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> |
9a78c92e | 5 | * 2015 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
b8aa1682 JD |
6 | * |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License, version 2 only, | |
9 | * as published by the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
14 | * more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; if not, write to the Free Software Foundation, Inc., | |
18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
19 | */ | |
20 | ||
21 | #define _GNU_SOURCE | |
22 | #include <getopt.h> | |
23 | #include <grp.h> | |
24 | #include <limits.h> | |
25 | #include <pthread.h> | |
26 | #include <signal.h> | |
27 | #include <stdio.h> | |
28 | #include <stdlib.h> | |
29 | #include <string.h> | |
30 | #include <sys/mman.h> | |
31 | #include <sys/mount.h> | |
32 | #include <sys/resource.h> | |
33 | #include <sys/socket.h> | |
34 | #include <sys/stat.h> | |
35 | #include <sys/types.h> | |
36 | #include <sys/wait.h> | |
173af62f | 37 | #include <inttypes.h> |
b8aa1682 JD |
38 | #include <urcu/futex.h> |
39 | #include <urcu/uatomic.h> | |
40 | #include <unistd.h> | |
41 | #include <fcntl.h> | |
42 | #include <config.h> | |
43 | ||
44 | #include <lttng/lttng.h> | |
45 | #include <common/common.h> | |
46 | #include <common/compat/poll.h> | |
47 | #include <common/compat/socket.h> | |
f263b7fd | 48 | #include <common/compat/endian.h> |
b8aa1682 | 49 | #include <common/defaults.h> |
3fd27398 | 50 | #include <common/daemonize.h> |
b8aa1682 JD |
51 | #include <common/futex.h> |
52 | #include <common/sessiond-comm/sessiond-comm.h> | |
53 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
54 | #include <common/sessiond-comm/relayd.h> |
55 | #include <common/uri.h> | |
a02de639 | 56 | #include <common/utils.h> |
cd60b05a | 57 | #include <common/config/config.h> |
9a78c92e | 58 | #include <urcu/rculist.h> |
b8aa1682 | 59 | |
0f907de1 | 60 | #include "cmd.h" |
d3e2ba59 | 61 | #include "ctf-trace.h" |
1c20f0e2 | 62 | #include "index.h" |
0f907de1 | 63 | #include "utils.h" |
b8aa1682 | 64 | #include "lttng-relayd.h" |
d3e2ba59 | 65 | #include "live.h" |
55706a7d | 66 | #include "health-relayd.h" |
9b5e0863 | 67 | #include "testpoint.h" |
2f8f53af | 68 | #include "viewer-stream.h" |
2a174661 DG |
69 | #include "session.h" |
70 | #include "stream.h" | |
58eb9381 | 71 | #include "connection.h" |
b8aa1682 JD |
72 | |
73 | /* command line options */ | |
0f907de1 | 74 | char *opt_output_path; |
b5218ffb | 75 | static int opt_daemon, opt_background; |
3fd27398 MD |
76 | |
77 | /* | |
78 | * We need to wait for listener and live listener threads, as well as | |
79 | * health check thread, before being ready to signal readiness. | |
80 | */ | |
81 | #define NR_LTTNG_RELAY_READY 3 | |
82 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
83 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ | |
84 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
85 | ||
095a4ae5 MD |
86 | static struct lttng_uri *control_uri; |
87 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 88 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
89 | |
90 | const char *progname; | |
b8aa1682 | 91 | |
65931c8b | 92 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
93 | static int tracing_group_name_override; |
94 | ||
95 | const char * const config_section_name = "relayd"; | |
65931c8b | 96 | |
b8aa1682 JD |
97 | /* |
98 | * Quit pipe for all threads. This permits a single cancellation point | |
99 | * for all threads when receiving an event on the pipe. | |
100 | */ | |
0b242f62 | 101 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
102 | |
103 | /* | |
104 | * This pipe is used to inform the worker thread that a command is queued and | |
105 | * ready to be processed. | |
106 | */ | |
58eb9381 | 107 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 108 | |
26c9d55e | 109 | /* Shared between threads */ |
b8aa1682 JD |
110 | static int dispatch_thread_exit; |
111 | ||
112 | static pthread_t listener_thread; | |
113 | static pthread_t dispatcher_thread; | |
114 | static pthread_t worker_thread; | |
65931c8b | 115 | static pthread_t health_thread; |
b8aa1682 | 116 | |
9a78c92e MD |
117 | /* |
118 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
119 | * atomicity on 32-bit architectures. | |
120 | */ | |
121 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 122 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
123 | |
124 | /* | |
125 | * Relay command queue. | |
126 | * | |
127 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
128 | * queue. | |
129 | */ | |
58eb9381 | 130 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 JD |
131 | |
132 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
133 | static char *data_buffer; |
134 | static unsigned int data_buffer_size; | |
b8aa1682 | 135 | |
d3e2ba59 JD |
136 | /* Global relay stream hash table. */ |
137 | struct lttng_ht *relay_streams_ht; | |
138 | ||
92c6ca54 DG |
139 | /* Global relay viewer stream hash table. */ |
140 | struct lttng_ht *viewer_streams_ht; | |
141 | ||
9a78c92e MD |
142 | /* Global relay sessions hash table. */ |
143 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 144 | |
55706a7d | 145 | /* Relayd health monitoring */ |
eea7556c | 146 | struct health_app *health_relayd; |
55706a7d | 147 | |
cd60b05a JG |
148 | static struct option long_options[] = { |
149 | { "control-port", 1, 0, 'C', }, | |
150 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 151 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 152 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 153 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
154 | { "group", 1, 0, 'g', }, |
155 | { "help", 0, 0, 'h', }, | |
156 | { "output", 1, 0, 'o', }, | |
157 | { "verbose", 0, 0, 'v', }, | |
158 | { "config", 1, 0, 'f' }, | |
159 | { NULL, 0, 0, 0, }, | |
160 | }; | |
161 | ||
162 | static const char *config_ignore_options[] = { "help", "config" }; | |
163 | ||
b8aa1682 JD |
164 | /* |
165 | * usage function on stderr | |
166 | */ | |
9a78c92e | 167 | static void usage(void) |
b8aa1682 JD |
168 | { |
169 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
170 | fprintf(stderr, " -h, --help Display this usage.\n"); |
171 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
b5218ffb | 172 | fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); |
994fa64f DG |
173 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); |
174 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
8d5c808e | 175 | fprintf(stderr, " -L, --live-port URL Live view port listening.\n"); |
994fa64f DG |
176 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); |
177 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 178 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 179 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
180 | } |
181 | ||
cd60b05a JG |
182 | /* |
183 | * Take an option from the getopt output and set it in the right variable to be | |
184 | * used later. | |
185 | * | |
186 | * Return 0 on success else a negative value. | |
187 | */ | |
9a78c92e | 188 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 189 | { |
cd60b05a JG |
190 | int ret; |
191 | ||
192 | switch (opt) { | |
193 | case 0: | |
194 | fprintf(stderr, "option %s", optname); | |
195 | if (arg) { | |
196 | fprintf(stderr, " with arg %s\n", arg); | |
197 | } | |
198 | break; | |
199 | case 'C': | |
200 | ret = uri_parse(arg, &control_uri); | |
201 | if (ret < 0) { | |
202 | ERR("Invalid control URI specified"); | |
203 | goto end; | |
204 | } | |
205 | if (control_uri->port == 0) { | |
206 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
207 | } | |
208 | break; | |
209 | case 'D': | |
210 | ret = uri_parse(arg, &data_uri); | |
211 | if (ret < 0) { | |
212 | ERR("Invalid data URI specified"); | |
213 | goto end; | |
214 | } | |
215 | if (data_uri->port == 0) { | |
216 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
217 | } | |
218 | break; | |
8d5c808e AM |
219 | case 'L': |
220 | ret = uri_parse(arg, &live_uri); | |
221 | if (ret < 0) { | |
222 | ERR("Invalid live URI specified"); | |
223 | goto end; | |
224 | } | |
225 | if (live_uri->port == 0) { | |
226 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
227 | } | |
228 | break; | |
cd60b05a JG |
229 | case 'd': |
230 | opt_daemon = 1; | |
231 | break; | |
b5218ffb MD |
232 | case 'b': |
233 | opt_background = 1; | |
234 | break; | |
cd60b05a JG |
235 | case 'g': |
236 | tracing_group_name = strdup(arg); | |
46cd2129 MD |
237 | if (tracing_group_name == NULL) { |
238 | ret = -errno; | |
239 | PERROR("strdup"); | |
240 | goto end; | |
241 | } | |
cd60b05a JG |
242 | tracing_group_name_override = 1; |
243 | break; | |
244 | case 'h': | |
245 | usage(); | |
246 | exit(EXIT_FAILURE); | |
247 | case 'o': | |
248 | ret = asprintf(&opt_output_path, "%s", arg); | |
249 | if (ret < 0) { | |
250 | ret = -errno; | |
251 | PERROR("asprintf opt_output_path"); | |
252 | goto end; | |
253 | } | |
254 | break; | |
255 | case 'v': | |
256 | /* Verbose level can increase using multiple -v */ | |
257 | if (arg) { | |
258 | lttng_opt_verbose = config_parse_value(arg); | |
259 | } else { | |
849e5b7b DG |
260 | /* Only 3 level of verbosity (-vvv). */ |
261 | if (lttng_opt_verbose < 3) { | |
262 | lttng_opt_verbose += 1; | |
263 | } | |
cd60b05a JG |
264 | } |
265 | break; | |
266 | default: | |
267 | /* Unknown option or other error. | |
268 | * Error is printed by getopt, just return */ | |
269 | ret = -1; | |
270 | goto end; | |
271 | } | |
272 | ||
273 | /* All good. */ | |
274 | ret = 0; | |
275 | ||
276 | end: | |
277 | return ret; | |
278 | } | |
279 | ||
280 | /* | |
281 | * config_entry_handler_cb used to handle options read from a config file. | |
282 | * See config_entry_handler_cb comment in common/config/config.h for the | |
283 | * return value conventions. | |
284 | */ | |
9a78c92e | 285 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
286 | { |
287 | int ret = 0, i; | |
288 | ||
289 | if (!entry || !entry->name || !entry->value) { | |
290 | ret = -EINVAL; | |
291 | goto end; | |
292 | } | |
293 | ||
294 | /* Check if the option is to be ignored */ | |
295 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
296 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
297 | goto end; | |
298 | } | |
299 | } | |
300 | ||
301 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
302 | /* Ignore if entry name is not fully matched. */ | |
303 | if (strcmp(entry->name, long_options[i].name)) { | |
304 | continue; | |
305 | } | |
306 | ||
307 | /* | |
9a78c92e MD |
308 | * If the option takes no argument on the command line, |
309 | * we have to check if the value is "true". We support | |
310 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
311 | */ |
312 | if (!long_options[i].has_arg) { | |
313 | ret = config_parse_value(entry->value); | |
314 | if (ret <= 0) { | |
315 | if (ret) { | |
316 | WARN("Invalid configuration value \"%s\" for option %s", | |
317 | entry->value, entry->name); | |
318 | } | |
319 | /* False, skip boolean config option. */ | |
320 | goto end; | |
321 | } | |
322 | } | |
323 | ||
324 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
325 | goto end; | |
326 | } | |
327 | ||
328 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
329 | entry->name); | |
330 | ||
331 | end: | |
332 | return ret; | |
333 | } | |
334 | ||
9a78c92e | 335 | static int set_options(int argc, char **argv) |
cd60b05a | 336 | { |
77c54fed | 337 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
338 | int orig_optopt = optopt, orig_optind = optind; |
339 | char *default_address, *optstring; | |
340 | const char *config_path = NULL; | |
341 | ||
342 | optstring = utils_generate_optstring(long_options, | |
343 | sizeof(long_options) / sizeof(struct option)); | |
344 | if (!optstring) { | |
77c54fed | 345 | retval = -ENOMEM; |
cd60b05a JG |
346 | goto exit; |
347 | } | |
348 | ||
349 | /* Check for the --config option */ | |
350 | ||
351 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
352 | &option_index)) != -1) { | |
353 | if (c == '?') { | |
77c54fed | 354 | retval = -EINVAL; |
cd60b05a JG |
355 | goto exit; |
356 | } else if (c != 'f') { | |
357 | continue; | |
358 | } | |
359 | ||
360 | config_path = utils_expand_path(optarg); | |
361 | if (!config_path) { | |
362 | ERR("Failed to resolve path: %s", optarg); | |
363 | } | |
364 | } | |
365 | ||
366 | ret = config_get_section_entries(config_path, config_section_name, | |
