_current; |
cef0f7d5 JD |
988 | vstream->close_write_flag = 1; |
989 | pthread_mutex_unlock(&stream->viewer_stream_rotation_lock); | |
94d49140 JD |
990 | } |
991 | ||
efc8a87f DG |
992 | /* Cleanup index of that stream. */ |
993 | relay_index_destroy_by_stream_id(stream->stream_handle); | |
994 | ||
94d49140 JD |
995 | iter.iter.node = &stream->stream_n.node; |
996 | delret = lttng_ht_del(relay_streams_ht, &iter); | |
997 | assert(!delret); | |
998 | iter.iter.node = &stream->ctf_trace_node.node; | |
c07d8a78 | 999 | delret = lttng_ht_del(stream->ctf_traces_ht, &iter); |
94d49140 | 1000 | assert(!delret); |
157df586 JD |
1001 | |
1002 | if (stream->ctf_trace) { | |
1003 | ctf_trace_try_destroy(stream->ctf_trace); | |
1004 | } | |
1005 | ||
94d49140 JD |
1006 | call_rcu(&stream->rcu_node, deferred_free_stream); |
1007 | DBG("Closed tracefile %d from close stream", stream->fd); | |
1008 | } | |
1009 | ||
b8aa1682 JD |
1010 | /* |
1011 | * relay_delete_session: Free all memory associated with a session and | |
1012 | * close all the FDs | |
1013 | */ | |
1014 | static | |
d3e2ba59 JD |
1015 | void relay_delete_session(struct relay_command *cmd, |
1016 | struct lttng_ht *sessions_ht) | |
b8aa1682 JD |
1017 | { |
1018 | struct lttng_ht_iter iter; | |
1019 | struct lttng_ht_node_ulong *node; | |
1020 | struct relay_stream *stream; | |
1021 | int ret; | |
1022 | ||
095a4ae5 | 1023 | if (!cmd->session) { |
b8aa1682 | 1024 | return; |
095a4ae5 | 1025 | } |
b8aa1682 | 1026 | |
77c7c900 | 1027 | DBG("Relay deleting session %" PRIu64, cmd->session->id); |
5b6d8097 | 1028 | |
9d1bbf21 | 1029 | rcu_read_lock(); |
d3e2ba59 | 1030 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, node, node) { |
b8aa1682 | 1031 | node = lttng_ht_iter_get_node_ulong(&iter); |
4c80d9a2 DG |
1032 | if (!node) { |
1033 | continue; | |
b8aa1682 | 1034 | } |
4c80d9a2 DG |
1035 | stream = caa_container_of(node, struct relay_stream, stream_n); |
1036 | if (stream->session == cmd->session) { | |
c07d8a78 | 1037 | destroy_stream(stream); |
87b576ec JD |
1038 | cmd->session->stream_count--; |
1039 | assert(cmd->session->stream_count >= 0); | |
4c80d9a2 | 1040 | } |
b8aa1682 | 1041 | } |
4c80d9a2 DG |
1042 | |
1043 | /* Make this session not visible anymore. */ | |
d3e2ba59 JD |
1044 | iter.iter.node = &cmd->session->session_n.node; |
1045 | ret = lttng_ht_del(sessions_ht, &iter); | |
1046 | assert(!ret); | |
4c80d9a2 | 1047 | call_rcu(&cmd->session->rcu_node, deferred_free_session); |
9d1bbf21 | 1048 | rcu_read_unlock(); |
b8aa1682 JD |
1049 | } |
1050 | ||
1c20f0e2 JD |
1051 | /* |
1052 | * Copy index data from the control port to a given index object. | |
1053 | */ | |
1054 | static void copy_index_control_data(struct relay_index *index, | |
1055 | struct lttcomm_relayd_index *data) | |
1056 | { | |
1057 | assert(index); | |
1058 | assert(data); | |
1059 | ||
1060 | /* | |
1061 | * The index on disk is encoded in big endian, so we don't need to convert | |
1062 | * the data received on the network. The data_offset value is NEVER | |
1063 | * modified here and is updated by the data thread. | |
1064 | */ | |
1065 | index->index_data.packet_size = data->packet_size; | |
1066 | index->index_data.content_size = data->content_size; | |
1067 | index->index_data.timestamp_begin = data->timestamp_begin; | |
1068 | index->index_data.timestamp_end = data->timestamp_end; | |
1069 | index->index_data.events_discarded = data->events_discarded; | |
1070 | index->index_data.stream_id = data->stream_id; | |
1071 | } | |
1072 | ||
c5b6f4f0 DG |
1073 | /* |
1074 | * Handle the RELAYD_CREATE_SESSION command. | |
1075 | * | |
1076 | * On success, send back the session id or else return a negative value. | |
1077 | */ | |
1078 | static | |
1079 | int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 JD |
1080 | struct relay_command *cmd, |
1081 | struct lttng_ht *sessions_ht) | |
c5b6f4f0 DG |
1082 | { |
1083 | int ret = 0, send_ret; | |
1084 | struct relay_session *session; | |
1085 | struct lttcomm_relayd_status_session reply; | |
1086 | ||
1087 | assert(recv_hdr); | |
1088 | assert(cmd); | |
1089 | ||
1090 | memset(&reply, 0, sizeof(reply)); | |
1091 | ||
1092 | session = zmalloc(sizeof(struct relay_session)); | |
1093 | if (session == NULL) { | |
1094 | PERROR("relay session zmalloc"); | |
1095 | ret = -1; | |
1096 | goto error; | |
1097 | } | |
1098 | ||
1099 | session->id = ++last_relay_session_id; | |
1100 | session->sock = cmd->sock; | |
7d2f7452 DG |
1101 | session->minor = cmd->minor; |
1102 | session->major = cmd->major; | |
c5b6f4f0 DG |
1103 | cmd->session = session; |
1104 | ||
1105 | reply.session_id = htobe64(session->id); | |
1106 | ||
d3e2ba59 JD |
1107 | switch (cmd->minor) { |
1108 | case 4: /* LTTng sessiond 2.4 */ | |
1109 | default: | |
1110 | ret = cmd_create_session_2_4(cmd, session); | |
1111 | break; | |
1112 | } | |
1113 | ||
1114 | lttng_ht_node_init_ulong(&session->session_n, | |
1115 | (unsigned long) session->id); | |
1116 | lttng_ht_add_unique_ulong(sessions_ht, | |
1117 | &session->session_n); | |
1118 | ||
c5b6f4f0 DG |
1119 | DBG("Created session %" PRIu64, session->id); |
1120 | ||
1121 | error: | |
1122 | if (ret < 0) { | |
1123 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1124 | } else { | |
1125 | reply.ret_code = htobe32(LTTNG_OK); | |
1126 | } | |
1127 | ||
1128 | send_ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, sizeof(reply), 0); | |
1129 | if (send_ret < 0) { | |
1130 | ERR("Relayd sending session id"); | |
4169f5ad | 1131 | ret = send_ret; |
c5b6f4f0 DG |
1132 | } |
1133 | ||
1134 | return ret; | |
1135 | } | |
1136 | ||
b8aa1682 JD |
1137 | /* |
1138 | * relay_add_stream: allocate a new stream for a session | |
1139 | */ | |
1140 | static | |
1141 | int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1142 | struct relay_command *cmd, struct lttng_ht *sessions_ht) |
b8aa1682 JD |
1143 | { |
1144 | struct relay_session *session = cmd->session; | |
b8aa1682 JD |
1145 | struct relay_stream *stream = NULL; |
1146 | struct lttcomm_relayd_status_stream reply; | |
b8aa1682 JD |
1147 | int ret, send_ret; |
1148 | ||
c5b6f4f0 | 1149 | if (!session || cmd->version_check_done == 0) { |
b8aa1682 JD |
1150 | ERR("Trying to add a stream before version check"); |
1151 | ret = -1; | |
1152 | goto end_no_session; | |
1153 | } | |
1154 | ||
b8aa1682 JD |
1155 | stream = zmalloc(sizeof(struct relay_stream)); |
1156 | if (stream == NULL) { | |
1157 | PERROR("relay stream zmalloc"); | |
1158 | ret = -1; | |
1159 | goto end_no_session; | |
1160 | } | |
1161 | ||
0f907de1 JD |
1162 | switch (cmd->minor) { |
1163 | case 1: /* LTTng sessiond 2.1 */ | |
1164 | ret = cmd_recv_stream_2_1(cmd, stream); | |
1165 | break; | |
1166 | case 2: /* LTTng sessiond 2.2 */ | |
1167 | default: | |
1168 | ret = cmd_recv_stream_2_2(cmd, stream); | |
1169 | break; | |
1170 | } | |
1171 | if (ret < 0) { | |
1172 | goto err_free_stream; | |
1173 | } | |
1174 | ||
9d1bbf21 | 1175 | rcu_read_lock(); |
b8aa1682 | 1176 | stream->stream_handle = ++last_relay_stream_id; |
173af62f | 1177 | stream->prev_seq = -1ULL; |
b8aa1682 | 1178 | stream->session = session; |
1c20f0e2 | 1179 | stream->index_fd = -1; |
d3e2ba59 JD |
1180 | stream->read_index_fd = -1; |
1181 | stream->ctf_trace = NULL; | |
1182 | pthread_mutex_init(&stream->lock, NULL); | |
b8aa1682 | 1183 | |
0f907de1 | 1184 | ret = utils_mkdir_recursive(stream->path_name, S_IRWXU | S_IRWXG); |
b8aa1682 | 1185 | if (ret < 0) { |
b8aa1682 JD |
1186 | ERR("relay creating output directory"); |
1187 | goto end; | |
1188 | } | |
1189 | ||
be96a7d1 DG |
1190 | /* |
1191 | * No need to use run_as API here because whatever we receives, the relayd | |
1192 | * uses its own credentials for the stream files. | |
1193 | */ | |
0f907de1 | 1194 | ret = utils_create_stream_file(stream->path_name, stream->channel_name, |
1c20f0e2 | 1195 | stream->tracefile_size, 0, relayd_uid, relayd_gid, NULL); |
b8aa1682 | 1196 | if (ret < 0) { |
0f907de1 | 1197 | ERR("Create output file"); |
b8aa1682 JD |
1198 | goto end; |
1199 | } | |
b8aa1682 | 1200 | stream->fd = ret; |
0f907de1 JD |
1201 | if (stream->tracefile_size) { |
1202 | DBG("Tracefile %s/%s_0 created", stream->path_name, stream->channel_name); | |
1203 | } else { | |
1204 | DBG("Tracefile %s/%s created", stream->path_name, stream->channel_name); | |
1205 | } | |
b8aa1682 | 1206 | |
d3e2ba59 JD |
1207 | if (!strncmp(stream->channel_name, DEFAULT_METADATA_NAME, NAME_MAX)) { |
1208 | stream->metadata_flag = 1; | |
1209 | /* | |
1210 | * When we receive a new metadata stream, we create a new | |
1211 | * ctf_trace and we assign this ctf_trace to all streams with | |
1212 | * the same path. | |
1213 | * | |
1214 | * If later on we receive a new stream for the same ctf_trace, | |
