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> |
5a693d30 | 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 | |
6c1c0768 | 22 | #define _LGPL_SOURCE |
b8aa1682 JD |
23 | #include <getopt.h> |
24 | #include <grp.h> | |
25 | #include <limits.h> | |
26 | #include <pthread.h> | |
27 | #include <signal.h> | |
28 | #include <stdio.h> | |
29 | #include <stdlib.h> | |
30 | #include <string.h> | |
31 | #include <sys/mman.h> | |
32 | #include <sys/mount.h> | |
33 | #include <sys/resource.h> | |
34 | #include <sys/socket.h> | |
35 | #include <sys/stat.h> | |
36 | #include <sys/types.h> | |
37 | #include <sys/wait.h> | |
173af62f | 38 | #include <inttypes.h> |
b8aa1682 JD |
39 | #include <urcu/futex.h> |
40 | #include <urcu/uatomic.h> | |
41 | #include <unistd.h> | |
42 | #include <fcntl.h> | |
43 | #include <config.h> | |
44 | ||
45 | #include <lttng/lttng.h> | |
46 | #include <common/common.h> | |
47 | #include <common/compat/poll.h> | |
48 | #include <common/compat/socket.h> | |
f263b7fd | 49 | #include <common/compat/endian.h> |
e8fa9fb0 | 50 | #include <common/compat/getenv.h> |
b8aa1682 | 51 | #include <common/defaults.h> |
3fd27398 | 52 | #include <common/daemonize.h> |
b8aa1682 JD |
53 | #include <common/futex.h> |
54 | #include <common/sessiond-comm/sessiond-comm.h> | |
55 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
56 | #include <common/sessiond-comm/relayd.h> |
57 | #include <common/uri.h> | |
a02de639 | 58 | #include <common/utils.h> |
cd60b05a | 59 | #include <common/config/config.h> |
5a693d30 | 60 | #include <urcu/rculist.h> |
b8aa1682 | 61 | |
0f907de1 | 62 | #include "cmd.h" |
d3e2ba59 | 63 | #include "ctf-trace.h" |
1c20f0e2 | 64 | #include "index.h" |
0f907de1 | 65 | #include "utils.h" |
b8aa1682 | 66 | #include "lttng-relayd.h" |
d3e2ba59 | 67 | #include "live.h" |
55706a7d | 68 | #include "health-relayd.h" |
9b5e0863 | 69 | #include "testpoint.h" |
2f8f53af | 70 | #include "viewer-stream.h" |
2a174661 DG |
71 | #include "session.h" |
72 | #include "stream.h" | |
58eb9381 | 73 | #include "connection.h" |
b8aa1682 JD |
74 | |
75 | /* command line options */ | |
0f907de1 | 76 | char *opt_output_path; |
b5218ffb | 77 | static int opt_daemon, opt_background; |
3fd27398 MD |
78 | |
79 | /* | |
80 | * We need to wait for listener and live listener threads, as well as | |
81 | * health check thread, before being ready to signal readiness. | |
82 | */ | |
83 | #define NR_LTTNG_RELAY_READY 3 | |
84 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
85 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ | |
86 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
87 | ||
095a4ae5 MD |
88 | static struct lttng_uri *control_uri; |
89 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 90 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
91 | |
92 | const char *progname; | |
b8aa1682 | 93 | |
65931c8b | 94 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
95 | static int tracing_group_name_override; |
96 | ||
97 | const char * const config_section_name = "relayd"; | |
65931c8b | 98 | |
b8aa1682 JD |
99 | /* |
100 | * Quit pipe for all threads. This permits a single cancellation point | |
101 | * for all threads when receiving an event on the pipe. | |
102 | */ | |
0b242f62 | 103 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
104 | |
105 | /* | |
106 | * This pipe is used to inform the worker thread that a command is queued and | |
107 | * ready to be processed. | |
108 | */ | |
58eb9381 | 109 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 110 | |
26c9d55e | 111 | /* Shared between threads */ |
b8aa1682 JD |
112 | static int dispatch_thread_exit; |
113 | ||
114 | static pthread_t listener_thread; | |
115 | static pthread_t dispatcher_thread; | |
116 | static pthread_t worker_thread; | |
65931c8b | 117 | static pthread_t health_thread; |
b8aa1682 | 118 | |
5a693d30 MD |
119 | /* |
120 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
121 | * atomicity on 32-bit architectures. | |
122 | */ | |
123 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 124 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
125 | |
126 | /* | |
127 | * Relay command queue. | |
128 | * | |
129 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
130 | * queue. | |
131 | */ | |
58eb9381 | 132 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 JD |
133 | |
134 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
135 | static char *data_buffer; |
136 | static unsigned int data_buffer_size; | |
b8aa1682 | 137 | |
d3e2ba59 JD |
138 | /* Global relay stream hash table. */ |
139 | struct lttng_ht *relay_streams_ht; | |
140 | ||
92c6ca54 DG |
141 | /* Global relay viewer stream hash table. */ |
142 | struct lttng_ht *viewer_streams_ht; | |
143 | ||
5a693d30 MD |
144 | /* Global relay sessions hash table. */ |
145 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 146 | |
55706a7d | 147 | /* Relayd health monitoring */ |
eea7556c | 148 | struct health_app *health_relayd; |
55706a7d | 149 | |
cd60b05a JG |
150 | static struct option long_options[] = { |
151 | { "control-port", 1, 0, 'C', }, | |
152 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 153 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 154 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 155 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
156 | { "group", 1, 0, 'g', }, |
157 | { "help", 0, 0, 'h', }, | |
158 | { "output", 1, 0, 'o', }, | |
159 | { "verbose", 0, 0, 'v', }, | |
160 | { "config", 1, 0, 'f' }, | |
161 | { NULL, 0, 0, 0, }, | |
162 | }; | |
163 | ||
164 | static const char *config_ignore_options[] = { "help", "config" }; | |
165 | ||
b8aa1682 JD |
166 | /* |
167 | * usage function on stderr | |
168 | */ | |
5a693d30 | 169 | static void usage(void) |
b8aa1682 JD |
170 | { |
171 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
172 | fprintf(stderr, " -h, --help Display this usage.\n"); |
173 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
b5218ffb | 174 | fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); |
994fa64f DG |
175 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); |
176 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
8d5c808e | 177 | fprintf(stderr, " -L, --live-port URL Live view port listening.\n"); |
994fa64f DG |
178 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); |
179 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 180 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 181 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
182 | } |
183 | ||
cd60b05a JG |
184 | /* |
185 | * Take an option from the getopt output and set it in the right variable to be | |
186 | * used later. | |
187 | * | |
188 | * Return 0 on success else a negative value. | |
189 | */ | |
5a693d30 | 190 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 191 | { |
cd60b05a JG |
192 | int ret; |
193 | ||
194 | switch (opt) { | |
195 | case 0: | |
196 | fprintf(stderr, "option %s", optname); | |
197 | if (arg) { | |
198 | fprintf(stderr, " with arg %s\n", arg); | |
199 | } | |
200 | break; | |
201 | case 'C': | |
e8fa9fb0 MD |
202 | if (lttng_is_setuid_setgid()) { |
203 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
204 | "-C, --control-port"); | |
205 | } else { | |
206 | ret = uri_parse(arg, &control_uri); | |
207 | if (ret < 0) { | |
208 | ERR("Invalid control URI specified"); | |
209 | goto end; | |
210 | } | |
211 | if (control_uri->port == 0) { | |
212 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
213 | } | |
cd60b05a JG |
214 | } |
215 | break; | |
216 | case 'D': | |
e8fa9fb0 MD |
217 | if (lttng_is_setuid_setgid()) { |
218 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
219 | "-D, -data-port"); | |
220 | } else { | |
221 | ret = uri_parse(arg, &data_uri); | |
222 | if (ret < 0) { | |
223 | ERR("Invalid data URI specified"); | |
224 | goto end; | |
225 | } | |
226 | if (data_uri->port == 0) { | |
227 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
228 | } | |
cd60b05a JG |
229 | } |
230 | break; | |
8d5c808e | 231 | case 'L': |
e8fa9fb0 MD |
232 | if (lttng_is_setuid_setgid()) { |
233 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
234 | "-L, -live-port"); | |
235 | } else { | |
236 | ret = uri_parse(arg, &live_uri); | |
237 | if (ret < 0) { | |
238 | ERR("Invalid live URI specified"); | |
239 | goto end; | |
240 | } | |
241 | if (live_uri->port == 0) { | |
242 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
243 | } | |
8d5c808e AM |
244 | } |
245 | break; | |
cd60b05a JG |
246 | case 'd': |
247 | opt_daemon = 1; | |
248 | break; | |
b5218ffb MD |
249 | case 'b': |
250 | opt_background = 1; | |
251 | break; | |
cd60b05a | 252 | case 'g': |
e8fa9fb0 MD |
253 | if (lttng_is_setuid_setgid()) { |
254 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
255 | "-g, --group"); | |
256 | } else { | |
257 | tracing_group_name = strdup(arg); | |
258 | if (tracing_group_name == NULL) { | |
259 | ret = -errno; | |
260 | PERROR("strdup"); | |
261 | goto end; | |
262 | } | |
263 | tracing_group_name_override = 1; | |
330a40bb | 264 | } |
cd60b05a JG |
265 | break; |
266 | case 'h': | |
267 | usage(); | |
268 | exit(EXIT_FAILURE); | |
269 | case 'o': | |
e8fa9fb0 MD |
270 | if (lttng_is_setuid_setgid()) { |
271 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
272 | "-o, --output"); | |
273 | } else { | |
274 | ret = asprintf(&opt_output_path, "%s", arg); | |
275 | if (ret < 0) { | |
276 | ret = -errno; | |
277 | PERROR("asprintf opt_output_path"); | |
278 | goto end; | |
279 | } | |
cd60b05a JG |
280 | } |
281 | break; | |
282 | case 'v': | |
283 | /* Verbose level can increase using multiple -v */ | |
284 | if (arg) { | |
285 | lttng_opt_verbose = config_parse_value(arg); | |
286 | } else { | |
849e5b7b DG |
287 | /* Only 3 level of verbosity (-vvv). */ |
288 | if (lttng_opt_verbose < 3) { | |
289 | lttng_opt_verbose += 1; | |
290 | } | |
cd60b05a JG |
291 | } |
292 | break; | |
293 | default: | |
294 | /* Unknown option or other error. | |
295 | * Error is printed by getopt, just return */ | |
296 | ret = -1; | |
297 | goto end; | |
298 | } | |
299 | ||
300 | /* All good. */ | |
301 | ret = 0; | |
302 | ||
303 | end: | |
304 | return ret; | |
305 | } | |
306 | ||
307 | /* | |
308 | * config_entry_handler_cb used to handle options read from a config file. | |
309 | * See config_entry_handler_cb comment in common/config/config.h for the | |
310 | * return value conventions. | |
311 | */ | |
5a693d30 | 312 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
313 | { |
314 | int ret = 0, i; | |
315 | ||
316 | if (!entry || !entry->name || !entry->value) { | |
317 | ret = -EINVAL; | |
318 | goto end; | |
319 | } | |
320 | ||
321 | /* Check if the option is to be ignored */ | |
322 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
323 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
324 | goto end; | |
325 | } | |
326 | } | |
327 | ||
328 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
329 | /* Ignore if entry name is not fully matched. */ | |
330 | if (strcmp(entry->name, long_options[i].name)) { | |
331 | continue; | |
332 | } | |
