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" |
c00978f1 | 72 | #include "tracefile-array.h" |
b8aa1682 JD |
73 | |
74 | /* command line options */ | |
0f907de1 | 75 | char *opt_output_path; |
b5218ffb | 76 | static int opt_daemon, opt_background; |
3fd27398 MD |
77 | |
78 | /* | |
79 | * We need to wait for listener and live listener threads, as well as | |
80 | * health check thread, before being ready to signal readiness. | |
81 | */ | |
82 | #define NR_LTTNG_RELAY_READY 3 | |
83 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
84 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ | |
85 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
86 | ||
095a4ae5 MD |
87 | static struct lttng_uri *control_uri; |
88 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 89 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
90 | |
91 | const char *progname; | |
b8aa1682 | 92 | |
65931c8b | 93 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
94 | static int tracing_group_name_override; |
95 | ||
96 | const char * const config_section_name = "relayd"; | |
65931c8b | 97 | |
b8aa1682 JD |
98 | /* |
99 | * Quit pipe for all threads. This permits a single cancellation point | |
100 | * for all threads when receiving an event on the pipe. | |
101 | */ | |
0b242f62 | 102 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
103 | |
104 | /* | |
105 | * This pipe is used to inform the worker thread that a command is queued and | |
106 | * ready to be processed. | |
107 | */ | |
58eb9381 | 108 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 109 | |
26c9d55e | 110 | /* Shared between threads */ |
b8aa1682 JD |
111 | static int dispatch_thread_exit; |
112 | ||
113 | static pthread_t listener_thread; | |
114 | static pthread_t dispatcher_thread; | |
115 | static pthread_t worker_thread; | |
65931c8b | 116 | static pthread_t health_thread; |
b8aa1682 | 117 | |
9a78c92e MD |
118 | /* |
119 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
120 | * atomicity on 32-bit architectures. | |
121 | */ | |
122 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 123 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
124 | |
125 | /* | |
126 | * Relay command queue. | |
127 | * | |
128 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
129 | * queue. | |
130 | */ | |
58eb9381 | 131 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 JD |
132 | |
133 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
134 | static char *data_buffer; |
135 | static unsigned int data_buffer_size; | |
b8aa1682 | 136 | |
d3e2ba59 JD |
137 | /* Global relay stream hash table. */ |
138 | struct lttng_ht *relay_streams_ht; | |
139 | ||
92c6ca54 DG |
140 | /* Global relay viewer stream hash table. */ |
141 | struct lttng_ht *viewer_streams_ht; | |
142 | ||
9a78c92e MD |
143 | /* Global relay sessions hash table. */ |
144 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 145 | |
55706a7d | 146 | /* Relayd health monitoring */ |
eea7556c | 147 | struct health_app *health_relayd; |
55706a7d | 148 | |
cd60b05a JG |
149 | static struct option long_options[] = { |
150 | { "control-port", 1, 0, 'C', }, | |
151 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 152 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 153 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 154 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
155 | { "group", 1, 0, 'g', }, |
156 | { "help", 0, 0, 'h', }, | |
157 | { "output", 1, 0, 'o', }, | |
158 | { "verbose", 0, 0, 'v', }, | |
159 | { "config", 1, 0, 'f' }, | |
160 | { NULL, 0, 0, 0, }, | |
161 | }; | |
162 | ||
163 | static const char *config_ignore_options[] = { "help", "config" }; | |
164 | ||
b8aa1682 JD |
165 | /* |
166 | * usage function on stderr | |
167 | */ | |
9a78c92e | 168 | static void usage(void) |
b8aa1682 JD |
169 | { |
170 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
171 | fprintf(stderr, " -h, --help Display this usage.\n"); |
172 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
b5218ffb | 173 | fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); |
994fa64f DG |
174 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); |
175 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
8d5c808e | 176 | fprintf(stderr, " -L, --live-port URL Live view port listening.\n"); |
994fa64f DG |
177 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); |
178 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 179 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 180 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
181 | } |
182 | ||
cd60b05a JG |
183 | /* |
184 | * Take an option from the getopt output and set it in the right variable to be | |
185 | * used later. | |
186 | * | |
187 | * Return 0 on success else a negative value. | |
188 | */ | |
9a78c92e | 189 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 190 | { |
cd60b05a JG |
191 | int ret; |
192 | ||
193 | switch (opt) { | |
194 | case 0: | |
195 | fprintf(stderr, "option %s", optname); | |
196 | if (arg) { | |
197 | fprintf(stderr, " with arg %s\n", arg); | |
198 | } | |
199 | break; | |
200 | case 'C': | |
201 | ret = uri_parse(arg, &control_uri); | |
202 | if (ret < 0) { | |
203 | ERR("Invalid control URI specified"); | |
204 | goto end; | |
205 | } | |
206 | if (control_uri->port == 0) { | |
207 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
208 | } | |
209 | break; | |
210 | case 'D': | |
211 | ret = uri_parse(arg, &data_uri); | |
212 | if (ret < 0) { | |
213 | ERR("Invalid data URI specified"); | |
214 | goto end; | |
215 | } | |
216 | if (data_uri->port == 0) { | |
217 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
218 | } | |
219 | break; | |
8d5c808e AM |
220 | case 'L': |
221 | ret = uri_parse(arg, &live_uri); | |
222 | if (ret < 0) { | |
223 | ERR("Invalid live URI specified"); | |
224 | goto end; | |
225 | } | |
226 | if (live_uri->port == 0) { | |
227 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
228 | } | |
229 | break; | |
cd60b05a JG |
230 | case 'd': |
231 | opt_daemon = 1; | |
232 | break; | |
b5218ffb MD |
233 | case 'b': |
234 | opt_background = 1; | |
235 | break; | |
cd60b05a JG |
236 | case 'g': |
237 | tracing_group_name = strdup(arg); | |
46cd2129 MD |
238 | if (tracing_group_name == NULL) { |
239 | ret = -errno; | |
240 | PERROR("strdup"); | |
241 | goto end; | |
242 | } | |
cd60b05a JG |
243 | tracing_group_name_override = 1; |
244 | break; | |
245 | case 'h': | |
246 | usage(); | |
247 | exit(EXIT_FAILURE); | |
248 | case 'o': | |
249 | ret = asprintf(&opt_output_path, "%s", arg); | |
250 | if (ret < 0) { | |
251 | ret = -errno; | |
252 | PERROR("asprintf opt_output_path"); | |
253 | goto end; | |
254 | } | |
255 | break; | |
256 | case 'v': | |
257 | /* Verbose level can increase using multiple -v */ | |
258 | if (arg) { | |
259 | lttng_opt_verbose = config_parse_value(arg); | |
260 | } else { | |
849e5b7b DG |
261 | /* Only 3 level of verbosity (-vvv). */ |
262 | if (lttng_opt_verbose < 3) { | |
263 | lttng_opt_verbose += 1; | |
264 | } | |
cd60b05a JG |
265 | } |
266 | break; | |
267 | default: | |
268 | /* Unknown option or other error. | |
269 | * Error is printed by getopt, just return */ | |
270 | ret = -1; | |
271 | goto end; | |
272 | } | |
273 | ||
274 | /* All good. */ | |
275 | ret = 0; | |
276 | ||
277 | end: | |
278 | return ret; | |
279 | } | |
280 | ||
281 | /* | |
282 | * config_entry_handler_cb used to handle options read from a config file. | |
283 | * See config_entry_handler_cb comment in common/config/config.h for the | |
284 | * return value conventions. | |
285 | */ | |
9a78c92e | 286 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
287 | { |
288 | int ret = 0, i; | |
289 | ||
290 | if (!entry || !entry->name || !entry->value) { | |
291 | ret = -EINVAL; | |
292 | goto end; | |
293 | } | |
294 | ||
295 | /* Check if the option is to be ignored */ | |
296 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
297 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
298 | goto end; | |
299 | } | |
300 | } | |
301 | ||
302 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
303 | /* Ignore if entry name is not fully matched. */ | |
304 | if (strcmp(entry->name, long_options[i].name)) { | |
305 | continue; | |
306 | } | |
307 | ||
308 | /* | |
9a78c92e MD |
309 | * If the option takes no argument on the command line, |
310 | * we have to check if the value is "true". We support | |
311 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
312 | */ |
313 | if (!long_options[i].has_arg) { | |
314 | ret = config_parse_value(entry->value); | |
315 | if (ret <= 0) { | |
316 | if (ret) { | |
317 | WARN("Invalid configuration value \"%s\" for option %s", | |
318 | entry->value, entry->name); | |
319 | } | |
320 | /* False, skip boolean config option. */ | |
321 | goto end; | |
322 | } | |
323 | } | |
324 | ||
325 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
326 | goto end; | |
327 | } | |
328 | ||
329 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
330 | entry->name); | |
331 | ||
332 | end: | |
333 | return ret; | |
334 | } | |
335 | ||
9a78c92e | 336 | static int set_options(int argc, char **argv) |
cd60b05a | 337 | { |
77c54fed | 338 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
339 | int orig_optopt = optopt, orig_optind = optind; |
340 | char *default_address, *optstring; | |
341 | const char *config_path = NULL; | |
342 | ||
343 | optstring = utils_generate_optstring(long_options, | |
344 | sizeof(long_options) / sizeof(struct option)); | |
345 | if (!optstring) { | |
77c54fed | 346 | retval = -ENOMEM; |
cd60b05a JG |
347 | goto exit; |
348 | } | |
349 | ||
350 | /* Check for the --config option */ | |
351 | ||
352 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
353 | &option_index)) != -1) { | |
354 | if (c == '?') { | |
77c54fed | 355 | retval = -EINVAL; |
cd60b05a JG |
356 | goto exit; |
357 | } else if (c != 'f') { | |
358 | continue; | |
359 | } | |
360 | ||
361 | config_path = utils_expand_path(optarg); | |
362 | if (!config_path) { | |
363 | ERR("Failed to resolve path: %s", optarg); | |
364 | } | |
365 | } | |
366 | ||
367 | ret = config_get_section_entries(config_path, config_section_name, | |
368 | config_entry_handler, NULL); | |
369 | if (ret) { | |
370 | if (ret > 0) { | |
371 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 372 | } |
77c54fed | 373 | retval = -1; |
cd60b05a JG |
374 | goto exit; |
375 | } | |
b8aa1682 | 376 | |
cd60b05a JG |
377 | /* Reset getopt's global state */ |
378 | optopt = orig_optopt; | |
379 | optind = orig_optind; | |
b8aa1682 | 380 | while (1) { |
cd60b05a | 381 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
382 | if (c == -1) { |
383 | break; | |
384 | } | |
385 | ||
cd60b05a JG |
386 | ret = set_option(c, optarg, long_options[option_index].name); |
387 | if (ret < 0) { | |
77c54fed | 388 | retval = -1; |
b8aa1682 JD |
389 | goto exit; |
390 | } | |
391 | } | |
392 | ||
393 | /* assign default values */ | |
394 | if (control_uri == NULL) { | |
fa91dc52 MD |
395 | ret = asprintf(&default_address, |
396 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
397 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
398 | if (ret < 0) { |
399 | PERROR("asprintf default data address"); | |
77c54fed | 400 | retval = -1; |
b8aa1682 JD |
401 | goto exit; |
402 | } | |
403 | ||
404 | ret = uri_parse(default_address, &control_uri); | |
405 | free(default_address); | |
406 | if (ret < 0) { | |
407 | ERR("Invalid control URI specified"); | |
77c54fed | 408 | retval = -1; |
b8aa1682 JD |
409 | goto exit; |
410 | } | |
411 | } | |
412 | if (data_uri == NULL) { | |
fa91dc52 MD |
413 | ret = asprintf(&default_address, |
414 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
415 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
416 | if (ret < 0) { |
417 | PERROR("asprintf default data address"); | |
77c54fed | 418 | retval = -1; |
b8aa1682 JD |
419 | goto exit; |
420 | } | |
421 | ||
422 | ret = uri_parse(default_address, &data_uri); | |
423 | free(default_address); | |
424 | if (ret < 0) { | |
425 | ERR("Invalid data URI specified"); | |
77c54fed | 426 | retval = -1; |
b8aa1682 JD |
427 | goto exit; |
428 | } | |
429 | } | |
d3e2ba59 | 430 | if (live_uri == NULL) { |