367 | config_entry_handler, NULL); | |
368 | if (ret) { | |
369 | if (ret > 0) { | |
370 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 371 | } |
77c54fed | 372 | retval = -1; |
cd60b05a JG |
373 | goto exit; |
374 | } | |
b8aa1682 | 375 | |
cd60b05a JG |
376 | /* Reset getopt's global state */ |
377 | optopt = orig_optopt; | |
378 | optind = orig_optind; | |
b8aa1682 | 379 | while (1) { |
cd60b05a | 380 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
381 | if (c == -1) { |
382 | break; | |
383 | } | |
384 | ||
cd60b05a JG |
385 | ret = set_option(c, optarg, long_options[option_index].name); |
386 | if (ret < 0) { | |
77c54fed | 387 | retval = -1; |
b8aa1682 JD |
388 | goto exit; |
389 | } | |
390 | } | |
391 | ||
392 | /* assign default values */ | |
393 | if (control_uri == NULL) { | |
fa91dc52 MD |
394 | ret = asprintf(&default_address, |
395 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
396 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
397 | if (ret < 0) { |
398 | PERROR("asprintf default data address"); | |
77c54fed | 399 | retval = -1; |
b8aa1682 JD |
400 | goto exit; |
401 | } | |
402 | ||
403 | ret = uri_parse(default_address, &control_uri); | |
404 | free(default_address); | |
405 | if (ret < 0) { | |
406 | ERR("Invalid control URI specified"); | |
77c54fed | 407 | retval = -1; |
b8aa1682 JD |
408 | goto exit; |
409 | } | |
410 | } | |
411 | if (data_uri == NULL) { | |
fa91dc52 MD |
412 | ret = asprintf(&default_address, |
413 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
414 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
415 | if (ret < 0) { |
416 | PERROR("asprintf default data address"); | |
77c54fed | 417 | retval = -1; |
b8aa1682 JD |
418 | goto exit; |
419 | } | |
420 | ||
421 | ret = uri_parse(default_address, &data_uri); | |
422 | free(default_address); | |
423 | if (ret < 0) { | |
424 | ERR("Invalid data URI specified"); | |
77c54fed | 425 | retval = -1; |
b8aa1682 JD |
426 | goto exit; |
427 | } | |
428 | } | |
d3e2ba59 | 429 | if (live_uri == NULL) { |
fa91dc52 MD |
430 | ret = asprintf(&default_address, |
431 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
432 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
433 | if (ret < 0) { |
434 | PERROR("asprintf default viewer control address"); | |
77c54fed | 435 | retval = -1; |
d3e2ba59 JD |
436 | goto exit; |
437 | } | |
438 | ||
439 | ret = uri_parse(default_address, &live_uri); | |
440 | free(default_address); | |
441 | if (ret < 0) { | |
442 | ERR("Invalid viewer control URI specified"); | |
77c54fed | 443 | retval = -1; |
d3e2ba59 JD |
444 | goto exit; |
445 | } | |
446 | } | |
b8aa1682 JD |
447 | |
448 | exit: | |
cd60b05a | 449 | free(optstring); |
77c54fed | 450 | return retval; |
b8aa1682 JD |
451 | } |
452 | ||
9a78c92e MD |
453 | static void print_global_objects(void) |
454 | { | |
455 | rcu_register_thread(); | |
456 | ||
457 | print_viewer_streams(); | |
458 | print_relay_streams(); | |
459 | print_sessions(); | |
460 | ||
461 | rcu_unregister_thread(); | |
462 | } | |
463 | ||
b8aa1682 JD |
464 | /* |
465 | * Cleanup the daemon | |
466 | */ | |
9a78c92e | 467 | static void relayd_cleanup(void) |
b8aa1682 | 468 | { |
9a78c92e MD |
469 | print_global_objects(); |
470 | ||
b8aa1682 JD |
471 | DBG("Cleaning up"); |
472 | ||
77c54fed MD |
473 | if (viewer_streams_ht) |
474 | lttng_ht_destroy(viewer_streams_ht); | |
475 | if (relay_streams_ht) | |
476 | lttng_ht_destroy(relay_streams_ht); | |
9a78c92e MD |
477 | if (sessions_ht) |
478 | lttng_ht_destroy(sessions_ht); | |
77c54fed | 479 | |
095a4ae5 MD |
480 | /* free the dynamically allocated opt_output_path */ |
481 | free(opt_output_path); | |
482 | ||
a02de639 CB |
483 | /* Close thread quit pipes */ |
484 | utils_close_pipe(thread_quit_pipe); | |
485 | ||
710c1f73 DG |
486 | uri_free(control_uri); |
487 | uri_free(data_uri); | |
8d5c808e | 488 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
489 | |
490 | if (tracing_group_name_override) { | |
491 | free((void *) tracing_group_name); | |
492 | } | |
b8aa1682 JD |
493 | } |
494 | ||
495 | /* | |
496 | * Write to writable pipe used to notify a thread. | |
497 | */ | |
9a78c92e | 498 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 499 | { |
6cd525e8 | 500 | ssize_t ret; |
b8aa1682 | 501 | |
6cd525e8 MD |
502 | ret = lttng_write(wpipe, "!", 1); |
503 | if (ret < 1) { | |
b8aa1682 | 504 | PERROR("write poll pipe"); |
770f96f4 | 505 | goto end; |
b8aa1682 | 506 | } |
770f96f4 MD |
507 | ret = 0; |
508 | end: | |
b8aa1682 JD |
509 | return ret; |
510 | } | |
511 | ||
9a78c92e | 512 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 513 | { |
6cd525e8 | 514 | ssize_t ret; |
65931c8b | 515 | |
6cd525e8 MD |
516 | ret = lttng_write(pipe[1], "4", 1); |
517 | if (ret < 1) { | |
65931c8b | 518 | PERROR("write relay health quit"); |
770f96f4 | 519 | goto end; |
65931c8b | 520 | } |
770f96f4 MD |
521 | ret = 0; |
522 | end: | |
523 | return ret; | |
65931c8b MD |
524 | } |
525 | ||
b8aa1682 | 526 | /* |
770f96f4 | 527 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 528 | */ |
770f96f4 | 529 | int lttng_relay_stop_threads(void) |
b8aa1682 | 530 | { |
770f96f4 | 531 | int retval = 0; |
b8aa1682 JD |
532 | |
533 | /* Stopping all threads */ | |
534 | DBG("Terminating all threads"); | |
770f96f4 | 535 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 536 | ERR("write error on thread quit pipe"); |
770f96f4 | 537 | retval = -1; |
b8aa1682 JD |
538 | } |
539 | ||
770f96f4 MD |
540 | if (notify_health_quit_pipe(health_quit_pipe)) { |
541 | ERR("write error on health quit pipe"); | |
542 | } | |
65931c8b | 543 | |
b8aa1682 | 544 | /* Dispatch thread */ |
26c9d55e | 545 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 546 | futex_nto1_wake(&relay_conn_queue.futex); |
77c54fed | 547 | |
770f96f4 | 548 | if (relayd_live_stop()) { |
77c54fed | 549 | ERR("Error stopping live threads"); |
770f96f4 | 550 | retval = -1; |
77c54fed | 551 | } |
770f96f4 | 552 | return retval; |
b8aa1682 JD |
553 | } |
554 | ||
555 | /* | |
556 | * Signal handler for the daemon | |
557 | * | |
558 | * Simply stop all worker threads, leaving main() return gracefully after | |
559 | * joining all threads and calling cleanup(). | |
560 | */ | |
9a78c92e | 561 | static void sighandler(int sig) |
b8aa1682 JD |
562 | { |
563 | switch (sig) { | |
564 | case SIGPIPE: | |
565 | DBG("SIGPIPE caught"); | |
566 | return; | |
567 | case SIGINT: | |
568 | DBG("SIGINT caught"); | |
770f96f4 MD |
569 | if (lttng_relay_stop_threads()) { |
570 | ERR("Error stopping threads"); | |
571 | } | |
b8aa1682 JD |
572 | break; |
573 | case SIGTERM: | |
574 | DBG("SIGTERM caught"); | |
770f96f4 MD |
575 | if (lttng_relay_stop_threads()) { |
576 | ERR("Error stopping threads"); | |
577 | } | |
b8aa1682 | 578 | break; |
3fd27398 MD |
579 | case SIGUSR1: |
580 | CMM_STORE_SHARED(recv_child_signal, 1); | |
581 | break; | |
b8aa1682 JD |
582 | default: |
583 | break; | |
584 | } | |
585 | } | |
586 | ||
587 | /* | |
588 | * Setup signal handler for : | |
589 | * SIGINT, SIGTERM, SIGPIPE | |
590 | */ | |
9a78c92e | 591 | static int set_signal_handler(void) |
b8aa1682 JD |
592 | { |
593 | int ret = 0; | |
594 | struct sigaction sa; | |
595 | sigset_t sigset; | |
596 | ||
597 | if ((ret = sigemptyset(&sigset)) < 0) { | |
598 | PERROR("sigemptyset"); | |
599 | return ret; | |
600 | } | |
601 | ||
602 | sa.sa_handler = sighandler; | |
603 | sa.sa_mask = sigset; | |
604 | sa.sa_flags = 0; | |
605 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
606 | PERROR("sigaction"); | |
607 | return ret; | |
608 | } | |
609 | ||
610 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
611 | PERROR("sigaction"); | |
612 | return ret; | |
613 | } | |
614 | ||
615 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
616 | PERROR("sigaction"); | |
617 | return ret; | |
618 | } | |
619 | ||
3fd27398 MD |
620 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
621 | PERROR("sigaction"); | |
622 | return ret; | |
623 | } | |
624 | ||
625 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
626 | |
627 | return ret; | |
628 | } | |
629 | ||
3fd27398 MD |
630 | void lttng_relay_notify_ready(void) |
631 | { | |
632 | /* Notify the parent of the fork() process that we are ready. */ | |
633 | if (opt_daemon || opt_background) { | |
634 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
635 | kill(child_ppid, SIGUSR1); | |
636 | } | |
637 | } | |
638 | } | |
639 | ||
b8aa1682 JD |
640 | /* |
641 | * Init thread quit pipe. | |
642 | * | |
643 | * Return -1 on error or 0 if all pipes are created. | |
644 | */ | |
9a78c92e | 645 | static int init_thread_quit_pipe(void) |
b8aa1682 | 646 | { |
a02de639 | 647 | int ret; |
b8aa1682 | 648 | |
a02de639 | 649 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 650 | |
b8aa1682 JD |
651 | return ret; |
652 | } | |
653 | ||
654 | /* | |
655 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
656 | */ | |
9a78c92e | 657 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
658 | { |
659 | int ret; | |
660 | ||
661 | if (events == NULL || size == 0) { | |
662 | ret = -1; | |
663 | goto error; | |
664 | } | |
665 | ||
666 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
667 | if (ret < 0) { | |
668 | goto error; | |
669 | } | |
670 | ||
671 | /* Add quit pipe */ | |
c7759e6a | 672 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
673 | if (ret < 0) { |
674 | goto error; | |
675 | } | |
676 | ||
677 | return 0; | |
678 | ||
679 | error: | |
680 | return ret; | |
681 | } | |
682 | ||
683 | /* | |
684 | * Check if the thread quit pipe was triggered. | |
685 | * | |
686 | * Return 1 if it was triggered else 0; | |
687 | */ | |
9a78c92e | 688 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
689 | { |
690 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
691 | return 1; | |
692 | } | |
693 | ||
694 | return 0; | |
695 | } | |
696 | ||
697 | /* | |
698 | * Create and init socket from uri. | |
699 | */ | |
9a78c92e | 700 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
701 | { |
702 | int ret; | |
703 | struct lttcomm_sock *sock = NULL; | |
704 | ||
705 | sock = lttcomm_alloc_sock_from_uri(uri); | |
706 | if (sock == NULL) { | |
707 | ERR("Allocating socket"); | |
708 | goto error; | |
709 | } | |
710 | ||
711 | ret = lttcomm_create_sock(sock); | |
712 | if (ret < 0) { | |
713 | goto error; | |
714 | } | |
715 | DBG("Listening on sock %d", sock->fd); | |
716 | ||
717 | ret = sock->ops->bind(sock); | |
718 | if (ret < 0) { | |
719 | goto error; | |
720 | } | |
721 | ||
722 | ret = sock->ops->listen(sock, -1); | |
723 | if (ret < 0) { | |
724 | goto error; | |
725 | ||
726 | } | |
727 | ||
728 | return sock; | |
729 | ||
730 | error: | |
731 | if (sock) { | |
732 | lttcomm_destroy_sock(sock); | |
733 | } | |
734 | return NULL; | |
735 | } | |
736 | ||
737 | /* | |
738 | * This thread manages the listening for new connections on the network | |
739 | */ | |
9a78c92e | 740 | static void *relay_thread_listener(void *data) |
b8aa1682 | 741 | { |
095a4ae5 | 742 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
743 | uint32_t revents, nb_fd; |
744 | struct lttng_poll_event events; | |
745 | struct lttcomm_sock *control_sock, *data_sock; | |
746 | ||
b8aa1682 JD |
747 | DBG("[thread] Relay listener started"); |
748 | ||
55706a7d MD |
749 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
750 | ||
f385ae0a MD |
751 | health_code_update(); |
752 | ||
9a78c92e | 753 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 754 | if (!control_sock) { |
095a4ae5 | 755 | goto error_sock_control; |
b8aa1682 JD |
756 | } |
757 | ||
9a78c92e | 758 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 759 | if (!data_sock) { |
095a4ae5 | 760 | goto error_sock_relay; |
b8aa1682 JD |
761 | } |
762 | ||
763 | /* | |
9a78c92e MD |
764 | * Pass 3 as size here for the thread quit pipe, control and |
765 | * data socket. | |
b8aa1682 JD |
766 | */ |
767 | ret = create_thread_poll_set(&events, 3); | |
768 | if (ret < 0) { | |
769 | goto error_create_poll; | |
770 | } | |
771 | ||
772 | /* Add the control socket */ | |
773 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
774 | if (ret < 0) { | |
775 | goto error_poll_add; | |
776 | } | |
777 | ||
778 | /* Add the data socket */ | |
779 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
780 | if (ret < 0) { | |
781 | goto error_poll_add; | |
782 | } | |
783 | ||
3fd27398 MD |