333 | ||
334 | /* | |
5a693d30 MD |
335 | * If the option takes no argument on the command line, |
336 | * we have to check if the value is "true". We support | |
337 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
338 | */ |
339 | if (!long_options[i].has_arg) { | |
340 | ret = config_parse_value(entry->value); | |
341 | if (ret <= 0) { | |
342 | if (ret) { | |
343 | WARN("Invalid configuration value \"%s\" for option %s", | |
344 | entry->value, entry->name); | |
345 | } | |
346 | /* False, skip boolean config option. */ | |
347 | goto end; | |
348 | } | |
349 | } | |
350 | ||
351 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
352 | goto end; | |
353 | } | |
354 | ||
355 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
356 | entry->name); | |
357 | ||
358 | end: | |
359 | return ret; | |
360 | } | |
361 | ||
5a693d30 | 362 | static int set_options(int argc, char **argv) |
cd60b05a | 363 | { |
178a0557 | 364 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
365 | int orig_optopt = optopt, orig_optind = optind; |
366 | char *default_address, *optstring; | |
367 | const char *config_path = NULL; | |
368 | ||
369 | optstring = utils_generate_optstring(long_options, | |
370 | sizeof(long_options) / sizeof(struct option)); | |
371 | if (!optstring) { | |
178a0557 | 372 | retval = -ENOMEM; |
cd60b05a JG |
373 | goto exit; |
374 | } | |
375 | ||
376 | /* Check for the --config option */ | |
377 | ||
378 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
379 | &option_index)) != -1) { | |
380 | if (c == '?') { | |
178a0557 | 381 | retval = -EINVAL; |
cd60b05a JG |
382 | goto exit; |
383 | } else if (c != 'f') { | |
384 | continue; | |
385 | } | |
386 | ||
e8fa9fb0 MD |
387 | if (lttng_is_setuid_setgid()) { |
388 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
389 | "-f, --config"); | |
390 | } else { | |
391 | config_path = utils_expand_path(optarg); | |
392 | if (!config_path) { | |
393 | ERR("Failed to resolve path: %s", optarg); | |
394 | } | |
cd60b05a JG |
395 | } |
396 | } | |
397 | ||
398 | ret = config_get_section_entries(config_path, config_section_name, | |
399 | config_entry_handler, NULL); | |
400 | if (ret) { | |
401 | if (ret > 0) { | |
402 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 403 | } |
178a0557 | 404 | retval = -1; |
cd60b05a JG |
405 | goto exit; |
406 | } | |
b8aa1682 | 407 | |
cd60b05a JG |
408 | /* Reset getopt's global state */ |
409 | optopt = orig_optopt; | |
410 | optind = orig_optind; | |
b8aa1682 | 411 | while (1) { |
cd60b05a | 412 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
413 | if (c == -1) { |
414 | break; | |
415 | } | |
416 | ||
cd60b05a JG |
417 | ret = set_option(c, optarg, long_options[option_index].name); |
418 | if (ret < 0) { | |
178a0557 | 419 | retval = -1; |
b8aa1682 JD |
420 | goto exit; |
421 | } | |
422 | } | |
423 | ||
424 | /* assign default values */ | |
425 | if (control_uri == NULL) { | |
fa91dc52 MD |
426 | ret = asprintf(&default_address, |
427 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
428 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
429 | if (ret < 0) { |
430 | PERROR("asprintf default data address"); | |
178a0557 | 431 | retval = -1; |
b8aa1682 JD |
432 | goto exit; |
433 | } | |
434 | ||
435 | ret = uri_parse(default_address, &control_uri); | |
436 | free(default_address); | |
437 | if (ret < 0) { | |
438 | ERR("Invalid control URI specified"); | |
178a0557 | 439 | retval = -1; |
b8aa1682 JD |
440 | goto exit; |
441 | } | |
442 | } | |
443 | if (data_uri == NULL) { | |
fa91dc52 MD |
444 | ret = asprintf(&default_address, |
445 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
446 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
447 | if (ret < 0) { |
448 | PERROR("asprintf default data address"); | |
178a0557 | 449 | retval = -1; |
b8aa1682 JD |
450 | goto exit; |
451 | } | |
452 | ||
453 | ret = uri_parse(default_address, &data_uri); | |
454 | free(default_address); | |
455 | if (ret < 0) { | |
456 | ERR("Invalid data URI specified"); | |
178a0557 | 457 | retval = -1; |
b8aa1682 JD |
458 | goto exit; |
459 | } | |
460 | } | |
d3e2ba59 | 461 | if (live_uri == NULL) { |
fa91dc52 MD |
462 | ret = asprintf(&default_address, |
463 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
464 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
465 | if (ret < 0) { |
466 | PERROR("asprintf default viewer control address"); | |
178a0557 | 467 | retval = -1; |
d3e2ba59 JD |
468 | goto exit; |
469 | } | |
470 | ||
471 | ret = uri_parse(default_address, &live_uri); | |
472 | free(default_address); | |
473 | if (ret < 0) { | |
474 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 475 | retval = -1; |
d3e2ba59 JD |
476 | goto exit; |
477 | } | |
478 | } | |
b8aa1682 JD |
479 | |
480 | exit: | |
cd60b05a | 481 | free(optstring); |
178a0557 | 482 | return retval; |
b8aa1682 JD |
483 | } |
484 | ||
5a693d30 MD |
485 | static void print_global_objects(void) |
486 | { | |
487 | rcu_register_thread(); | |
488 | ||
489 | print_viewer_streams(); | |
490 | print_relay_streams(); | |
491 | print_sessions(); | |
492 | ||
493 | rcu_unregister_thread(); | |
494 | } | |
495 | ||
b8aa1682 JD |
496 | /* |
497 | * Cleanup the daemon | |
498 | */ | |
5a693d30 | 499 | static void relayd_cleanup(void) |
b8aa1682 | 500 | { |
5a693d30 MD |
501 | print_global_objects(); |
502 | ||
b8aa1682 JD |
503 | DBG("Cleaning up"); |
504 | ||
178a0557 MD |
505 | if (viewer_streams_ht) |
506 | lttng_ht_destroy(viewer_streams_ht); | |
507 | if (relay_streams_ht) | |
508 | lttng_ht_destroy(relay_streams_ht); | |
5a693d30 MD |
509 | if (sessions_ht) |
510 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 511 | |
095a4ae5 MD |
512 | /* free the dynamically allocated opt_output_path */ |
513 | free(opt_output_path); | |
514 | ||
a02de639 CB |
515 | /* Close thread quit pipes */ |
516 | utils_close_pipe(thread_quit_pipe); | |
517 | ||
710c1f73 DG |
518 | uri_free(control_uri); |
519 | uri_free(data_uri); | |
8d5c808e | 520 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
521 | |
522 | if (tracing_group_name_override) { | |
523 | free((void *) tracing_group_name); | |
524 | } | |
b8aa1682 JD |
525 | } |
526 | ||
527 | /* | |
528 | * Write to writable pipe used to notify a thread. | |
529 | */ | |
5a693d30 | 530 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 531 | { |
6cd525e8 | 532 | ssize_t ret; |
b8aa1682 | 533 | |
6cd525e8 MD |
534 | ret = lttng_write(wpipe, "!", 1); |
535 | if (ret < 1) { | |
b8aa1682 | 536 | PERROR("write poll pipe"); |
b4aacfdc | 537 | goto end; |
b8aa1682 | 538 | } |
b4aacfdc MD |
539 | ret = 0; |
540 | end: | |
b8aa1682 JD |
541 | return ret; |
542 | } | |
543 | ||
5a693d30 | 544 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 545 | { |
6cd525e8 | 546 | ssize_t ret; |
65931c8b | 547 | |
6cd525e8 MD |
548 | ret = lttng_write(pipe[1], "4", 1); |
549 | if (ret < 1) { | |
65931c8b | 550 | PERROR("write relay health quit"); |
b4aacfdc | 551 | goto end; |
65931c8b | 552 | } |
b4aacfdc MD |
553 | ret = 0; |
554 | end: | |
555 | return ret; | |
65931c8b MD |
556 | } |
557 | ||
b8aa1682 | 558 | /* |
b4aacfdc | 559 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 560 | */ |
b4aacfdc | 561 | int lttng_relay_stop_threads(void) |
b8aa1682 | 562 | { |
b4aacfdc | 563 | int retval = 0; |
b8aa1682 JD |
564 | |
565 | /* Stopping all threads */ | |
566 | DBG("Terminating all threads"); | |
b4aacfdc | 567 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 568 | ERR("write error on thread quit pipe"); |
b4aacfdc | 569 | retval = -1; |
b8aa1682 JD |
570 | } |
571 | ||
b4aacfdc MD |
572 | if (notify_health_quit_pipe(health_quit_pipe)) { |
573 | ERR("write error on health quit pipe"); | |
574 | } | |
65931c8b | 575 | |
b8aa1682 | 576 | /* Dispatch thread */ |
26c9d55e | 577 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 578 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 579 | |
b4aacfdc | 580 | if (relayd_live_stop()) { |
178a0557 | 581 | ERR("Error stopping live threads"); |
b4aacfdc | 582 | retval = -1; |
178a0557 | 583 | } |
b4aacfdc | 584 | return retval; |
b8aa1682 JD |
585 | } |
586 | ||
587 | /* | |
588 | * Signal handler for the daemon | |
589 | * | |
590 | * Simply stop all worker threads, leaving main() return gracefully after | |
591 | * joining all threads and calling cleanup(). | |
592 | */ | |
5a693d30 | 593 | static void sighandler(int sig) |
b8aa1682 JD |
594 | { |
595 | switch (sig) { | |
596 | case SIGPIPE: | |
597 | DBG("SIGPIPE caught"); | |
598 | return; | |
599 | case SIGINT: | |
600 | DBG("SIGINT caught"); | |
b4aacfdc MD |
601 | if (lttng_relay_stop_threads()) { |
602 | ERR("Error stopping threads"); | |
603 | } | |
b8aa1682 JD |
604 | break; |
605 | case SIGTERM: | |
606 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
607 | if (lttng_relay_stop_threads()) { |
608 | ERR("Error stopping threads"); | |
609 | } | |
b8aa1682 | 610 | break; |
3fd27398 MD |
611 | case SIGUSR1: |
612 | CMM_STORE_SHARED(recv_child_signal, 1); | |
613 | break; | |
b8aa1682 JD |
614 | default: |
615 | break; | |
616 | } | |
617 | } | |
618 | ||
619 | /* | |
620 | * Setup signal handler for : | |
621 | * SIGINT, SIGTERM, SIGPIPE | |
622 | */ | |
5a693d30 | 623 | static int set_signal_handler(void) |
b8aa1682 JD |
624 | { |
625 | int ret = 0; | |
626 | struct sigaction sa; | |
627 | sigset_t sigset; | |
628 | ||
629 | if ((ret = sigemptyset(&sigset)) < 0) { | |
630 | PERROR("sigemptyset"); | |
631 | return ret; | |
632 | } | |
633 | ||
634 | sa.sa_handler = sighandler; | |
635 | sa.sa_mask = sigset; | |
636 | sa.sa_flags = 0; | |
637 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
638 | PERROR("sigaction"); | |
639 | return ret; | |
640 | } | |
641 | ||
642 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
643 | PERROR("sigaction"); | |
644 | return ret; | |
645 | } | |
646 | ||
647 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
648 | PERROR("sigaction"); | |
649 | return ret; | |
650 | } | |
651 | ||
3fd27398 MD |
652 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
653 | PERROR("sigaction"); | |
654 | return ret; | |
655 | } | |
656 | ||
657 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
658 | |
659 | return ret; | |
660 | } | |
661 | ||
3fd27398 MD |
662 | void lttng_relay_notify_ready(void) |
663 | { | |
664 | /* Notify the parent of the fork() process that we are ready. */ | |
665 | if (opt_daemon || opt_background) { | |
666 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
667 | kill(child_ppid, SIGUSR1); | |
668 | } | |
669 | } | |
670 | } | |
671 | ||
b8aa1682 JD |
672 | /* |
673 | * Init thread quit pipe. | |
674 | * | |
675 | * Return -1 on error or 0 if all pipes are created. | |
676 | */ | |
5a693d30 | 677 | static int init_thread_quit_pipe(void) |
b8aa1682 | 678 | { |
a02de639 | 679 | int ret; |
b8aa1682 | 680 | |
a02de639 | 681 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 682 | |
b8aa1682 JD |
683 | return ret; |
684 | } | |
685 | ||
686 | /* | |
687 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
688 | */ | |