fa91dc52 MD |
431 | ret = asprintf(&default_address, |
432 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
433 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
434 | if (ret < 0) { |
435 | PERROR("asprintf default viewer control address"); | |
77c54fed | 436 | retval = -1; |
d3e2ba59 JD |
437 | goto exit; |
438 | } | |
439 | ||
440 | ret = uri_parse(default_address, &live_uri); | |
441 | free(default_address); | |
442 | if (ret < 0) { | |
443 | ERR("Invalid viewer control URI specified"); | |
77c54fed | 444 | retval = -1; |
d3e2ba59 JD |
445 | goto exit; |
446 | } | |
447 | } | |
b8aa1682 JD |
448 | |
449 | exit: | |
cd60b05a | 450 | free(optstring); |
77c54fed | 451 | return retval; |
b8aa1682 JD |
452 | } |
453 | ||
9a78c92e MD |
454 | static void print_global_objects(void) |
455 | { | |
456 | rcu_register_thread(); | |
457 | ||
458 | print_viewer_streams(); | |
459 | print_relay_streams(); | |
460 | print_sessions(); | |
461 | ||
462 | rcu_unregister_thread(); | |
463 | } | |
464 | ||
b8aa1682 JD |
465 | /* |
466 | * Cleanup the daemon | |
467 | */ | |
9a78c92e | 468 | static void relayd_cleanup(void) |
b8aa1682 | 469 | { |
9a78c92e MD |
470 | print_global_objects(); |
471 | ||
b8aa1682 JD |
472 | DBG("Cleaning up"); |
473 | ||
77c54fed MD |
474 | if (viewer_streams_ht) |
475 | lttng_ht_destroy(viewer_streams_ht); | |
476 | if (relay_streams_ht) | |
477 | lttng_ht_destroy(relay_streams_ht); | |
9a78c92e MD |
478 | if (sessions_ht) |
479 | lttng_ht_destroy(sessions_ht); | |
77c54fed | 480 | |
095a4ae5 MD |
481 | /* free the dynamically allocated opt_output_path */ |
482 | free(opt_output_path); | |
483 | ||
a02de639 CB |
484 | /* Close thread quit pipes */ |
485 | utils_close_pipe(thread_quit_pipe); | |
486 | ||
710c1f73 DG |
487 | uri_free(control_uri); |
488 | uri_free(data_uri); | |
8d5c808e | 489 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
490 | |
491 | if (tracing_group_name_override) { | |
492 | free((void *) tracing_group_name); | |
493 | } | |
b8aa1682 JD |
494 | } |
495 | ||
496 | /* | |
497 | * Write to writable pipe used to notify a thread. | |
498 | */ | |
9a78c92e | 499 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 500 | { |
6cd525e8 | 501 | ssize_t ret; |
b8aa1682 | 502 | |
6cd525e8 MD |
503 | ret = lttng_write(wpipe, "!", 1); |
504 | if (ret < 1) { | |
b8aa1682 | 505 | PERROR("write poll pipe"); |
770f96f4 | 506 | goto end; |
b8aa1682 | 507 | } |
770f96f4 MD |
508 | ret = 0; |
509 | end: | |
b8aa1682 JD |
510 | return ret; |
511 | } | |
512 | ||
9a78c92e | 513 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 514 | { |
6cd525e8 | 515 | ssize_t ret; |
65931c8b | 516 | |
6cd525e8 MD |
517 | ret = lttng_write(pipe[1], "4", 1); |
518 | if (ret < 1) { | |
65931c8b | 519 | PERROR("write relay health quit"); |
770f96f4 | 520 | goto end; |
65931c8b | 521 | } |
770f96f4 MD |
522 | ret = 0; |
523 | end: | |
524 | return ret; | |
65931c8b MD |
525 | } |
526 | ||
b8aa1682 | 527 | /* |
770f96f4 | 528 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 529 | */ |
770f96f4 | 530 | int lttng_relay_stop_threads(void) |
b8aa1682 | 531 | { |
770f96f4 | 532 | int retval = 0; |
b8aa1682 JD |
533 | |
534 | /* Stopping all threads */ | |
535 | DBG("Terminating all threads"); | |
770f96f4 | 536 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 537 | ERR("write error on thread quit pipe"); |
770f96f4 | 538 | retval = -1; |
b8aa1682 JD |
539 | } |
540 | ||
770f96f4 MD |
541 | if (notify_health_quit_pipe(health_quit_pipe)) { |
542 | ERR("write error on health quit pipe"); | |
543 | } | |
65931c8b | 544 | |
b8aa1682 | 545 | /* Dispatch thread */ |
26c9d55e | 546 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 547 | futex_nto1_wake(&relay_conn_queue.futex); |
77c54fed | 548 | |
770f96f4 | 549 | if (relayd_live_stop()) { |
77c54fed | 550 | ERR("Error stopping live threads"); |
770f96f4 | 551 | retval = -1; |
77c54fed | 552 | } |
770f96f4 | 553 | return retval; |
b8aa1682 JD |
554 | } |
555 | ||
556 | /* | |
557 | * Signal handler for the daemon | |
558 | * | |
559 | * Simply stop all worker threads, leaving main() return gracefully after | |
560 | * joining all threads and calling cleanup(). | |
561 | */ | |
9a78c92e | 562 | static void sighandler(int sig) |
b8aa1682 JD |
563 | { |
564 | switch (sig) { | |
565 | case SIGPIPE: | |
566 | DBG("SIGPIPE caught"); | |
567 | return; | |
568 | case SIGINT: | |
569 | DBG("SIGINT caught"); | |
770f96f4 MD |
570 | if (lttng_relay_stop_threads()) { |
571 | ERR("Error stopping threads"); | |
572 | } | |
b8aa1682 JD |
573 | break; |
574 | case SIGTERM: | |
575 | DBG("SIGTERM caught"); | |
770f96f4 MD |
576 | if (lttng_relay_stop_threads()) { |
577 | ERR("Error stopping threads"); | |
578 | } | |
b8aa1682 | 579 | break; |
3fd27398 MD |
580 | case SIGUSR1: |
581 | CMM_STORE_SHARED(recv_child_signal, 1); | |
582 | break; | |
b8aa1682 JD |
583 | default: |
584 | break; | |
585 | } | |
586 | } | |
587 | ||
588 | /* | |
589 | * Setup signal handler for : | |
590 | * SIGINT, SIGTERM, SIGPIPE | |
591 | */ | |
9a78c92e | 592 | static int set_signal_handler(void) |
b8aa1682 JD |
593 | { |
594 | int ret = 0; | |
595 | struct sigaction sa; | |
596 | sigset_t sigset; | |
597 | ||
598 | if ((ret = sigemptyset(&sigset)) < 0) { | |
599 | PERROR("sigemptyset"); | |
600 | return ret; | |
601 | } | |
602 | ||
603 | sa.sa_handler = sighandler; | |
604 | sa.sa_mask = sigset; | |
605 | sa.sa_flags = 0; | |
606 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
607 | PERROR("sigaction"); | |
608 | return ret; | |
609 | } | |
610 | ||
611 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
612 | PERROR("sigaction"); | |
613 | return ret; | |
614 | } | |
615 | ||
616 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
617 | PERROR("sigaction"); | |
618 | return ret; | |
619 | } | |
620 | ||
3fd27398 MD |
621 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
622 | PERROR("sigaction"); | |
623 | return ret; | |
624 | } | |
625 | ||
626 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
627 | |
628 | return ret; | |
629 | } | |
630 | ||
3fd27398 MD |
631 | void lttng_relay_notify_ready(void) |
632 | { | |
633 | /* Notify the parent of the fork() process that we are ready. */ | |
634 | if (opt_daemon || opt_background) { | |
635 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
636 | kill(child_ppid, SIGUSR1); | |
637 | } | |
638 | } | |
639 | } | |
640 | ||
b8aa1682 JD |
641 | /* |
642 | * Init thread quit pipe. | |
643 | * | |
644 | * Return -1 on error or 0 if all pipes are created. | |
645 | */ | |
9a78c92e | 646 | static int init_thread_quit_pipe(void) |
b8aa1682 | 647 | { |
a02de639 | 648 | int ret; |
b8aa1682 | 649 | |
a02de639 | 650 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 651 | |
b8aa1682 JD |
652 | return ret; |
653 | } | |
654 | ||
655 | /* | |
656 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
657 | */ | |
9a78c92e | 658 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
659 | { |
660 | int ret; | |
661 | ||
662 | if (events == NULL || size == 0) { | |
663 | ret = -1; | |
664 | goto error; | |
665 | } | |
666 | ||
667 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
668 | if (ret < 0) { | |
669 | goto error; | |
670 | } | |
671 | ||
672 | /* Add quit pipe */ | |
c7759e6a | 673 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
674 | if (ret < 0) { |
675 | goto error; | |
676 | } | |
677 | ||
678 | return 0; | |
679 | ||
680 | error: | |
681 | return ret; | |
682 | } | |
683 | ||
684 | /* | |
685 | * Check if the thread quit pipe was triggered. | |
686 | * | |
687 | * Return 1 if it was triggered else 0; | |
688 | */ | |
9a78c92e | 689 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
690 | { |
691 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
692 | return 1; | |
693 | } | |
694 | ||
695 | return 0; | |
696 | } | |
697 | ||
698 | /* | |
699 | * Create and init socket from uri. | |
700 | */ | |
9a78c92e | 701 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
702 | { |
703 | int ret; | |
704 | struct lttcomm_sock *sock = NULL; | |
705 | ||
706 | sock = lttcomm_alloc_sock_from_uri(uri); | |
707 | if (sock == NULL) { | |
708 | ERR("Allocating socket"); | |
709 | goto error; | |
710 | } | |
711 | ||
712 | ret = lttcomm_create_sock(sock); | |
713 | if (ret < 0) { | |
714 | goto error; | |
715 | } | |
716 | DBG("Listening on sock %d", sock->fd); | |
717 | ||
718 | ret = sock->ops->bind(sock); | |
719 | if (ret < 0) { | |
720 | goto error; | |
721 | } | |
722 | ||
723 | ret = sock->ops->listen(sock, -1); | |
724 | if (ret < 0) { | |
725 | goto error; | |
726 | ||
727 | } | |
728 | ||
729 | return sock; | |
730 | ||
731 | error: | |
732 | if (sock) { | |
733 | lttcomm_destroy_sock(sock); | |
734 | } | |
735 | return NULL; | |
736 | } | |
737 | ||
738 | /* | |
739 | * This thread manages the listening for new connections on the network | |
740 | */ | |
9a78c92e | 741 | static void *relay_thread_listener(void *data) |
b8aa1682 | 742 | { |
095a4ae5 | 743 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
744 | uint32_t revents, nb_fd; |
745 | struct lttng_poll_event events; | |
746 | struct lttcomm_sock *control_sock, *data_sock; | |
747 | ||
b8aa1682 JD |
748 | DBG("[thread] Relay listener started"); |
749 | ||
55706a7d MD |
750 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
751 | ||
f385ae0a MD |
752 | health_code_update(); |
753 | ||
9a78c92e | 754 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 755 | if (!control_sock) { |
095a4ae5 | 756 | goto error_sock_control; |
b8aa1682 JD |
757 | } |
758 | ||
9a78c92e | 759 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 760 | if (!data_sock) { |
095a4ae5 | 761 | goto error_sock_relay; |
b8aa1682 JD |
762 | } |
763 | ||
764 | /* | |
9a78c92e MD |
765 | * Pass 3 as size here for the thread quit pipe, control and |
766 | * data socket. | |
b8aa1682 JD |
767 | */ |
768 | ret = create_thread_poll_set(&events, 3); | |
769 | if (ret < 0) { | |
770 | goto error_create_poll; | |
771 | } | |
772 | ||
773 | /* Add the control socket */ | |
774 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
775 | if (ret < 0) { | |
776 | goto error_poll_add; | |
777 | } | |
778 | ||
779 | /* Add the data socket */ | |
780 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
781 | if (ret < 0) { | |
782 | goto error_poll_add; | |
783 | } | |
784 | ||
3fd27398 MD |
785 | lttng_relay_notify_ready(); |
786 | ||
9b5e0863 MD |
787 | if (testpoint(relayd_thread_listener)) { |
788 | goto error_testpoint; | |
789 | } | |
790 | ||
b8aa1682 | 791 | while (1) { |
f385ae0a MD |
792 | health_code_update(); |
793 | ||
b8aa1682 JD |
794 | DBG("Listener accepting connections"); |
795 | ||
b8aa1682 | 796 | restart: |
f385ae0a | 797 | health_poll_entry(); |
b8aa1682 | 798 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 799 | health_poll_exit(); |
b8aa1682 JD |
800 | if (ret < 0) { |
801 | /* | |
802 | * Restart interrupted system call. | |
803 | */ | |
804 | if (errno == EINTR) { | |
805 | goto restart; | |
806 | } | |
807 | goto error; | |
808 | } | |
809 | ||
0d9c5d77 DG |
810 | nb_fd = ret; |
811 | ||
b8aa1682 JD |
812 | DBG("Relay new connection received"); |
813 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
814 | health_code_update(); |
815 | ||
b8aa1682 JD |
816 | /* Fetch once the poll data */ |
817 | revents = LTTNG_POLL_GETEV(&events, i); | |
818 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
819 | ||
e0077699 | 820 | if (!revents) { |
9a78c92e MD |
821 | /* |
822 | * No activity for this FD (poll | |
823 | * implementation). | |
824 | */ | |
e0077699 MD |
825 | continue; |
826 | } | |
827 | ||
b8aa1682 JD |
828 | /* Thread quit pipe has been closed. Killing thread. */ |
829 | ret = check_thread_quit_pipe(pollfd, revents); | |
830 | if (ret) { | |
095a4ae5 MD |
831 | err = 0; |
832 | goto exit; | |
b8aa1682 JD |
833 | } |
834 | ||
584fc280 | 835 | if (revents & LPOLLIN) { |
4b7f17b2 | 836 | /* |
9a78c92e MD |
837 | * A new connection is requested, therefore a |
838 | * sessiond/consumerd connection is allocated in | |
839 | * this thread, enqueued to a global queue and | |
840 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 841 | */ |
58eb9381 DG |
842 | int val = 1; |