1215 | * we copy the information from the first hit in the HT to the | |
1216 | * new stream. | |
1217 | */ | |
1218 | stream->ctf_trace = ctf_trace_create(); | |
1219 | if (!stream->ctf_trace) { | |
1220 | ret = -1; | |
1221 | goto end; | |
1222 | } | |
1223 | stream->ctf_trace->refcount++; | |
1224 | stream->ctf_trace->metadata_stream = stream; | |
1225 | } | |
1226 | ctf_trace_assign(cmd->ctf_traces_ht, stream); | |
c07d8a78 | 1227 | stream->ctf_traces_ht = cmd->ctf_traces_ht; |
d3e2ba59 | 1228 | |
b8aa1682 JD |
1229 | lttng_ht_node_init_ulong(&stream->stream_n, |
1230 | (unsigned long) stream->stream_handle); | |
d3e2ba59 | 1231 | lttng_ht_add_unique_ulong(relay_streams_ht, |
b8aa1682 JD |
1232 | &stream->stream_n); |
1233 | ||
d3e2ba59 JD |
1234 | lttng_ht_node_init_str(&stream->ctf_trace_node, stream->path_name); |
1235 | lttng_ht_add_str(cmd->ctf_traces_ht, &stream->ctf_trace_node); | |
87b576ec | 1236 | session->stream_count++; |
d3e2ba59 | 1237 | |
1c20f0e2 JD |
1238 | DBG("Relay new stream added %s with ID %" PRIu64, stream->channel_name, |
1239 | stream->stream_handle); | |
b8aa1682 JD |
1240 | |
1241 | end: | |
5af40280 | 1242 | reply.handle = htobe64(stream->stream_handle); |
b8aa1682 JD |
1243 | /* send the session id to the client or a negative return code on error */ |
1244 | if (ret < 0) { | |
f73fabfd | 1245 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
5af40280 CB |
1246 | /* stream was not properly added to the ht, so free it */ |
1247 | free(stream); | |
b8aa1682 | 1248 | } else { |
f73fabfd | 1249 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1250 | } |
5af40280 | 1251 | |
b8aa1682 JD |
1252 | send_ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, |
1253 | sizeof(struct lttcomm_relayd_status_stream), 0); | |
1254 | if (send_ret < 0) { | |
1255 | ERR("Relay sending stream id"); | |
4169f5ad | 1256 | ret = send_ret; |
b8aa1682 | 1257 | } |
9d1bbf21 | 1258 | rcu_read_unlock(); |
b8aa1682 JD |
1259 | |
1260 | end_no_session: | |
1261 | return ret; | |
0f907de1 JD |
1262 | |
1263 | err_free_stream: | |
1264 | free(stream->path_name); | |
1265 | free(stream->channel_name); | |
1266 | free(stream); | |
1267 | return ret; | |
b8aa1682 JD |
1268 | } |
1269 | ||
173af62f DG |
1270 | /* |
1271 | * relay_close_stream: close a specific stream | |
1272 | */ | |
1273 | static | |
1274 | int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, | |
92c6ca54 | 1275 | struct relay_command *cmd) |
173af62f | 1276 | { |
94d49140 | 1277 | int ret, send_ret; |
173af62f DG |
1278 | struct relay_session *session = cmd->session; |
1279 | struct lttcomm_relayd_close_stream stream_info; | |
1280 | struct lttcomm_relayd_generic_reply reply; | |
1281 | struct relay_stream *stream; | |
173af62f DG |
1282 | |
1283 | DBG("Close stream received"); | |
1284 | ||
c5b6f4f0 | 1285 | if (!session || cmd->version_check_done == 0) { |
173af62f DG |
1286 | ERR("Trying to close a stream before version check"); |
1287 | ret = -1; | |
1288 | goto end_no_session; | |
1289 | } | |
1290 | ||
1291 | ret = cmd->sock->ops->recvmsg(cmd->sock, &stream_info, | |
7c5aef62 | 1292 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1293 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1294 | if (ret == 0) { |
1295 | /* Orderly shutdown. Not necessary to print an error. */ | |
1296 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1297 | } else { | |
1298 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1299 | } | |
173af62f DG |
1300 | ret = -1; |
1301 | goto end_no_session; | |
1302 | } | |
1303 | ||
1304 | rcu_read_lock(); | |
d3e2ba59 | 1305 | stream = relay_stream_find_by_id(be64toh(stream_info.stream_id)); |
173af62f DG |
1306 | if (!stream) { |
1307 | ret = -1; | |
1308 | goto end_unlock; | |
1309 | } | |
1310 | ||
8e2583a4 | 1311 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
173af62f | 1312 | stream->close_flag = 1; |
87b576ec JD |
1313 | session->stream_count--; |
1314 | assert(session->stream_count >= 0); | |
173af62f DG |
1315 | |
1316 | if (close_stream_check(stream)) { | |
c07d8a78 | 1317 | destroy_stream(stream); |
173af62f DG |
1318 | } |
1319 | ||
1320 | end_unlock: | |
1321 | rcu_read_unlock(); | |
1322 | ||
1323 | if (ret < 0) { | |
f73fabfd | 1324 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1325 | } else { |
f73fabfd | 1326 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f DG |
1327 | } |
1328 | send_ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, | |
1329 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
1330 | if (send_ret < 0) { | |
1331 | ERR("Relay sending stream id"); | |
4169f5ad | 1332 | ret = send_ret; |
173af62f DG |
1333 | } |
1334 | ||
1335 | end_no_session: | |
1336 | return ret; | |
1337 | } | |
1338 | ||
b8aa1682 JD |
1339 | /* |
1340 | * relay_unknown_command: send -1 if received unknown command | |
1341 | */ | |
1342 | static | |
1343 | void relay_unknown_command(struct relay_command *cmd) | |
1344 | { | |
1345 | struct lttcomm_relayd_generic_reply reply; | |
1346 | int ret; | |
1347 | ||
f73fabfd | 1348 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1349 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, |
1350 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
1351 | if (ret < 0) { | |
1352 | ERR("Relay sending unknown command"); | |
1353 | } | |
1354 | } | |
1355 | ||
1356 | /* | |
1357 | * relay_start: send an acknowledgment to the client to tell if we are | |
1358 | * ready to receive data. We are ready if a session is established. | |
1359 | */ | |
1360 | static | |
1361 | int relay_start(struct lttcomm_relayd_hdr *recv_hdr, | |
1362 | struct relay_command *cmd) | |
1363 | { | |
f73fabfd | 1364 | int ret = htobe32(LTTNG_OK); |
b8aa1682 JD |
1365 | struct lttcomm_relayd_generic_reply reply; |
1366 | struct relay_session *session = cmd->session; | |
1367 | ||
1368 | if (!session) { | |
1369 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1370 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1371 | } |
1372 | ||
1373 | reply.ret_code = ret; | |
1374 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, | |
1375 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
1376 | if (ret < 0) { | |
1377 | ERR("Relay sending start ack"); | |
1378 | } | |
1379 | ||
1380 | return ret; | |
1381 | } | |
1382 | ||
1d4dfdef DG |
1383 | /* |
1384 | * Append padding to the file pointed by the file descriptor fd. | |
1385 | */ | |
1386 | static int write_padding_to_file(int fd, uint32_t size) | |
1387 | { | |
6cd525e8 | 1388 | ssize_t ret = 0; |
1d4dfdef DG |
1389 | char *zeros; |
1390 | ||
1391 | if (size == 0) { | |
1392 | goto end; | |
1393 | } | |
1394 | ||
1395 | zeros = zmalloc(size); | |
1396 | if (zeros == NULL) { | |
1397 | PERROR("zmalloc zeros for padding"); | |
1398 | ret = -1; | |
1399 | goto end; | |
1400 | } | |
1401 | ||
6cd525e8 MD |
1402 | ret = lttng_write(fd, zeros, size); |
1403 | if (ret < size) { | |
1d4dfdef DG |
1404 | PERROR("write padding to file"); |
1405 | } | |
1406 | ||
e986c7a1 DG |
1407 | free(zeros); |
1408 | ||
1d4dfdef DG |
1409 | end: |
1410 | return ret; | |
1411 | } | |
1412 | ||
b8aa1682 JD |
1413 | /* |
1414 | * relay_recv_metadata: receive the metada for the session. | |
1415 | */ | |
1416 | static | |
1417 | int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1418 | struct relay_command *cmd) |
b8aa1682 | 1419 | { |
f73fabfd | 1420 | int ret = htobe32(LTTNG_OK); |
6cd525e8 | 1421 | ssize_t size_ret; |
b8aa1682 JD |
1422 | struct relay_session *session = cmd->session; |
1423 | struct lttcomm_relayd_metadata_payload *metadata_struct; | |
1424 | struct relay_stream *metadata_stream; | |
1425 | uint64_t data_size, payload_size; | |
1426 | ||
1427 | if (!session) { | |
1428 | ERR("Metadata sent before version check"); | |
1429 | ret = -1; | |
1430 | goto end; | |
1431 | } | |
1432 | ||
f6416125 MD |
1433 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1434 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1435 | ERR("Incorrect data size"); | |
1436 | ret = -1; | |
1437 | goto end; | |
1438 | } | |
1439 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1440 | ||
b8aa1682 | 1441 | if (data_buffer_size < data_size) { |
d7b3776f | 1442 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1443 | char *tmp_data_ptr; |
1444 | ||
1445 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1446 | if (!tmp_data_ptr) { | |
b8aa1682 | 1447 | ERR("Allocating data buffer"); |
c617c0c6 | 1448 | free(data_buffer); |
b8aa1682 JD |
1449 | ret = -1; |
1450 | goto end; | |
1451 | } | |
c617c0c6 | 1452 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1453 | data_buffer_size = data_size; |
1454 | } | |
1455 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1456 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