5a693d30 | 689 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
690 | { |
691 | int ret; | |
692 | ||
693 | if (events == NULL || size == 0) { | |
694 | ret = -1; | |
695 | goto error; | |
696 | } | |
697 | ||
698 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
699 | if (ret < 0) { | |
700 | goto error; | |
701 | } | |
702 | ||
703 | /* Add quit pipe */ | |
c7759e6a | 704 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
705 | if (ret < 0) { |
706 | goto error; | |
707 | } | |
708 | ||
709 | return 0; | |
710 | ||
711 | error: | |
712 | return ret; | |
713 | } | |
714 | ||
715 | /* | |
716 | * Check if the thread quit pipe was triggered. | |
717 | * | |
718 | * Return 1 if it was triggered else 0; | |
719 | */ | |
5a693d30 | 720 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
721 | { |
722 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
723 | return 1; | |
724 | } | |
725 | ||
726 | return 0; | |
727 | } | |
728 | ||
729 | /* | |
730 | * Create and init socket from uri. | |
731 | */ | |
5a693d30 | 732 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
733 | { |
734 | int ret; | |
735 | struct lttcomm_sock *sock = NULL; | |
736 | ||
737 | sock = lttcomm_alloc_sock_from_uri(uri); | |
738 | if (sock == NULL) { | |
739 | ERR("Allocating socket"); | |
740 | goto error; | |
741 | } | |
742 | ||
743 | ret = lttcomm_create_sock(sock); | |
744 | if (ret < 0) { | |
745 | goto error; | |
746 | } | |
747 | DBG("Listening on sock %d", sock->fd); | |
748 | ||
749 | ret = sock->ops->bind(sock); | |
750 | if (ret < 0) { | |
751 | goto error; | |
752 | } | |
753 | ||
754 | ret = sock->ops->listen(sock, -1); | |
755 | if (ret < 0) { | |
756 | goto error; | |
757 | ||
758 | } | |
759 | ||
760 | return sock; | |
761 | ||
762 | error: | |
763 | if (sock) { | |
764 | lttcomm_destroy_sock(sock); | |
765 | } | |
766 | return NULL; | |
767 | } | |
768 | ||
769 | /* | |
770 | * This thread manages the listening for new connections on the network | |
771 | */ | |
5a693d30 | 772 | static void *relay_thread_listener(void *data) |
b8aa1682 | 773 | { |
095a4ae5 | 774 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
775 | uint32_t revents, nb_fd; |
776 | struct lttng_poll_event events; | |
777 | struct lttcomm_sock *control_sock, *data_sock; | |
778 | ||
b8aa1682 JD |
779 | DBG("[thread] Relay listener started"); |
780 | ||
55706a7d MD |
781 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
782 | ||
f385ae0a MD |
783 | health_code_update(); |
784 | ||
5a693d30 | 785 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 786 | if (!control_sock) { |
095a4ae5 | 787 | goto error_sock_control; |
b8aa1682 JD |
788 | } |
789 | ||
5a693d30 | 790 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 791 | if (!data_sock) { |
095a4ae5 | 792 | goto error_sock_relay; |
b8aa1682 JD |
793 | } |
794 | ||
795 | /* | |
5a693d30 MD |
796 | * Pass 3 as size here for the thread quit pipe, control and |
797 | * data socket. | |
b8aa1682 JD |
798 | */ |
799 | ret = create_thread_poll_set(&events, 3); | |
800 | if (ret < 0) { | |
801 | goto error_create_poll; | |
802 | } | |
803 | ||
804 | /* Add the control socket */ | |
805 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
806 | if (ret < 0) { | |
807 | goto error_poll_add; | |
808 | } | |
809 | ||
810 | /* Add the data socket */ | |
811 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
812 | if (ret < 0) { | |
813 | goto error_poll_add; | |
814 | } | |
815 | ||
3fd27398 MD |
816 | lttng_relay_notify_ready(); |
817 | ||
9b5e0863 MD |
818 | if (testpoint(relayd_thread_listener)) { |
819 | goto error_testpoint; | |
820 | } | |
821 | ||
b8aa1682 | 822 | while (1) { |
f385ae0a MD |
823 | health_code_update(); |
824 | ||
b8aa1682 JD |
825 | DBG("Listener accepting connections"); |
826 | ||
b8aa1682 | 827 | restart: |
f385ae0a | 828 | health_poll_entry(); |
b8aa1682 | 829 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 830 | health_poll_exit(); |
b8aa1682 JD |
831 | if (ret < 0) { |
832 | /* | |
833 | * Restart interrupted system call. | |
834 | */ | |
835 | if (errno == EINTR) { | |
836 | goto restart; | |
837 | } | |
838 | goto error; | |
839 | } | |
840 | ||
0d9c5d77 DG |
841 | nb_fd = ret; |
842 | ||
b8aa1682 JD |
843 | DBG("Relay new connection received"); |
844 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
845 | health_code_update(); |
846 | ||
b8aa1682 JD |
847 | /* Fetch once the poll data */ |
848 | revents = LTTNG_POLL_GETEV(&events, i); | |
849 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
850 | ||
fd20dac9 | 851 | if (!revents) { |
5a693d30 MD |
852 | /* |
853 | * No activity for this FD (poll | |
854 | * implementation). | |
855 | */ | |
fd20dac9 MD |
856 | continue; |
857 | } | |
858 | ||
b8aa1682 JD |
859 | /* Thread quit pipe has been closed. Killing thread. */ |
860 | ret = check_thread_quit_pipe(pollfd, revents); | |
861 | if (ret) { | |
095a4ae5 MD |
862 | err = 0; |
863 | goto exit; | |
b8aa1682 JD |
864 | } |
865 | ||
866 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
867 | ERR("socket poll error"); | |
868 | goto error; | |
869 | } else if (revents & LPOLLIN) { | |
4b7f17b2 | 870 | /* |
5a693d30 MD |
871 | * A new connection is requested, therefore a |
872 | * sessiond/consumerd connection is allocated in | |
873 | * this thread, enqueued to a global queue and | |
874 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 875 | */ |
58eb9381 DG |
876 | int val = 1; |
877 | struct relay_connection *new_conn; | |
4b7f17b2 | 878 | struct lttcomm_sock *newsock; |
5a693d30 | 879 | enum connection_type type; |
b8aa1682 JD |
880 | |
881 | if (pollfd == data_sock->fd) { | |
5a693d30 | 882 | type = RELAY_DATA; |
b8aa1682 | 883 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
884 | DBG("Relay data connection accepted, socket %d", |
885 | newsock->fd); | |
4b7f17b2 MD |
886 | } else { |
887 | assert(pollfd == control_sock->fd); | |
5a693d30 | 888 | type = RELAY_CONTROL; |
b8aa1682 | 889 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
890 | DBG("Relay control connection accepted, socket %d", |
891 | newsock->fd); | |
b8aa1682 | 892 | } |
58eb9381 DG |
893 | if (!newsock) { |
894 | PERROR("accepting sock"); | |
58eb9381 DG |
895 | goto error; |
896 | } | |
897 | ||
898 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
899 | sizeof(val)); | |
b8aa1682 JD |
900 | if (ret < 0) { |
901 | PERROR("setsockopt inet"); | |
4b7f17b2 | 902 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
903 | goto error; |
904 | } | |
5a693d30 MD |
905 | new_conn = connection_create(newsock, type); |
906 | if (!new_conn) { | |
907 | lttcomm_destroy_sock(newsock); | |
908 | goto error; | |
909 | } | |
58eb9381 DG |
910 | |
911 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
912 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
913 | &new_conn->qnode); | |
b8aa1682 JD |
914 | |
915 | /* | |
5a693d30 MD |
916 | * Wake the dispatch queue futex. |
917 | * Implicit memory barrier with the | |
918 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 919 | */ |
58eb9381 | 920 | futex_nto1_wake(&relay_conn_queue.futex); |
b8aa1682 JD |
921 | } |
922 | } | |
923 | } | |
924 | ||
095a4ae5 | 925 | exit: |
b8aa1682 JD |
926 | error: |
927 | error_poll_add: | |
9b5e0863 | 928 | error_testpoint: |
b8aa1682 JD |
929 | lttng_poll_clean(&events); |
930 | error_create_poll: | |
095a4ae5 MD |
931 | if (data_sock->fd >= 0) { |
932 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
933 | if (ret) { |
934 | PERROR("close"); | |
935 | } | |
b8aa1682 | 936 | } |
095a4ae5 MD |
937 | lttcomm_destroy_sock(data_sock); |
938 | error_sock_relay: | |
939 | if (control_sock->fd >= 0) { | |
940 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
941 | if (ret) { |
942 | PERROR("close"); | |
943 | } | |
b8aa1682 | 944 | } |
095a4ae5 MD |
945 | lttcomm_destroy_sock(control_sock); |
946 | error_sock_control: | |
947 | if (err) { | |
f385ae0a MD |
948 | health_error(); |
949 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 950 | } |
55706a7d | 951 | health_unregister(health_relayd); |
b8aa1682 | 952 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 953 | lttng_relay_stop_threads(); |
b8aa1682 JD |
954 | return NULL; |
955 | } | |
956 | ||
957 | /* | |
958 | * This thread manages the dispatching of the requests to worker threads | |
959 | */ | |
5a693d30 | 960 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 961 | { |
6cd525e8 MD |
962 | int err = -1; |
963 | ssize_t ret; | |
8bdee6e2 | 964 | struct cds_wfcq_node *node; |
58eb9381 | 965 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
966 | |
967 | DBG("[thread] Relay dispatcher started"); | |
968 | ||
55706a7d MD |
969 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
970 | ||
9b5e0863 MD |
971 | if (testpoint(relayd_thread_dispatcher)) { |
972 | goto error_testpoint; | |
973 | } | |
974 | ||
f385ae0a MD |
975 | health_code_update(); |
976 | ||
26c9d55e | 977 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
978 | health_code_update(); |
979 | ||
b8aa1682 | 980 | /* Atomically prepare the queue futex */ |
58eb9381 | 981 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
982 | |
983 | do { | |
f385ae0a MD |
984 | health_code_update(); |
985 | ||
b8aa1682 | 986 | /* Dequeue commands */ |
8bdee6e2 SM |
987 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
988 | &relay_conn_queue.tail); | |
b8aa1682 JD |
989 | if (node == NULL) { |
990 | DBG("Woken up but nothing in the relay command queue"); | |
991 | /* Continue thread execution */ | |
992 | break; | |
993 | } | |
58eb9381 | 994 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 995 | |
58eb9381 | 996 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
997 | |
998 | /* | |
5a693d30 MD |
999 | * Inform worker thread of the new request. This |
1000 | * call is blocking so we can be assured that | |
1001 | * the data will be read at some point in time | |
1002 | * or wait to the end of the world :) | |
b8aa1682 | 1003 | */ |
58eb9381 DG |
1004 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1005 | if (ret < 0) { | |
1006 | PERROR("write connection pipe"); | |
5a693d30 | 1007 | connection_put(new_conn); |
b8aa1682 JD |
1008 | goto error; |
1009 | } | |
1010 | } while (node != NULL); | |
1011 | ||
1012 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1013 | health_poll_entry(); |
58eb9381 | 1014 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1015 | health_poll_exit(); |
b8aa1682 JD |
1016 | } |
1017 | ||
f385ae0a MD |
1018 | /* Normal exit, no error */ |
1019 | err = 0; | |
1020 | ||
b8aa1682 | 1021 | error: |
9b5e0863 | 1022 | error_testpoint: |
f385ae0a MD |
1023 | if (err) { |
1024 | health_error(); | |
1025 | ERR("Health error occurred in %s", __func__); | |
1026 | } | |
55706a7d | 1027 | health_unregister(health_relayd); |
b8aa1682 | 1028 | DBG("Dispatch thread dying"); |
b4aacfdc | 1029 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1030 | return NULL; |
1031 | } | |
1032 | ||
b8aa1682 | 1033 | /* |
5a693d30 | 1034 | * Set index data from the control port to a given index object. |
b8aa1682 | 1035 | */ |
5a693d30 | 1036 | static int set_index_control_data(struct relay_index *index, |