843 | struct relay_connection *new_conn; | |
4b7f17b2 | 844 | struct lttcomm_sock *newsock; |
9a78c92e | 845 | enum connection_type type; |
b8aa1682 JD |
846 | |
847 | if (pollfd == data_sock->fd) { | |
9a78c92e | 848 | type = RELAY_DATA; |
b8aa1682 | 849 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
850 | DBG("Relay data connection accepted, socket %d", |
851 | newsock->fd); | |
4b7f17b2 MD |
852 | } else { |
853 | assert(pollfd == control_sock->fd); | |
9a78c92e | 854 | type = RELAY_CONTROL; |
b8aa1682 | 855 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
856 | DBG("Relay control connection accepted, socket %d", |
857 | newsock->fd); | |
b8aa1682 | 858 | } |
58eb9381 DG |
859 | if (!newsock) { |
860 | PERROR("accepting sock"); | |
58eb9381 DG |
861 | goto error; |
862 | } | |
863 | ||
864 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
865 | sizeof(val)); | |
b8aa1682 JD |
866 | if (ret < 0) { |
867 | PERROR("setsockopt inet"); | |
4b7f17b2 | 868 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
869 | goto error; |
870 | } | |
9a78c92e MD |
871 | new_conn = connection_create(newsock, type); |
872 | if (!new_conn) { | |
873 | lttcomm_destroy_sock(newsock); | |
874 | goto error; | |
875 | } | |
58eb9381 DG |
876 | |
877 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
878 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
879 | &new_conn->qnode); | |
b8aa1682 JD |
880 | |
881 | /* | |
9a78c92e MD |
882 | * Wake the dispatch queue futex. |
883 | * Implicit memory barrier with the | |
884 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 885 | */ |
58eb9381 | 886 | futex_nto1_wake(&relay_conn_queue.futex); |
584fc280 MD |
887 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
888 | ERR("socket poll error"); | |
889 | goto error; | |
890 | } else { | |
891 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
892 | goto error; | |
b8aa1682 JD |
893 | } |
894 | } | |
895 | } | |
896 | ||
095a4ae5 | 897 | exit: |
b8aa1682 JD |
898 | error: |
899 | error_poll_add: | |
9b5e0863 | 900 | error_testpoint: |
b8aa1682 JD |
901 | lttng_poll_clean(&events); |
902 | error_create_poll: | |
095a4ae5 MD |
903 | if (data_sock->fd >= 0) { |
904 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
905 | if (ret) { |
906 | PERROR("close"); | |
907 | } | |
b8aa1682 | 908 | } |
095a4ae5 MD |
909 | lttcomm_destroy_sock(data_sock); |
910 | error_sock_relay: | |
911 | if (control_sock->fd >= 0) { | |
912 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
913 | if (ret) { |
914 | PERROR("close"); | |
915 | } | |
b8aa1682 | 916 | } |
095a4ae5 MD |
917 | lttcomm_destroy_sock(control_sock); |
918 | error_sock_control: | |
919 | if (err) { | |
f385ae0a MD |
920 | health_error(); |
921 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 922 | } |
55706a7d | 923 | health_unregister(health_relayd); |
b8aa1682 | 924 | DBG("Relay listener thread cleanup complete"); |
770f96f4 | 925 | lttng_relay_stop_threads(); |
b8aa1682 JD |
926 | return NULL; |
927 | } | |
928 | ||
929 | /* | |
930 | * This thread manages the dispatching of the requests to worker threads | |
931 | */ | |
9a78c92e | 932 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 933 | { |
6cd525e8 MD |
934 | int err = -1; |
935 | ssize_t ret; | |
8bdee6e2 | 936 | struct cds_wfcq_node *node; |
58eb9381 | 937 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
938 | |
939 | DBG("[thread] Relay dispatcher started"); | |
940 | ||
55706a7d MD |
941 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
942 | ||
9b5e0863 MD |
943 | if (testpoint(relayd_thread_dispatcher)) { |
944 | goto error_testpoint; | |
945 | } | |
946 | ||
f385ae0a MD |
947 | health_code_update(); |
948 | ||
26c9d55e | 949 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
950 | health_code_update(); |
951 | ||
b8aa1682 | 952 | /* Atomically prepare the queue futex */ |
58eb9381 | 953 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
954 | |
955 | do { | |
f385ae0a MD |
956 | health_code_update(); |
957 | ||
b8aa1682 | 958 | /* Dequeue commands */ |
8bdee6e2 SM |
959 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
960 | &relay_conn_queue.tail); | |
b8aa1682 JD |
961 | if (node == NULL) { |
962 | DBG("Woken up but nothing in the relay command queue"); | |
963 | /* Continue thread execution */ | |
964 | break; | |
965 | } | |
58eb9381 | 966 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 967 | |
58eb9381 | 968 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
969 | |
970 | /* | |
9a78c92e MD |
971 | * Inform worker thread of the new request. This |
972 | * call is blocking so we can be assured that | |
973 | * the data will be read at some point in time | |
974 | * or wait to the end of the world :) | |
b8aa1682 | 975 | */ |
58eb9381 DG |
976 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
977 | if (ret < 0) { | |
978 | PERROR("write connection pipe"); | |
9a78c92e | 979 | connection_put(new_conn); |
b8aa1682 JD |
980 | goto error; |
981 | } | |
982 | } while (node != NULL); | |
983 | ||
984 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 985 | health_poll_entry(); |
58eb9381 | 986 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 987 | health_poll_exit(); |
b8aa1682 JD |
988 | } |
989 | ||
f385ae0a MD |
990 | /* Normal exit, no error */ |
991 | err = 0; | |
992 | ||
b8aa1682 | 993 | error: |
9b5e0863 | 994 | error_testpoint: |
f385ae0a MD |
995 | if (err) { |
996 | health_error(); | |
997 | ERR("Health error occurred in %s", __func__); | |
998 | } | |
55706a7d | 999 | health_unregister(health_relayd); |
b8aa1682 | 1000 | DBG("Dispatch thread dying"); |
770f96f4 | 1001 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1002 | return NULL; |
1003 | } | |
1004 | ||
1c20f0e2 | 1005 | /* |
9a78c92e | 1006 | * Set index data from the control port to a given index object. |
1c20f0e2 | 1007 | */ |
9a78c92e | 1008 | static int set_index_control_data(struct relay_index *index, |
1c20f0e2 JD |
1009 | struct lttcomm_relayd_index *data) |
1010 | { | |
9a78c92e | 1011 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1012 | |
1013 | /* | |
9a78c92e MD |
1014 | * The index on disk is encoded in big endian, so we don't need |
1015 | * to convert the data received on the network. The data_offset | |
1016 | * value is NEVER modified here and is updated by the data | |
1017 | * thread. | |
1c20f0e2 | 1018 | */ |
9a78c92e MD |
1019 | index_data.packet_size = data->packet_size; |
1020 | index_data.content_size = data->content_size; | |
1021 | index_data.timestamp_begin = data->timestamp_begin; | |
1022 | index_data.timestamp_end = data->timestamp_end; | |
1023 | index_data.events_discarded = data->events_discarded; | |
1024 | index_data.stream_id = data->stream_id; | |
1025 | return relay_index_set_data(index, &index_data); | |
1c20f0e2 JD |
1026 | } |
1027 | ||
c5b6f4f0 DG |
1028 | /* |
1029 | * Handle the RELAYD_CREATE_SESSION command. | |
1030 | * | |
1031 | * On success, send back the session id or else return a negative value. | |
1032 | */ | |
9a78c92e | 1033 | static int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1034 | struct relay_connection *conn) |
c5b6f4f0 DG |
1035 | { |
1036 | int ret = 0, send_ret; | |
1037 | struct relay_session *session; | |
1038 | struct lttcomm_relayd_status_session reply; | |
9a78c92e MD |
1039 | char session_name[NAME_MAX]; |
1040 | char hostname[HOST_NAME_MAX]; | |
1041 | uint32_t live_timer = 0; | |
1042 | bool snapshot = false; | |
c5b6f4f0 | 1043 | |
9a78c92e MD |
1044 | memset(session_name, 0, NAME_MAX); |
1045 | memset(hostname, 0, HOST_NAME_MAX); | |
c5b6f4f0 DG |
1046 | |
1047 | memset(&reply, 0, sizeof(reply)); | |
1048 | ||
58eb9381 | 1049 | switch (conn->minor) { |
2a174661 DG |
1050 | case 1: |
1051 | case 2: | |
1052 | case 3: | |
1053 | break; | |
1054 | case 4: /* LTTng sessiond 2.4 */ | |
1055 | default: | |
9a78c92e MD |
1056 | ret = cmd_create_session_2_4(conn, session_name, |
1057 | hostname, &live_timer, &snapshot); | |
1058 | } | |
1059 | if (ret < 0) { | |
1060 | goto send_reply; | |
d3e2ba59 JD |
1061 | } |
1062 | ||
9a78c92e MD |
1063 | session = session_create(session_name, hostname, live_timer, |
1064 | snapshot, conn->major, conn->minor); | |
1065 | if (!session) { | |
1066 | ret = -1; | |
1067 | goto send_reply; | |
1068 | } | |
1069 | assert(!conn->session); | |
1070 | conn->session = session; | |
c5b6f4f0 DG |
1071 | DBG("Created session %" PRIu64, session->id); |
1072 | ||
9a78c92e MD |
1073 | reply.session_id = htobe64(session->id); |
1074 | ||
1075 | send_reply: | |
c5b6f4f0 DG |
1076 | if (ret < 0) { |
1077 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1078 | } else { | |
1079 | reply.ret_code = htobe32(LTTNG_OK); | |
1080 | } | |
1081 | ||
58eb9381 | 1082 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1083 | if (send_ret < 0) { |
1084 | ERR("Relayd sending session id"); | |
4169f5ad | 1085 | ret = send_ret; |
c5b6f4f0 DG |
1086 | } |
1087 | ||
1088 | return ret; | |
1089 | } | |
1090 | ||
a4baae1b JD |
1091 | /* |
1092 | * When we have received all the streams and the metadata for a channel, | |
1093 | * we make them visible to the viewer threads. | |
1094 | */ | |
9a78c92e | 1095 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1096 | { |
9a78c92e MD |
1097 | struct relay_stream *stream; |
1098 | struct relay_session *session = conn->session; | |
a4baae1b | 1099 | |
9a78c92e MD |
1100 | /* |
1101 | * We publish all streams belonging to a session atomically wrt | |
1102 | * session lock. | |
1103 | */ | |
1104 | pthread_mutex_lock(&session->lock); | |
1105 | rcu_read_lock(); | |
1106 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1107 | recv_node) { | |
1108 | stream_publish(stream); | |
a4baae1b | 1109 | } |
9a78c92e | 1110 | rcu_read_unlock(); |
a4baae1b | 1111 | |
9a78c92e MD |
1112 | /* |
1113 | * Inform the viewer that there are new streams in the session. | |
1114 | */ | |
1115 | if (session->viewer_attached) { | |
1116 | uatomic_set(&session->new_streams, 1); | |
1117 | } | |
1118 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1119 | } |
1120 | ||
b8aa1682 JD |
1121 | /* |
1122 | * relay_add_stream: allocate a new stream for a session | |
1123 | */ | |
9a78c92e | 1124 | static int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1125 | struct relay_connection *conn) |
b8aa1682 | 1126 | { |
9a78c92e MD |
1127 | int ret; |
1128 | ssize_t send_ret; | |
58eb9381 | 1129 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1130 | struct relay_stream *stream = NULL; |
1131 | struct lttcomm_relayd_status_stream reply; | |
9a78c92e MD |
1132 | struct ctf_trace *trace = NULL; |
1133 | uint64_t stream_handle = -1ULL; | |
1134 | char *path_name = NULL, *channel_name = NULL; | |
1135 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1136 | |
58eb9381 | 1137 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1138 | ERR("Trying to add a stream before version check"); |
1139 | ret = -1; | |
1140 | goto end_no_session; | |
1141 | } | |
1142 | ||
9a78c92e MD |
1143 | switch (session->minor) { |
1144 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
1145 | ret = cmd_recv_stream_2_1(conn, &path_name, | |
1146 | &channel_name); | |
0f907de1 | 1147 | break; |
9a78c92e | 1148 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1149 | default: |
9a78c92e MD |
1150 | ret = cmd_recv_stream_2_2(conn, &path_name, |
1151 | &channel_name, &tracefile_size, &tracefile_count); | |
0f907de1 JD |
1152 | break; |
1153 | } | |
1154 | if (ret < 0) { | |
9a78c92e | 1155 | goto send_reply; |
0f907de1 JD |
1156 | } |
1157 | ||
9a78c92e | 1158 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1159 | if (!trace) { |
9a78c92e | 1160 | goto send_reply; |
2a174661 | 1161 | } |
9a78c92e | 1162 | /* This stream here has one reference on the trace. */ |
2a174661 | 1163 | |
9a78c92e MD |
1164 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1165 | stream_handle = ++last_relay_stream_id; | |
1166 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1167 | |
9a78c92e MD |
1168 | /* We pass ownership of path_name and channel_name. */ |
1169 | stream = stream_create(trace, stream_handle, path_name, | |
1170 | channel_name, tracefile_size, tracefile_count); | |
1171 | path_name = NULL; | |