7c5aef62 | 1457 | ret = cmd->sock->ops->recvmsg(cmd->sock, data_buffer, data_size, 0); |
b8aa1682 | 1458 | if (ret < 0 || ret != data_size) { |
a6cd2b97 DG |
1459 | if (ret == 0) { |
1460 | /* Orderly shutdown. Not necessary to print an error. */ | |
1461 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1462 | } else { | |
1463 | ERR("Relay didn't receive the whole metadata"); | |
1464 | } | |
b8aa1682 | 1465 | ret = -1; |
b8aa1682 JD |
1466 | goto end; |
1467 | } | |
1468 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 MD |
1469 | |
1470 | rcu_read_lock(); | |
d3e2ba59 JD |
1471 | metadata_stream = relay_stream_find_by_id( |
1472 | be64toh(metadata_struct->stream_id)); | |
b8aa1682 JD |
1473 | if (!metadata_stream) { |
1474 | ret = -1; | |
9d1bbf21 | 1475 | goto end_unlock; |
b8aa1682 JD |
1476 | } |
1477 | ||
6cd525e8 MD |
1478 | size_ret = lttng_write(metadata_stream->fd, metadata_struct->payload, |
1479 | payload_size); | |
1480 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1481 | ERR("Relay error writing metadata on file"); |
1482 | ret = -1; | |
9d1bbf21 | 1483 | goto end_unlock; |
b8aa1682 | 1484 | } |
1d4dfdef DG |
1485 | |
1486 | ret = write_padding_to_file(metadata_stream->fd, | |
1487 | be32toh(metadata_struct->padding_size)); | |
1488 | if (ret < 0) { | |
1489 | goto end_unlock; | |
1490 | } | |
d3e2ba59 JD |
1491 | metadata_stream->ctf_trace->metadata_received += |
1492 | payload_size + be32toh(metadata_struct->padding_size); | |
1d4dfdef | 1493 | |
b8aa1682 JD |
1494 | DBG2("Relay metadata written"); |
1495 | ||
9d1bbf21 | 1496 | end_unlock: |
6e3c5836 | 1497 | rcu_read_unlock(); |
b8aa1682 JD |
1498 | end: |
1499 | return ret; | |
1500 | } | |
1501 | ||
1502 | /* | |
1503 | * relay_send_version: send relayd version number | |
1504 | */ | |
1505 | static | |
1506 | int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1507 | struct relay_command *cmd, struct lttng_ht *sessions_ht) |
b8aa1682 | 1508 | { |
7f51dcba | 1509 | int ret; |
092b6259 | 1510 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1511 | |
c5b6f4f0 DG |
1512 | assert(cmd); |
1513 | ||
1514 | cmd->version_check_done = 1; | |
b8aa1682 | 1515 | |
092b6259 | 1516 | /* Get version from the other side. */ |
7c5aef62 | 1517 | ret = cmd->sock->ops->recvmsg(cmd->sock, &msg, sizeof(msg), 0); |
092b6259 | 1518 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1519 | if (ret == 0) { |
1520 | /* Orderly shutdown. Not necessary to print an error. */ | |
1521 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1522 | } else { | |
1523 | ERR("Relay failed to receive the version values."); | |
1524 | } | |
092b6259 | 1525 | ret = -1; |
092b6259 DG |
1526 | goto end; |
1527 | } | |
1528 | ||
d83a952c MD |
1529 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1530 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1531 | |
1532 | /* Major versions must be the same */ | |
1533 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1534 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1535 | reply.major, be32toh(msg.major)); | |
d3e2ba59 | 1536 | relay_delete_session(cmd, sessions_ht); |
d4519fa3 JD |
1537 | ret = 0; |
1538 | goto end; | |
1539 | } | |
1540 | ||
0f907de1 JD |
1541 | cmd->major = reply.major; |
1542 | /* We adapt to the lowest compatible version */ | |
1543 | if (reply.minor <= be32toh(msg.minor)) { | |
1544 | cmd->minor = reply.minor; | |
1545 | } else { | |
1546 | cmd->minor = be32toh(msg.minor); | |
1547 | } | |
1548 | ||
6151a90f JD |
1549 | reply.major = htobe32(reply.major); |
1550 | reply.minor = htobe32(reply.minor); | |
1551 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, | |
1552 | sizeof(struct lttcomm_relayd_version), 0); | |
1553 | if (ret < 0) { | |
1554 | ERR("Relay sending version"); | |
1555 | } | |
1556 | ||
0f907de1 JD |
1557 | DBG("Version check done using protocol %u.%u", cmd->major, |
1558 | cmd->minor); | |
b8aa1682 JD |
1559 | |
1560 | end: | |
1561 | return ret; | |
1562 | } | |
1563 | ||
c8f59ee5 | 1564 | /* |
6d805429 | 1565 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 DG |
1566 | */ |
1567 | static | |
6d805429 | 1568 | int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
d3e2ba59 | 1569 | struct relay_command *cmd) |
c8f59ee5 DG |
1570 | { |
1571 | struct relay_session *session = cmd->session; | |
6d805429 | 1572 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1573 | struct lttcomm_relayd_generic_reply reply; |
1574 | struct relay_stream *stream; | |
1575 | int ret; | |
c8f59ee5 DG |
1576 | uint64_t last_net_seq_num, stream_id; |
1577 | ||
6d805429 | 1578 | DBG("Data pending command received"); |
c8f59ee5 | 1579 | |
c5b6f4f0 | 1580 | if (!session || cmd->version_check_done == 0) { |
c8f59ee5 DG |
1581 | ERR("Trying to check for data before version check"); |
1582 | ret = -1; | |
1583 | goto end_no_session; | |
1584 | } | |
1585 | ||
7c5aef62 | 1586 | ret = cmd->sock->ops->recvmsg(cmd->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1587 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1588 | if (ret == 0) { |
1589 | /* Orderly shutdown. Not necessary to print an error. */ | |
1590 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1591 | } else { | |
1592 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1593 | ret); | |
1594 | } | |
c8f59ee5 DG |
1595 | ret = -1; |
1596 | goto end_no_session; | |
1597 | } | |
1598 | ||
1599 | stream_id = be64toh(msg.stream_id); | |
1600 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1601 | ||
1602 | rcu_read_lock(); | |
d3e2ba59 | 1603 | stream = relay_stream_find_by_id(stream_id); |
de91f48a | 1604 | if (stream == NULL) { |
c8f59ee5 DG |
1605 | ret = -1; |
1606 | goto end_unlock; | |
1607 | } | |
1608 | ||
6d805429 | 1609 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1610 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1611 | last_net_seq_num); | |
1612 | ||
33832e64 | 1613 | /* Avoid wrapping issue */ |
39df6d9f | 1614 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1615 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1616 | ret = 0; |
6d805429 DG |
1617 | } else { |
1618 | /* Data still being streamed thus pending */ | |
1619 | ret = 1; | |
c8f59ee5 DG |
1620 | } |
1621 | ||
f7079f67 DG |
1622 | /* Pending check is now done. */ |
1623 | stream->data_pending_check_done = 1; | |
1624 | ||
c8f59ee5 DG |
1625 | end_unlock: |
1626 | rcu_read_unlock(); | |
1627 | ||
1628 | reply.ret_code = htobe32(ret); | |
1629 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, sizeof(reply), 0); | |
1630 | if (ret < 0) { | |
6d805429 | 1631 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1632 | } |
1633 | ||
1634 | end_no_session: | |
1635 | return ret; | |
1636 | } | |
1637 | ||
1638 | /* | |
1639 | * Wait for the control socket to reach a quiescent state. | |
1640 | * | |
1641 | * Note that for now, when receiving this command from the session daemon, this | |
1642 | * means that every subsequent commands or data received on the control socket | |
1643 | * has been handled. So, this is why we simply return OK here. | |
1644 | */ | |
1645 | static | |
1646 | int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1647 | struct relay_command *cmd) |
c8f59ee5 DG |
1648 | { |
1649 | int ret; | |
ad7051c0 DG |
1650 | uint64_t stream_id; |
1651 | struct relay_stream *stream; | |
1652 | struct lttng_ht_iter iter; | |
1653 | struct lttcomm_relayd_quiescent_control msg; | |
c8f59ee5 DG |
1654 | struct lttcomm_relayd_generic_reply reply; |
1655 | ||
1656 | DBG("Checking quiescent state on control socket"); | |
1657 | ||
ad7051c0 DG |
1658 | if (!cmd->session || cmd->version_check_done == 0) { |
1659 | ERR("Trying to check for data before version check"); | |
1660 | ret = -1; | |
1661 | goto end_no_session; | |
1662 | } | |
1663 | ||
1664 | ret = cmd->sock->ops->recvmsg(cmd->sock, &msg, sizeof(msg), 0); | |
1665 | if (ret < sizeof(msg)) { | |
a6cd2b97 DG |
1666 | if (ret == 0) { |
1667 | /* Orderly shutdown. Not necessary to print an error. */ | |
1668 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1669 | } else { | |
1670 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1671 | ret); | |
1672 | } | |
ad7051c0 DG |
1673 | ret = -1; |
1674 | goto end_no_session; | |
1675 | } | |
1676 | ||
1677 | stream_id = be64toh(msg.stream_id); | |
1678 | ||
1679 | rcu_read_lock(); | |
d3e2ba59 JD |
1680 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
1681 | stream_n.node) { | |
ad7051c0 DG |
1682 | if (stream->stream_handle == stream_id) { |
1683 | stream->data_pending_check_done = 1; | |
1684 | DBG("Relay quiescent control pending flag set to %" PRIu64, | |
1685 | stream_id); | |
1686 | break; | |
1687 | } | |
1688 | } | |
1689 | rcu_read_unlock(); | |
1690 | ||
c8f59ee5 DG |
1691 | reply.ret_code = htobe32(LTTNG_OK); |