1c20f0e2 JD |
1037 | struct lttcomm_relayd_index *data) |
1038 | { | |
5a693d30 | 1039 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1040 | |
1041 | /* | |
5a693d30 MD |
1042 | * The index on disk is encoded in big endian, so we don't need |
1043 | * to convert the data received on the network. The data_offset | |
1044 | * value is NEVER modified here and is updated by the data | |
1045 | * thread. | |
1c20f0e2 | 1046 | */ |
5a693d30 MD |
1047 | index_data.packet_size = data->packet_size; |
1048 | index_data.content_size = data->content_size; | |
1049 | index_data.timestamp_begin = data->timestamp_begin; | |
1050 | index_data.timestamp_end = data->timestamp_end; | |
1051 | index_data.events_discarded = data->events_discarded; | |
1052 | index_data.stream_id = data->stream_id; | |
1053 | return relay_index_set_data(index, &index_data); | |
1c20f0e2 JD |
1054 | } |
1055 | ||
c5b6f4f0 DG |
1056 | /* |
1057 | * Handle the RELAYD_CREATE_SESSION command. | |
1058 | * | |
1059 | * On success, send back the session id or else return a negative value. | |
1060 | */ | |
5a693d30 | 1061 | static int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1062 | struct relay_connection *conn) |
c5b6f4f0 DG |
1063 | { |
1064 | int ret = 0, send_ret; | |
1065 | struct relay_session *session; | |
1066 | struct lttcomm_relayd_status_session reply; | |
5a693d30 MD |
1067 | char session_name[NAME_MAX]; |
1068 | char hostname[HOST_NAME_MAX]; | |
1069 | uint32_t live_timer = 0; | |
1070 | bool snapshot = false; | |
c5b6f4f0 | 1071 | |
5a693d30 MD |
1072 | memset(session_name, 0, NAME_MAX); |
1073 | memset(hostname, 0, HOST_NAME_MAX); | |
c5b6f4f0 DG |
1074 | |
1075 | memset(&reply, 0, sizeof(reply)); | |
1076 | ||
58eb9381 | 1077 | switch (conn->minor) { |
2a174661 DG |
1078 | case 1: |
1079 | case 2: | |
1080 | case 3: | |
1081 | break; | |
1082 | case 4: /* LTTng sessiond 2.4 */ | |
1083 | default: | |
5a693d30 MD |
1084 | ret = cmd_create_session_2_4(conn, session_name, |
1085 | hostname, &live_timer, &snapshot); | |
1086 | } | |
1087 | if (ret < 0) { | |
1088 | goto send_reply; | |
d3e2ba59 JD |
1089 | } |
1090 | ||
5a693d30 MD |
1091 | session = session_create(session_name, hostname, live_timer, |
1092 | snapshot, conn->major, conn->minor); | |
1093 | if (!session) { | |
1094 | ret = -1; | |
1095 | goto send_reply; | |
1096 | } | |
1097 | assert(!conn->session); | |
1098 | conn->session = session; | |
c5b6f4f0 DG |
1099 | DBG("Created session %" PRIu64, session->id); |
1100 | ||
5a693d30 MD |
1101 | reply.session_id = htobe64(session->id); |
1102 | ||
1103 | send_reply: | |
c5b6f4f0 DG |
1104 | if (ret < 0) { |
1105 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1106 | } else { | |
1107 | reply.ret_code = htobe32(LTTNG_OK); | |
1108 | } | |
1109 | ||
58eb9381 | 1110 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1111 | if (send_ret < 0) { |
1112 | ERR("Relayd sending session id"); | |
4169f5ad | 1113 | ret = send_ret; |
c5b6f4f0 DG |
1114 | } |
1115 | ||
1116 | return ret; | |
1117 | } | |
1118 | ||
a4baae1b JD |
1119 | /* |
1120 | * When we have received all the streams and the metadata for a channel, | |
1121 | * we make them visible to the viewer threads. | |
1122 | */ | |
5a693d30 | 1123 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1124 | { |
5a693d30 MD |
1125 | struct relay_stream *stream; |
1126 | struct relay_session *session = conn->session; | |
a4baae1b | 1127 | |
5a693d30 MD |
1128 | /* |
1129 | * We publish all streams belonging to a session atomically wrt | |
1130 | * session lock. | |
1131 | */ | |
1132 | pthread_mutex_lock(&session->lock); | |
1133 | rcu_read_lock(); | |
1134 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1135 | recv_node) { | |
1136 | stream_publish(stream); | |
a4baae1b | 1137 | } |
5a693d30 | 1138 | rcu_read_unlock(); |
a4baae1b | 1139 | |
5a693d30 MD |
1140 | /* |
1141 | * Inform the viewer that there are new streams in the session. | |
1142 | */ | |
1143 | if (session->viewer_attached) { | |
1144 | uatomic_set(&session->new_streams, 1); | |
1145 | } | |
1146 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1147 | } |
1148 | ||
b8aa1682 JD |
1149 | /* |
1150 | * relay_add_stream: allocate a new stream for a session | |
1151 | */ | |
5a693d30 | 1152 | static int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1153 | struct relay_connection *conn) |
b8aa1682 | 1154 | { |
5a693d30 MD |
1155 | int ret; |
1156 | ssize_t send_ret; | |
58eb9381 | 1157 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1158 | struct relay_stream *stream = NULL; |
1159 | struct lttcomm_relayd_status_stream reply; | |
80ec61a6 | 1160 | struct ctf_trace *trace = NULL; |
5a693d30 MD |
1161 | uint64_t stream_handle = -1ULL; |
1162 | char *path_name = NULL, *channel_name = NULL; | |
1163 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1164 | |
58eb9381 | 1165 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1166 | ERR("Trying to add a stream before version check"); |
1167 | ret = -1; | |
1168 | goto end_no_session; | |
1169 | } | |
1170 | ||
5a693d30 MD |
1171 | switch (session->minor) { |
1172 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
1173 | ret = cmd_recv_stream_2_1(conn, &path_name, | |
1174 | &channel_name); | |
0f907de1 | 1175 | break; |
5a693d30 | 1176 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1177 | default: |
5a693d30 MD |
1178 | ret = cmd_recv_stream_2_2(conn, &path_name, |
1179 | &channel_name, &tracefile_size, &tracefile_count); | |
0f907de1 JD |
1180 | break; |
1181 | } | |
1182 | if (ret < 0) { | |
5a693d30 | 1183 | goto send_reply; |
b8aa1682 JD |
1184 | } |
1185 | ||
5a693d30 | 1186 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1187 | if (!trace) { |
5a693d30 | 1188 | goto send_reply; |
2a174661 | 1189 | } |
5a693d30 | 1190 | /* This stream here has one reference on the trace. */ |
2a174661 | 1191 | |
5a693d30 MD |
1192 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1193 | stream_handle = ++last_relay_stream_id; | |
1194 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1195 | |
5a693d30 MD |
1196 | /* We pass ownership of path_name and channel_name. */ |
1197 | stream = stream_create(trace, stream_handle, path_name, | |
1198 | channel_name, tracefile_size, tracefile_count); | |
1199 | path_name = NULL; | |
1200 | channel_name = NULL; | |
a4baae1b | 1201 | |
2a174661 | 1202 | /* |
5a693d30 MD |
1203 | * Streams are the owners of their trace. Reference to trace is |
1204 | * kept within stream_create(). | |
2a174661 | 1205 | */ |
5a693d30 | 1206 | ctf_trace_put(trace); |
d3e2ba59 | 1207 | |
5a693d30 | 1208 | send_reply: |
53efb85a | 1209 | memset(&reply, 0, sizeof(reply)); |
5a693d30 MD |
1210 | reply.handle = htobe64(stream_handle); |
1211 | if (!stream) { | |
f73fabfd | 1212 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1213 | } else { |
f73fabfd | 1214 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1215 | } |
5af40280 | 1216 | |
58eb9381 | 1217 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1218 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1219 | if (send_ret < 0) { | |
1220 | ERR("Relay sending stream id"); | |
5a693d30 | 1221 | ret = (int) send_ret; |
b8aa1682 JD |
1222 | } |
1223 | ||
1224 | end_no_session: | |
5a693d30 MD |
1225 | free(path_name); |
1226 | free(channel_name); | |
0f907de1 | 1227 | return ret; |
b8aa1682 JD |
1228 | } |
1229 | ||
173af62f DG |
1230 | /* |
1231 | * relay_close_stream: close a specific stream | |
1232 | */ | |
5a693d30 | 1233 | static int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1234 | struct relay_connection *conn) |
173af62f | 1235 | { |
94d49140 | 1236 | int ret, send_ret; |
58eb9381 | 1237 | struct relay_session *session = conn->session; |
173af62f DG |
1238 | struct lttcomm_relayd_close_stream stream_info; |
1239 | struct lttcomm_relayd_generic_reply reply; | |
1240 | struct relay_stream *stream; | |
173af62f DG |
1241 | |
1242 | DBG("Close stream received"); | |
1243 | ||
58eb9381 | 1244 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1245 | ERR("Trying to close a stream before version check"); |
1246 | ret = -1; | |
1247 | goto end_no_session; | |
1248 | } | |
1249 | ||
58eb9381 | 1250 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1251 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1252 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1253 | if (ret == 0) { |
1254 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1255 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1256 | } else { |
1257 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1258 | } | |
173af62f DG |
1259 | ret = -1; |
1260 | goto end_no_session; | |
1261 | } | |
1262 | ||
5a693d30 | 1263 | stream = stream_get_by_id(be64toh(stream_info.stream_id)); |
173af62f DG |
1264 | if (!stream) { |
1265 | ret = -1; | |
5a693d30 | 1266 | goto end; |
173af62f | 1267 | } |
5a693d30 MD |
1268 | pthread_mutex_lock(&stream->lock); |
1269 | stream->closed = true; | |
8e2583a4 | 1270 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
5a693d30 MD |
1271 | if (stream->is_metadata) { |
1272 | struct relay_viewer_stream *vstream; | |
173af62f | 1273 | |
5a693d30 MD |
1274 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1275 | if (vstream) { | |
1276 | if (vstream->metadata_sent == stream->metadata_received) { | |
1277 | /* | |
1278 | * Since all the metadata has been sent to the | |
1279 | * viewer and that we have a request to close | |
1280 | * its stream, we can safely teardown the | |
1281 | * corresponding metadata viewer stream. | |
1282 | */ | |
1283 | viewer_stream_put(vstream); | |
1284 | } | |
1285 | /* Put local reference. */ | |
1286 | viewer_stream_put(vstream); | |
1287 | } | |
1288 | } | |
1289 | pthread_mutex_unlock(&stream->lock); | |
1290 | stream_put(stream); | |
173af62f | 1291 | |
5a693d30 | 1292 | end: |
53efb85a | 1293 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1294 | if (ret < 0) { |
f73fabfd | 1295 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1296 | } else { |
f73fabfd | 1297 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1298 | } |
58eb9381 | 1299 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1300 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1301 | if (send_ret < 0) { | |
1302 | ERR("Relay sending stream id"); | |
4169f5ad | 1303 | ret = send_ret; |
173af62f DG |
1304 | } |
1305 | ||
1306 | end_no_session: | |
1307 | return ret; | |
1308 | } | |
1309 | ||
b8aa1682 JD |
1310 | /* |
1311 | * relay_unknown_command: send -1 if received unknown command | |
1312 | */ | |
5a693d30 | 1313 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1314 | { |
1315 | struct lttcomm_relayd_generic_reply reply; | |
1316 | int ret; | |
1317 | ||
53efb85a | 1318 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1319 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
58eb9381 | 1320 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1321 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1322 | if (ret < 0) { | |
1323 | ERR("Relay sending unknown command"); | |
1324 | } | |
1325 | } | |
1326 | ||