784 | lttng_relay_notify_ready(); |
785 | ||
9b5e0863 MD |
786 | if (testpoint(relayd_thread_listener)) { |
787 | goto error_testpoint; | |
788 | } | |
789 | ||
b8aa1682 | 790 | while (1) { |
f385ae0a MD |
791 | health_code_update(); |
792 | ||
b8aa1682 JD |
793 | DBG("Listener accepting connections"); |
794 | ||
b8aa1682 | 795 | restart: |
f385ae0a | 796 | health_poll_entry(); |
b8aa1682 | 797 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 798 | health_poll_exit(); |
b8aa1682 JD |
799 | if (ret < 0) { |
800 | /* | |
801 | * Restart interrupted system call. | |
802 | */ | |
803 | if (errno == EINTR) { | |
804 | goto restart; | |
805 | } | |
806 | goto error; | |
807 | } | |
808 | ||
0d9c5d77 DG |
809 | nb_fd = ret; |
810 | ||
b8aa1682 JD |
811 | DBG("Relay new connection received"); |
812 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
813 | health_code_update(); |
814 | ||
b8aa1682 JD |
815 | /* Fetch once the poll data */ |
816 | revents = LTTNG_POLL_GETEV(&events, i); | |
817 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
818 | ||
e0077699 | 819 | if (!revents) { |
9a78c92e MD |
820 | /* |
821 | * No activity for this FD (poll | |
822 | * implementation). | |
823 | */ | |
e0077699 MD |
824 | continue; |
825 | } | |
826 | ||
b8aa1682 JD |
827 | /* Thread quit pipe has been closed. Killing thread. */ |
828 | ret = check_thread_quit_pipe(pollfd, revents); | |
829 | if (ret) { | |
095a4ae5 MD |
830 | err = 0; |
831 | goto exit; | |
b8aa1682 JD |
832 | } |
833 | ||
834 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
835 | ERR("socket poll error"); | |
836 | goto error; | |
837 | } else if (revents & LPOLLIN) { | |
4b7f17b2 | 838 | /* |
9a78c92e MD |
839 | * A new connection is requested, therefore a |
840 | * sessiond/consumerd connection is allocated in | |
841 | * this thread, enqueued to a global queue and | |
842 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 843 | */ |
58eb9381 DG |
844 | int val = 1; |
845 | struct relay_connection *new_conn; | |
4b7f17b2 | 846 | struct lttcomm_sock *newsock; |
9a78c92e | 847 | enum connection_type type; |
b8aa1682 JD |
848 | |
849 | if (pollfd == data_sock->fd) { | |
9a78c92e | 850 | type = RELAY_DATA; |
b8aa1682 | 851 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
852 | DBG("Relay data connection accepted, socket %d", |
853 | newsock->fd); | |
4b7f17b2 MD |
854 | } else { |
855 | assert(pollfd == control_sock->fd); | |
9a78c92e | 856 | type = RELAY_CONTROL; |
b8aa1682 | 857 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
858 | DBG("Relay control connection accepted, socket %d", |
859 | newsock->fd); | |
b8aa1682 | 860 | } |
58eb9381 DG |
861 | if (!newsock) { |
862 | PERROR("accepting sock"); | |
58eb9381 DG |
863 | goto error; |
864 | } | |
865 | ||
866 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
867 | sizeof(val)); | |
b8aa1682 JD |
868 | if (ret < 0) { |
869 | PERROR("setsockopt inet"); | |
4b7f17b2 | 870 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
871 | goto error; |
872 | } | |
9a78c92e MD |
873 | new_conn = connection_create(newsock, type); |
874 | if (!new_conn) { | |
875 | lttcomm_destroy_sock(newsock); | |
876 | goto error; | |
877 | } | |
58eb9381 DG |
878 | |
879 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
880 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
881 | &new_conn->qnode); | |
b8aa1682 JD |
882 | |
883 | /* | |
9a78c92e MD |
884 | * Wake the dispatch queue futex. |
885 | * Implicit memory barrier with the | |
886 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 887 | */ |
58eb9381 | 888 | futex_nto1_wake(&relay_conn_queue.futex); |
b8aa1682 JD |
889 | } |
890 | } | |
891 | } | |
892 | ||
095a4ae5 | 893 | exit: |
b8aa1682 JD |
894 | error: |
895 | error_poll_add: | |
9b5e0863 | 896 | error_testpoint: |
b8aa1682 JD |
897 | lttng_poll_clean(&events); |
898 | error_create_poll: | |
095a4ae5 MD |
899 | if (data_sock->fd >= 0) { |
900 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
901 | if (ret) { |
902 | PERROR("close"); | |
903 | } | |
b8aa1682 | 904 | } |
095a4ae5 MD |
905 | lttcomm_destroy_sock(data_sock); |
906 | error_sock_relay: | |
907 | if (control_sock->fd >= 0) { | |
908 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
909 | if (ret) { |
910 | PERROR("close"); | |
911 | } | |
b8aa1682 | 912 | } |
095a4ae5 MD |
913 | lttcomm_destroy_sock(control_sock); |
914 | error_sock_control: | |
915 | if (err) { | |
f385ae0a MD |
916 | health_error(); |
917 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 918 | } |
55706a7d | 919 | health_unregister(health_relayd); |
b8aa1682 | 920 | DBG("Relay listener thread cleanup complete"); |
770f96f4 | 921 | lttng_relay_stop_threads(); |
b8aa1682 JD |
922 | return NULL; |
923 | } | |
924 | ||
925 | /* | |
926 | * This thread manages the dispatching of the requests to worker threads | |
927 | */ | |
9a78c92e | 928 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 929 | { |
6cd525e8 MD |
930 | int err = -1; |
931 | ssize_t ret; | |
8bdee6e2 | 932 | struct cds_wfcq_node *node; |
58eb9381 | 933 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
934 | |
935 | DBG("[thread] Relay dispatcher started"); | |
936 | ||
55706a7d MD |
937 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
938 | ||
9b5e0863 MD |
939 | if (testpoint(relayd_thread_dispatcher)) { |
940 | goto error_testpoint; | |
941 | } | |
942 | ||
f385ae0a MD |
943 | health_code_update(); |
944 | ||
26c9d55e | 945 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
946 | health_code_update(); |
947 | ||
b8aa1682 | 948 | /* Atomically prepare the queue futex */ |
58eb9381 | 949 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
950 | |
951 | do { | |
f385ae0a MD |
952 | health_code_update(); |
953 | ||
b8aa1682 | 954 | /* Dequeue commands */ |
8bdee6e2 SM |
955 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
956 | &relay_conn_queue.tail); | |
b8aa1682 JD |
957 | if (node == NULL) { |
958 | DBG("Woken up but nothing in the relay command queue"); | |
959 | /* Continue thread execution */ | |
960 | break; | |
961 | } | |
58eb9381 | 962 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 963 | |
58eb9381 | 964 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
965 | |
966 | /* | |
9a78c92e MD |
967 | * Inform worker thread of the new request. This |
968 | * call is blocking so we can be assured that | |
969 | * the data will be read at some point in time | |
970 | * or wait to the end of the world :) | |
b8aa1682 | 971 | */ |
58eb9381 DG |
972 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
973 | if (ret < 0) { | |
974 | PERROR("write connection pipe"); | |
9a78c92e | 975 | connection_put(new_conn); |
b8aa1682 JD |
976 | goto error; |
977 | } | |
978 | } while (node != NULL); | |
979 | ||
980 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 981 | health_poll_entry(); |
58eb9381 | 982 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 983 | health_poll_exit(); |
b8aa1682 JD |
984 | } |
985 | ||
f385ae0a MD |
986 | /* Normal exit, no error */ |
987 | err = 0; | |
988 | ||
b8aa1682 | 989 | error: |
9b5e0863 | 990 | error_testpoint: |
f385ae0a MD |
991 | if (err) { |
992 | health_error(); | |
993 | ERR("Health error occurred in %s", __func__); | |
994 | } | |
55706a7d | 995 | health_unregister(health_relayd); |
b8aa1682 | 996 | DBG("Dispatch thread dying"); |
770f96f4 | 997 | lttng_relay_stop_threads(); |
b8aa1682 JD |
998 | return NULL; |
999 | } | |
1000 | ||
1c20f0e2 | 1001 | /* |
9a78c92e | 1002 | * Set index data from the control port to a given index object. |
1c20f0e2 | 1003 | */ |
9a78c92e | 1004 | static int set_index_control_data(struct relay_index *index, |
1c20f0e2 JD |
1005 | struct lttcomm_relayd_index *data) |
1006 | { | |
9a78c92e | 1007 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1008 | |
1009 | /* | |
9a78c92e MD |
1010 | * The index on disk is encoded in big endian, so we don't need |
1011 | * to convert the data received on the network. The data_offset | |
1012 | * value is NEVER modified here and is updated by the data | |
1013 | * thread. | |
1c20f0e2 | 1014 | */ |
9a78c92e MD |
1015 | index_data.packet_size = data->packet_size; |
1016 | index_data.content_size = data->content_size; | |
1017 | index_data.timestamp_begin = data->timestamp_begin; | |
1018 | index_data.timestamp_end = data->timestamp_end; | |
1019 | index_data.events_discarded = data->events_discarded; | |
1020 | index_data.stream_id = data->stream_id; | |
1021 | return relay_index_set_data(index, &index_data); | |
1c20f0e2 JD |
1022 | } |
1023 | ||
c5b6f4f0 DG |
1024 | /* |
1025 | * Handle the RELAYD_CREATE_SESSION command. | |
1026 | * | |
1027 | * On success, send back the session id or else return a negative value. | |
1028 | */ | |
9a78c92e | 1029 | static int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1030 | struct relay_connection *conn) |
c5b6f4f0 DG |
1031 | { |
1032 | int ret = 0, send_ret; | |
1033 | struct relay_session *session; | |
1034 | struct lttcomm_relayd_status_session reply; | |
9a78c92e MD |
1035 | char session_name[NAME_MAX]; |
1036 | char hostname[HOST_NAME_MAX]; | |
1037 | uint32_t live_timer = 0; | |
1038 | bool snapshot = false; | |
c5b6f4f0 | 1039 | |
9a78c92e MD |
1040 | memset(session_name, 0, NAME_MAX); |
1041 | memset(hostname, 0, HOST_NAME_MAX); | |
c5b6f4f0 DG |
1042 | |
1043 | memset(&reply, 0, sizeof(reply)); | |
1044 | ||
58eb9381 | 1045 | switch (conn->minor) { |
2a174661 DG |
1046 | case 1: |
1047 | case 2: | |
1048 | case 3: | |
1049 | break; | |
1050 | case 4: /* LTTng sessiond 2.4 */ | |
1051 | default: | |
9a78c92e MD |
1052 | ret = cmd_create_session_2_4(conn, session_name, |
1053 | hostname, &live_timer, &snapshot); | |
1054 | } | |
1055 | if (ret < 0) { | |
1056 | goto send_reply; | |
d3e2ba59 JD |
1057 | } |
1058 | ||
9a78c92e MD |
1059 | session = session_create(session_name, hostname, live_timer, |
1060 | snapshot, conn->major, conn->minor); | |
1061 | if (!session) { | |
1062 | ret = -1; | |
1063 | goto send_reply; | |
1064 | } | |
1065 | assert(!conn->session); | |
1066 | conn->session = session; | |
c5b6f4f0 DG |
1067 | DBG("Created session %" PRIu64, session->id); |
1068 | ||
9a78c92e MD |
1069 | reply.session_id = htobe64(session->id); |
1070 | ||
1071 | send_reply: | |
c5b6f4f0 DG |
1072 | if (ret < 0) { |
1073 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1074 | } else { | |
1075 | reply.ret_code = htobe32(LTTNG_OK); | |
1076 | } | |
1077 | ||
58eb9381 | 1078 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1079 | if (send_ret < 0) { |
1080 | ERR("Relayd sending session id"); | |
4169f5ad | 1081 | ret = send_ret; |
c5b6f4f0 DG |
1082 | } |
1083 | ||
1084 | return ret; | |
1085 | } | |
1086 | ||
a4baae1b JD |
1087 | /* |
1088 | * When we have received all the streams and the metadata for a channel, | |
1089 | * we make them visible to the viewer threads. | |
1090 | */ | |
9a78c92e | 1091 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1092 | { |
9a78c92e MD |
1093 | struct relay_stream *stream; |
1094 | struct relay_session *session = conn->session; | |
a4baae1b | 1095 | |
9a78c92e MD |
1096 | /* |
1097 | * We publish all streams belonging to a session atomically wrt | |
1098 | * session lock. | |
1099 | */ | |
1100 | pthread_mutex_lock(&session->lock); | |
1101 | rcu_read_lock(); | |
1102 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1103 | recv_node) { | |
1104 | stream_publish(stream); | |
a4baae1b | 1105 | } |
9a78c92e | 1106 | rcu_read_unlock(); |
a4baae1b | 1107 | |
9a78c92e MD |
1108 | /* |
1109 | * Inform the viewer that there are new streams in the session. | |
1110 | */ | |
1111 | if (session->viewer_attached) { | |
1112 | uatomic_set(&session->new_streams, 1); | |
1113 | } | |
1114 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1115 | } |
1116 | ||
b8aa1682 JD |
1117 | /* |
1118 | * relay_add_stream: allocate a new stream for a session | |
1119 | */ | |
9a78c92e | 1120 | static int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1121 | struct relay_connection *conn) |
b8aa1682 | 1122 | { |
9a78c92e MD |
1123 | int ret; |
1124 | ssize_t send_ret; | |
58eb9381 | 1125 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1126 | struct relay_stream *stream = NULL; |
1127 | struct lttcomm_relayd_status_stream reply; | |
9a78c92e MD |
1128 | struct ctf_trace *trace = NULL; |
1129 | uint64_t stream_handle = -1ULL; | |
1130 | char *path_name = NULL, *channel_name = NULL; | |