1692 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, sizeof(reply), 0); | |
1693 | if (ret < 0) { | |
6d805429 | 1694 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1695 | } |
1696 | ||
ad7051c0 | 1697 | end_no_session: |
c8f59ee5 DG |
1698 | return ret; |
1699 | } | |
1700 | ||
f7079f67 DG |
1701 | /* |
1702 | * Initialize a data pending command. This means that a client is about to ask | |
1703 | * for data pending for each stream he/she holds. Simply iterate over all | |
1704 | * streams of a session and set the data_pending_check_done flag. | |
1705 | * | |
1706 | * This command returns to the client a LTTNG_OK code. | |
1707 | */ | |
1708 | static | |
1709 | int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1710 | struct relay_command *cmd) |
f7079f67 DG |
1711 | { |
1712 | int ret; | |
1713 | struct lttng_ht_iter iter; | |
1714 | struct lttcomm_relayd_begin_data_pending msg; | |
1715 | struct lttcomm_relayd_generic_reply reply; | |
1716 | struct relay_stream *stream; | |
1717 | uint64_t session_id; | |
1718 | ||
1719 | assert(recv_hdr); | |
1720 | assert(cmd); | |
f7079f67 DG |
1721 | |
1722 | DBG("Init streams for data pending"); | |
1723 | ||
1724 | if (!cmd->session || cmd->version_check_done == 0) { | |
1725 | ERR("Trying to check for data before version check"); | |
1726 | ret = -1; | |
1727 | goto end_no_session; | |
1728 | } | |
1729 | ||
1730 | ret = cmd->sock->ops->recvmsg(cmd->sock, &msg, sizeof(msg), 0); | |
1731 | if (ret < sizeof(msg)) { | |
a6cd2b97 DG |
1732 | if (ret == 0) { |
1733 | /* Orderly shutdown. Not necessary to print an error. */ | |
1734 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1735 | } else { | |
1736 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1737 | ret); | |
1738 | } | |
f7079f67 DG |
1739 | ret = -1; |
1740 | goto end_no_session; | |
1741 | } | |
1742 | ||
1743 | session_id = be64toh(msg.session_id); | |
1744 | ||
1745 | /* | |
1746 | * Iterate over all streams to set the begin data pending flag. For now, the | |
1747 | * streams are indexed by stream handle so we have to iterate over all | |
1748 | * streams to find the one associated with the right session_id. | |
1749 | */ | |
1750 | rcu_read_lock(); | |
d3e2ba59 JD |
1751 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
1752 | stream_n.node) { | |
f7079f67 DG |
1753 | if (stream->session->id == session_id) { |
1754 | stream->data_pending_check_done = 0; | |
1755 | DBG("Set begin data pending flag to stream %" PRIu64, | |
1756 | stream->stream_handle); | |
1757 | } | |
1758 | } | |
1759 | rcu_read_unlock(); | |
1760 | ||
1761 | /* All good, send back reply. */ | |
1762 | reply.ret_code = htobe32(LTTNG_OK); | |
1763 | ||
1764 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, sizeof(reply), 0); | |
1765 | if (ret < 0) { | |
1766 | ERR("Relay begin data pending send reply failed"); | |
1767 | } | |
1768 | ||
1769 | end_no_session: | |
1770 | return ret; | |
1771 | } | |
1772 | ||
1773 | /* | |
1774 | * End data pending command. This will check, for a given session id, if each | |
1775 | * stream associated with it has its data_pending_check_done flag set. If not, | |
1776 | * this means that the client lost track of the stream but the data is still | |
1777 | * being streamed on our side. In this case, we inform the client that data is | |
1778 | * inflight. | |
1779 | * | |
1780 | * Return to the client if there is data in flight or not with a ret_code. | |
1781 | */ | |
1782 | static | |
1783 | int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1784 | struct relay_command *cmd) |
f7079f67 DG |
1785 | { |
1786 | int ret; | |
1787 | struct lttng_ht_iter iter; | |
1788 | struct lttcomm_relayd_end_data_pending msg; | |
1789 | struct lttcomm_relayd_generic_reply reply; | |
1790 | struct relay_stream *stream; | |
1791 | uint64_t session_id; | |
1792 | uint32_t is_data_inflight = 0; | |
1793 | ||
1794 | assert(recv_hdr); | |
1795 | assert(cmd); | |
f7079f67 DG |
1796 | |
1797 | DBG("End data pending command"); | |
1798 | ||
1799 | if (!cmd->session || cmd->version_check_done == 0) { | |
1800 | ERR("Trying to check for data before version check"); | |
1801 | ret = -1; | |
1802 | goto end_no_session; | |
1803 | } | |
1804 | ||
1805 | ret = cmd->sock->ops->recvmsg(cmd->sock, &msg, sizeof(msg), 0); | |
1806 | if (ret < sizeof(msg)) { | |
a6cd2b97 DG |
1807 | if (ret == 0) { |
1808 | /* Orderly shutdown. Not necessary to print an error. */ | |
1809 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1810 | } else { | |
1811 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1812 | ret); | |
1813 | } | |
f7079f67 DG |
1814 | ret = -1; |
1815 | goto end_no_session; | |
1816 | } | |
1817 | ||
1818 | session_id = be64toh(msg.session_id); | |
1819 | ||
1820 | /* Iterate over all streams to see if the begin data pending flag is set. */ | |
1821 | rcu_read_lock(); | |
d3e2ba59 JD |
1822 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
1823 | stream_n.node) { | |
f7079f67 DG |
1824 | if (stream->session->id == session_id && |
1825 | !stream->data_pending_check_done) { | |
1826 | is_data_inflight = 1; | |
1827 | DBG("Data is still in flight for stream %" PRIu64, | |
1828 | stream->stream_handle); | |
1829 | break; | |
1830 | } | |
1831 | } | |
1832 | rcu_read_unlock(); | |
1833 | ||
1834 | /* All good, send back reply. */ | |
1835 | reply.ret_code = htobe32(is_data_inflight); | |
1836 | ||
1837 | ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, sizeof(reply), 0); | |
1838 | if (ret < 0) { | |
1839 | ERR("Relay end data pending send reply failed"); | |
1840 | } | |
1841 | ||
1842 | end_no_session: | |
1843 | return ret; | |
1844 | } | |
1845 | ||
1c20f0e2 JD |
1846 | /* |
1847 | * Receive an index for a specific stream. | |
1848 | * | |
1849 | * Return 0 on success else a negative value. | |
1850 | */ | |
1851 | static | |
1852 | int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, | |
0a6518b0 | 1853 | struct relay_command *cmd) |
1c20f0e2 JD |
1854 | { |
1855 | int ret, send_ret, index_created = 0; | |
1856 | struct relay_session *session = cmd->session; | |
1857 | struct lttcomm_relayd_index index_info; | |
1858 | struct relay_index *index, *wr_index = NULL; | |
1859 | struct lttcomm_relayd_generic_reply reply; | |
1860 | struct relay_stream *stream; | |
1861 | uint64_t net_seq_num; | |
1862 | ||
1863 | assert(cmd); | |
1c20f0e2 JD |
1864 | |
1865 | DBG("Relay receiving index"); | |
1866 | ||
1867 | if (!session || cmd->version_check_done == 0) { | |
1868 | ERR("Trying to close a stream before version check"); | |
1869 | ret = -1; | |
1870 | goto end_no_session; | |
1871 | } | |
1872 | ||
1873 | ret = cmd->sock->ops->recvmsg(cmd->sock, &index_info, | |
1874 | sizeof(index_info), 0); | |
1875 | if (ret < sizeof(index_info)) { | |
1876 | if (ret == 0) { | |
1877 | /* Orderly shutdown. Not necessary to print an error. */ | |
1878 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
1879 | } else { | |
1880 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1881 | } | |
1882 | ret = -1; | |
1883 | goto end_no_session; | |
1884 | } | |
1885 | ||
1886 | net_seq_num = be64toh(index_info.net_seq_num); | |
1887 | ||
1888 | rcu_read_lock(); | |
d3e2ba59 | 1889 | stream = relay_stream_find_by_id(be64toh(index_info.relay_stream_id)); |
1c20f0e2 JD |
1890 | if (!stream) { |
1891 | ret = -1; | |
1892 | goto end_rcu_unlock; | |
1893 | } | |
1894 | ||
d3e2ba59 JD |
1895 | /* Live beacon handling */ |
1896 | if (index_info.packet_size == 0) { | |
1897 | DBG("Received live beacon for stream %" PRIu64, stream->stream_handle); | |
1898 | ||
1899 | /* | |
1900 | * Only flag a stream inactive when it has already received data. | |
1901 | */ | |
1902 | if (stream->total_index_received > 0) { | |
1903 | stream->beacon_ts_end = be64toh(index_info.timestamp_end); | |
1904 | } | |
1905 | ret = 0; | |
1906 | goto end_rcu_unlock; | |
1907 | } else { | |
1908 | stream->beacon_ts_end = -1ULL; | |
1909 | } | |
1910 | ||
0a6518b0 | 1911 | index = relay_index_find(stream->stream_handle, net_seq_num); |
1c20f0e2 JD |
1912 | if (!index) { |
1913 | /* A successful creation will add the object to the HT. */ | |
1914 | index = relay_index_create(stream->stream_handle, net_seq_num); | |
1915 | if (!index) { | |
1916 | goto end_rcu_unlock; | |
1917 | } | |
1918 | index_created = 1; | |
1919 | } | |
1920 | ||
1921 | copy_index_control_data(index, &index_info); | |
1922 | ||
1923 | if (index_created) { | |
1924 | /* | |
1925 | * Try to add the relay index object to the hash table. If an object | |
1926 | * already exist, destroy back the index created, set the data in this | |
1927 | * object and write it on disk. | |
1928 | */ | |
0a6518b0 | 1929 | relay_index_add(index, &wr_index); |
1c20f0e2 JD |
1930 | if (wr_index) { |
1931 | copy_index_control_data(wr_index, &index_info); | |