1327 | /* | |
1328 | * relay_start: send an acknowledgment to the client to tell if we are | |
1329 | * ready to receive data. We are ready if a session is established. | |
1330 | */ | |
5a693d30 | 1331 | static int relay_start(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1332 | struct relay_connection *conn) |
b8aa1682 | 1333 | { |
f73fabfd | 1334 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1335 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1336 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1337 | |
1338 | if (!session) { | |
1339 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1340 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1341 | } |
1342 | ||
53efb85a | 1343 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1344 | reply.ret_code = ret; |
58eb9381 | 1345 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1346 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1347 | if (ret < 0) { | |
1348 | ERR("Relay sending start ack"); | |
1349 | } | |
1350 | ||
1351 | return ret; | |
1352 | } | |
1353 | ||
1d4dfdef DG |
1354 | /* |
1355 | * Append padding to the file pointed by the file descriptor fd. | |
1356 | */ | |
1357 | static int write_padding_to_file(int fd, uint32_t size) | |
1358 | { | |
6cd525e8 | 1359 | ssize_t ret = 0; |
1d4dfdef DG |
1360 | char *zeros; |
1361 | ||
1362 | if (size == 0) { | |
1363 | goto end; | |
1364 | } | |
1365 | ||
1366 | zeros = zmalloc(size); | |
1367 | if (zeros == NULL) { | |
1368 | PERROR("zmalloc zeros for padding"); | |
1369 | ret = -1; | |
1370 | goto end; | |
1371 | } | |
1372 | ||
6cd525e8 MD |
1373 | ret = lttng_write(fd, zeros, size); |
1374 | if (ret < size) { | |
1d4dfdef DG |
1375 | PERROR("write padding to file"); |
1376 | } | |
1377 | ||
e986c7a1 DG |
1378 | free(zeros); |
1379 | ||
1d4dfdef DG |
1380 | end: |
1381 | return ret; | |
1382 | } | |
1383 | ||
b8aa1682 | 1384 | /* |
5a693d30 | 1385 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1386 | */ |
5a693d30 | 1387 | static int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1388 | struct relay_connection *conn) |
b8aa1682 | 1389 | { |
f73fabfd | 1390 | int ret = htobe32(LTTNG_OK); |
6cd525e8 | 1391 | ssize_t size_ret; |
58eb9381 | 1392 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1393 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1394 | struct relay_stream *metadata_stream; | |
1395 | uint64_t data_size, payload_size; | |
1396 | ||
1397 | if (!session) { | |
1398 | ERR("Metadata sent before version check"); | |
1399 | ret = -1; | |
1400 | goto end; | |
1401 | } | |
1402 | ||
f6416125 MD |
1403 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1404 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1405 | ERR("Incorrect data size"); | |
1406 | ret = -1; | |
1407 | goto end; | |
1408 | } | |
1409 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1410 | ||
b8aa1682 | 1411 | if (data_buffer_size < data_size) { |
d7b3776f | 1412 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1413 | char *tmp_data_ptr; |
1414 | ||
1415 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1416 | if (!tmp_data_ptr) { | |
b8aa1682 | 1417 | ERR("Allocating data buffer"); |
c617c0c6 | 1418 | free(data_buffer); |
b8aa1682 JD |
1419 | ret = -1; |
1420 | goto end; | |
1421 | } | |
c617c0c6 | 1422 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1423 | data_buffer_size = data_size; |
1424 | } | |
1425 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1426 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
58eb9381 | 1427 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 1428 | if (ret < 0 || ret != data_size) { |
a6cd2b97 DG |
1429 | if (ret == 0) { |
1430 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1431 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1432 | } else { |
1433 | ERR("Relay didn't receive the whole metadata"); | |
1434 | } | |
b8aa1682 | 1435 | ret = -1; |
b8aa1682 JD |
1436 | goto end; |
1437 | } | |
1438 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 | 1439 | |
5a693d30 | 1440 | metadata_stream = stream_get_by_id(be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1441 | if (!metadata_stream) { |
1442 | ret = -1; | |
5a693d30 | 1443 | goto end; |
b8aa1682 JD |
1444 | } |
1445 | ||
5a693d30 MD |
1446 | pthread_mutex_lock(&metadata_stream->lock); |
1447 | ||
1448 | size_ret = lttng_write(metadata_stream->stream_fd->fd, metadata_struct->payload, | |
6cd525e8 MD |
1449 | payload_size); |
1450 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1451 | ERR("Relay error writing metadata on file"); |
1452 | ret = -1; | |
5a693d30 | 1453 | goto end_put; |
b8aa1682 | 1454 | } |
1d4dfdef | 1455 | |
5a693d30 | 1456 | ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
1d4dfdef DG |
1457 | be32toh(metadata_struct->padding_size)); |
1458 | if (ret < 0) { | |
5a693d30 | 1459 | goto end_put; |
1d4dfdef | 1460 | } |
2a174661 | 1461 | |
5a693d30 | 1462 | metadata_stream->metadata_received += |
d3e2ba59 | 1463 | payload_size + be32toh(metadata_struct->padding_size); |
5a693d30 MD |
1464 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1465 | metadata_stream->metadata_received); | |
1d4dfdef | 1466 | |
5a693d30 MD |
1467 | end_put: |
1468 | pthread_mutex_unlock(&metadata_stream->lock); | |
1469 | stream_put(metadata_stream); | |
b8aa1682 JD |
1470 | |
1471 | end: | |
1472 | return ret; | |
1473 | } | |
1474 | ||
1475 | /* | |
1476 | * relay_send_version: send relayd version number | |
1477 | */ | |
5a693d30 | 1478 | static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1479 | struct relay_connection *conn) |
b8aa1682 | 1480 | { |
7f51dcba | 1481 | int ret; |
092b6259 | 1482 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1483 | |
58eb9381 | 1484 | conn->version_check_done = 1; |
b8aa1682 | 1485 | |
092b6259 | 1486 | /* Get version from the other side. */ |
58eb9381 | 1487 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1488 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1489 | if (ret == 0) { |
1490 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1491 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1492 | } else { |
1493 | ERR("Relay failed to receive the version values."); | |
1494 | } | |
092b6259 | 1495 | ret = -1; |
092b6259 DG |
1496 | goto end; |
1497 | } | |
1498 | ||
53efb85a | 1499 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1500 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1501 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1502 | |
1503 | /* Major versions must be the same */ | |
1504 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1505 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1506 | reply.major, be32toh(msg.major)); | |
5a693d30 | 1507 | connection_put(conn); |
d4519fa3 JD |
1508 | ret = 0; |
1509 | goto end; | |
1510 | } | |
1511 | ||
58eb9381 | 1512 | conn->major = reply.major; |
0f907de1 JD |
1513 | /* We adapt to the lowest compatible version */ |
1514 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1515 | conn->minor = reply.minor; |
0f907de1 | 1516 | } else { |
58eb9381 | 1517 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1518 | } |
1519 | ||
6151a90f JD |
1520 | reply.major = htobe32(reply.major); |
1521 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1522 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1523 | sizeof(struct lttcomm_relayd_version), 0); |
1524 | if (ret < 0) { | |
1525 | ERR("Relay sending version"); | |
1526 | } | |
1527 | ||
58eb9381 DG |
1528 | DBG("Version check done using protocol %u.%u", conn->major, |
1529 | conn->minor); | |
b8aa1682 JD |
1530 | |
1531 | end: | |
1532 | return ret; | |
1533 | } | |
1534 | ||
c8f59ee5 | 1535 | /* |
6d805429 | 1536 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1537 | */ |
5a693d30 | 1538 | static int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1539 | struct relay_connection *conn) |
c8f59ee5 | 1540 | { |
58eb9381 | 1541 | struct relay_session *session = conn->session; |
6d805429 | 1542 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1543 | struct lttcomm_relayd_generic_reply reply; |
1544 | struct relay_stream *stream; | |
1545 | int ret; | |
c8f59ee5 DG |
1546 | uint64_t last_net_seq_num, stream_id; |
1547 | ||
6d805429 | 1548 | DBG("Data pending command received"); |
c8f59ee5 | 1549 | |
58eb9381 | 1550 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1551 | ERR("Trying to check for data before version check"); |
1552 | ret = -1; | |
1553 | goto end_no_session; | |
1554 | } | |
1555 | ||
58eb9381 | 1556 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1557 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1558 | if (ret == 0) { |
1559 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1560 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1561 | } else { |
1562 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1563 | ret); | |
1564 | } | |
c8f59ee5 DG |
1565 | ret = -1; |
1566 | goto end_no_session; | |
1567 | } | |
1568 | ||
1569 | stream_id = be64toh(msg.stream_id); | |
1570 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1571 | ||
5a693d30 | 1572 | stream = stream_get_by_id(stream_id); |
de91f48a | 1573 | if (stream == NULL) { |
c8f59ee5 | 1574 | ret = -1; |
5a693d30 | 1575 | goto end; |
c8f59ee5 DG |
1576 | } |
1577 | ||
5a693d30 MD |
1578 | pthread_mutex_lock(&stream->lock); |
1579 | ||
6d805429 | 1580 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1581 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1582 | last_net_seq_num); | |
1583 | ||
33832e64 | 1584 | /* Avoid wrapping issue */ |
39df6d9f | 1585 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1586 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1587 | ret = 0; |
6d805429 DG |
1588 | } else { |
1589 | /* Data still being streamed thus pending */ | |
1590 | ret = 1; | |
c8f59ee5 DG |
1591 | } |
1592 | ||
5a693d30 MD |
1593 | stream->data_pending_check_done = true; |
1594 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1595 | |
5a693d30 MD |
1596 | stream_put(stream); |
1597 | end: | |
c8f59ee5 | 1598 | |
53efb85a | 1599 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1600 | reply.ret_code = htobe32(ret); |
58eb9381 | 1601 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1602 | if (ret < 0) { |
6d805429 | 1603 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1604 | } |
1605 | ||
1606 | end_no_session: | |
1607 | return ret; | |
1608 | } | |
1609 | ||
1610 | /* | |
1611 | * Wait for the control socket to reach a quiescent state. | |
1612 | * | |
5a693d30 MD |
1613 | * Note that for now, when receiving this command from the session |
1614 | * daemon, this means that every subsequent commands or data received on | |
1615 | * the control socket has been handled. So, this is why we simply return | |
1616 | * OK here. | |
c8f59ee5 | 1617 | */ |
5a693d30 | 1618 | static int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1619 | struct relay_connection *conn) |
c8f59ee5 DG |
1620 | { |
1621 | int ret; | |
ad7051c0 DG |
1622 | uint64_t stream_id; |
1623 | struct relay_stream *stream; | |