1172 | channel_name = NULL; | |
a4baae1b | 1173 | |
2a174661 | 1174 | /* |
9a78c92e MD |
1175 | * Streams are the owners of their trace. Reference to trace is |
1176 | * kept within stream_create(). | |
2a174661 | 1177 | */ |
9a78c92e | 1178 | ctf_trace_put(trace); |
d3e2ba59 | 1179 | |
9a78c92e | 1180 | send_reply: |
53efb85a | 1181 | memset(&reply, 0, sizeof(reply)); |
9a78c92e MD |
1182 | reply.handle = htobe64(stream_handle); |
1183 | if (!stream) { | |
f73fabfd | 1184 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1185 | } else { |
f73fabfd | 1186 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1187 | } |
5af40280 | 1188 | |
58eb9381 | 1189 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1190 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1191 | if (send_ret < 0) { | |
1192 | ERR("Relay sending stream id"); | |
9a78c92e | 1193 | ret = (int) send_ret; |
b8aa1682 JD |
1194 | } |
1195 | ||
1196 | end_no_session: | |
9a78c92e MD |
1197 | free(path_name); |
1198 | free(channel_name); | |
0f907de1 | 1199 | return ret; |
b8aa1682 JD |
1200 | } |
1201 | ||
173af62f DG |
1202 | /* |
1203 | * relay_close_stream: close a specific stream | |
1204 | */ | |
9a78c92e | 1205 | static int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1206 | struct relay_connection *conn) |
173af62f | 1207 | { |
94d49140 | 1208 | int ret, send_ret; |
58eb9381 | 1209 | struct relay_session *session = conn->session; |
173af62f DG |
1210 | struct lttcomm_relayd_close_stream stream_info; |
1211 | struct lttcomm_relayd_generic_reply reply; | |
1212 | struct relay_stream *stream; | |
173af62f DG |
1213 | |
1214 | DBG("Close stream received"); | |
1215 | ||
58eb9381 | 1216 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1217 | ERR("Trying to close a stream before version check"); |
1218 | ret = -1; | |
1219 | goto end_no_session; | |
1220 | } | |
1221 | ||
58eb9381 | 1222 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1223 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1224 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1225 | if (ret == 0) { |
1226 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1227 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1228 | } else { |
1229 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1230 | } | |
173af62f DG |
1231 | ret = -1; |
1232 | goto end_no_session; | |
1233 | } | |
1234 | ||
9a78c92e | 1235 | stream = stream_get_by_id(be64toh(stream_info.stream_id)); |
173af62f DG |
1236 | if (!stream) { |
1237 | ret = -1; | |
9a78c92e | 1238 | goto end; |
173af62f | 1239 | } |
ffc70591 MD |
1240 | |
1241 | /* | |
1242 | * Set last_net_seq_num before the close flag. Required by data | |
1243 | * pending check. | |
1244 | */ | |
9a78c92e | 1245 | pthread_mutex_lock(&stream->lock); |
8e2583a4 | 1246 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
ffc70591 MD |
1247 | pthread_mutex_unlock(&stream->lock); |
1248 | ||
2a5a6d16 JG |
1249 | /* |
1250 | * This is one of the conditions which may trigger a stream close | |
1251 | * with the others being: | |
1252 | * 1) A close command is received for a stream | |
1253 | * 2) The control connection owning the stream is closed | |
1254 | * 3) We have received all of the stream's data _after_ a close | |
1255 | * request. | |
1256 | */ | |
1257 | try_stream_close(stream); | |
9a78c92e MD |
1258 | if (stream->is_metadata) { |
1259 | struct relay_viewer_stream *vstream; | |
173af62f | 1260 | |
9a78c92e MD |
1261 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1262 | if (vstream) { | |
1263 | if (vstream->metadata_sent == stream->metadata_received) { | |
1264 | /* | |
1265 | * Since all the metadata has been sent to the | |
1266 | * viewer and that we have a request to close | |
1267 | * its stream, we can safely teardown the | |
1268 | * corresponding metadata viewer stream. | |
1269 | */ | |
1270 | viewer_stream_put(vstream); | |
1271 | } | |
1272 | /* Put local reference. */ | |
1273 | viewer_stream_put(vstream); | |
1274 | } | |
1275 | } | |
9a78c92e | 1276 | stream_put(stream); |
173af62f | 1277 | |
9a78c92e | 1278 | end: |
53efb85a | 1279 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1280 | if (ret < 0) { |
f73fabfd | 1281 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1282 | } else { |
f73fabfd | 1283 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1284 | } |
58eb9381 | 1285 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1286 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1287 | if (send_ret < 0) { | |
1288 | ERR("Relay sending stream id"); | |
4169f5ad | 1289 | ret = send_ret; |
173af62f DG |
1290 | } |
1291 | ||
1292 | end_no_session: | |
1293 | return ret; | |
1294 | } | |
1295 | ||
b8aa1682 JD |
1296 | /* |
1297 | * relay_unknown_command: send -1 if received unknown command | |
1298 | */ | |
9a78c92e | 1299 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1300 | { |
1301 | struct lttcomm_relayd_generic_reply reply; | |
1302 | int ret; | |
1303 | ||
53efb85a | 1304 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1305 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
58eb9381 | 1306 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1307 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1308 | if (ret < 0) { | |
1309 | ERR("Relay sending unknown command"); | |
1310 | } | |
1311 | } | |
1312 | ||
1313 | /* | |
1314 | * relay_start: send an acknowledgment to the client to tell if we are | |
1315 | * ready to receive data. We are ready if a session is established. | |
1316 | */ | |
9a78c92e | 1317 | static int relay_start(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1318 | struct relay_connection *conn) |
b8aa1682 | 1319 | { |
f73fabfd | 1320 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1321 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1322 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1323 | |
1324 | if (!session) { | |
1325 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1326 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1327 | } |
1328 | ||
53efb85a | 1329 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1330 | reply.ret_code = ret; |
58eb9381 | 1331 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1332 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1333 | if (ret < 0) { | |
1334 | ERR("Relay sending start ack"); | |
1335 | } | |
1336 | ||
1337 | return ret; | |
1338 | } | |
1339 | ||
1d4dfdef DG |
1340 | /* |
1341 | * Append padding to the file pointed by the file descriptor fd. | |
1342 | */ | |
1343 | static int write_padding_to_file(int fd, uint32_t size) | |
1344 | { | |
6cd525e8 | 1345 | ssize_t ret = 0; |
1d4dfdef DG |
1346 | char *zeros; |
1347 | ||
1348 | if (size == 0) { | |
1349 | goto end; | |
1350 | } | |
1351 | ||
1352 | zeros = zmalloc(size); | |
1353 | if (zeros == NULL) { | |
1354 | PERROR("zmalloc zeros for padding"); | |
1355 | ret = -1; | |
1356 | goto end; | |
1357 | } | |
1358 | ||
6cd525e8 MD |
1359 | ret = lttng_write(fd, zeros, size); |
1360 | if (ret < size) { | |
1d4dfdef DG |
1361 | PERROR("write padding to file"); |
1362 | } | |
1363 | ||
e986c7a1 DG |
1364 | free(zeros); |
1365 | ||
1d4dfdef DG |
1366 | end: |
1367 | return ret; | |
1368 | } | |
1369 | ||
b8aa1682 | 1370 | /* |
9a78c92e | 1371 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1372 | */ |
9a78c92e | 1373 | static int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1374 | struct relay_connection *conn) |
b8aa1682 | 1375 | { |
fcf4d551 | 1376 | int ret = 0; |
6cd525e8 | 1377 | ssize_t size_ret; |
58eb9381 | 1378 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1379 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1380 | struct relay_stream *metadata_stream; | |
1381 | uint64_t data_size, payload_size; | |
1382 | ||
1383 | if (!session) { | |
1384 | ERR("Metadata sent before version check"); | |
1385 | ret = -1; | |
1386 | goto end; | |
1387 | } | |
1388 | ||
f6416125 MD |
1389 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1390 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1391 | ERR("Incorrect data size"); | |
1392 | ret = -1; | |
1393 | goto end; | |
1394 | } | |
1395 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1396 | ||
b8aa1682 | 1397 | if (data_buffer_size < data_size) { |
d7b3776f | 1398 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1399 | char *tmp_data_ptr; |
1400 | ||
1401 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1402 | if (!tmp_data_ptr) { | |
b8aa1682 | 1403 | ERR("Allocating data buffer"); |
c617c0c6 | 1404 | free(data_buffer); |
b8aa1682 JD |
1405 | ret = -1; |
1406 | goto end; | |
1407 | } | |
c617c0c6 | 1408 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1409 | data_buffer_size = data_size; |
1410 | } | |
1411 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1412 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
fcf4d551 JG |
1413 | size_ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
1414 | if (size_ret < 0 || size_ret != data_size) { | |
1415 | if (size_ret == 0) { | |
a6cd2b97 | 1416 | /* Orderly shutdown. Not necessary to print an error. */ |
58eb9381 | 1417 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1418 | } else { |
1419 | ERR("Relay didn't receive the whole metadata"); | |
1420 | } | |
b8aa1682 | 1421 | ret = -1; |
b8aa1682 JD |
1422 | goto end; |
1423 | } | |
1424 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 | 1425 | |
9a78c92e | 1426 | metadata_stream = stream_get_by_id(be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1427 | if (!metadata_stream) { |
1428 | ret = -1; | |
9a78c92e | 1429 | goto end; |
b8aa1682 JD |
1430 | } |
1431 | ||
9a78c92e MD |
1432 | pthread_mutex_lock(&metadata_stream->lock); |
1433 | ||
1434 | size_ret = lttng_write(metadata_stream->stream_fd->fd, metadata_struct->payload, | |
6cd525e8 MD |
1435 | payload_size); |
1436 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1437 | ERR("Relay error writing metadata on file"); |
1438 | ret = -1; | |
9a78c92e | 1439 | goto end_put; |
b8aa1682 | 1440 | } |
1d4dfdef | 1441 | |
fcf4d551 | 1442 | size_ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
1d4dfdef | 1443 | be32toh(metadata_struct->padding_size)); |
fcf4d551 | 1444 | if (size_ret < 0) { |
9a78c92e | 1445 | goto end_put; |
1d4dfdef | 1446 | } |
2a174661 | 1447 | |
9a78c92e | 1448 | metadata_stream->metadata_received += |
d3e2ba59 | 1449 | payload_size + be32toh(metadata_struct->padding_size); |
9a78c92e MD |
1450 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1451 | metadata_stream->metadata_received); | |
1d4dfdef | 1452 | |
9a78c92e MD |
1453 | end_put: |
1454 | pthread_mutex_unlock(&metadata_stream->lock); | |
1455 | stream_put(metadata_stream); | |
b8aa1682 JD |
1456 | end: |
1457 | return ret; | |
1458 | } | |
1459 | ||
1460 | /* | |
1461 | * relay_send_version: send relayd version number | |
1462 | */ | |
9a78c92e | 1463 | static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1464 | struct relay_connection *conn) |
b8aa1682 | 1465 | { |
7f51dcba | 1466 | int ret; |
092b6259 | 1467 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1468 | |
58eb9381 | 1469 | conn->version_check_done = 1; |
b8aa1682 | 1470 | |
092b6259 | 1471 | /* Get version from the other side. */ |
58eb9381 | 1472 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1473 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1474 | if (ret == 0) { |
1475 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1476 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1477 | } else { |
1478 | ERR("Relay failed to receive the version values."); | |
1479 | } | |
092b6259 | 1480 | ret = -1; |
092b6259 DG |
1481 | goto end; |
1482 | } | |
1483 | ||
53efb85a | 1484 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1485 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1486 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1487 | |
1488 | /* Major versions must be the same */ | |
1489 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1490 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1491 | reply.major, be32toh(msg.major)); | |
9a78c92e | 1492 | connection_put(conn); |
d4519fa3 JD |
1493 | ret = 0; |
1494 | goto end; | |
1495 | } | |
1496 | ||
58eb9381 | 1497 | conn->major = reply.major; |
0f907de1 JD |