1131 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1132 | |
58eb9381 | 1133 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1134 | ERR("Trying to add a stream before version check"); |
1135 | ret = -1; | |
1136 | goto end_no_session; | |
1137 | } | |
1138 | ||
9a78c92e MD |
1139 | switch (session->minor) { |
1140 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
1141 | ret = cmd_recv_stream_2_1(conn, &path_name, | |
1142 | &channel_name); | |
0f907de1 | 1143 | break; |
9a78c92e | 1144 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1145 | default: |
9a78c92e MD |
1146 | ret = cmd_recv_stream_2_2(conn, &path_name, |
1147 | &channel_name, &tracefile_size, &tracefile_count); | |
0f907de1 JD |
1148 | break; |
1149 | } | |
1150 | if (ret < 0) { | |
9a78c92e | 1151 | goto send_reply; |
0f907de1 JD |
1152 | } |
1153 | ||
9a78c92e | 1154 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1155 | if (!trace) { |
9a78c92e | 1156 | goto send_reply; |
2a174661 | 1157 | } |
9a78c92e | 1158 | /* This stream here has one reference on the trace. */ |
2a174661 | 1159 | |
9a78c92e MD |
1160 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1161 | stream_handle = ++last_relay_stream_id; | |
1162 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1163 | |
9a78c92e MD |
1164 | /* We pass ownership of path_name and channel_name. */ |
1165 | stream = stream_create(trace, stream_handle, path_name, | |
1166 | channel_name, tracefile_size, tracefile_count); | |
1167 | path_name = NULL; | |
1168 | channel_name = NULL; | |
a4baae1b | 1169 | |
2a174661 | 1170 | /* |
9a78c92e MD |
1171 | * Streams are the owners of their trace. Reference to trace is |
1172 | * kept within stream_create(). | |
2a174661 | 1173 | */ |
9a78c92e | 1174 | ctf_trace_put(trace); |
d3e2ba59 | 1175 | |
9a78c92e | 1176 | send_reply: |
53efb85a | 1177 | memset(&reply, 0, sizeof(reply)); |
9a78c92e MD |
1178 | reply.handle = htobe64(stream_handle); |
1179 | if (!stream) { | |
f73fabfd | 1180 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1181 | } else { |
f73fabfd | 1182 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1183 | } |
5af40280 | 1184 | |
58eb9381 | 1185 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1186 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1187 | if (send_ret < 0) { | |
1188 | ERR("Relay sending stream id"); | |
9a78c92e | 1189 | ret = (int) send_ret; |
b8aa1682 JD |
1190 | } |
1191 | ||
1192 | end_no_session: | |
9a78c92e MD |
1193 | free(path_name); |
1194 | free(channel_name); | |
0f907de1 | 1195 | return ret; |
b8aa1682 JD |
1196 | } |
1197 | ||
173af62f DG |
1198 | /* |
1199 | * relay_close_stream: close a specific stream | |
1200 | */ | |
9a78c92e | 1201 | static int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1202 | struct relay_connection *conn) |
173af62f | 1203 | { |
94d49140 | 1204 | int ret, send_ret; |
58eb9381 | 1205 | struct relay_session *session = conn->session; |
173af62f DG |
1206 | struct lttcomm_relayd_close_stream stream_info; |
1207 | struct lttcomm_relayd_generic_reply reply; | |
1208 | struct relay_stream *stream; | |
173af62f DG |
1209 | |
1210 | DBG("Close stream received"); | |
1211 | ||
58eb9381 | 1212 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1213 | ERR("Trying to close a stream before version check"); |
1214 | ret = -1; | |
1215 | goto end_no_session; | |
1216 | } | |
1217 | ||
58eb9381 | 1218 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1219 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1220 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1221 | if (ret == 0) { |
1222 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1223 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1224 | } else { |
1225 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1226 | } | |
173af62f DG |
1227 | ret = -1; |
1228 | goto end_no_session; | |
1229 | } | |
1230 | ||
9a78c92e | 1231 | stream = stream_get_by_id(be64toh(stream_info.stream_id)); |
173af62f DG |
1232 | if (!stream) { |
1233 | ret = -1; | |
9a78c92e | 1234 | goto end; |
173af62f | 1235 | } |
9a78c92e MD |
1236 | pthread_mutex_lock(&stream->lock); |
1237 | stream->closed = true; | |
8e2583a4 | 1238 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
9a78c92e MD |
1239 | if (stream->is_metadata) { |
1240 | struct relay_viewer_stream *vstream; | |
173af62f | 1241 | |
9a78c92e MD |
1242 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1243 | if (vstream) { | |
1244 | if (vstream->metadata_sent == stream->metadata_received) { | |
1245 | /* | |
1246 | * Since all the metadata has been sent to the | |
1247 | * viewer and that we have a request to close | |
1248 | * its stream, we can safely teardown the | |
1249 | * corresponding metadata viewer stream. | |
1250 | */ | |
1251 | viewer_stream_put(vstream); | |
1252 | } | |
1253 | /* Put local reference. */ | |
1254 | viewer_stream_put(vstream); | |
1255 | } | |
1256 | } | |
1257 | pthread_mutex_unlock(&stream->lock); | |
1258 | stream_put(stream); | |
173af62f | 1259 | |
9a78c92e | 1260 | end: |
53efb85a | 1261 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1262 | if (ret < 0) { |
f73fabfd | 1263 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1264 | } else { |
f73fabfd | 1265 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1266 | } |
58eb9381 | 1267 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1268 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1269 | if (send_ret < 0) { | |
1270 | ERR("Relay sending stream id"); | |
4169f5ad | 1271 | ret = send_ret; |
173af62f DG |
1272 | } |
1273 | ||
1274 | end_no_session: | |
1275 | return ret; | |
1276 | } | |
1277 | ||
b8aa1682 JD |
1278 | /* |
1279 | * relay_unknown_command: send -1 if received unknown command | |
1280 | */ | |
9a78c92e | 1281 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1282 | { |
1283 | struct lttcomm_relayd_generic_reply reply; | |
1284 | int ret; | |
1285 | ||
53efb85a | 1286 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1287 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
58eb9381 | 1288 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1289 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1290 | if (ret < 0) { | |
1291 | ERR("Relay sending unknown command"); | |
1292 | } | |
1293 | } | |
1294 | ||
1295 | /* | |
1296 | * relay_start: send an acknowledgment to the client to tell if we are | |
1297 | * ready to receive data. We are ready if a session is established. | |
1298 | */ | |
9a78c92e | 1299 | static int relay_start(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1300 | struct relay_connection *conn) |
b8aa1682 | 1301 | { |
f73fabfd | 1302 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1303 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1304 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1305 | |
1306 | if (!session) { | |
1307 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1308 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1309 | } |
1310 | ||
53efb85a | 1311 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1312 | reply.ret_code = ret; |
58eb9381 | 1313 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1314 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1315 | if (ret < 0) { | |
1316 | ERR("Relay sending start ack"); | |
1317 | } | |
1318 | ||
1319 | return ret; | |
1320 | } | |
1321 | ||
1d4dfdef DG |
1322 | /* |
1323 | * Append padding to the file pointed by the file descriptor fd. | |
1324 | */ | |
1325 | static int write_padding_to_file(int fd, uint32_t size) | |
1326 | { | |
6cd525e8 | 1327 | ssize_t ret = 0; |
1d4dfdef DG |
1328 | char *zeros; |
1329 | ||
1330 | if (size == 0) { | |
1331 | goto end; | |
1332 | } | |
1333 | ||
1334 | zeros = zmalloc(size); | |
1335 | if (zeros == NULL) { | |
1336 | PERROR("zmalloc zeros for padding"); | |
1337 | ret = -1; | |
1338 | goto end; | |
1339 | } | |
1340 | ||
6cd525e8 MD |
1341 | ret = lttng_write(fd, zeros, size); |
1342 | if (ret < size) { | |
1d4dfdef DG |
1343 | PERROR("write padding to file"); |
1344 | } | |
1345 | ||
e986c7a1 DG |
1346 | free(zeros); |
1347 | ||
1d4dfdef DG |
1348 | end: |
1349 | return ret; | |
1350 | } | |
1351 | ||
b8aa1682 | 1352 | /* |
9a78c92e | 1353 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1354 | */ |
9a78c92e | 1355 | static int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1356 | struct relay_connection *conn) |
b8aa1682 | 1357 | { |
f73fabfd | 1358 | int ret = htobe32(LTTNG_OK); |
6cd525e8 | 1359 | ssize_t size_ret; |
58eb9381 | 1360 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1361 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1362 | struct relay_stream *metadata_stream; | |
1363 | uint64_t data_size, payload_size; | |
1364 | ||
1365 | if (!session) { | |
1366 | ERR("Metadata sent before version check"); | |
1367 | ret = -1; | |
1368 | goto end; | |
1369 | } | |
1370 | ||
f6416125 MD |
1371 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1372 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1373 | ERR("Incorrect data size"); | |
1374 | ret = -1; | |
1375 | goto end; | |
1376 | } | |
1377 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1378 | ||
b8aa1682 | 1379 | if (data_buffer_size < data_size) { |
d7b3776f | 1380 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1381 | char *tmp_data_ptr; |
1382 | ||
1383 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1384 | if (!tmp_data_ptr) { | |
b8aa1682 | 1385 | ERR("Allocating data buffer"); |
c617c0c6 | 1386 | free(data_buffer); |
b8aa1682 JD |
1387 | ret = -1; |
1388 | goto end; | |
1389 | } | |
c617c0c6 | 1390 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1391 | data_buffer_size = data_size; |
1392 | } | |
1393 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1394 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
58eb9381 | 1395 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 1396 | if (ret < 0 || ret != data_size) { |
a6cd2b97 DG |
1397 | if (ret == 0) { |
1398 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1399 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1400 | } else { |
1401 | ERR("Relay didn't receive the whole metadata"); | |
1402 | } | |
b8aa1682 | 1403 | ret = -1; |
b8aa1682 JD |
1404 | goto end; |
1405 | } | |
1406 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 | 1407 | |
9a78c92e | 1408 | metadata_stream = stream_get_by_id(be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1409 | if (!metadata_stream) { |
1410 | ret = -1; | |
9a78c92e | 1411 | goto end; |
b8aa1682 JD |
1412 | } |
1413 | ||
9a78c92e MD |
1414 | pthread_mutex_lock(&metadata_stream->lock); |
1415 | ||
1416 | size_ret = lttng_write(metadata_stream->stream_fd->fd, metadata_struct->payload, | |
6cd525e8 MD |
1417 | payload_size); |
1418 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1419 | ERR("Relay error writing metadata on file"); |
1420 | ret = -1; | |
9a78c92e | 1421 | goto end_put; |
b8aa1682 | 1422 | } |
1d4dfdef | 1423 | |
9a78c92e | 1424 | ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
1d4dfdef DG |
1425 | be32toh(metadata_struct->padding_size)); |
1426 | if (ret < 0) { | |
9a78c92e | 1427 | goto end_put; |
1d4dfdef | 1428 | } |
2a174661 | 1429 | |
9a78c92e | 1430 | metadata_stream->metadata_received += |
d3e2ba59 | 1431 | payload_size + be32toh(metadata_struct->padding_size); |
9a78c92e MD |
1432 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1433 | metadata_stream->metadata_received); | |
1d4dfdef | 1434 | |
9a78c92e MD |
1435 | end_put: |
1436 | pthread_mutex_unlock(&metadata_stream->lock); | |
1437 | stream_put(metadata_stream); | |
b8aa1682 JD |
1438 | |
1439 | end: | |
1440 | return ret; | |
1441 | } | |
1442 | ||
1443 | /* | |
1444 | * relay_send_version: send relayd version number | |
1445 | */ | |
9a78c92e | 1446 | static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1447 | struct relay_connection *conn) |
b8aa1682 | 1448 | { |
7f51dcba | 1449 | int ret; |
092b6259 | 1450 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1451 | |
58eb9381 | 1452 | conn->version_check_done = 1; |
b8aa1682 | 1453 | |
092b6259 | 1454 | /* Get version from the other side. */ |
58eb9381 | 1455 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1456 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1457 | if (ret == 0) { |
1458 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1459 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1460 | } else { |