1932 | free(index); | |
1933 | } | |
1934 | } else { | |
1935 | /* The index already exists so write it on disk. */ | |
1936 | wr_index = index; | |
1937 | } | |
1938 | ||
1939 | /* Do we have a writable ready index to write on disk. */ | |
1940 | if (wr_index) { | |
1941 | /* Starting at 2.4, create the index file if none available. */ | |
1942 | if (cmd->minor >= 4 && stream->index_fd < 0) { | |
1943 | ret = index_create_file(stream->path_name, stream->channel_name, | |
1944 | relayd_uid, relayd_gid, stream->tracefile_size, | |
1945 | stream->tracefile_count_current); | |
1946 | if (ret < 0) { | |
1947 | goto end_rcu_unlock; | |
1948 | } | |
1949 | stream->index_fd = ret; | |
1950 | } | |
1951 | ||
0a6518b0 | 1952 | ret = relay_index_write(wr_index->fd, wr_index); |
1c20f0e2 JD |
1953 | if (ret < 0) { |
1954 | goto end_rcu_unlock; | |
1955 | } | |
d3e2ba59 | 1956 | stream->total_index_received++; |
1c20f0e2 JD |
1957 | } |
1958 | ||
1959 | end_rcu_unlock: | |
1960 | rcu_read_unlock(); | |
1961 | ||
1962 | if (ret < 0) { | |
1963 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1964 | } else { | |
1965 | reply.ret_code = htobe32(LTTNG_OK); | |
1966 | } | |
1967 | send_ret = cmd->sock->ops->sendmsg(cmd->sock, &reply, sizeof(reply), 0); | |
1968 | if (send_ret < 0) { | |
1969 | ERR("Relay sending close index id reply"); | |
1970 | ret = send_ret; | |
1971 | } | |
1972 | ||
1973 | end_no_session: | |
1974 | return ret; | |
1975 | } | |
1976 | ||
b8aa1682 | 1977 | /* |
d3e2ba59 | 1978 | * Process the commands received on the control socket |
b8aa1682 JD |
1979 | */ |
1980 | static | |
1981 | int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, | |
d3e2ba59 | 1982 | struct relay_command *cmd, struct relay_local_data *ctx) |
b8aa1682 JD |
1983 | { |
1984 | int ret = 0; | |
1985 | ||
1986 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 1987 | case RELAYD_CREATE_SESSION: |
d3e2ba59 | 1988 | ret = relay_create_session(recv_hdr, cmd, ctx->sessions_ht); |
b8aa1682 | 1989 | break; |
b8aa1682 | 1990 | case RELAYD_ADD_STREAM: |
d3e2ba59 | 1991 | ret = relay_add_stream(recv_hdr, cmd, ctx->sessions_ht); |
b8aa1682 JD |
1992 | break; |
1993 | case RELAYD_START_DATA: | |
1994 | ret = relay_start(recv_hdr, cmd); | |
1995 | break; | |
1996 | case RELAYD_SEND_METADATA: | |
d3e2ba59 | 1997 | ret = relay_recv_metadata(recv_hdr, cmd); |
b8aa1682 JD |
1998 | break; |
1999 | case RELAYD_VERSION: | |
d3e2ba59 | 2000 | ret = relay_send_version(recv_hdr, cmd, ctx->sessions_ht); |
b8aa1682 | 2001 | break; |
173af62f | 2002 | case RELAYD_CLOSE_STREAM: |
92c6ca54 | 2003 | ret = relay_close_stream(recv_hdr, cmd); |
173af62f | 2004 | break; |
6d805429 | 2005 | case RELAYD_DATA_PENDING: |
d3e2ba59 | 2006 | ret = relay_data_pending(recv_hdr, cmd); |
c8f59ee5 DG |
2007 | break; |
2008 | case RELAYD_QUIESCENT_CONTROL: | |
d3e2ba59 | 2009 | ret = relay_quiescent_control(recv_hdr, cmd); |
c8f59ee5 | 2010 | break; |
f7079f67 | 2011 | case RELAYD_BEGIN_DATA_PENDING: |
d3e2ba59 | 2012 | ret = relay_begin_data_pending(recv_hdr, cmd); |
f7079f67 DG |
2013 | break; |
2014 | case RELAYD_END_DATA_PENDING: | |
d3e2ba59 | 2015 | ret = relay_end_data_pending(recv_hdr, cmd); |
f7079f67 | 2016 | break; |
1c20f0e2 | 2017 | case RELAYD_SEND_INDEX: |
0a6518b0 | 2018 | ret = relay_recv_index(recv_hdr, cmd); |
1c20f0e2 | 2019 | break; |
b8aa1682 JD |
2020 | case RELAYD_UPDATE_SYNC_INFO: |
2021 | default: | |
2022 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
2023 | relay_unknown_command(cmd); | |
2024 | ret = -1; | |
2025 | goto end; | |
2026 | } | |
2027 | ||
2028 | end: | |
2029 | return ret; | |
2030 | } | |
2031 | ||
7d2f7452 DG |
2032 | /* |
2033 | * Handle index for a data stream. | |
2034 | * | |
2035 | * RCU read side lock MUST be acquired. | |
2036 | * | |
2037 | * Return 0 on success else a negative value. | |
2038 | */ | |
2039 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2040 | int rotate_index) | |
2041 | { | |
2042 | int ret = 0, index_created = 0; | |
2043 | uint64_t stream_id, data_offset; | |
2044 | struct relay_index *index, *wr_index = NULL; | |
2045 | ||
2046 | assert(stream); | |
2047 | ||
2048 | stream_id = stream->stream_handle; | |
2049 | /* Get data offset because we are about to update the index. */ | |
2050 | data_offset = htobe64(stream->tracefile_size_current); | |
2051 | ||
2052 | /* | |
2053 | * Lookup for an existing index for that stream id/sequence number. If on | |
2054 | * exists, the control thread already received the data for it thus we need | |
2055 | * to write it on disk. | |
2056 | */ | |
2057 | index = relay_index_find(stream_id, net_seq_num); | |
2058 | if (!index) { | |
2059 | /* A successful creation will add the object to the HT. */ | |
2060 | index = relay_index_create(stream_id, net_seq_num); | |
2061 | if (!index) { | |
2062 | ret = -1; | |
2063 | goto error; | |
2064 | } | |
2065 | index_created = 1; | |
2066 | } | |
2067 | ||
2068 | if (rotate_index || stream->index_fd < 0) { | |
2069 | index->to_close_fd = stream->index_fd; | |
2070 | ret = index_create_file(stream->path_name, stream->channel_name, | |
2071 | relayd_uid, relayd_gid, stream->tracefile_size, | |
2072 | stream->tracefile_count_current); | |
2073 | if (ret < 0) { | |
2074 | /* This will close the stream's index fd if one. */ | |
2075 | relay_index_free_safe(index); | |
2076 | goto error; | |
2077 | } | |
2078 | stream->index_fd = ret; | |
2079 | } | |
2080 | index->fd = stream->index_fd; | |
2081 | index->index_data.offset = data_offset; | |
2082 | ||
2083 | if (index_created) { | |
2084 | /* | |
2085 | * Try to add the relay index object to the hash table. If an object | |
2086 | * already exist, destroy back the index created and set the data. | |
2087 | */ | |
2088 | relay_index_add(index, &wr_index); | |
2089 | if (wr_index) { | |
2090 | /* Copy back data from the created index. */ | |
2091 | wr_index->fd = index->fd; | |
2092 | wr_index->to_close_fd = index->to_close_fd; | |
2093 | wr_index->index_data.offset = data_offset; | |
2094 | free(index); | |
2095 | } | |
2096 | } else { | |
2097 | /* The index already exists so write it on disk. */ | |
2098 | wr_index = index; | |
2099 | } | |
2100 | ||
2101 | /* Do we have a writable ready index to write on disk. */ | |
2102 | if (wr_index) { | |
2103 | ret = relay_index_write(wr_index->fd, wr_index); | |
2104 | if (ret < 0) { | |
2105 | goto error; | |
2106 | } | |
2107 | stream->total_index_received++; | |
2108 | } | |
2109 | ||
2110 | error: | |
2111 | return ret; | |
2112 | } | |
2113 | ||
b8aa1682 JD |
2114 | /* |
2115 | * relay_process_data: Process the data received on the data socket | |
2116 | */ | |
2117 | static | |
0a6518b0 | 2118 | int relay_process_data(struct relay_command *cmd) |
b8aa1682 | 2119 | { |
7d2f7452 | 2120 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2121 | ssize_t size_ret; |
b8aa1682 JD |
2122 | struct relay_stream *stream; |
2123 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2124 | uint64_t stream_id; |
173af62f | 2125 | uint64_t net_seq_num; |
b8aa1682 JD |
2126 | uint32_t data_size; |
2127 | ||
2128 | ret = cmd->sock->ops->recvmsg(cmd->sock, &data_hdr, | |
7c5aef62 | 2129 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2130 | if (ret <= 0) { |
a6cd2b97 DG |
2131 | if (ret == 0) { |
2132 | /* Orderly shutdown. Not necessary to print an error. */ | |
2133 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
2134 | } else { | |
2135 | ERR("Unable to receive data header on sock %d", cmd->sock->fd); | |
2136 | } | |
b8aa1682 JD |
2137 | ret = -1; |
2138 | goto end; | |
2139 | } | |
2140 | ||
2141 | stream_id = be64toh(data_hdr.stream_id); | |
9d1bbf21 MD |
2142 | |
2143 | rcu_read_lock(); | |
d3e2ba59 | 2144 | stream = relay_stream_find_by_id(stream_id); |
b8aa1682 JD |
2145 | if (!stream) { |
2146 | ret = -1; | |
1c20f0e2 | 2147 | goto end_rcu_unlock; |
b8aa1682 JD |
2148 | } |
2149 | ||
2150 | data_size = be32toh(data_hdr.data_size); | |
2151 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2152 | char *tmp_data_ptr; |
2153 | ||
2154 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2155 | if (!tmp_data_ptr) { | |
b8aa1682 | 2156 | ERR("Allocating data buffer"); |
c617c0c6 | 2157 | free(data_buffer); |
b8aa1682 | 2158 | ret = -1; |
1c20f0e2 | 2159 | goto end_rcu_unlock; |
b8aa1682 | 2160 | } |
c617c0c6 | 2161 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2162 | data_buffer_size = data_size; |
2163 | } | |
2164 | memset(data_buffer, 0, data_size); | |
2165 | ||
173af62f DG |
2166 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2167 | ||
77c7c900 | 2168 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2169 | data_size, stream_id, net_seq_num); |
7c5aef62 | 2170 | ret = cmd->sock->ops->recvmsg(cmd->sock, data_buffer, data_size, 0); |