ad7051c0 | 1624 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1625 | struct lttcomm_relayd_generic_reply reply; |
1626 | ||
1627 | DBG("Checking quiescent state on control socket"); | |
1628 | ||
58eb9381 | 1629 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1630 | ERR("Trying to check for data before version check"); |
1631 | ret = -1; | |
1632 | goto end_no_session; | |
1633 | } | |
1634 | ||
58eb9381 | 1635 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1636 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1637 | if (ret == 0) { |
1638 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1639 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1640 | } else { |
1641 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1642 | ret); | |
1643 | } | |
ad7051c0 DG |
1644 | ret = -1; |
1645 | goto end_no_session; | |
1646 | } | |
1647 | ||
1648 | stream_id = be64toh(msg.stream_id); | |
5a693d30 MD |
1649 | stream = stream_get_by_id(stream_id); |
1650 | if (!stream) { | |
1651 | goto reply; | |
1652 | } | |
1653 | pthread_mutex_lock(&stream->lock); | |
1654 | stream->data_pending_check_done = true; | |
1655 | pthread_mutex_unlock(&stream->lock); | |
1656 | DBG("Relay quiescent control pending flag set to %" PRIu64, stream_id); | |
1657 | stream_put(stream); | |
1658 | reply: | |
53efb85a | 1659 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1660 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1661 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1662 | if (ret < 0) { |
6d805429 | 1663 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1664 | } |
1665 | ||
ad7051c0 | 1666 | end_no_session: |
c8f59ee5 DG |
1667 | return ret; |
1668 | } | |
1669 | ||
f7079f67 | 1670 | /* |
5a693d30 MD |
1671 | * Initialize a data pending command. This means that a consumer is about |
1672 | * to ask for data pending for each stream it holds. Simply iterate over | |
1673 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1674 | * |
1675 | * This command returns to the client a LTTNG_OK code. | |
1676 | */ | |
5a693d30 | 1677 | static int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1678 | struct relay_connection *conn) |
f7079f67 DG |
1679 | { |
1680 | int ret; | |
1681 | struct lttng_ht_iter iter; | |
1682 | struct lttcomm_relayd_begin_data_pending msg; | |
1683 | struct lttcomm_relayd_generic_reply reply; | |
1684 | struct relay_stream *stream; | |
1685 | uint64_t session_id; | |
1686 | ||
1687 | assert(recv_hdr); | |
58eb9381 | 1688 | assert(conn); |
f7079f67 DG |
1689 | |
1690 | DBG("Init streams for data pending"); | |
1691 | ||
58eb9381 | 1692 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1693 | ERR("Trying to check for data before version check"); |
1694 | ret = -1; | |
1695 | goto end_no_session; | |
1696 | } | |
1697 | ||
58eb9381 | 1698 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1699 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1700 | if (ret == 0) { |
1701 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1702 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1703 | } else { |
1704 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1705 | ret); | |
1706 | } | |
f7079f67 DG |
1707 | ret = -1; |
1708 | goto end_no_session; | |
1709 | } | |
1710 | ||
1711 | session_id = be64toh(msg.session_id); | |
1712 | ||
1713 | /* | |
5a693d30 MD |
1714 | * Iterate over all streams to set the begin data pending flag. |
1715 | * For now, the streams are indexed by stream handle so we have | |
1716 | * to iterate over all streams to find the one associated with | |
1717 | * the right session_id. | |
f7079f67 DG |
1718 | */ |
1719 | rcu_read_lock(); | |
d3e2ba59 | 1720 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1721 | node.node) { |
5a693d30 MD |
1722 | if (!stream_get(stream)) { |
1723 | continue; | |
1724 | } | |
1725 | if (stream->trace->session->id == session_id) { | |
1726 | pthread_mutex_lock(&stream->lock); | |
1727 | stream->data_pending_check_done = false; | |
1728 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1729 | DBG("Set begin data pending flag to stream %" PRIu64, |
1730 | stream->stream_handle); | |
1731 | } | |
5a693d30 | 1732 | stream_put(stream); |
f7079f67 DG |
1733 | } |
1734 | rcu_read_unlock(); | |
1735 | ||
53efb85a | 1736 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1737 | /* All good, send back reply. */ |
1738 | reply.ret_code = htobe32(LTTNG_OK); | |
1739 | ||
58eb9381 | 1740 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1741 | if (ret < 0) { |
1742 | ERR("Relay begin data pending send reply failed"); | |
1743 | } | |
1744 | ||
1745 | end_no_session: | |
1746 | return ret; | |
1747 | } | |
1748 | ||
1749 | /* | |
5a693d30 MD |
1750 | * End data pending command. This will check, for a given session id, if |
1751 | * each stream associated with it has its data_pending_check_done flag | |
1752 | * set. If not, this means that the client lost track of the stream but | |
1753 | * the data is still being streamed on our side. In this case, we inform | |
1754 | * the client that data is in flight. | |
f7079f67 DG |
1755 | * |
1756 | * Return to the client if there is data in flight or not with a ret_code. | |
1757 | */ | |
5a693d30 | 1758 | static int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1759 | struct relay_connection *conn) |
f7079f67 DG |
1760 | { |
1761 | int ret; | |
1762 | struct lttng_ht_iter iter; | |
1763 | struct lttcomm_relayd_end_data_pending msg; | |
1764 | struct lttcomm_relayd_generic_reply reply; | |
1765 | struct relay_stream *stream; | |
1766 | uint64_t session_id; | |
1767 | uint32_t is_data_inflight = 0; | |
1768 | ||
f7079f67 DG |
1769 | DBG("End data pending command"); |
1770 | ||
58eb9381 | 1771 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1772 | ERR("Trying to check for data before version check"); |
1773 | ret = -1; | |
1774 | goto end_no_session; | |
1775 | } | |
1776 | ||
58eb9381 | 1777 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1778 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1779 | if (ret == 0) { |
1780 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1781 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1782 | } else { |
1783 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1784 | ret); | |
1785 | } | |
f7079f67 DG |
1786 | ret = -1; |
1787 | goto end_no_session; | |
1788 | } | |
1789 | ||
1790 | session_id = be64toh(msg.session_id); | |
1791 | ||
5a693d30 MD |
1792 | /* |
1793 | * Iterate over all streams to see if the begin data pending | |
1794 | * flag is set. | |
1795 | */ | |
f7079f67 | 1796 | rcu_read_lock(); |
d3e2ba59 | 1797 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1798 | node.node) { |
5a693d30 MD |
1799 | if (!stream_get(stream)) { |
1800 | continue; | |
1801 | } | |
1802 | if (stream->trace->session->id != session_id) { | |
1803 | stream_put(stream); | |
1804 | continue; | |
1805 | } | |
1806 | pthread_mutex_lock(&stream->lock); | |
1807 | if (!stream->data_pending_check_done) { | |
1808 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1809 | is_data_inflight = 1; | |
1810 | DBG("Data is still in flight for stream %" PRIu64, | |
1811 | stream->stream_handle); | |
1812 | pthread_mutex_unlock(&stream->lock); | |
1813 | stream_put(stream); | |
1814 | break; | |
1815 | } | |
f7079f67 | 1816 | } |
5a693d30 MD |
1817 | pthread_mutex_unlock(&stream->lock); |
1818 | stream_put(stream); | |
f7079f67 DG |
1819 | } |
1820 | rcu_read_unlock(); | |
1821 | ||
53efb85a | 1822 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1823 | /* All good, send back reply. */ |
1824 | reply.ret_code = htobe32(is_data_inflight); | |
1825 | ||
58eb9381 | 1826 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1827 | if (ret < 0) { |
1828 | ERR("Relay end data pending send reply failed"); | |
1829 | } | |
1830 | ||
1831 | end_no_session: | |
1832 | return ret; | |
1833 | } | |
1834 | ||
1c20f0e2 JD |
1835 | /* |
1836 | * Receive an index for a specific stream. | |
1837 | * | |
1838 | * Return 0 on success else a negative value. | |
1839 | */ | |
5a693d30 | 1840 | static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1841 | struct relay_connection *conn) |
1c20f0e2 | 1842 | { |
5a693d30 | 1843 | int ret, send_ret; |
58eb9381 | 1844 | struct relay_session *session = conn->session; |
1c20f0e2 | 1845 | struct lttcomm_relayd_index index_info; |
5a693d30 | 1846 | struct relay_index *index; |
1c20f0e2 JD |
1847 | struct lttcomm_relayd_generic_reply reply; |
1848 | struct relay_stream *stream; | |
1849 | uint64_t net_seq_num; | |
1850 | ||
58eb9381 | 1851 | assert(conn); |
1c20f0e2 JD |
1852 | |
1853 | DBG("Relay receiving index"); | |
1854 | ||
58eb9381 | 1855 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1856 | ERR("Trying to close a stream before version check"); |
1857 | ret = -1; | |
1858 | goto end_no_session; | |
1859 | } | |
1860 | ||
58eb9381 | 1861 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1862 | sizeof(index_info), 0); |
1863 | if (ret < sizeof(index_info)) { | |
1864 | if (ret == 0) { | |
1865 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1866 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1867 | } else { |
1868 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1869 | } | |
1870 | ret = -1; | |
1871 | goto end_no_session; | |
1872 | } | |
1873 | ||
1874 | net_seq_num = be64toh(index_info.net_seq_num); | |
1875 | ||
5a693d30 | 1876 | stream = stream_get_by_id(be64toh(index_info.relay_stream_id)); |
1c20f0e2 | 1877 | if (!stream) { |
5a693d30 | 1878 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 1879 | ret = -1; |
5a693d30 | 1880 | goto end; |
1c20f0e2 | 1881 | } |
5a693d30 | 1882 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 1883 | |
d3e2ba59 JD |
1884 | /* Live beacon handling */ |
1885 | if (index_info.packet_size == 0) { | |
5a693d30 MD |
1886 | DBG("Received live beacon for stream %" PRIu64, |
1887 | stream->stream_handle); | |
d3e2ba59 JD |
1888 | |
1889 | /* | |
5a693d30 MD |
1890 | * Only flag a stream inactive when it has already |
1891 | * received data and no indexes are in flight. | |
d3e2ba59 | 1892 | */ |
5a693d30 MD |
1893 | if (stream->total_index_received > 0 |
1894 | && stream->indexes_in_flight == 0) { | |
1895 | stream->beacon_ts_end = | |
1896 | be64toh(index_info.timestamp_end); | |
d3e2ba59 JD |
1897 | } |
1898 | ret = 0; | |
5a693d30 | 1899 | goto end_stream_put; |
d3e2ba59 JD |
1900 | } else { |
1901 | stream->beacon_ts_end = -1ULL; | |
1902 | } | |
1903 | ||
528f2ffa JD |
1904 | if (stream->ctf_stream_id == -1ULL) { |
1905 | stream->ctf_stream_id = be64toh(index_info.stream_id); | |
1906 | } | |
5a693d30 MD |
1907 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
1908 | if (!index) { | |
1909 | ret = -1; | |
1910 | ERR("relay_index_get_by_id_or_create index NULL"); | |
1911 | goto end_stream_put; | |
1c20f0e2 | 1912 | } |
5a693d30 MD |
1913 | if (set_index_control_data(index, &index_info)) { |
1914 | ERR("set_index_control_data error"); | |
1915 | relay_index_put(index); | |
1916 | ret = -1; | |
1917 | goto end_stream_put; | |
1918 | } | |
1919 | ret = relay_index_try_flush(index); | |
1920 | if (ret == 0) { | |