1498 | /* We adapt to the lowest compatible version */ |
1499 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1500 | conn->minor = reply.minor; |
0f907de1 | 1501 | } else { |
58eb9381 | 1502 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1503 | } |
1504 | ||
6151a90f JD |
1505 | reply.major = htobe32(reply.major); |
1506 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1507 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1508 | sizeof(struct lttcomm_relayd_version), 0); |
1509 | if (ret < 0) { | |
1510 | ERR("Relay sending version"); | |
1511 | } | |
1512 | ||
58eb9381 DG |
1513 | DBG("Version check done using protocol %u.%u", conn->major, |
1514 | conn->minor); | |
b8aa1682 JD |
1515 | |
1516 | end: | |
1517 | return ret; | |
1518 | } | |
1519 | ||
c8f59ee5 | 1520 | /* |
6d805429 | 1521 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1522 | */ |
9a78c92e | 1523 | static int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1524 | struct relay_connection *conn) |
c8f59ee5 | 1525 | { |
58eb9381 | 1526 | struct relay_session *session = conn->session; |
6d805429 | 1527 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1528 | struct lttcomm_relayd_generic_reply reply; |
1529 | struct relay_stream *stream; | |
1530 | int ret; | |
c8f59ee5 DG |
1531 | uint64_t last_net_seq_num, stream_id; |
1532 | ||
6d805429 | 1533 | DBG("Data pending command received"); |
c8f59ee5 | 1534 | |
58eb9381 | 1535 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1536 | ERR("Trying to check for data before version check"); |
1537 | ret = -1; | |
1538 | goto end_no_session; | |
1539 | } | |
1540 | ||
58eb9381 | 1541 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1542 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1543 | if (ret == 0) { |
1544 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1545 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1546 | } else { |
1547 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1548 | ret); | |
1549 | } | |
c8f59ee5 DG |
1550 | ret = -1; |
1551 | goto end_no_session; | |
1552 | } | |
1553 | ||
1554 | stream_id = be64toh(msg.stream_id); | |
1555 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1556 | ||
9a78c92e | 1557 | stream = stream_get_by_id(stream_id); |
de91f48a | 1558 | if (stream == NULL) { |
c8f59ee5 | 1559 | ret = -1; |
9a78c92e | 1560 | goto end; |
c8f59ee5 DG |
1561 | } |
1562 | ||
9a78c92e MD |
1563 | pthread_mutex_lock(&stream->lock); |
1564 | ||
6d805429 | 1565 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1566 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1567 | last_net_seq_num); | |
1568 | ||
33832e64 | 1569 | /* Avoid wrapping issue */ |
39df6d9f | 1570 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1571 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1572 | ret = 0; |
6d805429 DG |
1573 | } else { |
1574 | /* Data still being streamed thus pending */ | |
1575 | ret = 1; | |
c8f59ee5 DG |
1576 | } |
1577 | ||
9a78c92e MD |
1578 | stream->data_pending_check_done = true; |
1579 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1580 | |
9a78c92e MD |
1581 | stream_put(stream); |
1582 | end: | |
c8f59ee5 | 1583 | |
53efb85a | 1584 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1585 | reply.ret_code = htobe32(ret); |
58eb9381 | 1586 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1587 | if (ret < 0) { |
6d805429 | 1588 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1589 | } |
1590 | ||
1591 | end_no_session: | |
1592 | return ret; | |
1593 | } | |
1594 | ||
1595 | /* | |
1596 | * Wait for the control socket to reach a quiescent state. | |
1597 | * | |
9a78c92e MD |
1598 | * Note that for now, when receiving this command from the session |
1599 | * daemon, this means that every subsequent commands or data received on | |
1600 | * the control socket has been handled. So, this is why we simply return | |
1601 | * OK here. | |
c8f59ee5 | 1602 | */ |
9a78c92e | 1603 | static int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1604 | struct relay_connection *conn) |
c8f59ee5 DG |
1605 | { |
1606 | int ret; | |
ad7051c0 DG |
1607 | uint64_t stream_id; |
1608 | struct relay_stream *stream; | |
ad7051c0 | 1609 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1610 | struct lttcomm_relayd_generic_reply reply; |
1611 | ||
1612 | DBG("Checking quiescent state on control socket"); | |
1613 | ||
58eb9381 | 1614 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1615 | ERR("Trying to check for data before version check"); |
1616 | ret = -1; | |
1617 | goto end_no_session; | |
1618 | } | |
1619 | ||
58eb9381 | 1620 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1621 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1622 | if (ret == 0) { |
1623 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1624 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1625 | } else { |
1626 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1627 | ret); | |
1628 | } | |
ad7051c0 DG |
1629 | ret = -1; |
1630 | goto end_no_session; | |
1631 | } | |
1632 | ||
1633 | stream_id = be64toh(msg.stream_id); | |
9a78c92e MD |
1634 | stream = stream_get_by_id(stream_id); |
1635 | if (!stream) { | |
1636 | goto reply; | |
1637 | } | |
1638 | pthread_mutex_lock(&stream->lock); | |
1639 | stream->data_pending_check_done = true; | |
1640 | pthread_mutex_unlock(&stream->lock); | |
1641 | DBG("Relay quiescent control pending flag set to %" PRIu64, stream_id); | |
1642 | stream_put(stream); | |
1643 | reply: | |
53efb85a | 1644 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1645 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1646 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1647 | if (ret < 0) { |
6d805429 | 1648 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1649 | } |
1650 | ||
ad7051c0 | 1651 | end_no_session: |
c8f59ee5 DG |
1652 | return ret; |
1653 | } | |
1654 | ||
f7079f67 | 1655 | /* |
9a78c92e MD |
1656 | * Initialize a data pending command. This means that a consumer is about |
1657 | * to ask for data pending for each stream it holds. Simply iterate over | |
1658 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1659 | * |
1660 | * This command returns to the client a LTTNG_OK code. | |
1661 | */ | |
9a78c92e | 1662 | static int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1663 | struct relay_connection *conn) |
f7079f67 DG |
1664 | { |
1665 | int ret; | |
1666 | struct lttng_ht_iter iter; | |
1667 | struct lttcomm_relayd_begin_data_pending msg; | |
1668 | struct lttcomm_relayd_generic_reply reply; | |
1669 | struct relay_stream *stream; | |
1670 | uint64_t session_id; | |
1671 | ||
1672 | assert(recv_hdr); | |
58eb9381 | 1673 | assert(conn); |
f7079f67 DG |
1674 | |
1675 | DBG("Init streams for data pending"); | |
1676 | ||
58eb9381 | 1677 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1678 | ERR("Trying to check for data before version check"); |
1679 | ret = -1; | |
1680 | goto end_no_session; | |
1681 | } | |
1682 | ||
58eb9381 | 1683 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1684 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1685 | if (ret == 0) { |
1686 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1687 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1688 | } else { |
1689 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1690 | ret); | |
1691 | } | |
f7079f67 DG |
1692 | ret = -1; |
1693 | goto end_no_session; | |
1694 | } | |
1695 | ||
1696 | session_id = be64toh(msg.session_id); | |
1697 | ||
1698 | /* | |
9a78c92e MD |
1699 | * Iterate over all streams to set the begin data pending flag. |
1700 | * For now, the streams are indexed by stream handle so we have | |
1701 | * to iterate over all streams to find the one associated with | |
1702 | * the right session_id. | |
f7079f67 DG |
1703 | */ |
1704 | rcu_read_lock(); | |
d3e2ba59 | 1705 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1706 | node.node) { |
9a78c92e MD |
1707 | if (!stream_get(stream)) { |
1708 | continue; | |
1709 | } | |
1710 | if (stream->trace->session->id == session_id) { | |
1711 | pthread_mutex_lock(&stream->lock); | |
1712 | stream->data_pending_check_done = false; | |
1713 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1714 | DBG("Set begin data pending flag to stream %" PRIu64, |
1715 | stream->stream_handle); | |
1716 | } | |
9a78c92e | 1717 | stream_put(stream); |
f7079f67 DG |
1718 | } |
1719 | rcu_read_unlock(); | |
1720 | ||
53efb85a | 1721 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1722 | /* All good, send back reply. */ |
1723 | reply.ret_code = htobe32(LTTNG_OK); | |
1724 | ||
58eb9381 | 1725 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1726 | if (ret < 0) { |
1727 | ERR("Relay begin data pending send reply failed"); | |
1728 | } | |
1729 | ||
1730 | end_no_session: | |
1731 | return ret; | |
1732 | } | |
1733 | ||
1734 | /* | |
9a78c92e MD |
1735 | * End data pending command. This will check, for a given session id, if |
1736 | * each stream associated with it has its data_pending_check_done flag | |
1737 | * set. If not, this means that the client lost track of the stream but | |
1738 | * the data is still being streamed on our side. In this case, we inform | |
1739 | * the client that data is in flight. | |
f7079f67 DG |
1740 | * |
1741 | * Return to the client if there is data in flight or not with a ret_code. | |
1742 | */ | |
9a78c92e | 1743 | static int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1744 | struct relay_connection *conn) |
f7079f67 DG |
1745 | { |
1746 | int ret; | |
1747 | struct lttng_ht_iter iter; | |
1748 | struct lttcomm_relayd_end_data_pending msg; | |
1749 | struct lttcomm_relayd_generic_reply reply; | |
1750 | struct relay_stream *stream; | |
1751 | uint64_t session_id; | |
1752 | uint32_t is_data_inflight = 0; | |
1753 | ||
f7079f67 DG |
1754 | DBG("End data pending command"); |
1755 | ||
58eb9381 | 1756 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1757 | ERR("Trying to check for data before version check"); |
1758 | ret = -1; | |
1759 | goto end_no_session; | |
1760 | } | |
1761 | ||
58eb9381 | 1762 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1763 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1764 | if (ret == 0) { |
1765 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1766 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1767 | } else { |
1768 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1769 | ret); | |
1770 | } | |
f7079f67 DG |
1771 | ret = -1; |
1772 | goto end_no_session; | |
1773 | } | |
1774 | ||
1775 | session_id = be64toh(msg.session_id); | |
1776 | ||
9a78c92e MD |
1777 | /* |
1778 | * Iterate over all streams to see if the begin data pending | |
1779 | * flag is set. | |
1780 | */ | |
f7079f67 | 1781 | rcu_read_lock(); |
d3e2ba59 | 1782 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1783 | node.node) { |
9a78c92e MD |
1784 | if (!stream_get(stream)) { |
1785 | continue; | |
1786 | } | |
1787 | if (stream->trace->session->id != session_id) { | |
1788 | stream_put(stream); | |
1789 | continue; | |
1790 | } | |
1791 | pthread_mutex_lock(&stream->lock); | |
1792 | if (!stream->data_pending_check_done) { | |
1793 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1794 | is_data_inflight = 1; | |
1795 | DBG("Data is still in flight for stream %" PRIu64, | |
1796 | stream->stream_handle); | |
1797 | pthread_mutex_unlock(&stream->lock); | |
1798 | stream_put(stream); | |
1799 | break; | |
1800 | } | |
f7079f67 | 1801 | } |
9a78c92e MD |
1802 | pthread_mutex_unlock(&stream->lock); |
1803 | stream_put(stream); | |
f7079f67 DG |
1804 | } |
1805 | rcu_read_unlock(); | |
1806 | ||
53efb85a | 1807 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1808 | /* All good, send back reply. */ |
1809 | reply.ret_code = htobe32(is_data_inflight); | |
1810 | ||
58eb9381 | 1811 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1812 | if (ret < 0) { |
1813 | ERR("Relay end data pending send reply failed"); | |
1814 | } | |
1815 | ||
1816 | end_no_session: | |
1817 | return ret; | |
1818 | } | |
1819 | ||
1c20f0e2 JD |
1820 | /* |
1821 | * Receive an index for a specific stream. | |
1822 | * | |
1823 | * Return 0 on success else a negative value. | |
1824 | */ | |