b8aa1682 | 2171 | if (ret <= 0) { |
a6cd2b97 DG |
2172 | if (ret == 0) { |
2173 | /* Orderly shutdown. Not necessary to print an error. */ | |
2174 | DBG("Socket %d did an orderly shutdown", cmd->sock->fd); | |
2175 | } | |
b8aa1682 | 2176 | ret = -1; |
1c20f0e2 | 2177 | goto end_rcu_unlock; |
b8aa1682 JD |
2178 | } |
2179 | ||
1c20f0e2 | 2180 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2181 | if (stream->tracefile_size > 0 && |
2182 | (stream->tracefile_size_current + data_size) > | |
2183 | stream->tracefile_size) { | |
6b6b9a5a JD |
2184 | struct relay_viewer_stream *vstream; |
2185 | uint64_t new_id; | |
2186 | ||
2187 | new_id = (stream->tracefile_count_current + 1) % | |
2188 | stream->tracefile_count; | |
2189 | /* | |
2190 | * When we wrap-around back to 0, we start overwriting old | |
2191 | * trace data. | |
2192 | */ | |
2193 | if (!stream->tracefile_overwrite && new_id == 0) { | |
2194 | stream->tracefile_overwrite = 1; | |
2195 | } | |
2196 | pthread_mutex_lock(&stream->viewer_stream_rotation_lock); | |
2197 | if (stream->tracefile_overwrite) { | |
2198 | stream->oldest_tracefile_id = | |
2199 | (stream->oldest_tracefile_id + 1) % | |
2200 | stream->tracefile_count; | |
2201 | } | |
2202 | vstream = live_find_viewer_stream_by_id(stream->stream_handle); | |
2203 | if (vstream) { | |
2204 | /* | |
2205 | * The viewer is reading a file about to be | |
2206 | * overwritten. Close the FDs it is | |
2207 | * currently using and let it handle the fault. | |
2208 | */ | |
2209 | if (vstream->tracefile_count_current == new_id) { | |
cef0f7d5 | 2210 | pthread_mutex_lock(&vstream->overwrite_lock); |
6b6b9a5a | 2211 | vstream->abort_flag = 1; |
cef0f7d5 | 2212 | pthread_mutex_unlock(&vstream->overwrite_lock); |
6b6b9a5a JD |
2213 | DBG("Streaming side setting abort_flag on stream %s_%lu\n", |
2214 | stream->channel_name, new_id); | |
2215 | } else if (vstream->tracefile_count_current == | |
2216 | stream->tracefile_count_current) { | |
2217 | /* | |
2218 | * The reader and writer were in the | |
2219 | * same trace file, inform the viewer | |
2220 | * that no new index will ever be added | |
2221 | * to this file. | |
2222 | */ | |
2223 | vstream->close_write_flag = 1; | |
2224 | } | |
2225 | } | |
1c20f0e2 JD |
2226 | ret = utils_rotate_stream_file(stream->path_name, stream->channel_name, |
2227 | stream->tracefile_size, stream->tracefile_count, | |
2228 | relayd_uid, relayd_gid, stream->fd, | |
2229 | &(stream->tracefile_count_current), &stream->fd); | |
cef0f7d5 | 2230 | stream->total_index_received = 0; |
6b6b9a5a | 2231 | pthread_mutex_unlock(&stream->viewer_stream_rotation_lock); |
0f907de1 | 2232 | if (ret < 0) { |
1c20f0e2 JD |
2233 | ERR("Rotating stream output file"); |
2234 | goto end_rcu_unlock; | |
0f907de1 | 2235 | } |
a6976990 DG |
2236 | /* Reset current size because we just perform a stream rotation. */ |
2237 | stream->tracefile_size_current = 0; | |
1c20f0e2 JD |
2238 | rotate_index = 1; |
2239 | } | |
2240 | ||
1c20f0e2 | 2241 | /* |
7d2f7452 DG |
2242 | * Index are handled in protocol version 2.4 and above. Also, snapshot and |
2243 | * index are NOT supported. | |
1c20f0e2 | 2244 | */ |
7d2f7452 DG |
2245 | if (stream->session->minor >= 4 && !stream->session->snapshot) { |
2246 | ret = handle_index_data(stream, net_seq_num, rotate_index); | |
1c20f0e2 | 2247 | if (ret < 0) { |
1c20f0e2 JD |
2248 | goto end_rcu_unlock; |
2249 | } | |
1c20f0e2 JD |
2250 | } |
2251 | ||
7d2f7452 | 2252 | /* Write data to stream output fd. */ |
6cd525e8 MD |
2253 | size_ret = lttng_write(stream->fd, data_buffer, data_size); |
2254 | if (size_ret < data_size) { | |
b8aa1682 JD |
2255 | ERR("Relay error writing data to file"); |
2256 | ret = -1; | |
1c20f0e2 | 2257 | goto end_rcu_unlock; |
b8aa1682 | 2258 | } |
1d4dfdef | 2259 | |
5ab7344e JD |
2260 | DBG2("Relay wrote %d bytes to tracefile for stream id %" PRIu64, |
2261 | ret, stream->stream_handle); | |
2262 | ||
1d4dfdef DG |
2263 | ret = write_padding_to_file(stream->fd, be32toh(data_hdr.padding_size)); |
2264 | if (ret < 0) { | |
1c20f0e2 | 2265 | goto end_rcu_unlock; |
1d4dfdef | 2266 | } |
1c20f0e2 | 2267 | stream->tracefile_size_current += data_size + be32toh(data_hdr.padding_size); |
1d4dfdef | 2268 | |
173af62f DG |
2269 | stream->prev_seq = net_seq_num; |
2270 | ||
2271 | /* Check if we need to close the FD */ | |
2272 | if (close_stream_check(stream)) { | |
c07d8a78 | 2273 | destroy_stream(stream); |
173af62f DG |
2274 | } |
2275 | ||
1c20f0e2 | 2276 | end_rcu_unlock: |
9d1bbf21 | 2277 | rcu_read_unlock(); |
b8aa1682 JD |
2278 | end: |
2279 | return ret; | |
2280 | } | |
2281 | ||
2282 | static | |
9d1bbf21 | 2283 | void relay_cleanup_poll_connection(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2284 | { |
2285 | int ret; | |
2286 | ||
b8aa1682 JD |
2287 | lttng_poll_del(events, pollfd); |
2288 | ||
2289 | ret = close(pollfd); | |
2290 | if (ret < 0) { | |
2291 | ERR("Closing pollfd %d", pollfd); | |
2292 | } | |
2293 | } | |
2294 | ||
2295 | static | |
2296 | int relay_add_connection(int fd, struct lttng_poll_event *events, | |
2297 | struct lttng_ht *relay_connections_ht) | |
2298 | { | |
b8aa1682 | 2299 | struct relay_command *relay_connection; |
6cd525e8 | 2300 | ssize_t ret; |
b8aa1682 JD |
2301 | |
2302 | relay_connection = zmalloc(sizeof(struct relay_command)); | |
2303 | if (relay_connection == NULL) { | |
2304 | PERROR("Relay command zmalloc"); | |
9d1bbf21 | 2305 | goto error; |
b8aa1682 | 2306 | } |
6cd525e8 MD |
2307 | ret = lttng_read(fd, relay_connection, sizeof(struct relay_command)); |
2308 | if (ret < sizeof(struct relay_command)) { | |
b8aa1682 | 2309 | PERROR("read relay cmd pipe"); |
9d1bbf21 | 2310 | goto error_read; |
b8aa1682 JD |
2311 | } |
2312 | ||
c07d8a78 DG |
2313 | /* |
2314 | * Only used by the control side and the reference is copied inside each | |
2315 | * stream from that connection. Thus a destroy HT must be done after every | |
2316 | * stream has been destroyed. | |
2317 | */ | |
2318 | if (relay_connection->type == RELAY_CONTROL) { | |
2319 | relay_connection->ctf_traces_ht = lttng_ht_new(0, | |
2320 | LTTNG_HT_TYPE_STRING); | |
2321 | if (!relay_connection->ctf_traces_ht) { | |
2322 | goto error_read; | |
2323 | } | |
d3e2ba59 JD |
2324 | } |
2325 | ||
b8aa1682 JD |
2326 | lttng_ht_node_init_ulong(&relay_connection->sock_n, |
2327 | (unsigned long) relay_connection->sock->fd); | |
9d1bbf21 | 2328 | rcu_read_lock(); |
b8aa1682 JD |
2329 | lttng_ht_add_unique_ulong(relay_connections_ht, |
2330 | &relay_connection->sock_n); | |
9d1bbf21 MD |
2331 | rcu_read_unlock(); |
2332 | return lttng_poll_add(events, | |
b8aa1682 JD |
2333 | relay_connection->sock->fd, |
2334 | LPOLLIN | LPOLLRDHUP); | |
2335 | ||
9d1bbf21 MD |
2336 | error_read: |
2337 | free(relay_connection); | |
2338 | error: | |
2339 | return -1; | |
2340 | } | |
2341 | ||
2342 | static | |
2343 | void deferred_free_connection(struct rcu_head *head) | |
2344 | { | |
2345 | struct relay_command *relay_connection = | |
2346 | caa_container_of(head, struct relay_command, rcu_node); | |
5b6d8097 DG |
2347 | |
2348 | lttcomm_destroy_sock(relay_connection->sock); | |
9d1bbf21 MD |
2349 | free(relay_connection); |
2350 | } | |
2351 | ||
2352 | static | |
2353 | void relay_del_connection(struct lttng_ht *relay_connections_ht, | |
d3e2ba59 JD |
2354 | struct lttng_ht_iter *iter, struct relay_command *relay_connection, |
2355 | struct lttng_ht *sessions_ht) | |
9d1bbf21 MD |
2356 | { |
2357 | int ret; | |
2358 | ||
2359 | ret = lttng_ht_del(relay_connections_ht, iter); | |
2360 | assert(!ret); | |
c07d8a78 | 2361 | |
9d1bbf21 | 2362 | if (relay_connection->type == RELAY_CONTROL) { |
d3e2ba59 | 2363 | relay_delete_session(relay_connection, sessions_ht); |
c07d8a78 | 2364 | lttng_ht_destroy(relay_connection->ctf_traces_ht); |
9d1bbf21 | 2365 | } |
5b6d8097 | 2366 | |
c07d8a78 | 2367 | call_rcu(&relay_connection->rcu_node, deferred_free_connection); |
b8aa1682 JD |
2368 | } |
2369 | ||
2370 | /* | |
2371 | * This thread does the actual work | |
2372 | */ | |
2373 | static | |
2374 | void *relay_thread_worker(void *data) | |
2375 | { | |
beaad64c DG |
2376 | int ret, err = -1, last_seen_data_fd = -1; |
2377 | uint32_t nb_fd; | |
b8aa1682 JD |
2378 | struct relay_command *relay_connection; |
2379 | struct lttng_poll_event events; | |
2380 | struct lttng_ht *relay_connections_ht; | |
2381 | struct lttng_ht_node_ulong *node; | |
2382 | struct lttng_ht_iter iter; | |
b8aa1682 | 2383 | struct lttcomm_relayd_hdr recv_hdr; |
d3e2ba59 JD |
2384 | struct relay_local_data *relay_ctx = (struct relay_local_data *) data; |
2385 | struct lttng_ht *sessions_ht = relay_ctx->sessions_ht; | |
b8aa1682 JD |
2386 | |
2387 | DBG("[thread] Relay worker started"); | |