d3e2ba59 | 1921 | stream->total_index_received++; |
5a693d30 MD |
1922 | } else if (ret > 0) { |
1923 | /* no flush. */ | |
1924 | ret = 0; | |
1925 | } else { | |
1926 | ERR("relay_index_try_flush error %d", ret); | |
1927 | relay_index_put(index); | |
1928 | ret = -1; | |
1c20f0e2 JD |
1929 | } |
1930 | ||
5a693d30 MD |
1931 | end_stream_put: |
1932 | pthread_mutex_unlock(&stream->lock); | |
1933 | stream_put(stream); | |
1934 | ||
1935 | end: | |
1c20f0e2 | 1936 | |
53efb85a | 1937 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
1938 | if (ret < 0) { |
1939 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1940 | } else { | |
1941 | reply.ret_code = htobe32(LTTNG_OK); | |
1942 | } | |
58eb9381 | 1943 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
1944 | if (send_ret < 0) { |
1945 | ERR("Relay sending close index id reply"); | |
1946 | ret = send_ret; | |
1947 | } | |
1948 | ||
1949 | end_no_session: | |
1950 | return ret; | |
1951 | } | |
1952 | ||
a4baae1b JD |
1953 | /* |
1954 | * Receive the streams_sent message. | |
1955 | * | |
1956 | * Return 0 on success else a negative value. | |
1957 | */ | |
5a693d30 | 1958 | static int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1959 | struct relay_connection *conn) |
a4baae1b JD |
1960 | { |
1961 | int ret, send_ret; | |
1962 | struct lttcomm_relayd_generic_reply reply; | |
1963 | ||
58eb9381 | 1964 | assert(conn); |
a4baae1b JD |
1965 | |
1966 | DBG("Relay receiving streams_sent"); | |
1967 | ||
58eb9381 | 1968 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
1969 | ERR("Trying to close a stream before version check"); |
1970 | ret = -1; | |
1971 | goto end_no_session; | |
1972 | } | |
1973 | ||
1974 | /* | |
5a693d30 MD |
1975 | * Publish every pending stream in the connection recv list which are |
1976 | * now ready to be used by the viewer. | |
4a9daf17 | 1977 | */ |
5a693d30 | 1978 | publish_connection_local_streams(conn); |
4a9daf17 | 1979 | |
53efb85a | 1980 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 1981 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1982 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
1983 | if (send_ret < 0) { |
1984 | ERR("Relay sending sent_stream reply"); | |
1985 | ret = send_ret; | |
1986 | } else { | |
1987 | /* Success. */ | |
1988 | ret = 0; | |
1989 | } | |
1990 | ||
1991 | end_no_session: | |
1992 | return ret; | |
1993 | } | |
1994 | ||
b8aa1682 | 1995 | /* |
d3e2ba59 | 1996 | * Process the commands received on the control socket |
b8aa1682 | 1997 | */ |
5a693d30 | 1998 | static int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1999 | struct relay_connection *conn) |
b8aa1682 JD |
2000 | { |
2001 | int ret = 0; | |
2002 | ||
2003 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 2004 | case RELAYD_CREATE_SESSION: |
58eb9381 | 2005 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 2006 | break; |
b8aa1682 | 2007 | case RELAYD_ADD_STREAM: |
58eb9381 | 2008 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
2009 | break; |
2010 | case RELAYD_START_DATA: | |
58eb9381 | 2011 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
2012 | break; |
2013 | case RELAYD_SEND_METADATA: | |
58eb9381 | 2014 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
2015 | break; |
2016 | case RELAYD_VERSION: | |
58eb9381 | 2017 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 2018 | break; |
173af62f | 2019 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 2020 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 2021 | break; |
6d805429 | 2022 | case RELAYD_DATA_PENDING: |
58eb9381 | 2023 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
2024 | break; |
2025 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 2026 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 2027 | break; |
f7079f67 | 2028 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 2029 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
2030 | break; |
2031 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2032 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2033 | break; |
1c20f0e2 | 2034 | case RELAYD_SEND_INDEX: |
58eb9381 | 2035 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2036 | break; |
a4baae1b | 2037 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2038 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2039 | break; |
b8aa1682 JD |
2040 | case RELAYD_UPDATE_SYNC_INFO: |
2041 | default: | |
2042 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2043 | relay_unknown_command(conn); |
b8aa1682 JD |
2044 | ret = -1; |
2045 | goto end; | |
2046 | } | |
2047 | ||
2048 | end: | |
2049 | return ret; | |
2050 | } | |
2051 | ||
7d2f7452 DG |
2052 | /* |
2053 | * Handle index for a data stream. | |
2054 | * | |
5a693d30 | 2055 | * Called with the stream lock held. |
7d2f7452 DG |
2056 | * |
2057 | * Return 0 on success else a negative value. | |
2058 | */ | |
2059 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2060 | int rotate_index) | |
2061 | { | |
5a693d30 MD |
2062 | int ret = 0; |
2063 | uint64_t data_offset; | |
2064 | struct relay_index *index; | |
7d2f7452 | 2065 | |
7d2f7452 DG |
2066 | /* Get data offset because we are about to update the index. */ |
2067 | data_offset = htobe64(stream->tracefile_size_current); | |
2068 | ||
5a693d30 MD |
2069 | DBG("handle_index_data: stream %" PRIu64 " data offset %" PRIu64, |
2070 | stream->stream_handle, stream->tracefile_size_current); | |
2071 | ||
7d2f7452 | 2072 | /* |
5a693d30 MD |
2073 | * Lookup for an existing index for that stream id/sequence |
2074 | * number. If it exists, the control thread has already received the | |
2075 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2076 | */ |
5a693d30 | 2077 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2078 | if (!index) { |
5a693d30 MD |
2079 | ret = -1; |
2080 | goto end; | |
7d2f7452 DG |
2081 | } |
2082 | ||
5a693d30 MD |
2083 | if (rotate_index || !stream->index_fd) { |
2084 | int fd; | |
2085 | ||
2086 | /* Put ref on previous index_fd. */ | |
2087 | if (stream->index_fd) { | |
2088 | stream_fd_put(stream->index_fd); | |
2089 | stream->index_fd = NULL; | |
7d2f7452 | 2090 | } |
7d2f7452 | 2091 | |
5a693d30 MD |
2092 | fd = index_create_file(stream->path_name, stream->channel_name, |
2093 | -1, -1, stream->tracefile_size, | |
2094 | stream->current_tracefile_id); | |
2095 | if (fd < 0) { | |
2096 | ret = -1; | |
2097 | /* Put self-ref for this index due to error. */ | |
2098 | relay_index_put(index); | |
2099 | goto end; | |
2100 | } | |
2101 | stream->index_fd = stream_fd_create(fd); | |
2102 | if (!stream->index_fd) { | |
2103 | ret = -1; | |
2104 | if (close(fd)) { | |
2105 | PERROR("Error closing FD %d", fd); | |
2106 | } | |
2107 | /* Put self-ref for this index due to error. */ | |
2108 | relay_index_put(index); | |
2109 | /* Will put the local ref. */ | |
2110 | goto end; | |
7d2f7452 | 2111 | } |
7d2f7452 DG |
2112 | } |
2113 | ||
5a693d30 MD |
2114 | if (relay_index_set_fd(index, stream->index_fd, data_offset)) { |
2115 | ret = -1; | |
2116 | /* Put self-ref for this index due to error. */ | |
2117 | relay_index_put(index); | |
2118 | goto end; | |
7d2f7452 DG |
2119 | } |
2120 | ||
5a693d30 MD |
2121 | ret = relay_index_try_flush(index); |
2122 | if (ret == 0) { | |
2123 | stream->total_index_received++; | |
2124 | } else if (ret > 0) { | |
2125 | /* No flush. */ | |
2126 | ret = 0; | |
2127 | } else { | |
2128 | /* Put self-ref for this index due to error. */ | |
2129 | relay_index_put(index); | |
2130 | ret = -1; | |
2131 | } | |
2132 | end: | |
7d2f7452 DG |
2133 | return ret; |
2134 | } | |
2135 | ||
b8aa1682 JD |
2136 | /* |
2137 | * relay_process_data: Process the data received on the data socket | |
2138 | */ | |
5a693d30 | 2139 | static int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2140 | { |
7d2f7452 | 2141 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2142 | ssize_t size_ret; |
b8aa1682 JD |
2143 | struct relay_stream *stream; |
2144 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2145 | uint64_t stream_id; |
173af62f | 2146 | uint64_t net_seq_num; |
b8aa1682 | 2147 | uint32_t data_size; |
2a174661 | 2148 | struct relay_session *session; |
b8aa1682 | 2149 | |
58eb9381 | 2150 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, |
7c5aef62 | 2151 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2152 | if (ret <= 0) { |
a6cd2b97 DG |
2153 | if (ret == 0) { |
2154 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2155 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2156 | } else { |
58eb9381 | 2157 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2158 | } |
b8aa1682 JD |
2159 | ret = -1; |
2160 | goto end; | |
2161 | } | |
2162 | ||
2163 | stream_id = be64toh(data_hdr.stream_id); | |
5a693d30 | 2164 | stream = stream_get_by_id(stream_id); |
b8aa1682 JD |
2165 | if (!stream) { |
2166 | ret = -1; | |
5a693d30 | 2167 | goto end; |
b8aa1682 | 2168 | } |
5a693d30 | 2169 | session = stream->trace->session; |
b8aa1682 JD |
2170 | data_size = be32toh(data_hdr.data_size); |
2171 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2172 | char *tmp_data_ptr; |
2173 | ||
2174 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2175 | if (!tmp_data_ptr) { | |
b8aa1682 | 2176 | ERR("Allocating data buffer"); |
c617c0c6 | 2177 | free(data_buffer); |
b8aa1682 | 2178 | ret = -1; |
5a693d30 | 2179 | goto end_stream_put; |
b8aa1682 | 2180 | } |
c617c0c6 | 2181 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2182 | data_buffer_size = data_size; |
2183 | } | |
2184 | memset(data_buffer, 0, data_size); | |
2185 | ||
173af62f DG |
2186 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2187 | ||
77c7c900 | 2188 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2189 | data_size, stream_id, net_seq_num); |
58eb9381 | 2190 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 2191 | if (ret <= 0) { |
a6cd2b97 DG |
2192 | if (ret == 0) { |
2193 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2194 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2195 | } |
b8aa1682 | 2196 | ret = -1; |
5a693d30 | 2197 | goto end_stream_put; |
b8aa1682 JD |
2198 | } |
2199 | ||
5a693d30 MD |
2200 | pthread_mutex_lock(&stream->lock); |
2201 | ||
1c20f0e2 | 2202 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2203 | if (stream->tracefile_size > 0 && |
2204 | (stream->tracefile_size_current + data_size) > | |
2205 | stream->tracefile_size) { | |
6b6b9a5a JD |
2206 | uint64_t new_id; |
2207 | ||
5a693d30 | 2208 | new_id = (stream->current_tracefile_id + 1) % |
6b6b9a5a JD |
2209 | stream->tracefile_count; |
2210 | /* | |
5a693d30 MD |
2211 | * Move viewer oldest available data position forward if |
2212 | * we are overwriting a tracefile. | |
6b6b9a5a | 2213 | */ |
5a693d30 | 2214 | if (new_id == stream->oldest_tracefile_id) { |
6b6b9a5a JD |
2215 | stream->oldest_tracefile_id = |
2216 | (stream->oldest_tracefile_id + 1) % | |
2217 | stream->tracefile_count; | |
2218 | } | |
5a693d30 MD |
2219 | ret = utils_rotate_stream_file(stream->path_name, |
2220 | stream->channel_name, stream->tracefile_size, | |
2221 | stream->tracefile_count, -1, | |