9a78c92e | 1825 | static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1826 | struct relay_connection *conn) |
1c20f0e2 | 1827 | { |
9a78c92e | 1828 | int ret, send_ret; |
58eb9381 | 1829 | struct relay_session *session = conn->session; |
1c20f0e2 | 1830 | struct lttcomm_relayd_index index_info; |
9a78c92e | 1831 | struct relay_index *index; |
1c20f0e2 JD |
1832 | struct lttcomm_relayd_generic_reply reply; |
1833 | struct relay_stream *stream; | |
1834 | uint64_t net_seq_num; | |
1835 | ||
58eb9381 | 1836 | assert(conn); |
1c20f0e2 JD |
1837 | |
1838 | DBG("Relay receiving index"); | |
1839 | ||
58eb9381 | 1840 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1841 | ERR("Trying to close a stream before version check"); |
1842 | ret = -1; | |
1843 | goto end_no_session; | |
1844 | } | |
1845 | ||
58eb9381 | 1846 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1847 | sizeof(index_info), 0); |
1848 | if (ret < sizeof(index_info)) { | |
1849 | if (ret == 0) { | |
1850 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1851 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1852 | } else { |
1853 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1854 | } | |
1855 | ret = -1; | |
1856 | goto end_no_session; | |
1857 | } | |
1858 | ||
1859 | net_seq_num = be64toh(index_info.net_seq_num); | |
1860 | ||
9a78c92e | 1861 | stream = stream_get_by_id(be64toh(index_info.relay_stream_id)); |
1c20f0e2 | 1862 | if (!stream) { |
9a78c92e | 1863 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 1864 | ret = -1; |
9a78c92e | 1865 | goto end; |
1c20f0e2 | 1866 | } |
9a78c92e | 1867 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 1868 | |
d3e2ba59 JD |
1869 | /* Live beacon handling */ |
1870 | if (index_info.packet_size == 0) { | |
9a78c92e MD |
1871 | DBG("Received live beacon for stream %" PRIu64, |
1872 | stream->stream_handle); | |
d3e2ba59 JD |
1873 | |
1874 | /* | |
9a78c92e MD |
1875 | * Only flag a stream inactive when it has already |
1876 | * received data and no indexes are in flight. | |
d3e2ba59 | 1877 | */ |
c00978f1 | 1878 | if (stream->index_received_seqcount > 0 |
9a78c92e MD |
1879 | && stream->indexes_in_flight == 0) { |
1880 | stream->beacon_ts_end = | |
1881 | be64toh(index_info.timestamp_end); | |
d3e2ba59 JD |
1882 | } |
1883 | ret = 0; | |
9a78c92e | 1884 | goto end_stream_put; |
d3e2ba59 JD |
1885 | } else { |
1886 | stream->beacon_ts_end = -1ULL; | |
1887 | } | |
1888 | ||
528f2ffa JD |
1889 | if (stream->ctf_stream_id == -1ULL) { |
1890 | stream->ctf_stream_id = be64toh(index_info.stream_id); | |
1891 | } | |
9a78c92e MD |
1892 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
1893 | if (!index) { | |
1894 | ret = -1; | |
1895 | ERR("relay_index_get_by_id_or_create index NULL"); | |
1896 | goto end_stream_put; | |
1c20f0e2 | 1897 | } |
9a78c92e MD |
1898 | if (set_index_control_data(index, &index_info)) { |
1899 | ERR("set_index_control_data error"); | |
1900 | relay_index_put(index); | |
1901 | ret = -1; | |
1902 | goto end_stream_put; | |
1903 | } | |
1904 | ret = relay_index_try_flush(index); | |
1905 | if (ret == 0) { | |
c00978f1 MD |
1906 | tracefile_array_commit_seq(stream->tfa); |
1907 | stream->index_received_seqcount++; | |
9a78c92e MD |
1908 | } else if (ret > 0) { |
1909 | /* no flush. */ | |
1910 | ret = 0; | |
1911 | } else { | |
1912 | ERR("relay_index_try_flush error %d", ret); | |
1913 | relay_index_put(index); | |
1914 | ret = -1; | |
1c20f0e2 JD |
1915 | } |
1916 | ||
9a78c92e MD |
1917 | end_stream_put: |
1918 | pthread_mutex_unlock(&stream->lock); | |
1919 | stream_put(stream); | |
1920 | ||
1921 | end: | |
1c20f0e2 | 1922 | |
53efb85a | 1923 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
1924 | if (ret < 0) { |
1925 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1926 | } else { | |
1927 | reply.ret_code = htobe32(LTTNG_OK); | |
1928 | } | |
58eb9381 | 1929 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
1930 | if (send_ret < 0) { |
1931 | ERR("Relay sending close index id reply"); | |
1932 | ret = send_ret; | |
1933 | } | |
1934 | ||
1935 | end_no_session: | |
1936 | return ret; | |
1937 | } | |
1938 | ||
a4baae1b JD |
1939 | /* |
1940 | * Receive the streams_sent message. | |
1941 | * | |
1942 | * Return 0 on success else a negative value. | |
1943 | */ | |
9a78c92e | 1944 | static int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1945 | struct relay_connection *conn) |
a4baae1b JD |
1946 | { |
1947 | int ret, send_ret; | |
1948 | struct lttcomm_relayd_generic_reply reply; | |
1949 | ||
58eb9381 | 1950 | assert(conn); |
a4baae1b JD |
1951 | |
1952 | DBG("Relay receiving streams_sent"); | |
1953 | ||
58eb9381 | 1954 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
1955 | ERR("Trying to close a stream before version check"); |
1956 | ret = -1; | |
1957 | goto end_no_session; | |
1958 | } | |
1959 | ||
1960 | /* | |
9a78c92e MD |
1961 | * Publish every pending stream in the connection recv list which are |
1962 | * now ready to be used by the viewer. | |
a4baae1b | 1963 | */ |
9a78c92e | 1964 | publish_connection_local_streams(conn); |
4a9daf17 | 1965 | |
53efb85a | 1966 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 1967 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1968 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
1969 | if (send_ret < 0) { |
1970 | ERR("Relay sending sent_stream reply"); | |
1971 | ret = send_ret; | |
1972 | } else { | |
1973 | /* Success. */ | |
1974 | ret = 0; | |
1975 | } | |
1976 | ||
1977 | end_no_session: | |
1978 | return ret; | |
1979 | } | |
1980 | ||
b8aa1682 | 1981 | /* |
d3e2ba59 | 1982 | * Process the commands received on the control socket |
b8aa1682 | 1983 | */ |
9a78c92e | 1984 | static int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1985 | struct relay_connection *conn) |
b8aa1682 JD |
1986 | { |
1987 | int ret = 0; | |
1988 | ||
1989 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 1990 | case RELAYD_CREATE_SESSION: |
58eb9381 | 1991 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 1992 | break; |
b8aa1682 | 1993 | case RELAYD_ADD_STREAM: |
58eb9381 | 1994 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
1995 | break; |
1996 | case RELAYD_START_DATA: | |
58eb9381 | 1997 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
1998 | break; |
1999 | case RELAYD_SEND_METADATA: | |
58eb9381 | 2000 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
2001 | break; |
2002 | case RELAYD_VERSION: | |
58eb9381 | 2003 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 2004 | break; |
173af62f | 2005 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 2006 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 2007 | break; |
6d805429 | 2008 | case RELAYD_DATA_PENDING: |
58eb9381 | 2009 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
2010 | break; |
2011 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 2012 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 2013 | break; |
f7079f67 | 2014 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 2015 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
2016 | break; |
2017 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2018 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2019 | break; |
1c20f0e2 | 2020 | case RELAYD_SEND_INDEX: |
58eb9381 | 2021 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2022 | break; |
a4baae1b | 2023 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2024 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2025 | break; |
b8aa1682 JD |
2026 | case RELAYD_UPDATE_SYNC_INFO: |
2027 | default: | |
2028 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2029 | relay_unknown_command(conn); |
b8aa1682 JD |
2030 | ret = -1; |
2031 | goto end; | |
2032 | } | |
2033 | ||
2034 | end: | |
2035 | return ret; | |
2036 | } | |
2037 | ||
7d2f7452 DG |
2038 | /* |
2039 | * Handle index for a data stream. | |
2040 | * | |
9a78c92e | 2041 | * Called with the stream lock held. |
7d2f7452 DG |
2042 | * |
2043 | * Return 0 on success else a negative value. | |
2044 | */ | |
2045 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2046 | int rotate_index) | |
2047 | { | |
9a78c92e MD |
2048 | int ret = 0; |
2049 | uint64_t data_offset; | |
2050 | struct relay_index *index; | |
7d2f7452 | 2051 | |
7d2f7452 DG |
2052 | /* Get data offset because we are about to update the index. */ |
2053 | data_offset = htobe64(stream->tracefile_size_current); | |
2054 | ||
9a78c92e MD |
2055 | DBG("handle_index_data: stream %" PRIu64 " data offset %" PRIu64, |
2056 | stream->stream_handle, stream->tracefile_size_current); | |
2057 | ||
7d2f7452 | 2058 | /* |
9a78c92e MD |
2059 | * Lookup for an existing index for that stream id/sequence |
2060 | * number. If it exists, the control thread has already received the | |
2061 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2062 | */ |
9a78c92e | 2063 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2064 | if (!index) { |
9a78c92e MD |
2065 | ret = -1; |
2066 | goto end; | |
7d2f7452 DG |
2067 | } |
2068 | ||
9a78c92e MD |
2069 | if (rotate_index || !stream->index_fd) { |
2070 | int fd; | |
2071 | ||
2072 | /* Put ref on previous index_fd. */ | |
2073 | if (stream->index_fd) { | |
2074 | stream_fd_put(stream->index_fd); | |
2075 | stream->index_fd = NULL; | |
7d2f7452 | 2076 | } |
7d2f7452 | 2077 | |
9a78c92e MD |
2078 | fd = index_create_file(stream->path_name, stream->channel_name, |
2079 | -1, -1, stream->tracefile_size, | |
c00978f1 | 2080 | tracefile_array_get_file_index_head(stream->tfa)); |
9a78c92e MD |
2081 | if (fd < 0) { |
2082 | ret = -1; | |
2083 | /* Put self-ref for this index due to error. */ | |
2084 | relay_index_put(index); | |
2085 | goto end; | |
2086 | } | |
2087 | stream->index_fd = stream_fd_create(fd); | |
2088 | if (!stream->index_fd) { | |
2089 | ret = -1; | |
2090 | if (close(fd)) { | |
2091 | PERROR("Error closing FD %d", fd); | |
2092 | } | |
2093 | /* Put self-ref for this index due to error. */ | |
2094 | relay_index_put(index); | |
2095 | /* Will put the local ref. */ | |
2096 | goto end; | |
7d2f7452 | 2097 | } |
7d2f7452 DG |
2098 | } |
2099 | ||
9a78c92e MD |
2100 | if (relay_index_set_fd(index, stream->index_fd, data_offset)) { |
2101 | ret = -1; | |
2102 | /* Put self-ref for this index due to error. */ | |
2103 | relay_index_put(index); | |
2104 | goto end; | |
7d2f7452 DG |
2105 | } |
2106 | ||
9a78c92e MD |
2107 | ret = relay_index_try_flush(index); |
2108 | if (ret == 0) { | |
c00978f1 MD |
2109 | tracefile_array_commit_seq(stream->tfa); |
2110 | stream->index_received_seqcount++; | |
9a78c92e MD |
2111 | } else if (ret > 0) { |
2112 | /* No flush. */ | |
2113 | ret = 0; | |
2114 | } else { | |
2115 | /* Put self-ref for this index due to error. */ | |
2116 | relay_index_put(index); | |
2117 | ret = -1; | |
2118 | } | |
2119 | end: | |
7d2f7452 DG |
2120 | return ret; |
2121 | } | |
2122 | ||
b8aa1682 JD |
2123 | /* |
2124 | * relay_process_data: Process the data received on the data socket | |
2125 | */ | |
9a78c92e | 2126 | static int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2127 | { |
7d2f7452 | 2128 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2129 | ssize_t size_ret; |
b8aa1682 JD |
2130 | struct relay_stream *stream; |
2131 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2132 | uint64_t stream_id; |
173af62f | 2133 | uint64_t net_seq_num; |
b8aa1682 | 2134 | uint32_t data_size; |
2a174661 | 2135 | struct relay_session *session; |
2a5a6d16 | 2136 | bool new_stream = false, close_requested = false; |
b8aa1682 | 2137 | |
58eb9381 | 2138 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, |
7c5aef62 | 2139 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2140 | if (ret <= 0) { |
a6cd2b97 DG |
2141 | if (ret == 0) { |
2142 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2143 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2144 | } else { |
58eb9381 | 2145 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2146 | } |
b8aa1682 JD |
2147 | ret = -1; |
2148 | goto end; | |
2149 | } | |
2150 | ||
2151 | stream_id = be64toh(data_hdr.stream_id); | |
9a78c92e | 2152 | stream = stream_get_by_id(stream_id); |
b8aa1682 JD |
2153 | if (!stream) { |
2154 | ret = -1; | |