2388 | ||
9d1bbf21 MD |
2389 | rcu_register_thread(); |
2390 | ||
55706a7d MD |
2391 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2392 | ||
f385ae0a MD |
2393 | health_code_update(); |
2394 | ||
b8aa1682 JD |
2395 | /* table of connections indexed on socket */ |
2396 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2397 | if (!relay_connections_ht) { |
2398 | goto relay_connections_ht_error; | |
2399 | } | |
b8aa1682 | 2400 | |
1c20f0e2 JD |
2401 | /* Tables of received indexes indexed by index handle and net_seq_num. */ |
2402 | indexes_ht = lttng_ht_new(0, LTTNG_HT_TYPE_TWO_U64); | |
2403 | if (!indexes_ht) { | |
2404 | goto indexes_ht_error; | |
2405 | } | |
2406 | ||
b8aa1682 JD |
2407 | ret = create_thread_poll_set(&events, 2); |
2408 | if (ret < 0) { | |
2409 | goto error_poll_create; | |
2410 | } | |
2411 | ||
2412 | ret = lttng_poll_add(&events, relay_cmd_pipe[0], LPOLLIN | LPOLLRDHUP); | |
2413 | if (ret < 0) { | |
2414 | goto error; | |
2415 | } | |
2416 | ||
beaad64c | 2417 | restart: |
b8aa1682 | 2418 | while (1) { |
beaad64c DG |
2419 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2420 | ||
f385ae0a MD |
2421 | health_code_update(); |
2422 | ||
b8aa1682 | 2423 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2424 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2425 | health_poll_entry(); |
b8aa1682 | 2426 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2427 | health_poll_exit(); |
b8aa1682 JD |
2428 | if (ret < 0) { |
2429 | /* | |
2430 | * Restart interrupted system call. | |
2431 | */ | |
2432 | if (errno == EINTR) { | |
2433 | goto restart; | |
2434 | } | |
2435 | goto error; | |
2436 | } | |
2437 | ||
0d9c5d77 DG |
2438 | nb_fd = ret; |
2439 | ||
beaad64c DG |
2440 | /* |
2441 | * Process control. The control connection is prioritised so we don't | |
2442 | * starve it with high throughout put tracing data on the data | |
2443 | * connection. | |
2444 | */ | |
b8aa1682 JD |
2445 | for (i = 0; i < nb_fd; i++) { |
2446 | /* Fetch once the poll data */ | |
beaad64c DG |
2447 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2448 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2449 | |
f385ae0a MD |
2450 | health_code_update(); |
2451 | ||
b8aa1682 JD |
2452 | /* Thread quit pipe has been closed. Killing thread. */ |
2453 | ret = check_thread_quit_pipe(pollfd, revents); | |
2454 | if (ret) { | |
095a4ae5 MD |
2455 | err = 0; |
2456 | goto exit; | |
b8aa1682 JD |
2457 | } |
2458 | ||
2459 | /* Inspect the relay cmd pipe for new connection */ | |
2460 | if (pollfd == relay_cmd_pipe[0]) { | |
2461 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2462 | ERR("Relay pipe error"); | |
2463 | goto error; | |
2464 | } else if (revents & LPOLLIN) { | |
2465 | DBG("Relay command received"); | |
2466 | ret = relay_add_connection(relay_cmd_pipe[0], | |
2467 | &events, relay_connections_ht); | |
2468 | if (ret < 0) { | |
2469 | goto error; | |
2470 | } | |
2471 | } | |
beaad64c | 2472 | } else if (revents) { |
9d1bbf21 | 2473 | rcu_read_lock(); |
b8aa1682 JD |
2474 | lttng_ht_lookup(relay_connections_ht, |
2475 | (void *)((unsigned long) pollfd), | |
2476 | &iter); | |
2477 | node = lttng_ht_iter_get_node_ulong(&iter); | |
2478 | if (node == NULL) { | |
2479 | DBG2("Relay sock %d not found", pollfd); | |
9d1bbf21 | 2480 | rcu_read_unlock(); |
b8aa1682 JD |
2481 | goto error; |
2482 | } | |
2483 | relay_connection = caa_container_of(node, | |
2484 | struct relay_command, sock_n); | |
2485 | ||
2486 | if (revents & (LPOLLERR)) { | |
2487 | ERR("POLL ERROR"); | |
9d1bbf21 MD |
2488 | relay_cleanup_poll_connection(&events, pollfd); |
2489 | relay_del_connection(relay_connections_ht, | |
d3e2ba59 | 2490 | &iter, relay_connection, sessions_ht); |
beaad64c DG |
2491 | if (last_seen_data_fd == pollfd) { |
2492 | last_seen_data_fd = last_notdel_data_fd; | |
2493 | } | |
b8aa1682 JD |
2494 | } else if (revents & (LPOLLHUP | LPOLLRDHUP)) { |
2495 | DBG("Socket %d hung up", pollfd); | |
9d1bbf21 MD |
2496 | relay_cleanup_poll_connection(&events, pollfd); |
2497 | relay_del_connection(relay_connections_ht, | |
d3e2ba59 | 2498 | &iter, relay_connection, sessions_ht); |
beaad64c DG |
2499 | if (last_seen_data_fd == pollfd) { |
2500 | last_seen_data_fd = last_notdel_data_fd; | |
2501 | } | |
b8aa1682 JD |
2502 | } else if (revents & LPOLLIN) { |
2503 | /* control socket */ | |
2504 | if (relay_connection->type == RELAY_CONTROL) { | |
2505 | ret = relay_connection->sock->ops->recvmsg( | |
2506 | relay_connection->sock, &recv_hdr, | |
7c5aef62 | 2507 | sizeof(struct lttcomm_relayd_hdr), 0); |
b8aa1682 JD |
2508 | /* connection closed */ |
2509 | if (ret <= 0) { | |
9d1bbf21 MD |
2510 | relay_cleanup_poll_connection(&events, pollfd); |
2511 | relay_del_connection(relay_connections_ht, | |
d3e2ba59 | 2512 | &iter, relay_connection, sessions_ht); |
b8aa1682 JD |
2513 | DBG("Control connection closed with %d", pollfd); |
2514 | } else { | |
2515 | if (relay_connection->session) { | |
77c7c900 | 2516 | DBG2("Relay worker receiving data for session : %" PRIu64, |
b8aa1682 JD |
2517 | relay_connection->session->id); |
2518 | } | |
2519 | ret = relay_process_control(&recv_hdr, | |
d3e2ba59 | 2520 | relay_connection, relay_ctx); |
b8aa1682 | 2521 | if (ret < 0) { |
beaad64c | 2522 | /* Clear the session on error. */ |
9d1bbf21 MD |
2523 | relay_cleanup_poll_connection(&events, pollfd); |
2524 | relay_del_connection(relay_connections_ht, | |
d3e2ba59 | 2525 | &iter, relay_connection, sessions_ht); |
b8aa1682 JD |
2526 | DBG("Connection closed with %d", pollfd); |
2527 | } | |
beaad64c | 2528 | seen_control = 1; |
b8aa1682 | 2529 | } |
beaad64c DG |
2530 | } else { |
2531 | /* | |
2532 | * Flag the last seen data fd not deleted. It will be | |
2533 | * used as the last seen fd if any fd gets deleted in | |
2534 | * this first loop. | |
2535 | */ | |
2536 | last_notdel_data_fd = pollfd; | |
2537 | } | |
2538 | } | |
2539 | rcu_read_unlock(); | |
2540 | } | |
2541 | } | |
2542 | ||
2543 | /* | |
2544 | * The last loop handled a control request, go back to poll to make | |
2545 | * sure we prioritise the control socket. | |
2546 | */ | |
2547 | if (seen_control) { | |
2548 | continue; | |
2549 | } | |
2550 | ||
2551 | if (last_seen_data_fd >= 0) { | |
2552 | for (i = 0; i < nb_fd; i++) { | |
2553 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2554 | |
2555 | health_code_update(); | |
2556 | ||
beaad64c DG |
2557 | if (last_seen_data_fd == pollfd) { |
2558 | idx = i; | |
2559 | break; | |
2560 | } | |
2561 | } | |
2562 | } | |
2563 | ||
2564 | /* Process data connection. */ | |
2565 | for (i = idx + 1; i < nb_fd; i++) { | |
2566 | /* Fetch the poll data. */ | |
2567 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2568 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
2569 | ||
f385ae0a MD |
2570 | health_code_update(); |
2571 | ||
beaad64c DG |
2572 | /* Skip the command pipe. It's handled in the first loop. */ |
2573 | if (pollfd == relay_cmd_pipe[0]) { | |
2574 | continue; | |
2575 | } | |
2576 | ||
2577 | if (revents) { | |
2578 | rcu_read_lock(); | |
2579 | lttng_ht_lookup(relay_connections_ht, | |
2580 | (void *)((unsigned long) pollfd), | |
2581 | &iter); | |
2582 | node = lttng_ht_iter_get_node_ulong(&iter); | |
2583 | if (node == NULL) { | |
2584 | /* Skip it. Might be removed before. */ | |
2585 | rcu_read_unlock(); | |
2586 | continue; | |
2587 | } | |
2588 | relay_connection = caa_container_of(node, | |
2589 | struct relay_command, sock_n); | |
2590 | ||
2591 | if (revents & LPOLLIN) { | |
2592 | if (relay_connection->type != RELAY_DATA) { | |
2593 | continue; | |
2594 | } | |
2595 | ||
0a6518b0 | 2596 | ret = relay_process_data(relay_connection); |
beaad64c DG |
2597 | /* connection closed */ |
2598 | if (ret < 0) { | |
2599 | relay_cleanup_poll_connection(&events, pollfd); | |
2600 | relay_del_connection(relay_connections_ht, | |
d3e2ba59 | 2601 | &iter, relay_connection, sessions_ht); |
beaad64c DG |
2602 | DBG("Data connection closed with %d", pollfd); |
2603 | /* | |
2604 | * Every goto restart call sets the last seen fd where | |
2605 | * here we don't really care since we gracefully | |
2606 | * continue the loop after the connection is deleted. | |
2607 | */ | |
2608 | } else { | |
2609 | /* Keep last seen port. */ | |
2610 | last_seen_data_fd = pollfd; | |
2611 | rcu_read_unlock(); | |
2612 | goto restart; | |
b8aa1682 JD |
2613 | } |
2614 | } | |
9d1bbf21 | 2615 | rcu_read_unlock(); |
b8aa1682 JD |
2616 | } |
2617 | } | |
beaad64c | 2618 | last_seen_data_fd = -1; |
b8aa1682 JD |
2619 | } |
2620 | ||
f385ae0a MD |
2621 | /* Normal exit, no error */ |
2622 | ret = 0; | |
2623 | ||
095a4ae5 | 2624 | exit: |
b8aa1682 JD |
2625 | error: |
2626 | lttng_poll_clean(&events); | |
2627 | ||