1461 | ERR("Relay failed to receive the version values."); | |
1462 | } | |
092b6259 | 1463 | ret = -1; |
092b6259 DG |
1464 | goto end; |
1465 | } | |
1466 | ||
53efb85a | 1467 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1468 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1469 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1470 | |
1471 | /* Major versions must be the same */ | |
1472 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1473 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1474 | reply.major, be32toh(msg.major)); | |
9a78c92e | 1475 | connection_put(conn); |
d4519fa3 JD |
1476 | ret = 0; |
1477 | goto end; | |
1478 | } | |
1479 | ||
58eb9381 | 1480 | conn->major = reply.major; |
0f907de1 JD |
1481 | /* We adapt to the lowest compatible version */ |
1482 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1483 | conn->minor = reply.minor; |
0f907de1 | 1484 | } else { |
58eb9381 | 1485 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1486 | } |
1487 | ||
6151a90f JD |
1488 | reply.major = htobe32(reply.major); |
1489 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1490 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1491 | sizeof(struct lttcomm_relayd_version), 0); |
1492 | if (ret < 0) { | |
1493 | ERR("Relay sending version"); | |
1494 | } | |
1495 | ||
58eb9381 DG |
1496 | DBG("Version check done using protocol %u.%u", conn->major, |
1497 | conn->minor); | |
b8aa1682 JD |
1498 | |
1499 | end: | |
1500 | return ret; | |
1501 | } | |
1502 | ||
c8f59ee5 | 1503 | /* |
6d805429 | 1504 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1505 | */ |
9a78c92e | 1506 | static int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1507 | struct relay_connection *conn) |
c8f59ee5 | 1508 | { |
58eb9381 | 1509 | struct relay_session *session = conn->session; |
6d805429 | 1510 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1511 | struct lttcomm_relayd_generic_reply reply; |
1512 | struct relay_stream *stream; | |
1513 | int ret; | |
c8f59ee5 DG |
1514 | uint64_t last_net_seq_num, stream_id; |
1515 | ||
6d805429 | 1516 | DBG("Data pending command received"); |
c8f59ee5 | 1517 | |
58eb9381 | 1518 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1519 | ERR("Trying to check for data before version check"); |
1520 | ret = -1; | |
1521 | goto end_no_session; | |
1522 | } | |
1523 | ||
58eb9381 | 1524 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1525 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1526 | if (ret == 0) { |
1527 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1528 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1529 | } else { |
1530 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1531 | ret); | |
1532 | } | |
c8f59ee5 DG |
1533 | ret = -1; |
1534 | goto end_no_session; | |
1535 | } | |
1536 | ||
1537 | stream_id = be64toh(msg.stream_id); | |
1538 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1539 | ||
9a78c92e | 1540 | stream = stream_get_by_id(stream_id); |
de91f48a | 1541 | if (stream == NULL) { |
c8f59ee5 | 1542 | ret = -1; |
9a78c92e | 1543 | goto end; |
c8f59ee5 DG |
1544 | } |
1545 | ||
9a78c92e MD |
1546 | pthread_mutex_lock(&stream->lock); |
1547 | ||
6d805429 | 1548 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1549 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1550 | last_net_seq_num); | |
1551 | ||
33832e64 | 1552 | /* Avoid wrapping issue */ |
39df6d9f | 1553 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1554 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1555 | ret = 0; |
6d805429 DG |
1556 | } else { |
1557 | /* Data still being streamed thus pending */ | |
1558 | ret = 1; | |
c8f59ee5 DG |
1559 | } |
1560 | ||
9a78c92e MD |
1561 | stream->data_pending_check_done = true; |
1562 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1563 | |
9a78c92e MD |
1564 | stream_put(stream); |
1565 | end: | |
c8f59ee5 | 1566 | |
53efb85a | 1567 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1568 | reply.ret_code = htobe32(ret); |
58eb9381 | 1569 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1570 | if (ret < 0) { |
6d805429 | 1571 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1572 | } |
1573 | ||
1574 | end_no_session: | |
1575 | return ret; | |
1576 | } | |
1577 | ||
1578 | /* | |
1579 | * Wait for the control socket to reach a quiescent state. | |
1580 | * | |
9a78c92e MD |
1581 | * Note that for now, when receiving this command from the session |
1582 | * daemon, this means that every subsequent commands or data received on | |
1583 | * the control socket has been handled. So, this is why we simply return | |
1584 | * OK here. | |
c8f59ee5 | 1585 | */ |
9a78c92e | 1586 | static int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1587 | struct relay_connection *conn) |
c8f59ee5 DG |
1588 | { |
1589 | int ret; | |
ad7051c0 DG |
1590 | uint64_t stream_id; |
1591 | struct relay_stream *stream; | |
ad7051c0 | 1592 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1593 | struct lttcomm_relayd_generic_reply reply; |
1594 | ||
1595 | DBG("Checking quiescent state on control socket"); | |
1596 | ||
58eb9381 | 1597 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1598 | ERR("Trying to check for data before version check"); |
1599 | ret = -1; | |
1600 | goto end_no_session; | |
1601 | } | |
1602 | ||
58eb9381 | 1603 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1604 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1605 | if (ret == 0) { |
1606 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1607 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1608 | } else { |
1609 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1610 | ret); | |
1611 | } | |
ad7051c0 DG |
1612 | ret = -1; |
1613 | goto end_no_session; | |
1614 | } | |
1615 | ||
1616 | stream_id = be64toh(msg.stream_id); | |
9a78c92e MD |
1617 | stream = stream_get_by_id(stream_id); |
1618 | if (!stream) { | |
1619 | goto reply; | |
1620 | } | |
1621 | pthread_mutex_lock(&stream->lock); | |
1622 | stream->data_pending_check_done = true; | |
1623 | pthread_mutex_unlock(&stream->lock); | |
1624 | DBG("Relay quiescent control pending flag set to %" PRIu64, stream_id); | |
1625 | stream_put(stream); | |
1626 | reply: | |
53efb85a | 1627 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1628 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1629 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1630 | if (ret < 0) { |
6d805429 | 1631 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1632 | } |
1633 | ||
ad7051c0 | 1634 | end_no_session: |
c8f59ee5 DG |
1635 | return ret; |
1636 | } | |
1637 | ||
f7079f67 | 1638 | /* |
9a78c92e MD |
1639 | * Initialize a data pending command. This means that a consumer is about |
1640 | * to ask for data pending for each stream it holds. Simply iterate over | |
1641 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1642 | * |
1643 | * This command returns to the client a LTTNG_OK code. | |
1644 | */ | |
9a78c92e | 1645 | static int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1646 | struct relay_connection *conn) |
f7079f67 DG |
1647 | { |
1648 | int ret; | |
1649 | struct lttng_ht_iter iter; | |
1650 | struct lttcomm_relayd_begin_data_pending msg; | |
1651 | struct lttcomm_relayd_generic_reply reply; | |
1652 | struct relay_stream *stream; | |
1653 | uint64_t session_id; | |
1654 | ||
1655 | assert(recv_hdr); | |
58eb9381 | 1656 | assert(conn); |
f7079f67 DG |
1657 | |
1658 | DBG("Init streams for data pending"); | |
1659 | ||
58eb9381 | 1660 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1661 | ERR("Trying to check for data before version check"); |
1662 | ret = -1; | |
1663 | goto end_no_session; | |
1664 | } | |
1665 | ||
58eb9381 | 1666 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1667 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1668 | if (ret == 0) { |
1669 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1670 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1671 | } else { |
1672 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1673 | ret); | |
1674 | } | |
f7079f67 DG |
1675 | ret = -1; |
1676 | goto end_no_session; | |
1677 | } | |
1678 | ||
1679 | session_id = be64toh(msg.session_id); | |
1680 | ||
1681 | /* | |
9a78c92e MD |
1682 | * Iterate over all streams to set the begin data pending flag. |
1683 | * For now, the streams are indexed by stream handle so we have | |
1684 | * to iterate over all streams to find the one associated with | |
1685 | * the right session_id. | |
f7079f67 DG |
1686 | */ |
1687 | rcu_read_lock(); | |
d3e2ba59 | 1688 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1689 | node.node) { |
9a78c92e MD |
1690 | if (!stream_get(stream)) { |
1691 | continue; | |
1692 | } | |
1693 | if (stream->trace->session->id == session_id) { | |
1694 | pthread_mutex_lock(&stream->lock); | |
1695 | stream->data_pending_check_done = false; | |
1696 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1697 | DBG("Set begin data pending flag to stream %" PRIu64, |
1698 | stream->stream_handle); | |
1699 | } | |
9a78c92e | 1700 | stream_put(stream); |
f7079f67 DG |
1701 | } |
1702 | rcu_read_unlock(); | |
1703 | ||
53efb85a | 1704 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1705 | /* All good, send back reply. */ |
1706 | reply.ret_code = htobe32(LTTNG_OK); | |
1707 | ||
58eb9381 | 1708 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1709 | if (ret < 0) { |
1710 | ERR("Relay begin data pending send reply failed"); | |
1711 | } | |
1712 | ||
1713 | end_no_session: | |
1714 | return ret; | |
1715 | } | |
1716 | ||
1717 | /* | |
9a78c92e MD |
1718 | * End data pending command. This will check, for a given session id, if |
1719 | * each stream associated with it has its data_pending_check_done flag | |
1720 | * set. If not, this means that the client lost track of the stream but | |
1721 | * the data is still being streamed on our side. In this case, we inform | |
1722 | * the client that data is in flight. | |
f7079f67 DG |
1723 | * |
1724 | * Return to the client if there is data in flight or not with a ret_code. | |
1725 | */ | |
9a78c92e | 1726 | static int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1727 | struct relay_connection *conn) |
f7079f67 DG |
1728 | { |
1729 | int ret; | |
1730 | struct lttng_ht_iter iter; | |
1731 | struct lttcomm_relayd_end_data_pending msg; | |
1732 | struct lttcomm_relayd_generic_reply reply; | |
1733 | struct relay_stream *stream; | |
1734 | uint64_t session_id; | |
1735 | uint32_t is_data_inflight = 0; | |
1736 | ||
f7079f67 DG |
1737 | DBG("End data pending command"); |
1738 | ||
58eb9381 | 1739 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1740 | ERR("Trying to check for data before version check"); |
1741 | ret = -1; | |
1742 | goto end_no_session; | |
1743 | } | |
1744 | ||
58eb9381 | 1745 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1746 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1747 | if (ret == 0) { |
1748 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1749 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1750 | } else { |
1751 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1752 | ret); | |
1753 | } | |
f7079f67 DG |
1754 | ret = -1; |
1755 | goto end_no_session; | |
1756 | } | |
1757 | ||
1758 | session_id = be64toh(msg.session_id); | |
1759 | ||
9a78c92e MD |
1760 | /* |
1761 | * Iterate over all streams to see if the begin data pending | |
1762 | * flag is set. | |
1763 | */ | |
f7079f67 | 1764 | rcu_read_lock(); |
d3e2ba59 | 1765 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1766 | node.node) { |
9a78c92e MD |
1767 | if (!stream_get(stream)) { |
1768 | continue; | |
1769 | } | |
1770 | if (stream->trace->session->id != session_id) { | |
1771 | stream_put(stream); | |
1772 | continue; | |
1773 | } | |
1774 | pthread_mutex_lock(&stream->lock); | |
1775 | if (!stream->data_pending_check_done) { | |
1776 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1777 | is_data_inflight = 1; | |
1778 | DBG("Data is still in flight for stream %" PRIu64, | |
1779 | stream->stream_handle); | |
1780 | pthread_mutex_unlock(&stream->lock); | |
1781 | stream_put(stream); | |
1782 | break; | |
1783 | } | |
f7079f67 | 1784 | } |
9a78c92e MD |
1785 | pthread_mutex_unlock(&stream->lock); |
1786 | stream_put(stream); | |
f7079f67 DG |
1787 | } |
1788 | rcu_read_unlock(); | |
1789 | ||
53efb85a | 1790 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1791 | /* All good, send back reply. */ |