9a78c92e | 2155 | goto end; |
b8aa1682 | 2156 | } |
9a78c92e | 2157 | session = stream->trace->session; |
b8aa1682 JD |
2158 | data_size = be32toh(data_hdr.data_size); |
2159 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2160 | char *tmp_data_ptr; |
2161 | ||
2162 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2163 | if (!tmp_data_ptr) { | |
b8aa1682 | 2164 | ERR("Allocating data buffer"); |
c617c0c6 | 2165 | free(data_buffer); |
b8aa1682 | 2166 | ret = -1; |
9a78c92e | 2167 | goto end_stream_put; |
b8aa1682 | 2168 | } |
c617c0c6 | 2169 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2170 | data_buffer_size = data_size; |
2171 | } | |
2172 | memset(data_buffer, 0, data_size); | |
2173 | ||
173af62f DG |
2174 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2175 | ||
77c7c900 | 2176 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2177 | data_size, stream_id, net_seq_num); |
58eb9381 | 2178 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 2179 | if (ret <= 0) { |
a6cd2b97 DG |
2180 | if (ret == 0) { |
2181 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2182 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2183 | } |
b8aa1682 | 2184 | ret = -1; |
9a78c92e | 2185 | goto end_stream_put; |
b8aa1682 JD |
2186 | } |
2187 | ||
9a78c92e MD |
2188 | pthread_mutex_lock(&stream->lock); |
2189 | ||
1c20f0e2 | 2190 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2191 | if (stream->tracefile_size > 0 && |
2192 | (stream->tracefile_size_current + data_size) > | |
2193 | stream->tracefile_size) { | |
c00978f1 MD |
2194 | uint64_t old_id, new_id; |
2195 | ||
2196 | old_id = tracefile_array_get_file_index_head(stream->tfa); | |
2197 | tracefile_array_file_rotate(stream->tfa); | |
2198 | ||
2199 | /* new_id is updated by utils_rotate_stream_file. */ | |
2200 | new_id = old_id; | |
6b6b9a5a | 2201 | |
9a78c92e MD |
2202 | ret = utils_rotate_stream_file(stream->path_name, |
2203 | stream->channel_name, stream->tracefile_size, | |
2204 | stream->tracefile_count, -1, | |
2205 | -1, stream->stream_fd->fd, | |
c00978f1 | 2206 | &new_id, &stream->stream_fd->fd); |
0f907de1 | 2207 | if (ret < 0) { |
1c20f0e2 | 2208 | ERR("Rotating stream output file"); |
9a78c92e | 2209 | goto end_stream_unlock; |
0f907de1 | 2210 | } |
9a78c92e MD |
2211 | /* |
2212 | * Reset current size because we just performed a stream | |
2213 | * rotation. | |
2214 | */ | |
a6976990 | 2215 | stream->tracefile_size_current = 0; |
1c20f0e2 JD |
2216 | rotate_index = 1; |
2217 | } | |
2218 | ||
1c20f0e2 | 2219 | /* |
9a78c92e MD |
2220 | * Index are handled in protocol version 2.4 and above. Also, |
2221 | * snapshot and index are NOT supported. | |
1c20f0e2 | 2222 | */ |
2a174661 | 2223 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2224 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2225 | if (ret < 0) { |
9a78c92e | 2226 | goto end_stream_unlock; |
1c20f0e2 | 2227 | } |
1c20f0e2 JD |
2228 | } |
2229 | ||
7d2f7452 | 2230 | /* Write data to stream output fd. */ |
9a78c92e | 2231 | size_ret = lttng_write(stream->stream_fd->fd, data_buffer, data_size); |
6cd525e8 | 2232 | if (size_ret < data_size) { |
b8aa1682 JD |
2233 | ERR("Relay error writing data to file"); |
2234 | ret = -1; | |
9a78c92e | 2235 | goto end_stream_unlock; |
b8aa1682 | 2236 | } |
1d4dfdef | 2237 | |
9a78c92e MD |
2238 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, |
2239 | size_ret, stream->stream_handle); | |
5ab7344e | 2240 | |
9a78c92e MD |
2241 | ret = write_padding_to_file(stream->stream_fd->fd, |
2242 | be32toh(data_hdr.padding_size)); | |
1d4dfdef | 2243 | if (ret < 0) { |
9a78c92e | 2244 | goto end_stream_unlock; |
1d4dfdef | 2245 | } |
9a78c92e MD |
2246 | stream->tracefile_size_current += |
2247 | data_size + be32toh(data_hdr.padding_size); | |
b8ee61ff MD |
2248 | if (stream->prev_seq == -1ULL) { |
2249 | new_stream = true; | |
2250 | } | |
2251 | ||
173af62f DG |
2252 | stream->prev_seq = net_seq_num; |
2253 | ||
9a78c92e | 2254 | end_stream_unlock: |
2a5a6d16 | 2255 | close_requested = stream->close_requested; |
9a78c92e | 2256 | pthread_mutex_unlock(&stream->lock); |
2a5a6d16 JG |
2257 | if (close_requested) { |
2258 | try_stream_close(stream); | |
2259 | } | |
2260 | ||
b8ee61ff MD |
2261 | if (new_stream) { |
2262 | pthread_mutex_lock(&session->lock); | |
2263 | uatomic_set(&session->new_streams, 1); | |
2264 | pthread_mutex_unlock(&session->lock); | |
2265 | } | |
9a78c92e MD |
2266 | end_stream_put: |
2267 | stream_put(stream); | |
b8aa1682 JD |
2268 | end: |
2269 | return ret; | |
2270 | } | |
2271 | ||
9a78c92e | 2272 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2273 | { |
2274 | int ret; | |
2275 | ||
58eb9381 | 2276 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2277 | |
2278 | ret = close(pollfd); | |
2279 | if (ret < 0) { | |
2280 | ERR("Closing pollfd %d", pollfd); | |
2281 | } | |
2282 | } | |
2283 | ||
9a78c92e MD |
2284 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2285 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2286 | { |
9a78c92e | 2287 | const char *type_str; |
2a174661 | 2288 | |
9a78c92e MD |
2289 | switch (conn->type) { |
2290 | case RELAY_DATA: | |
2291 | type_str = "Data"; | |
2292 | break; | |
2293 | case RELAY_CONTROL: | |
2294 | type_str = "Control"; | |
2295 | break; | |
2296 | case RELAY_VIEWER_COMMAND: | |
2297 | type_str = "Viewer Command"; | |
2298 | break; | |
2299 | case RELAY_VIEWER_NOTIFICATION: | |
2300 | type_str = "Viewer Notification"; | |
2301 | break; | |
2302 | default: | |
2303 | type_str = "Unknown"; | |
9d1bbf21 | 2304 | } |
9a78c92e MD |
2305 | cleanup_connection_pollfd(events, pollfd); |
2306 | connection_put(conn); | |
2307 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2308 | } |
2309 | ||
2310 | /* | |
2311 | * This thread does the actual work | |
2312 | */ | |
9a78c92e | 2313 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2314 | { |
beaad64c DG |
2315 | int ret, err = -1, last_seen_data_fd = -1; |
2316 | uint32_t nb_fd; | |
b8aa1682 JD |
2317 | struct lttng_poll_event events; |
2318 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2319 | struct lttng_ht_iter iter; |
b8aa1682 | 2320 | struct lttcomm_relayd_hdr recv_hdr; |
9a78c92e | 2321 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2322 | |
2323 | DBG("[thread] Relay worker started"); | |
2324 | ||
9d1bbf21 MD |
2325 | rcu_register_thread(); |
2326 | ||
55706a7d MD |
2327 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2328 | ||
9b5e0863 MD |
2329 | if (testpoint(relayd_thread_worker)) { |
2330 | goto error_testpoint; | |
2331 | } | |
2332 | ||
f385ae0a MD |
2333 | health_code_update(); |
2334 | ||
b8aa1682 JD |
2335 | /* table of connections indexed on socket */ |
2336 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2337 | if (!relay_connections_ht) { |
2338 | goto relay_connections_ht_error; | |
2339 | } | |
b8aa1682 | 2340 | |
b8aa1682 JD |
2341 | ret = create_thread_poll_set(&events, 2); |
2342 | if (ret < 0) { | |
2343 | goto error_poll_create; | |
2344 | } | |
2345 | ||
58eb9381 | 2346 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2347 | if (ret < 0) { |
2348 | goto error; | |
2349 | } | |
2350 | ||
beaad64c | 2351 | restart: |
b8aa1682 | 2352 | while (1) { |
beaad64c DG |
2353 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2354 | ||
f385ae0a MD |
2355 | health_code_update(); |
2356 | ||
b8aa1682 | 2357 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2358 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2359 | health_poll_entry(); |
b8aa1682 | 2360 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2361 | health_poll_exit(); |
b8aa1682 JD |
2362 | if (ret < 0) { |
2363 | /* | |
2364 | * Restart interrupted system call. | |
2365 | */ | |
2366 | if (errno == EINTR) { | |
2367 | goto restart; | |
2368 | } | |
2369 | goto error; | |
2370 | } | |
2371 | ||
0d9c5d77 DG |
2372 | nb_fd = ret; |
2373 | ||
beaad64c | 2374 | /* |
9a78c92e MD |
2375 | * Process control. The control connection is |
2376 | * prioritized so we don't starve it with high | |
2377 | * throughput tracing data on the data connection. | |
beaad64c | 2378 | */ |
b8aa1682 JD |
2379 | for (i = 0; i < nb_fd; i++) { |
2380 | /* Fetch once the poll data */ | |
beaad64c DG |
2381 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2382 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2383 | |
f385ae0a MD |
2384 | health_code_update(); |
2385 | ||
e0077699 | 2386 | if (!revents) { |
9a78c92e MD |
2387 | /* |
2388 | * No activity for this FD (poll | |
2389 | * implementation). | |
2390 | */ | |
e0077699 MD |
2391 | continue; |
2392 | } | |
2393 | ||
b8aa1682 JD |
2394 | /* Thread quit pipe has been closed. Killing thread. */ |
2395 | ret = check_thread_quit_pipe(pollfd, revents); | |
2396 | if (ret) { | |
095a4ae5 MD |
2397 | err = 0; |
2398 | goto exit; | |
b8aa1682 JD |
2399 | } |
2400 | ||
58eb9381 DG |
2401 | /* Inspect the relay conn pipe for new connection */ |
2402 | if (pollfd == relay_conn_pipe[0]) { | |
9a78c92e MD |
2403 | if (revents & LPOLLIN) { |
2404 | struct relay_connection *conn; | |
2405 | ||
58eb9381 | 2406 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2407 | if (ret < 0) { |
2408 | goto error; | |
2409 | } | |
58eb9381 DG |
2410 | lttng_poll_add(&events, conn->sock->fd, |
2411 | LPOLLIN | LPOLLRDHUP); | |
9a78c92e | 2412 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2413 | DBG("Connection socket %d added", conn->sock->fd); |
9a78c92e MD |
2414 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2415 | ERR("Relay connection pipe error"); | |
2416 | goto error; | |
2417 | } else { | |
2418 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
2419 | goto error; | |
b8aa1682 | 2420 | } |
58eb9381 | 2421 | } else { |
9a78c92e MD |
2422 | struct relay_connection *ctrl_conn; |
2423 | ||
2424 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); | |
58eb9381 | 2425 | /* If not found, there is a synchronization issue. */ |
9a78c92e | 2426 | assert(ctrl_conn); |
58eb9381 | 2427 | |
9a78c92e MD |
2428 | if (ctrl_conn->type == RELAY_DATA) { |
2429 | if (revents & LPOLLIN) { | |
beaad64c DG |
2430 | /* |
2431 | * Flag the last seen data fd not deleted. It will be | |
2432 | * used as the last seen fd if any fd gets deleted in | |
2433 | * this first loop. | |
2434 | */ | |
2435 | last_notdel_data_fd = pollfd; | |
2436 | } | |
9a78c92e MD |
2437 | goto put_ctrl_connection; |
2438 | } | |
2439 | assert(ctrl_conn->type == RELAY_CONTROL); | |
2440 | ||
2441 | if (revents & LPOLLIN) { | |
2442 | ret = ctrl_conn->sock->ops->recvmsg(ctrl_conn->sock, | |
2443 | &recv_hdr, sizeof(recv_hdr), 0); | |
2444 | if (ret <= 0) { | |
2445 | /* Connection closed */ | |
2446 | relay_thread_close_connection(&events, pollfd, | |
2447 | ctrl_conn); | |
2448 | } else { | |
2449 | ret = relay_process_control(&recv_hdr, ctrl_conn); | |
2450 | if (ret < 0) { | |
2451 | /* Clear the session on error. */ | |
2452 | relay_thread_close_connection(&events, | |
2453 | pollfd, ctrl_conn); | |
2454 | } | |
2455 | seen_control = 1; | |
2456 | } | |
2457 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2458 | relay_thread_close_connection(&events, | |
2459 | pollfd, ctrl_conn); | |
2460 | if (last_seen_data_fd == pollfd) { | |
2461 | last_seen_data_fd = last_notdel_data_fd; | |
2462 | } | |
58eb9381 | 2463 | } else { |
9a78c92e MD |
2464 | ERR("Unexpected poll events %u for control sock %d", |
2465 | revents, pollfd); | |
2466 | connection_put(ctrl_conn); | |
2467 | goto error; | |
beaad64c | 2468 | } |
9a78c92e MD |
2469 | put_ctrl_connection: |
2470 | connection_put(ctrl_conn); | |
beaad64c DG |
2471 | } |
2472 | } | |
2473 | ||
2474 | /* | |
2475 | * The last loop handled a control request, go back to poll to make | |
2476 | * sure we prioritise the control socket. | |
2477 | */ | |
2478 | if (seen_control) { | |
2479 | continue; | |
2480 | } | |
2481 | ||
2482 | if (last_seen_data_fd >= 0) { | |
2483 | for (i = 0; i < nb_fd; i++) { | |
2484 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2485 | |
2486 | health_code_update(); | |
2487 | ||
beaad64c DG |
2488 | if (last_seen_data_fd == pollfd) { |
2489 | idx = i; | |
2490 | break; | |
2491 | } | |
2492 | } | |
2493 | } | |
2494 | ||
2495 | /* Process data connection. */ | |