2628 | /* empty the hash table and free the memory */ | |
9d1bbf21 | 2629 | rcu_read_lock(); |
b8aa1682 | 2630 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, node, node) { |
f385ae0a MD |
2631 | health_code_update(); |
2632 | ||
b8aa1682 JD |
2633 | node = lttng_ht_iter_get_node_ulong(&iter); |
2634 | if (node) { | |
2635 | relay_connection = caa_container_of(node, | |
2636 | struct relay_command, sock_n); | |
9d1bbf21 | 2637 | relay_del_connection(relay_connections_ht, |
d3e2ba59 | 2638 | &iter, relay_connection, sessions_ht); |
b8aa1682 | 2639 | } |
b8aa1682 | 2640 | } |
94d49140 | 2641 | rcu_read_unlock(); |
7d2f7452 DG |
2642 | error_poll_create: |
2643 | lttng_ht_destroy(indexes_ht); | |
1c20f0e2 | 2644 | indexes_ht_error: |
b8aa1682 | 2645 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2646 | relay_connections_ht_error: |
6620da75 DG |
2647 | /* Close relay cmd pipes */ |
2648 | utils_close_pipe(relay_cmd_pipe); | |
095a4ae5 MD |
2649 | if (err) { |
2650 | DBG("Thread exited with error"); | |
2651 | } | |
b8aa1682 | 2652 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2653 | free(data_buffer); |
f385ae0a MD |
2654 | if (err) { |
2655 | health_error(); | |
2656 | ERR("Health error occurred in %s", __func__); | |
2657 | } | |
2658 | health_unregister(health_relayd); | |
9d1bbf21 | 2659 | rcu_unregister_thread(); |
f385ae0a | 2660 | stop_threads(); |
b8aa1682 JD |
2661 | return NULL; |
2662 | } | |
2663 | ||
2664 | /* | |
2665 | * Create the relay command pipe to wake thread_manage_apps. | |
2666 | * Closed in cleanup(). | |
2667 | */ | |
2668 | static int create_relay_cmd_pipe(void) | |
2669 | { | |
a02de639 | 2670 | int ret; |
b8aa1682 | 2671 | |
a02de639 | 2672 | ret = utils_create_pipe_cloexec(relay_cmd_pipe); |
b8aa1682 | 2673 | |
b8aa1682 JD |
2674 | return ret; |
2675 | } | |
2676 | ||
2677 | /* | |
2678 | * main | |
2679 | */ | |
2680 | int main(int argc, char **argv) | |
2681 | { | |
2682 | int ret = 0; | |
2683 | void *status; | |
d3e2ba59 | 2684 | struct relay_local_data *relay_ctx; |
b8aa1682 JD |
2685 | |
2686 | /* Create thread quit pipe */ | |
2687 | if ((ret = init_thread_quit_pipe()) < 0) { | |
2688 | goto error; | |
2689 | } | |
2690 | ||
2691 | /* Parse arguments */ | |
2692 | progname = argv[0]; | |
cd60b05a | 2693 | if ((ret = set_options(argc, argv)) < 0) { |
a02de639 | 2694 | goto exit; |
b8aa1682 JD |
2695 | } |
2696 | ||
2697 | if ((ret = set_signal_handler()) < 0) { | |
2698 | goto exit; | |
2699 | } | |
2700 | ||
4d513a50 DG |
2701 | /* Try to create directory if -o, --output is specified. */ |
2702 | if (opt_output_path) { | |
994fa64f DG |
2703 | if (*opt_output_path != '/') { |
2704 | ERR("Please specify an absolute path for -o, --output PATH"); | |
2705 | goto exit; | |
2706 | } | |
2707 | ||
4d513a50 DG |
2708 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG); |
2709 | if (ret < 0) { | |
2710 | ERR("Unable to create %s", opt_output_path); | |
2711 | goto exit; | |
2712 | } | |
2713 | } | |
2714 | ||
b8aa1682 JD |
2715 | /* Daemonize */ |
2716 | if (opt_daemon) { | |
2717 | ret = daemon(0, 0); | |
2718 | if (ret < 0) { | |
2719 | PERROR("daemon"); | |
a02de639 | 2720 | goto exit; |
b8aa1682 JD |
2721 | } |
2722 | } | |
2723 | ||
1c20f0e2 JD |
2724 | /* We need those values for the file/dir creation. */ |
2725 | relayd_uid = getuid(); | |
2726 | relayd_gid = getgid(); | |
b8aa1682 | 2727 | |
1c20f0e2 JD |
2728 | /* Check if daemon is UID = 0 */ |
2729 | if (relayd_uid == 0) { | |
b8aa1682 JD |
2730 | if (control_uri->port < 1024 || data_uri->port < 1024) { |
2731 | ERR("Need to be root to use ports < 1024"); | |
2732 | ret = -1; | |
a02de639 | 2733 | goto exit; |
b8aa1682 JD |
2734 | } |
2735 | } | |
2736 | ||
2737 | /* Setup the thread apps communication pipe. */ | |
2738 | if ((ret = create_relay_cmd_pipe()) < 0) { | |
2739 | goto exit; | |
2740 | } | |
2741 | ||
2742 | /* Init relay command queue. */ | |
2743 | cds_wfq_init(&relay_cmd_queue.queue); | |
2744 | ||
2745 | /* Set up max poll set size */ | |
2746 | lttng_poll_set_max_size(); | |
2747 | ||
554831e7 MD |
2748 | /* Initialize communication library */ |
2749 | lttcomm_init(); | |
2750 | ||
d3e2ba59 JD |
2751 | relay_ctx = zmalloc(sizeof(struct relay_local_data)); |
2752 | if (!relay_ctx) { | |
2753 | PERROR("relay_ctx"); | |
2754 | goto exit; | |
2755 | } | |
2756 | ||
2757 | /* tables of sessions indexed by session ID */ | |
2758 | relay_ctx->sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
2759 | if (!relay_ctx->sessions_ht) { | |
2760 | goto exit_relay_ctx_sessions; | |
2761 | } | |
2762 | ||
2763 | /* tables of streams indexed by stream ID */ | |
2764 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
2765 | if (!relay_streams_ht) { | |
2766 | goto exit_relay_ctx_streams; | |
2767 | } | |
2768 | ||
2769 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2770 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2771 | if (!viewer_streams_ht) { | |
d3e2ba59 JD |
2772 | goto exit_relay_ctx_viewer_streams; |
2773 | } | |
2774 | ||
55706a7d MD |
2775 | /* Initialize thread health monitoring */ |
2776 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2777 | if (!health_relayd) { | |
2778 | PERROR("health_app_create error"); | |
2779 | goto exit_health_app_create; | |
2780 | } | |
2781 | ||
65931c8b MD |
2782 | ret = utils_create_pipe(health_quit_pipe); |
2783 | if (ret < 0) { | |
2784 | goto error_health_pipe; | |
2785 | } | |
2786 | ||
2787 | /* Create thread to manage the client socket */ | |
2788 | ret = pthread_create(&health_thread, NULL, | |
2789 | thread_manage_health, (void *) NULL); | |
2790 | if (ret != 0) { | |
2791 | PERROR("pthread_create health"); | |
2792 | goto health_error; | |
2793 | } | |
2794 | ||
b8aa1682 JD |
2795 | /* Setup the dispatcher thread */ |
2796 | ret = pthread_create(&dispatcher_thread, NULL, | |
2797 | relay_thread_dispatcher, (void *) NULL); | |
2798 | if (ret != 0) { | |
2799 | PERROR("pthread_create dispatcher"); | |
2800 | goto exit_dispatcher; | |
2801 | } | |
2802 | ||
2803 | /* Setup the worker thread */ | |
2804 | ret = pthread_create(&worker_thread, NULL, | |
d3e2ba59 | 2805 | relay_thread_worker, (void *) relay_ctx); |
b8aa1682 JD |
2806 | if (ret != 0) { |
2807 | PERROR("pthread_create worker"); | |
2808 | goto exit_worker; | |
2809 | } | |
2810 | ||
2811 | /* Setup the listener thread */ | |
2812 | ret = pthread_create(&listener_thread, NULL, | |
2813 | relay_thread_listener, (void *) NULL); | |
2814 | if (ret != 0) { | |
2815 | PERROR("pthread_create listener"); | |
2816 | goto exit_listener; | |
2817 | } | |
2818 | ||
42415026 | 2819 | ret = live_start_threads(live_uri, relay_ctx, thread_quit_pipe); |
d3e2ba59 JD |
2820 | if (ret != 0) { |
2821 | ERR("Starting live viewer threads"); | |
50138f51 | 2822 | goto exit_live; |
d3e2ba59 JD |
2823 | } |
2824 | ||
50138f51 | 2825 | exit_live: |
b8aa1682 JD |
2826 | ret = pthread_join(listener_thread, &status); |
2827 | if (ret != 0) { | |
2828 | PERROR("pthread_join"); | |
2829 | goto error; /* join error, exit without cleanup */ | |
2830 | } | |
2831 | ||
50138f51 | 2832 | exit_listener: |
b8aa1682 JD |
2833 | ret = pthread_join(worker_thread, &status); |
2834 | if (ret != 0) { | |
2835 | PERROR("pthread_join"); | |
2836 | goto error; /* join error, exit without cleanup */ | |
2837 | } | |
2838 | ||
50138f51 | 2839 | exit_worker: |
b8aa1682 JD |
2840 | ret = pthread_join(dispatcher_thread, &status); |
2841 | if (ret != 0) { | |
2842 | PERROR("pthread_join"); | |
2843 | goto error; /* join error, exit without cleanup */ | |
2844 | } | |
42415026 | 2845 | |
50138f51 | 2846 | exit_dispatcher: |
65931c8b MD |
2847 | ret = pthread_join(health_thread, &status); |
2848 | if (ret != 0) { | |
2849 | PERROR("pthread_join health thread"); | |
2850 | goto error; /* join error, exit without cleanup */ | |
2851 | } | |
2852 | ||
bd11e201 MD |
2853 | /* |
2854 | * Stop live threads only after joining other threads. | |
2855 | */ | |
2856 | live_stop_threads(); | |
2857 | ||
65931c8b MD |
2858 | health_error: |
2859 | utils_close_pipe(health_quit_pipe); | |
2860 | ||
2861 | error_health_pipe: | |
55706a7d MD |
2862 | health_app_destroy(health_relayd); |
2863 | ||
2864 | exit_health_app_create: | |
92c6ca54 | 2865 | lttng_ht_destroy(viewer_streams_ht); |
d3e2ba59 JD |
2866 | |
2867 | exit_relay_ctx_viewer_streams: | |
2868 | lttng_ht_destroy(relay_streams_ht); | |
2869 | ||
2870 | exit_relay_ctx_streams: | |
2871 | lttng_ht_destroy(relay_ctx->sessions_ht); | |
2872 | ||
2873 | exit_relay_ctx_sessions: | |
2874 | free(relay_ctx); | |
b8aa1682 JD |
2875 | |
2876 | exit: | |
2877 | cleanup(); | |
2878 | if (!ret) { | |
2879 | exit(EXIT_SUCCESS); | |
2880 | } | |
a02de639 | 2881 | |
b8aa1682 JD |
2882 | error: |
2883 | exit(EXIT_FAILURE); | |
2884 | } |