1792 | reply.ret_code = htobe32(is_data_inflight); | |
1793 | ||
58eb9381 | 1794 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1795 | if (ret < 0) { |
1796 | ERR("Relay end data pending send reply failed"); | |
1797 | } | |
1798 | ||
1799 | end_no_session: | |
1800 | return ret; | |
1801 | } | |
1802 | ||
1c20f0e2 JD |
1803 | /* |
1804 | * Receive an index for a specific stream. | |
1805 | * | |
1806 | * Return 0 on success else a negative value. | |
1807 | */ | |
9a78c92e | 1808 | static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1809 | struct relay_connection *conn) |
1c20f0e2 | 1810 | { |
9a78c92e | 1811 | int ret, send_ret; |
58eb9381 | 1812 | struct relay_session *session = conn->session; |
1c20f0e2 | 1813 | struct lttcomm_relayd_index index_info; |
9a78c92e | 1814 | struct relay_index *index; |
1c20f0e2 JD |
1815 | struct lttcomm_relayd_generic_reply reply; |
1816 | struct relay_stream *stream; | |
1817 | uint64_t net_seq_num; | |
1818 | ||
58eb9381 | 1819 | assert(conn); |
1c20f0e2 JD |
1820 | |
1821 | DBG("Relay receiving index"); | |
1822 | ||
58eb9381 | 1823 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1824 | ERR("Trying to close a stream before version check"); |
1825 | ret = -1; | |
1826 | goto end_no_session; | |
1827 | } | |
1828 | ||
58eb9381 | 1829 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1830 | sizeof(index_info), 0); |
1831 | if (ret < sizeof(index_info)) { | |
1832 | if (ret == 0) { | |
1833 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1834 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1835 | } else { |
1836 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1837 | } | |
1838 | ret = -1; | |
1839 | goto end_no_session; | |
1840 | } | |
1841 | ||
1842 | net_seq_num = be64toh(index_info.net_seq_num); | |
1843 | ||
9a78c92e | 1844 | stream = stream_get_by_id(be64toh(index_info.relay_stream_id)); |
1c20f0e2 | 1845 | if (!stream) { |
9a78c92e | 1846 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 1847 | ret = -1; |
9a78c92e | 1848 | goto end; |
1c20f0e2 | 1849 | } |
9a78c92e | 1850 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 1851 | |
d3e2ba59 JD |
1852 | /* Live beacon handling */ |
1853 | if (index_info.packet_size == 0) { | |
9a78c92e MD |
1854 | DBG("Received live beacon for stream %" PRIu64, |
1855 | stream->stream_handle); | |
d3e2ba59 JD |
1856 | |
1857 | /* | |
9a78c92e MD |
1858 | * Only flag a stream inactive when it has already |
1859 | * received data and no indexes are in flight. | |
d3e2ba59 | 1860 | */ |
9a78c92e MD |
1861 | if (stream->total_index_received > 0 |
1862 | && stream->indexes_in_flight == 0) { | |
1863 | stream->beacon_ts_end = | |
1864 | be64toh(index_info.timestamp_end); | |
d3e2ba59 JD |
1865 | } |
1866 | ret = 0; | |
9a78c92e | 1867 | goto end_stream_put; |
d3e2ba59 JD |
1868 | } else { |
1869 | stream->beacon_ts_end = -1ULL; | |
1870 | } | |
1871 | ||
528f2ffa JD |
1872 | if (stream->ctf_stream_id == -1ULL) { |
1873 | stream->ctf_stream_id = be64toh(index_info.stream_id); | |
1874 | } | |
9a78c92e MD |
1875 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
1876 | if (!index) { | |
1877 | ret = -1; | |
1878 | ERR("relay_index_get_by_id_or_create index NULL"); | |
1879 | goto end_stream_put; | |
1c20f0e2 | 1880 | } |
9a78c92e MD |
1881 | if (set_index_control_data(index, &index_info)) { |
1882 | ERR("set_index_control_data error"); | |
1883 | relay_index_put(index); | |
1884 | ret = -1; | |
1885 | goto end_stream_put; | |
1886 | } | |
1887 | ret = relay_index_try_flush(index); | |
1888 | if (ret == 0) { | |
d3e2ba59 | 1889 | stream->total_index_received++; |
9a78c92e MD |
1890 | } else if (ret > 0) { |
1891 | /* no flush. */ | |
1892 | ret = 0; | |
1893 | } else { | |
1894 | ERR("relay_index_try_flush error %d", ret); | |
1895 | relay_index_put(index); | |
1896 | ret = -1; | |
1c20f0e2 JD |
1897 | } |
1898 | ||
9a78c92e MD |
1899 | end_stream_put: |
1900 | pthread_mutex_unlock(&stream->lock); | |
1901 | stream_put(stream); | |
1902 | ||
1903 | end: | |
1c20f0e2 | 1904 | |
53efb85a | 1905 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
1906 | if (ret < 0) { |
1907 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1908 | } else { | |
1909 | reply.ret_code = htobe32(LTTNG_OK); | |
1910 | } | |
58eb9381 | 1911 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
1912 | if (send_ret < 0) { |
1913 | ERR("Relay sending close index id reply"); | |
1914 | ret = send_ret; | |
1915 | } | |
1916 | ||
1917 | end_no_session: | |
1918 | return ret; | |
1919 | } | |
1920 | ||
a4baae1b JD |
1921 | /* |
1922 | * Receive the streams_sent message. | |
1923 | * | |
1924 | * Return 0 on success else a negative value. | |
1925 | */ | |
9a78c92e | 1926 | static int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1927 | struct relay_connection *conn) |
a4baae1b JD |
1928 | { |
1929 | int ret, send_ret; | |
1930 | struct lttcomm_relayd_generic_reply reply; | |
1931 | ||
58eb9381 | 1932 | assert(conn); |
a4baae1b JD |
1933 | |
1934 | DBG("Relay receiving streams_sent"); | |
1935 | ||
58eb9381 | 1936 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
1937 | ERR("Trying to close a stream before version check"); |
1938 | ret = -1; | |
1939 | goto end_no_session; | |
1940 | } | |
1941 | ||
1942 | /* | |
9a78c92e MD |
1943 | * Publish every pending stream in the connection recv list which are |
1944 | * now ready to be used by the viewer. | |
a4baae1b | 1945 | */ |
9a78c92e | 1946 | publish_connection_local_streams(conn); |
4a9daf17 | 1947 | |
53efb85a | 1948 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 1949 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1950 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
1951 | if (send_ret < 0) { |
1952 | ERR("Relay sending sent_stream reply"); | |
1953 | ret = send_ret; | |
1954 | } else { | |
1955 | /* Success. */ | |
1956 | ret = 0; | |
1957 | } | |
1958 | ||
1959 | end_no_session: | |
1960 | return ret; | |
1961 | } | |
1962 | ||
b8aa1682 | 1963 | /* |
d3e2ba59 | 1964 | * Process the commands received on the control socket |
b8aa1682 | 1965 | */ |
9a78c92e | 1966 | static int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1967 | struct relay_connection *conn) |
b8aa1682 JD |
1968 | { |
1969 | int ret = 0; | |
1970 | ||
1971 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 1972 | case RELAYD_CREATE_SESSION: |
58eb9381 | 1973 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 1974 | break; |
b8aa1682 | 1975 | case RELAYD_ADD_STREAM: |
58eb9381 | 1976 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
1977 | break; |
1978 | case RELAYD_START_DATA: | |
58eb9381 | 1979 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
1980 | break; |
1981 | case RELAYD_SEND_METADATA: | |
58eb9381 | 1982 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
1983 | break; |
1984 | case RELAYD_VERSION: | |
58eb9381 | 1985 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 1986 | break; |
173af62f | 1987 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 1988 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 1989 | break; |
6d805429 | 1990 | case RELAYD_DATA_PENDING: |
58eb9381 | 1991 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
1992 | break; |
1993 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 1994 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 1995 | break; |
f7079f67 | 1996 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 1997 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
1998 | break; |
1999 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2000 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2001 | break; |
1c20f0e2 | 2002 | case RELAYD_SEND_INDEX: |
58eb9381 | 2003 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2004 | break; |
a4baae1b | 2005 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2006 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2007 | break; |
b8aa1682 JD |
2008 | case RELAYD_UPDATE_SYNC_INFO: |
2009 | default: | |
2010 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2011 | relay_unknown_command(conn); |
b8aa1682 JD |
2012 | ret = -1; |
2013 | goto end; | |
2014 | } | |
2015 | ||
2016 | end: | |
2017 | return ret; | |
2018 | } | |
2019 | ||
7d2f7452 DG |
2020 | /* |
2021 | * Handle index for a data stream. | |
2022 | * | |
9a78c92e | 2023 | * Called with the stream lock held. |
7d2f7452 DG |
2024 | * |
2025 | * Return 0 on success else a negative value. | |
2026 | */ | |
2027 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2028 | int rotate_index) | |
2029 | { | |
9a78c92e MD |
2030 | int ret = 0; |
2031 | uint64_t data_offset; | |
2032 | struct relay_index *index; | |
7d2f7452 | 2033 | |
7d2f7452 DG |
2034 | /* Get data offset because we are about to update the index. */ |
2035 | data_offset = htobe64(stream->tracefile_size_current); | |
2036 | ||
9a78c92e MD |
2037 | DBG("handle_index_data: stream %" PRIu64 " data offset %" PRIu64, |
2038 | stream->stream_handle, stream->tracefile_size_current); | |
2039 | ||
7d2f7452 | 2040 | /* |
9a78c92e MD |
2041 | * Lookup for an existing index for that stream id/sequence |
2042 | * number. If it exists, the control thread has already received the | |
2043 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2044 | */ |
9a78c92e | 2045 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2046 | if (!index) { |
9a78c92e MD |
2047 | ret = -1; |
2048 | goto end; | |
7d2f7452 DG |
2049 | } |
2050 | ||
9a78c92e MD |
2051 | if (rotate_index || !stream->index_fd) { |
2052 | int fd; | |
2053 | ||
2054 | /* Put ref on previous index_fd. */ | |
2055 | if (stream->index_fd) { | |
2056 | stream_fd_put(stream->index_fd); | |
2057 | stream->index_fd = NULL; | |
7d2f7452 | 2058 | } |
7d2f7452 | 2059 | |
9a78c92e MD |
2060 | fd = index_create_file(stream->path_name, stream->channel_name, |
2061 | -1, -1, stream->tracefile_size, | |
2062 | stream->current_tracefile_id); | |
2063 | if (fd < 0) { | |
2064 | ret = -1; | |
2065 | /* Put self-ref for this index due to error. */ | |
2066 | relay_index_put(index); | |
2067 | goto end; | |
2068 | } | |
2069 | stream->index_fd = stream_fd_create(fd); | |
2070 | if (!stream->index_fd) { | |
2071 | ret = -1; | |
2072 | if (close(fd)) { | |
2073 | PERROR("Error closing FD %d", fd); | |
2074 | } | |
2075 | /* Put self-ref for this index due to error. */ | |
2076 | relay_index_put(index); | |
2077 | /* Will put the local ref. */ | |
2078 | goto end; | |
7d2f7452 | 2079 | } |
7d2f7452 DG |
2080 | } |
2081 | ||
9a78c92e MD |
2082 | if (relay_index_set_fd(index, stream->index_fd, data_offset)) { |
2083 | ret = -1; | |
2084 | /* Put self-ref for this index due to error. */ | |
2085 | relay_index_put(index); | |
2086 | goto end; | |
7d2f7452 DG |
2087 | } |
2088 | ||
9a78c92e MD |
2089 | ret = relay_index_try_flush(index); |
2090 | if (ret == 0) { | |
2091 | stream->total_index_received++; | |
2092 | } else if (ret > 0) { | |
2093 | /* No flush. */ | |
2094 | ret = 0; | |
2095 | } else { | |
2096 | /* Put self-ref for this index due to error. */ | |
2097 | relay_index_put(index); | |
2098 | ret = -1; | |
2099 | } | |
2100 | end: | |
7d2f7452 DG |
2101 | return ret; |
2102 | } | |
2103 | ||
b8aa1682 JD |
2104 | /* |
2105 | * relay_process_data: Process the data received on the data socket | |
2106 | */ | |
9a78c92e | 2107 | static int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2108 | { |
7d2f7452 | 2109 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2110 | ssize_t size_ret; |
b8aa1682 JD |
2111 | struct relay_stream *stream; |
2112 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2113 | uint64_t stream_id; |
173af62f | 2114 | uint64_t net_seq_num; |
b8aa1682 | 2115 | uint32_t data_size; |
2a174661 | 2116 | struct relay_session *session; |
b8ee61ff | 2117 | bool new_stream = false; |
b8aa1682 | 2118 | |
58eb9381 | 2119 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, |
7c5aef62 | 2120 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2121 | if (ret <= 0) { |
a6cd2b97 DG |
2122 | if (ret == 0) { |
2123 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2124 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2125 | } else { |
58eb9381 | 2126 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2127 | } |
b8aa1682 JD |
2128 | ret = -1; |
2129 | goto end; | |
2130 | } | |
2131 | ||
2132 | stream_id = be64toh(data_hdr.stream_id); | |