2496 | for (i = idx + 1; i < nb_fd; i++) { | |
2497 | /* Fetch the poll data. */ | |
2498 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2499 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
9a78c92e | 2500 | struct relay_connection *data_conn; |
beaad64c | 2501 | |
f385ae0a MD |
2502 | health_code_update(); |
2503 | ||
e0077699 MD |
2504 | if (!revents) { |
2505 | /* No activity for this FD (poll implementation). */ | |
2506 | continue; | |
2507 | } | |
2508 | ||
beaad64c | 2509 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2510 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2511 | continue; |
2512 | } | |
2513 | ||
9a78c92e MD |
2514 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
2515 | if (!data_conn) { | |
e0077699 | 2516 | /* Skip it. Might be removed before. */ |
e0077699 MD |
2517 | continue; |
2518 | } | |
9a78c92e MD |
2519 | if (data_conn->type == RELAY_CONTROL) { |
2520 | goto put_data_connection; | |
2521 | } | |
2522 | assert(data_conn->type == RELAY_DATA); | |
e0077699 MD |
2523 | |
2524 | if (revents & LPOLLIN) { | |
9a78c92e | 2525 | ret = relay_process_data(data_conn); |
e0077699 MD |
2526 | /* Connection closed */ |
2527 | if (ret < 0) { | |
9a78c92e MD |
2528 | relay_thread_close_connection(&events, pollfd, |
2529 | data_conn); | |
e0077699 MD |
2530 | /* |
2531 | * Every goto restart call sets the last seen fd where | |
2532 | * here we don't really care since we gracefully | |
2533 | * continue the loop after the connection is deleted. | |
2534 | */ | |
2535 | } else { | |
2536 | /* Keep last seen port. */ | |
2537 | last_seen_data_fd = pollfd; | |
9a78c92e | 2538 | connection_put(data_conn); |
e0077699 | 2539 | goto restart; |
b8aa1682 | 2540 | } |
9a78c92e MD |
2541 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2542 | relay_thread_close_connection(&events, pollfd, | |
2543 | data_conn); | |
2544 | } else { | |
2545 | ERR("Unknown poll events %u for data sock %d", | |
2546 | revents, pollfd); | |
b8aa1682 | 2547 | } |
9a78c92e MD |
2548 | put_data_connection: |
2549 | connection_put(data_conn); | |
b8aa1682 | 2550 | } |
beaad64c | 2551 | last_seen_data_fd = -1; |
b8aa1682 JD |
2552 | } |
2553 | ||
f385ae0a MD |
2554 | /* Normal exit, no error */ |
2555 | ret = 0; | |
2556 | ||
095a4ae5 | 2557 | exit: |
b8aa1682 | 2558 | error: |
58eb9381 | 2559 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2560 | rcu_read_lock(); |
9a78c92e MD |
2561 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
2562 | destroy_conn, | |
58eb9381 | 2563 | sock_n.node) { |
f385ae0a | 2564 | health_code_update(); |
9a78c92e MD |
2565 | /* |
2566 | * No need to grab another ref, because we own | |
2567 | * destroy_conn. | |
2568 | */ | |
2569 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
2570 | destroy_conn); | |
b8aa1682 | 2571 | } |
94d49140 | 2572 | rcu_read_unlock(); |
9a78c92e MD |
2573 | |
2574 | lttng_poll_clean(&events); | |
7d2f7452 | 2575 | error_poll_create: |
b8aa1682 | 2576 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2577 | relay_connections_ht_error: |
58eb9381 DG |
2578 | /* Close relay conn pipes */ |
2579 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2580 | if (err) { |
2581 | DBG("Thread exited with error"); | |
2582 | } | |
b8aa1682 | 2583 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2584 | free(data_buffer); |
9b5e0863 | 2585 | error_testpoint: |
f385ae0a MD |
2586 | if (err) { |
2587 | health_error(); | |
2588 | ERR("Health error occurred in %s", __func__); | |
2589 | } | |
2590 | health_unregister(health_relayd); | |
9d1bbf21 | 2591 | rcu_unregister_thread(); |
770f96f4 | 2592 | lttng_relay_stop_threads(); |
b8aa1682 JD |
2593 | return NULL; |
2594 | } | |
2595 | ||
2596 | /* | |
2597 | * Create the relay command pipe to wake thread_manage_apps. | |
2598 | * Closed in cleanup(). | |
2599 | */ | |
58eb9381 | 2600 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2601 | { |
a02de639 | 2602 | int ret; |
b8aa1682 | 2603 | |
58eb9381 | 2604 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2605 | |
b8aa1682 JD |
2606 | return ret; |
2607 | } | |
2608 | ||
2609 | /* | |
2610 | * main | |
2611 | */ | |
2612 | int main(int argc, char **argv) | |
2613 | { | |
77c54fed | 2614 | int ret = 0, retval = 0; |
b8aa1682 JD |
2615 | void *status; |
2616 | ||
b8aa1682 JD |
2617 | /* Parse arguments */ |
2618 | progname = argv[0]; | |
77c54fed MD |
2619 | if (set_options(argc, argv)) { |
2620 | retval = -1; | |
2621 | goto exit_options; | |
b8aa1682 JD |
2622 | } |
2623 | ||
77c54fed MD |
2624 | if (set_signal_handler()) { |
2625 | retval = -1; | |
2626 | goto exit_options; | |
b8aa1682 JD |
2627 | } |
2628 | ||
4d513a50 DG |
2629 | /* Try to create directory if -o, --output is specified. */ |
2630 | if (opt_output_path) { | |
994fa64f DG |
2631 | if (*opt_output_path != '/') { |
2632 | ERR("Please specify an absolute path for -o, --output PATH"); | |
77c54fed MD |
2633 | retval = -1; |
2634 | goto exit_options; | |
994fa64f DG |
2635 | } |
2636 | ||
0d083fb8 JG |
2637 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
2638 | -1, -1); | |
4d513a50 DG |
2639 | if (ret < 0) { |
2640 | ERR("Unable to create %s", opt_output_path); | |
77c54fed MD |
2641 | retval = -1; |
2642 | goto exit_options; | |
4d513a50 DG |
2643 | } |
2644 | } | |
2645 | ||
b8aa1682 | 2646 | /* Daemonize */ |
b5218ffb | 2647 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2648 | int i; |
2649 | ||
2650 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2651 | !opt_background); | |
b8aa1682 | 2652 | if (ret < 0) { |
77c54fed MD |
2653 | retval = -1; |
2654 | goto exit_options; | |
b8aa1682 | 2655 | } |
3fd27398 MD |
2656 | |
2657 | /* | |
2658 | * We are in the child. Make sure all other file | |
2659 | * descriptors are closed, in case we are called with | |
2660 | * more opened file descriptors than the standard ones. | |
2661 | */ | |
2662 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2663 | (void) close(i); | |
2664 | } | |
2665 | } | |
2666 | ||
77c54fed MD |
2667 | |
2668 | /* Initialize thread health monitoring */ | |
2669 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2670 | if (!health_relayd) { | |
2671 | PERROR("health_app_create error"); | |
2672 | retval = -1; | |
2673 | goto exit_health_app_create; | |
2674 | } | |
2675 | ||
3fd27398 | 2676 | /* Create thread quit pipe */ |
77c54fed MD |
2677 | if (init_thread_quit_pipe()) { |
2678 | retval = -1; | |
2679 | goto exit_init_data; | |
b8aa1682 JD |
2680 | } |
2681 | ||
1c20f0e2 | 2682 | /* Check if daemon is UID = 0 */ |
9a78c92e | 2683 | if (!getuid()) { |
8d5c808e | 2684 | if (control_uri->port < 1024 || data_uri->port < 1024 || live_uri->port < 1024) { |
b8aa1682 | 2685 | ERR("Need to be root to use ports < 1024"); |
77c54fed MD |
2686 | retval = -1; |
2687 | goto exit_init_data; | |
b8aa1682 JD |
2688 | } |
2689 | } | |
2690 | ||
2691 | /* Setup the thread apps communication pipe. */ | |
77c54fed MD |
2692 | if (create_relay_conn_pipe()) { |
2693 | retval = -1; | |
2694 | goto exit_init_data; | |
b8aa1682 JD |
2695 | } |
2696 | ||
2697 | /* Init relay command queue. */ | |
8bdee6e2 | 2698 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 JD |
2699 | |
2700 | /* Set up max poll set size */ | |
2701 | lttng_poll_set_max_size(); | |
2702 | ||
554831e7 MD |
2703 | /* Initialize communication library */ |
2704 | lttcomm_init(); | |
87e45c13 | 2705 | lttcomm_inet_init(); |
554831e7 | 2706 | |
d3e2ba59 | 2707 | /* tables of sessions indexed by session ID */ |
9a78c92e MD |
2708 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2709 | if (!sessions_ht) { | |
77c54fed MD |
2710 | retval = -1; |
2711 | goto exit_init_data; | |
d3e2ba59 JD |
2712 | } |
2713 | ||
2714 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2715 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 2716 | if (!relay_streams_ht) { |
77c54fed MD |
2717 | retval = -1; |
2718 | goto exit_init_data; | |
d3e2ba59 JD |
2719 | } |
2720 | ||
2721 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2722 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2723 | if (!viewer_streams_ht) { | |
77c54fed MD |
2724 | retval = -1; |
2725 | goto exit_init_data; | |
55706a7d MD |
2726 | } |
2727 | ||
65931c8b | 2728 | ret = utils_create_pipe(health_quit_pipe); |
77c54fed MD |
2729 | if (ret) { |
2730 | retval = -1; | |
2731 | goto exit_health_quit_pipe; | |
65931c8b MD |
2732 | } |
2733 | ||
2734 | /* Create thread to manage the client socket */ | |
2735 | ret = pthread_create(&health_thread, NULL, | |
2736 | thread_manage_health, (void *) NULL); | |
77c54fed MD |
2737 | if (ret) { |
2738 | errno = ret; | |
65931c8b | 2739 | PERROR("pthread_create health"); |
77c54fed MD |
2740 | retval = -1; |
2741 | goto exit_health_thread; | |
65931c8b MD |
2742 | } |
2743 | ||
b8aa1682 JD |
2744 | /* Setup the dispatcher thread */ |
2745 | ret = pthread_create(&dispatcher_thread, NULL, | |
2746 | relay_thread_dispatcher, (void *) NULL); | |
77c54fed MD |
2747 | if (ret) { |
2748 | errno = ret; | |
b8aa1682 | 2749 | PERROR("pthread_create dispatcher"); |
77c54fed MD |
2750 | retval = -1; |
2751 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
2752 | } |
2753 | ||
2754 | /* Setup the worker thread */ | |
2755 | ret = pthread_create(&worker_thread, NULL, | |
9a78c92e | 2756 | relay_thread_worker, NULL); |
77c54fed MD |
2757 | if (ret) { |
2758 | errno = ret; | |
b8aa1682 | 2759 | PERROR("pthread_create worker"); |
77c54fed MD |
2760 | retval = -1; |
2761 | goto exit_worker_thread; | |
b8aa1682 JD |
2762 | } |
2763 | ||
2764 | /* Setup the listener thread */ | |
2765 | ret = pthread_create(&listener_thread, NULL, | |
2766 | relay_thread_listener, (void *) NULL); | |
77c54fed MD |
2767 | if (ret) { |
2768 | errno = ret; | |
b8aa1682 | 2769 | PERROR("pthread_create listener"); |
77c54fed MD |
2770 | retval = -1; |
2771 | goto exit_listener_thread; | |
b8aa1682 JD |
2772 | } |
2773 | ||
9a78c92e | 2774 | ret = relayd_live_create(live_uri); |
77c54fed | 2775 | if (ret) { |
d3e2ba59 | 2776 | ERR("Starting live viewer threads"); |
77c54fed | 2777 | retval = -1; |
50138f51 | 2778 | goto exit_live; |
d3e2ba59 JD |
2779 | } |
2780 | ||
77c54fed MD |
2781 | /* |
2782 | * This is where we start awaiting program completion (e.g. through | |
2783 | * signal that asks threads to teardown). | |
2784 | */ | |
2785 | ||
2786 | ret = relayd_live_join(); | |
2787 | if (ret) { | |
2788 | retval = -1; | |
2789 | } | |
50138f51 | 2790 | exit_live: |
77c54fed | 2791 | |
b8aa1682 | 2792 | ret = pthread_join(listener_thread, &status); |
77c54fed MD |
2793 | if (ret) { |
2794 | errno = ret; | |
2795 | PERROR("pthread_join listener_thread"); | |
2796 | retval = -1; | |
b8aa1682 JD |
2797 | } |
2798 | ||
77c54fed | 2799 | exit_listener_thread: |
b8aa1682 | 2800 | ret = pthread_join(worker_thread, &status); |
77c54fed MD |
2801 | if (ret) { |
2802 | errno = ret; | |
2803 | PERROR("pthread_join worker_thread"); | |
2804 | retval = -1; | |
b8aa1682 JD |
2805 | } |
2806 | ||
77c54fed | 2807 | exit_worker_thread: |
b8aa1682 | 2808 | ret = pthread_join(dispatcher_thread, &status); |
77c54fed MD |
2809 | if (ret) { |
2810 | errno = ret; | |
2811 | PERROR("pthread_join dispatcher_thread"); | |
2812 | retval = -1; | |
b8aa1682 | 2813 | } |
77c54fed | 2814 | exit_dispatcher_thread: |
42415026 | 2815 | |
65931c8b | 2816 | ret = pthread_join(health_thread, &status); |
77c54fed MD |
2817 | if (ret) { |
2818 | errno = ret; | |
2819 | PERROR("pthread_join health_thread"); | |
2820 | retval = -1; | |
65931c8b | 2821 | } |
77c54fed | 2822 | exit_health_thread: |
65931c8b | 2823 | |
65931c8b | 2824 | utils_close_pipe(health_quit_pipe); |
77c54fed | 2825 | exit_health_quit_pipe: |
65931c8b | 2826 | |
77c54fed | 2827 | exit_init_data: |
55706a7d | 2828 | health_app_destroy(health_relayd); |
55706a7d | 2829 | exit_health_app_create: |
77c54fed | 2830 | exit_options: |
9a78c92e MD |
2831 | relayd_cleanup(); |
2832 | ||
2833 | /* Ensure all prior call_rcu are done. */ | |
2834 | rcu_barrier(); | |
d3e2ba59 | 2835 | |
77c54fed | 2836 | if (!retval) { |
b8aa1682 | 2837 | exit(EXIT_SUCCESS); |
77c54fed MD |
2838 | } else { |
2839 | exit(EXIT_FAILURE); | |
b8aa1682 | 2840 | } |
b8aa1682 | 2841 | } |