9a78c92e | 2133 | stream = stream_get_by_id(stream_id); |
b8aa1682 JD |
2134 | if (!stream) { |
2135 | ret = -1; | |
9a78c92e | 2136 | goto end; |
b8aa1682 | 2137 | } |
9a78c92e | 2138 | session = stream->trace->session; |
b8aa1682 JD |
2139 | data_size = be32toh(data_hdr.data_size); |
2140 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2141 | char *tmp_data_ptr; |
2142 | ||
2143 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2144 | if (!tmp_data_ptr) { | |
b8aa1682 | 2145 | ERR("Allocating data buffer"); |
c617c0c6 | 2146 | free(data_buffer); |
b8aa1682 | 2147 | ret = -1; |
9a78c92e | 2148 | goto end_stream_put; |
b8aa1682 | 2149 | } |
c617c0c6 | 2150 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2151 | data_buffer_size = data_size; |
2152 | } | |
2153 | memset(data_buffer, 0, data_size); | |
2154 | ||
173af62f DG |
2155 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2156 | ||
77c7c900 | 2157 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2158 | data_size, stream_id, net_seq_num); |
58eb9381 | 2159 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 2160 | if (ret <= 0) { |
a6cd2b97 DG |
2161 | if (ret == 0) { |
2162 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2163 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2164 | } |
b8aa1682 | 2165 | ret = -1; |
9a78c92e | 2166 | goto end_stream_put; |
b8aa1682 JD |
2167 | } |
2168 | ||
9a78c92e MD |
2169 | pthread_mutex_lock(&stream->lock); |
2170 | ||
1c20f0e2 | 2171 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2172 | if (stream->tracefile_size > 0 && |
2173 | (stream->tracefile_size_current + data_size) > | |
2174 | stream->tracefile_size) { | |
6b6b9a5a JD |
2175 | uint64_t new_id; |
2176 | ||
9a78c92e | 2177 | new_id = (stream->current_tracefile_id + 1) % |
6b6b9a5a JD |
2178 | stream->tracefile_count; |
2179 | /* | |
9a78c92e MD |
2180 | * Move viewer oldest available data position forward if |
2181 | * we are overwriting a tracefile. | |
6b6b9a5a | 2182 | */ |
9a78c92e | 2183 | if (new_id == stream->oldest_tracefile_id) { |
6b6b9a5a JD |
2184 | stream->oldest_tracefile_id = |
2185 | (stream->oldest_tracefile_id + 1) % | |
2186 | stream->tracefile_count; | |
2187 | } | |
9a78c92e MD |
2188 | ret = utils_rotate_stream_file(stream->path_name, |
2189 | stream->channel_name, stream->tracefile_size, | |
2190 | stream->tracefile_count, -1, | |
2191 | -1, stream->stream_fd->fd, | |
2192 | &stream->current_tracefile_id, | |
2193 | &stream->stream_fd->fd); | |
0f907de1 | 2194 | if (ret < 0) { |
1c20f0e2 | 2195 | ERR("Rotating stream output file"); |
9a78c92e | 2196 | goto end_stream_unlock; |
0f907de1 | 2197 | } |
9a78c92e MD |
2198 | stream->current_tracefile_seq++; |
2199 | if (stream->current_tracefile_seq | |
2200 | - stream->oldest_tracefile_seq >= | |
2201 | stream->tracefile_count) { | |
2202 | stream->oldest_tracefile_seq++; | |
2203 | } | |
2204 | /* | |
2205 | * Reset current size because we just performed a stream | |
2206 | * rotation. | |
2207 | */ | |
a6976990 | 2208 | stream->tracefile_size_current = 0; |
1c20f0e2 JD |
2209 | rotate_index = 1; |
2210 | } | |
2211 | ||
1c20f0e2 | 2212 | /* |
9a78c92e MD |
2213 | * Index are handled in protocol version 2.4 and above. Also, |
2214 | * snapshot and index are NOT supported. | |
1c20f0e2 | 2215 | */ |
2a174661 | 2216 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2217 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2218 | if (ret < 0) { |
9a78c92e | 2219 | goto end_stream_unlock; |
1c20f0e2 | 2220 | } |
1c20f0e2 JD |
2221 | } |
2222 | ||
7d2f7452 | 2223 | /* Write data to stream output fd. */ |
9a78c92e | 2224 | size_ret = lttng_write(stream->stream_fd->fd, data_buffer, data_size); |
6cd525e8 | 2225 | if (size_ret < data_size) { |
b8aa1682 JD |
2226 | ERR("Relay error writing data to file"); |
2227 | ret = -1; | |
9a78c92e | 2228 | goto end_stream_unlock; |
b8aa1682 | 2229 | } |
1d4dfdef | 2230 | |
9a78c92e MD |
2231 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, |
2232 | size_ret, stream->stream_handle); | |
5ab7344e | 2233 | |
9a78c92e MD |
2234 | ret = write_padding_to_file(stream->stream_fd->fd, |
2235 | be32toh(data_hdr.padding_size)); | |
1d4dfdef | 2236 | if (ret < 0) { |
9a78c92e | 2237 | goto end_stream_unlock; |
1d4dfdef | 2238 | } |
9a78c92e MD |
2239 | stream->tracefile_size_current += |
2240 | data_size + be32toh(data_hdr.padding_size); | |
b8ee61ff MD |
2241 | if (stream->prev_seq == -1ULL) { |
2242 | new_stream = true; | |
2243 | } | |
2244 | ||
173af62f DG |
2245 | stream->prev_seq = net_seq_num; |
2246 | ||
9a78c92e MD |
2247 | end_stream_unlock: |
2248 | pthread_mutex_unlock(&stream->lock); | |
b8ee61ff MD |
2249 | if (new_stream) { |
2250 | pthread_mutex_lock(&session->lock); | |
2251 | uatomic_set(&session->new_streams, 1); | |
2252 | pthread_mutex_unlock(&session->lock); | |
2253 | } | |
9a78c92e MD |
2254 | end_stream_put: |
2255 | stream_put(stream); | |
b8aa1682 JD |
2256 | end: |
2257 | return ret; | |
2258 | } | |
2259 | ||
9a78c92e | 2260 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2261 | { |
2262 | int ret; | |
2263 | ||
58eb9381 | 2264 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2265 | |
2266 | ret = close(pollfd); | |
2267 | if (ret < 0) { | |
2268 | ERR("Closing pollfd %d", pollfd); | |
2269 | } | |
2270 | } | |
2271 | ||
9a78c92e MD |
2272 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2273 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2274 | { |
9a78c92e | 2275 | const char *type_str; |
2a174661 | 2276 | |
9a78c92e MD |
2277 | switch (conn->type) { |
2278 | case RELAY_DATA: | |
2279 | type_str = "Data"; | |
2280 | break; | |
2281 | case RELAY_CONTROL: | |
2282 | type_str = "Control"; | |
2283 | break; | |
2284 | case RELAY_VIEWER_COMMAND: | |
2285 | type_str = "Viewer Command"; | |
2286 | break; | |
2287 | case RELAY_VIEWER_NOTIFICATION: | |
2288 | type_str = "Viewer Notification"; | |
2289 | break; | |
2290 | default: | |
2291 | type_str = "Unknown"; | |
9d1bbf21 | 2292 | } |
9a78c92e MD |
2293 | cleanup_connection_pollfd(events, pollfd); |
2294 | connection_put(conn); | |
2295 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2296 | } |
2297 | ||
2298 | /* | |
2299 | * This thread does the actual work | |
2300 | */ | |
9a78c92e | 2301 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2302 | { |
beaad64c DG |
2303 | int ret, err = -1, last_seen_data_fd = -1; |
2304 | uint32_t nb_fd; | |
b8aa1682 JD |
2305 | struct lttng_poll_event events; |
2306 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2307 | struct lttng_ht_iter iter; |
b8aa1682 | 2308 | struct lttcomm_relayd_hdr recv_hdr; |
9a78c92e | 2309 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2310 | |
2311 | DBG("[thread] Relay worker started"); | |
2312 | ||
9d1bbf21 MD |
2313 | rcu_register_thread(); |
2314 | ||
55706a7d MD |
2315 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2316 | ||
9b5e0863 MD |
2317 | if (testpoint(relayd_thread_worker)) { |
2318 | goto error_testpoint; | |
2319 | } | |
2320 | ||
f385ae0a MD |
2321 | health_code_update(); |
2322 | ||
b8aa1682 JD |
2323 | /* table of connections indexed on socket */ |
2324 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2325 | if (!relay_connections_ht) { |
2326 | goto relay_connections_ht_error; | |
2327 | } | |
b8aa1682 | 2328 | |
b8aa1682 JD |
2329 | ret = create_thread_poll_set(&events, 2); |
2330 | if (ret < 0) { | |
2331 | goto error_poll_create; | |
2332 | } | |
2333 | ||
58eb9381 | 2334 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2335 | if (ret < 0) { |
2336 | goto error; | |
2337 | } | |
2338 | ||
beaad64c | 2339 | restart: |
b8aa1682 | 2340 | while (1) { |
beaad64c DG |
2341 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2342 | ||
f385ae0a MD |
2343 | health_code_update(); |
2344 | ||
b8aa1682 | 2345 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2346 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2347 | health_poll_entry(); |
b8aa1682 | 2348 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2349 | health_poll_exit(); |
b8aa1682 JD |
2350 | if (ret < 0) { |
2351 | /* | |
2352 | * Restart interrupted system call. | |
2353 | */ | |
2354 | if (errno == EINTR) { | |
2355 | goto restart; | |
2356 | } | |
2357 | goto error; | |
2358 | } | |
2359 | ||
0d9c5d77 DG |
2360 | nb_fd = ret; |
2361 | ||
beaad64c | 2362 | /* |
9a78c92e MD |
2363 | * Process control. The control connection is |
2364 | * prioritized so we don't starve it with high | |
2365 | * throughput tracing data on the data connection. | |
beaad64c | 2366 | */ |
b8aa1682 JD |
2367 | for (i = 0; i < nb_fd; i++) { |
2368 | /* Fetch once the poll data */ | |
beaad64c DG |
2369 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2370 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2371 | |
f385ae0a MD |
2372 | health_code_update(); |
2373 | ||
e0077699 | 2374 | if (!revents) { |
9a78c92e MD |
2375 | /* |
2376 | * No activity for this FD (poll | |
2377 | * implementation). | |
2378 | */ | |
e0077699 MD |
2379 | continue; |
2380 | } | |
2381 | ||
b8aa1682 JD |
2382 | /* Thread quit pipe has been closed. Killing thread. */ |
2383 | ret = check_thread_quit_pipe(pollfd, revents); | |
2384 | if (ret) { | |
095a4ae5 MD |
2385 | err = 0; |
2386 | goto exit; | |
b8aa1682 JD |
2387 | } |
2388 | ||
58eb9381 DG |
2389 | /* Inspect the relay conn pipe for new connection */ |
2390 | if (pollfd == relay_conn_pipe[0]) { | |
9a78c92e MD |
2391 | if (revents & LPOLLIN) { |
2392 | struct relay_connection *conn; | |
2393 | ||
58eb9381 | 2394 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2395 | if (ret < 0) { |
2396 | goto error; | |
2397 | } | |
58eb9381 DG |
2398 | lttng_poll_add(&events, conn->sock->fd, |
2399 | LPOLLIN | LPOLLRDHUP); | |
9a78c92e | 2400 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2401 | DBG("Connection socket %d added", conn->sock->fd); |
9a78c92e MD |
2402 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2403 | ERR("Relay connection pipe error"); | |
2404 | goto error; | |
2405 | } else { | |
2406 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
2407 | goto error; | |
b8aa1682 | 2408 | } |
58eb9381 | 2409 | } else { |
9a78c92e MD |
2410 | struct relay_connection *ctrl_conn; |
2411 | ||
2412 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); | |
58eb9381 | 2413 | /* If not found, there is a synchronization issue. */ |
9a78c92e | 2414 | assert(ctrl_conn); |
58eb9381 | 2415 | |
9a78c92e MD |
2416 | if (ctrl_conn->type == RELAY_DATA) { |
2417 | if (revents & LPOLLIN) { | |
beaad64c DG |
2418 | /* |
2419 | * Flag the last seen data fd not deleted. It will be | |
2420 | * used as the last seen fd if any fd gets deleted in | |
2421 | * this first loop. | |
2422 | */ | |
2423 | last_notdel_data_fd = pollfd; | |
2424 | } | |
9a78c92e MD |
2425 | goto put_ctrl_connection; |
2426 | } | |
2427 | assert(ctrl_conn->type == RELAY_CONTROL); | |
2428 | ||
2429 | if (revents & LPOLLIN) { | |
2430 | ret = ctrl_conn->sock->ops->recvmsg(ctrl_conn->sock, | |
2431 | &recv_hdr, sizeof(recv_hdr), 0); | |
2432 | if (ret <= 0) { | |
2433 | /* Connection closed */ | |
2434 | relay_thread_close_connection(&events, pollfd, | |
2435 | ctrl_conn); | |
2436 | } else { | |
2437 | ret = relay_process_control(&recv_hdr, ctrl_conn); | |
2438 | if (ret < 0) { | |
2439 | /* Clear the session on error. */ | |
2440 | relay_thread_close_connection(&events, | |
2441 | pollfd, ctrl_conn); | |
2442 | } | |
2443 | seen_control = 1; | |
2444 | } | |
2445 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2446 | relay_thread_close_connection(&events, | |
2447 | pollfd, ctrl_conn); | |
2448 | if (last_seen_data_fd == pollfd) { | |
2449 | last_seen_data_fd = last_notdel_data_fd; | |
2450 | } | |
58eb9381 | 2451 | } else { |
9a78c92e MD |
2452 | ERR("Unexpected poll events %u for control sock %d", |
2453 | revents, pollfd); | |
2454 | connection_put(ctrl_conn); | |
2455 | goto error; | |
beaad64c | 2456 | } |
9a78c92e MD |
2457 | put_ctrl_connection: |
2458 | connection_put(ctrl_conn); | |
beaad64c DG |
2459 | } |
2460 | } | |
2461 | ||
2462 | /* | |
2463 | * The last loop handled a control request, go back to poll to make | |
2464 | * sure we prioritise the control socket. | |
2465 | */ | |
2466 | if (seen_control) { | |
2467 | continue; | |
2468 | } | |
2469 | ||
2470 | if (last_seen_data_fd >= 0) { | |
2471 | for (i = 0; i < nb_fd; i++) { | |
2472 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2473 | |
2474 | health_code_update(); | |
2475 | ||
beaad64c DG |
2476 | if (last_seen_data_fd == pollfd) { |
2477 | idx = i; | |
2478 | break; | |
2479 | } | |
2480 | } | |
2481 | } | |
2482 | ||
2483 | /* Process data connection. */ | |
2484 | for (i = idx + 1; i < nb_fd; i++) { | |
2485 | /* Fetch the poll data. */ | |
2486 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2487 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
9a78c92e | 2488 | struct relay_connection *data_conn; |
beaad64c | 2489 | |
f385ae0a MD |
2490 | health_code_update(); |
2491 | ||
e0077699 MD |
2492 | if (!revents) { |
2493 | /* No activity for this FD (poll implementation). */ | |
2494 | continue; | |
2495 | } | |
2496 | ||
beaad64c | 2497 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2498 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2499 | continue; |
2500 | } | |
2501 | ||
9a78c92e MD |
2502 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
2503 | if (!data_conn) { | |
e0077699 | 2504 | /* Skip it. Might be removed before. */ |
e0077699 MD |
2505 | continue; |
2506 | } | |
9a78c92e MD |
2507 | if (data_conn->type == RELAY_CONTROL) { |
2508 | goto put_data_connection; | |
2509 | } | |
2510 | assert(data_conn->type == RELAY_DATA); | |
e0077699 MD |
2511 | |
2512 | if (revents & LPOLLIN) { | |
9a78c92e | 2513 | ret = relay_process_data(data_conn); |
e0077699 MD |
2514 | /* Connection closed */ |
2515 | if (ret < 0) { | |
9a78c92e MD |
2516 | relay_thread_close_connection(&events, pollfd, |
2517 | data_conn); | |
e0077699 MD |
2518 | /* |
2519 | * Every goto restart call sets the last seen fd where | |
2520 | * here we don't really care since we gracefully | |
2521 | * continue the loop after the connection is deleted. | |
2522 | */ | |
2523 | } else { | |
2524 | /* Keep last seen port. */ | |
2525 | last_seen_data_fd = pollfd; | |
9a78c92e | 2526 | connection_put(data_conn); |
e0077699 | 2527 | goto restart; |
b8aa1682 | 2528 | } |
9a78c92e MD |
2529 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2530 | relay_thread_close_connection(&events, pollfd, | |
2531 | data_conn); | |
2532 | } else { | |
2533 | ERR("Unknown poll events %u for data sock %d", | |
2534 | revents, pollfd); | |
b8aa1682 | 2535 | } |
9a78c92e MD |
2536 | put_data_connection: |
2537 | connection_put(data_conn); | |
b8aa1682 | 2538 | } |
beaad64c | 2539 | last_seen_data_fd = -1; |
b8aa1682 JD |
2540 | } |
2541 | ||
f385ae0a MD |
2542 | /* Normal exit, no error */ |
2543 | ret = 0; | |
2544 | ||
095a4ae5 | 2545 | exit: |
b8aa1682 | 2546 | error: |
58eb9381 | 2547 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2548 | rcu_read_lock(); |
9a78c92e MD |
2549 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
2550 | destroy_conn, | |
58eb9381 | 2551 | sock_n.node) { |
f385ae0a | 2552 | health_code_update(); |
9a78c92e MD |
2553 | /* |
2554 | * No need to grab another ref, because we own | |
2555 | * destroy_conn. | |
2556 | */ | |
2557 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
2558 | destroy_conn); | |
b8aa1682 | 2559 | } |
94d49140 | 2560 | rcu_read_unlock(); |
9a78c92e MD |
2561 | |
2562 | lttng_poll_clean(&events); | |
7d2f7452 | 2563 | error_poll_create: |
b8aa1682 | 2564 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2565 | relay_connections_ht_error: |
58eb9381 DG |
2566 | /* Close relay conn pipes */ |
2567 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2568 | if (err) { |
2569 | DBG("Thread exited with error"); | |
2570 | } | |
b8aa1682 | 2571 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2572 | free(data_buffer); |
9b5e0863 | 2573 | error_testpoint: |
f385ae0a MD |
2574 | if (err) { |
2575 | health_error(); | |
2576 | ERR("Health error occurred in %s", __func__); | |
2577 | } | |
2578 | health_unregister(health_relayd); | |
9d1bbf21 | 2579 | rcu_unregister_thread(); |
770f96f4 | 2580 | lttng_relay_stop_threads(); |
b8aa1682 JD |
2581 | return NULL; |
2582 | } | |
2583 | ||
2584 | /* | |
2585 | * Create the relay command pipe to wake thread_manage_apps. | |
2586 | * Closed in cleanup(). | |
2587 | */ | |
58eb9381 | 2588 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2589 | { |
a02de639 | 2590 | int ret; |
b8aa1682 | 2591 | |
58eb9381 | 2592 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2593 | |
b8aa1682 JD |
2594 | return ret; |
2595 | } | |
2596 | ||
2597 | /* | |
2598 | * main | |
2599 | */ | |
2600 | int main(int argc, char **argv) | |
2601 | { | |
77c54fed | 2602 | int ret = 0, retval = 0; |
b8aa1682 JD |
2603 | void *status; |
2604 | ||
b8aa1682 JD |
2605 | /* Parse arguments */ |
2606 | progname = argv[0]; | |
77c54fed MD |
2607 | if (set_options(argc, argv)) { |
2608 | retval = -1; | |
2609 | goto exit_options; | |
b8aa1682 JD |
2610 | } |
2611 | ||
77c54fed MD |
2612 | if (set_signal_handler()) { |
2613 | retval = -1; | |
2614 | goto exit_options; | |
b8aa1682 JD |
2615 | } |
2616 | ||
4d513a50 DG |
2617 | /* Try to create directory if -o, --output is specified. */ |
2618 | if (opt_output_path) { | |
994fa64f DG |
2619 | if (*opt_output_path != '/') { |
2620 | ERR("Please specify an absolute path for -o, --output PATH"); | |
77c54fed MD |
2621 | retval = -1; |
2622 | goto exit_options; | |
994fa64f DG |
2623 | } |
2624 | ||
0d083fb8 JG |
2625 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
2626 | -1, -1); | |
4d513a50 DG |
2627 | if (ret < 0) { |
2628 | ERR("Unable to create %s", opt_output_path); | |
77c54fed MD |
2629 | retval = -1; |
2630 | goto exit_options; | |
4d513a50 DG |
2631 | } |
2632 | } | |
2633 | ||
b8aa1682 | 2634 | /* Daemonize */ |
b5218ffb | 2635 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2636 | int i; |
2637 | ||
2638 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2639 | !opt_background); | |
b8aa1682 | 2640 | if (ret < 0) { |
77c54fed MD |
2641 | retval = -1; |
2642 | goto exit_options; | |
b8aa1682 | 2643 | } |
3fd27398 MD |
2644 | |
2645 | /* | |
2646 | * We are in the child. Make sure all other file | |
2647 | * descriptors are closed, in case we are called with | |
2648 | * more opened file descriptors than the standard ones. | |
2649 | */ | |
2650 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2651 | (void) close(i); | |
2652 | } | |
2653 | } | |
2654 | ||
77c54fed MD |
2655 | |
2656 | /* Initialize thread health monitoring */ | |
2657 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2658 | if (!health_relayd) { | |
2659 | PERROR("health_app_create error"); | |
2660 | retval = -1; | |
2661 | goto exit_health_app_create; | |
2662 | } | |
2663 | ||
3fd27398 | 2664 | /* Create thread quit pipe */ |
77c54fed MD |
2665 | if (init_thread_quit_pipe()) { |
2666 | retval = -1; | |
2667 | goto exit_init_data; | |
b8aa1682 JD |
2668 | } |
2669 | ||
1c20f0e2 | 2670 | /* Check if daemon is UID = 0 */ |
9a78c92e | 2671 | if (!getuid()) { |
8d5c808e | 2672 | if (control_uri->port < 1024 || data_uri->port < 1024 || live_uri->port < 1024) { |
b8aa1682 | 2673 | ERR("Need to be root to use ports < 1024"); |
77c54fed MD |
2674 | retval = -1; |
2675 | goto exit_init_data; | |
b8aa1682 JD |
2676 | } |
2677 | } | |
2678 | ||
2679 | /* Setup the thread apps communication pipe. */ | |
77c54fed MD |
2680 | if (create_relay_conn_pipe()) { |
2681 | retval = -1; | |
2682 | goto exit_init_data; | |
b8aa1682 JD |
2683 | } |
2684 | ||
2685 | /* Init relay command queue. */ | |
8bdee6e2 | 2686 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 JD |
2687 | |
2688 | /* Set up max poll set size */ | |
2689 | lttng_poll_set_max_size(); | |
2690 | ||
554831e7 MD |
2691 | /* Initialize communication library */ |
2692 | lttcomm_init(); | |
87e45c13 | 2693 | lttcomm_inet_init(); |
554831e7 | 2694 | |
d3e2ba59 | 2695 | /* tables of sessions indexed by session ID */ |
9a78c92e MD |
2696 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2697 | if (!sessions_ht) { | |
77c54fed MD |
2698 | retval = -1; |
2699 | goto exit_init_data; | |
d3e2ba59 JD |
2700 | } |
2701 | ||
2702 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2703 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 2704 | if (!relay_streams_ht) { |
77c54fed MD |
2705 | retval = -1; |
2706 | goto exit_init_data; | |
d3e2ba59 JD |
2707 | } |
2708 | ||
2709 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2710 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2711 | if (!viewer_streams_ht) { | |
77c54fed MD |
2712 | retval = -1; |
2713 | goto exit_init_data; | |
55706a7d MD |
2714 | } |
2715 | ||
65931c8b | 2716 | ret = utils_create_pipe(health_quit_pipe); |
77c54fed MD |
2717 | if (ret) { |
2718 | retval = -1; | |
2719 | goto exit_health_quit_pipe; | |
65931c8b MD |
2720 | } |
2721 | ||
2722 | /* Create thread to manage the client socket */ | |
2723 | ret = pthread_create(&health_thread, NULL, | |
2724 | thread_manage_health, (void *) NULL); | |
77c54fed MD |
2725 | if (ret) { |
2726 | errno = ret; | |
65931c8b | 2727 | PERROR("pthread_create health"); |
77c54fed MD |
2728 | retval = -1; |
2729 | goto exit_health_thread; | |
65931c8b MD |
2730 | } |
2731 | ||
b8aa1682 JD |
2732 | /* Setup the dispatcher thread */ |
2733 | ret = pthread_create(&dispatcher_thread, NULL, | |
2734 | relay_thread_dispatcher, (void *) NULL); | |
77c54fed MD |
2735 | if (ret) { |
2736 | errno = ret; | |
b8aa1682 | 2737 | PERROR("pthread_create dispatcher"); |
77c54fed MD |
2738 | retval = -1; |
2739 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
2740 | } |
2741 | ||
2742 | /* Setup the worker thread */ | |
2743 | ret = pthread_create(&worker_thread, NULL, | |
9a78c92e | 2744 | relay_thread_worker, NULL); |
77c54fed MD |
2745 | if (ret) { |
2746 | errno = ret; | |
b8aa1682 | 2747 | PERROR("pthread_create worker"); |
77c54fed MD |
2748 | retval = -1; |
2749 | goto exit_worker_thread; | |
b8aa1682 JD |
2750 | } |
2751 | ||
2752 | /* Setup the listener thread */ | |
2753 | ret = pthread_create(&listener_thread, NULL, | |
2754 | relay_thread_listener, (void *) NULL); | |
77c54fed MD |
2755 | if (ret) { |
2756 | errno = ret; | |
b8aa1682 | 2757 | PERROR("pthread_create listener"); |
77c54fed MD |
2758 | retval = -1; |
2759 | goto exit_listener_thread; | |
b8aa1682 JD |
2760 | } |
2761 | ||
9a78c92e | 2762 | ret = relayd_live_create(live_uri); |
77c54fed | 2763 | if (ret) { |
d3e2ba59 | 2764 | ERR("Starting live viewer threads"); |
77c54fed | 2765 | retval = -1; |
50138f51 | 2766 | goto exit_live; |
d3e2ba59 JD |
2767 | } |
2768 | ||
77c54fed MD |
2769 | /* |
2770 | * This is where we start awaiting program completion (e.g. through | |
2771 | * signal that asks threads to teardown). | |
2772 | */ | |
2773 | ||
2774 | ret = relayd_live_join(); | |
2775 | if (ret) { | |
2776 | retval = -1; | |
2777 | } | |
50138f51 | 2778 | exit_live: |
77c54fed | 2779 | |
b8aa1682 | 2780 | ret = pthread_join(listener_thread, &status); |
77c54fed MD |
2781 | if (ret) { |
2782 | errno = ret; | |
2783 | PERROR("pthread_join listener_thread"); | |
2784 | retval = -1; | |
b8aa1682 JD |
2785 | } |
2786 | ||
77c54fed | 2787 | exit_listener_thread: |
b8aa1682 | 2788 | ret = pthread_join(worker_thread, &status); |
77c54fed MD |
2789 | if (ret) { |
2790 | errno = ret; | |
2791 | PERROR("pthread_join worker_thread"); | |
2792 | retval = -1; | |
b8aa1682 JD |
2793 | } |
2794 | ||
77c54fed | 2795 | exit_worker_thread: |
b8aa1682 | 2796 | ret = pthread_join(dispatcher_thread, &status); |
77c54fed MD |
2797 | if (ret) { |
2798 | errno = ret; | |
2799 | PERROR("pthread_join dispatcher_thread"); | |
2800 | retval = -1; | |
b8aa1682 | 2801 | } |
77c54fed | 2802 | exit_dispatcher_thread: |
42415026 | 2803 | |
65931c8b | 2804 | ret = pthread_join(health_thread, &status); |
77c54fed MD |
2805 | if (ret) { |
2806 | errno = ret; | |
2807 | PERROR("pthread_join health_thread"); | |
2808 | retval = -1; | |
65931c8b | 2809 | } |
77c54fed | 2810 | exit_health_thread: |
65931c8b | 2811 | |
65931c8b | 2812 | utils_close_pipe(health_quit_pipe); |
77c54fed | 2813 | exit_health_quit_pipe: |
65931c8b | 2814 | |
77c54fed | 2815 | exit_init_data: |
55706a7d | 2816 | health_app_destroy(health_relayd); |
55706a7d | 2817 | exit_health_app_create: |
77c54fed | 2818 | exit_options: |
9a78c92e MD |
2819 | relayd_cleanup(); |
2820 | ||
2821 | /* Ensure all prior call_rcu are done. */ | |
2822 | rcu_barrier(); | |
d3e2ba59 | 2823 | |
77c54fed | 2824 | if (!retval) { |
b8aa1682 | 2825 | exit(EXIT_SUCCESS); |
77c54fed MD |
2826 | } else { |
2827 | exit(EXIT_FAILURE); | |
b8aa1682 | 2828 | } |
b8aa1682 | 2829 | } |