2222 | -1, stream->stream_fd->fd, | |
2223 | &stream->current_tracefile_id, | |
2224 | &stream->stream_fd->fd); | |
0f907de1 | 2225 | if (ret < 0) { |
1c20f0e2 | 2226 | ERR("Rotating stream output file"); |
5a693d30 MD |
2227 | goto end_stream_unlock; |
2228 | } | |
2229 | stream->current_tracefile_seq++; | |
2230 | if (stream->current_tracefile_seq | |
2231 | - stream->oldest_tracefile_seq >= | |
2232 | stream->tracefile_count) { | |
2233 | stream->oldest_tracefile_seq++; | |
0f907de1 | 2234 | } |
5a693d30 MD |
2235 | /* |
2236 | * Reset current size because we just performed a stream | |
2237 | * rotation. | |
2238 | */ | |
a6976990 | 2239 | stream->tracefile_size_current = 0; |
1c20f0e2 JD |
2240 | rotate_index = 1; |
2241 | } | |
2242 | ||
1c20f0e2 | 2243 | /* |
5a693d30 MD |
2244 | * Index are handled in protocol version 2.4 and above. Also, |
2245 | * snapshot and index are NOT supported. | |
1c20f0e2 | 2246 | */ |
2a174661 | 2247 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2248 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2249 | if (ret < 0) { |
5a693d30 | 2250 | goto end_stream_unlock; |
1c20f0e2 | 2251 | } |
1c20f0e2 JD |
2252 | } |
2253 | ||
7d2f7452 | 2254 | /* Write data to stream output fd. */ |
5a693d30 | 2255 | size_ret = lttng_write(stream->stream_fd->fd, data_buffer, data_size); |
6cd525e8 | 2256 | if (size_ret < data_size) { |
b8aa1682 JD |
2257 | ERR("Relay error writing data to file"); |
2258 | ret = -1; | |
5a693d30 | 2259 | goto end_stream_unlock; |
b8aa1682 | 2260 | } |
1d4dfdef | 2261 | |
5a693d30 MD |
2262 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, |
2263 | size_ret, stream->stream_handle); | |
5ab7344e | 2264 | |
5a693d30 MD |
2265 | ret = write_padding_to_file(stream->stream_fd->fd, |
2266 | be32toh(data_hdr.padding_size)); | |
1d4dfdef | 2267 | if (ret < 0) { |
5a693d30 | 2268 | goto end_stream_unlock; |
1d4dfdef | 2269 | } |
5a693d30 MD |
2270 | stream->tracefile_size_current += |
2271 | data_size + be32toh(data_hdr.padding_size); | |
173af62f DG |
2272 | stream->prev_seq = net_seq_num; |
2273 | ||
5a693d30 MD |
2274 | end_stream_unlock: |
2275 | pthread_mutex_unlock(&stream->lock); | |
2276 | end_stream_put: | |
2277 | stream_put(stream); | |
b8aa1682 JD |
2278 | end: |
2279 | return ret; | |
2280 | } | |
2281 | ||
5a693d30 | 2282 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2283 | { |
2284 | int ret; | |
2285 | ||
58eb9381 | 2286 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2287 | |
2288 | ret = close(pollfd); | |
2289 | if (ret < 0) { | |
2290 | ERR("Closing pollfd %d", pollfd); | |
2291 | } | |
2292 | } | |
2293 | ||
5a693d30 MD |
2294 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2295 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2296 | { |
5a693d30 | 2297 | const char *type_str; |
2a174661 | 2298 | |
5a693d30 MD |
2299 | switch (conn->type) { |
2300 | case RELAY_DATA: | |
2301 | type_str = "Data"; | |
2302 | break; | |
2303 | case RELAY_CONTROL: | |
2304 | type_str = "Control"; | |
2305 | break; | |
2306 | case RELAY_VIEWER_COMMAND: | |
2307 | type_str = "Viewer Command"; | |
2308 | break; | |
2309 | case RELAY_VIEWER_NOTIFICATION: | |
2310 | type_str = "Viewer Notification"; | |
2311 | break; | |
2312 | default: | |
2313 | type_str = "Unknown"; | |
9d1bbf21 | 2314 | } |
5a693d30 MD |
2315 | cleanup_connection_pollfd(events, pollfd); |
2316 | connection_put(conn); | |
2317 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2318 | } |
2319 | ||
2320 | /* | |
2321 | * This thread does the actual work | |
2322 | */ | |
5a693d30 | 2323 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2324 | { |
beaad64c DG |
2325 | int ret, err = -1, last_seen_data_fd = -1; |
2326 | uint32_t nb_fd; | |
b8aa1682 JD |
2327 | struct lttng_poll_event events; |
2328 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2329 | struct lttng_ht_iter iter; |
b8aa1682 | 2330 | struct lttcomm_relayd_hdr recv_hdr; |
90e7d72f | 2331 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2332 | |
2333 | DBG("[thread] Relay worker started"); | |
2334 | ||
9d1bbf21 MD |
2335 | rcu_register_thread(); |
2336 | ||
55706a7d MD |
2337 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2338 | ||
9b5e0863 MD |
2339 | if (testpoint(relayd_thread_worker)) { |
2340 | goto error_testpoint; | |
2341 | } | |
2342 | ||
f385ae0a MD |
2343 | health_code_update(); |
2344 | ||
b8aa1682 JD |
2345 | /* table of connections indexed on socket */ |
2346 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2347 | if (!relay_connections_ht) { |
2348 | goto relay_connections_ht_error; | |
2349 | } | |
b8aa1682 | 2350 | |
b8aa1682 JD |
2351 | ret = create_thread_poll_set(&events, 2); |
2352 | if (ret < 0) { | |
2353 | goto error_poll_create; | |
2354 | } | |
2355 | ||
58eb9381 | 2356 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2357 | if (ret < 0) { |
2358 | goto error; | |
2359 | } | |
2360 | ||
beaad64c | 2361 | restart: |
b8aa1682 | 2362 | while (1) { |
beaad64c DG |
2363 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2364 | ||
f385ae0a MD |
2365 | health_code_update(); |
2366 | ||
b8aa1682 | 2367 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2368 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2369 | health_poll_entry(); |
b8aa1682 | 2370 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2371 | health_poll_exit(); |
b8aa1682 JD |
2372 | if (ret < 0) { |
2373 | /* | |
2374 | * Restart interrupted system call. | |
2375 | */ | |
2376 | if (errno == EINTR) { | |
2377 | goto restart; | |
2378 | } | |
2379 | goto error; | |
2380 | } | |
2381 | ||
0d9c5d77 DG |
2382 | nb_fd = ret; |
2383 | ||
beaad64c | 2384 | /* |
5a693d30 MD |
2385 | * Process control. The control connection is |
2386 | * prioritized so we don't starve it with high | |
2387 | * throughput tracing data on the data connection. | |
beaad64c | 2388 | */ |
b8aa1682 JD |
2389 | for (i = 0; i < nb_fd; i++) { |
2390 | /* Fetch once the poll data */ | |
beaad64c DG |
2391 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2392 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2393 | |
f385ae0a MD |
2394 | health_code_update(); |
2395 | ||
fd20dac9 | 2396 | if (!revents) { |
5a693d30 MD |
2397 | /* |
2398 | * No activity for this FD (poll | |
2399 | * implementation). | |
2400 | */ | |
fd20dac9 MD |
2401 | continue; |
2402 | } | |
2403 | ||
b8aa1682 JD |
2404 | /* Thread quit pipe has been closed. Killing thread. */ |
2405 | ret = check_thread_quit_pipe(pollfd, revents); | |
2406 | if (ret) { | |
095a4ae5 MD |
2407 | err = 0; |
2408 | goto exit; | |
b8aa1682 JD |
2409 | } |
2410 | ||
58eb9381 DG |
2411 | /* Inspect the relay conn pipe for new connection */ |
2412 | if (pollfd == relay_conn_pipe[0]) { | |
b8aa1682 | 2413 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
58eb9381 | 2414 | ERR("Relay connection pipe error"); |
b8aa1682 JD |
2415 | goto error; |
2416 | } else if (revents & LPOLLIN) { | |
90e7d72f JG |
2417 | struct relay_connection *conn; |
2418 | ||
58eb9381 | 2419 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2420 | if (ret < 0) { |
2421 | goto error; | |
2422 | } | |
58eb9381 DG |
2423 | lttng_poll_add(&events, conn->sock->fd, |
2424 | LPOLLIN | LPOLLRDHUP); | |
5a693d30 | 2425 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2426 | DBG("Connection socket %d added", conn->sock->fd); |
b8aa1682 | 2427 | } |
58eb9381 | 2428 | } else { |
90e7d72f JG |
2429 | struct relay_connection *ctrl_conn; |
2430 | ||
5a693d30 | 2431 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 2432 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 2433 | assert(ctrl_conn); |
58eb9381 DG |
2434 | |
2435 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
5a693d30 | 2436 | relay_thread_close_connection(&events, pollfd, ctrl_conn); |
beaad64c DG |
2437 | if (last_seen_data_fd == pollfd) { |
2438 | last_seen_data_fd = last_notdel_data_fd; | |
2439 | } | |
b8aa1682 | 2440 | } else if (revents & LPOLLIN) { |
90e7d72f JG |
2441 | if (ctrl_conn->type == RELAY_CONTROL) { |
2442 | ret = ctrl_conn->sock->ops->recvmsg(ctrl_conn->sock, &recv_hdr, | |
58eb9381 | 2443 | sizeof(recv_hdr), 0); |
b8aa1682 | 2444 | if (ret <= 0) { |
58eb9381 | 2445 | /* Connection closed */ |
5a693d30 MD |
2446 | relay_thread_close_connection(&events, pollfd, |
2447 | ctrl_conn); | |
b8aa1682 | 2448 | } else { |
90e7d72f | 2449 | ret = relay_process_control(&recv_hdr, ctrl_conn); |
b8aa1682 | 2450 | if (ret < 0) { |
beaad64c | 2451 | /* Clear the session on error. */ |
5a693d30 MD |
2452 | relay_thread_close_connection(&events, pollfd, |
2453 | ctrl_conn); | |
b8aa1682 | 2454 | } |
beaad64c | 2455 | seen_control = 1; |
b8aa1682 | 2456 | } |
beaad64c DG |
2457 | } else { |
2458 | /* | |
2459 | * Flag the last seen data fd not deleted. It will be | |
2460 | * used as the last seen fd if any fd gets deleted in | |
2461 | * this first loop. | |
2462 | */ | |
2463 | last_notdel_data_fd = pollfd; | |
2464 | } | |
58eb9381 DG |
2465 | } else { |
2466 | ERR("Unknown poll events %u for sock %d", revents, pollfd); | |
beaad64c | 2467 | } |
5a693d30 | 2468 | connection_put(ctrl_conn); |
beaad64c DG |
2469 | } |
2470 | } | |
2471 | ||
2472 | /* | |
2473 | * The last loop handled a control request, go back to poll to make | |
2474 | * sure we prioritise the control socket. | |
2475 | */ | |
2476 | if (seen_control) { | |
2477 | continue; | |
2478 | } | |
2479 | ||
2480 | if (last_seen_data_fd >= 0) { | |
2481 | for (i = 0; i < nb_fd; i++) { | |
2482 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2483 | |
2484 | health_code_update(); | |
2485 | ||
beaad64c DG |
2486 | if (last_seen_data_fd == pollfd) { |
2487 | idx = i; | |
2488 | break; | |
2489 | } | |
2490 | } | |
2491 | } | |
2492 | ||
2493 | /* Process data connection. */ | |
2494 | for (i = idx + 1; i < nb_fd; i++) { | |
2495 | /* Fetch the poll data. */ | |
2496 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2497 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 2498 | struct relay_connection *data_conn; |
beaad64c | 2499 | |
f385ae0a MD |
2500 | health_code_update(); |
2501 | ||
fd20dac9 MD |
2502 | if (!revents) { |
2503 | /* No activity for this FD (poll implementation). */ | |
2504 | continue; | |
2505 | } | |
2506 | ||
beaad64c | 2507 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2508 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2509 | continue; |
2510 | } | |
2511 | ||
5a693d30 | 2512 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 2513 | if (!data_conn) { |
fd20dac9 | 2514 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
2515 | continue; |
2516 | } | |
2517 | ||
2518 | if (revents & LPOLLIN) { | |
90e7d72f | 2519 | if (data_conn->type != RELAY_DATA) { |
5a693d30 | 2520 | goto put_connection; |
beaad64c | 2521 | } |
beaad64c | 2522 | |
90e7d72f | 2523 | ret = relay_process_data(data_conn); |
fd20dac9 MD |
2524 | /* Connection closed */ |
2525 | if (ret < 0) { | |
5a693d30 MD |
2526 | relay_thread_close_connection(&events, pollfd, |
2527 | data_conn); | |
fd20dac9 MD |
2528 | /* |
2529 | * Every goto restart call sets the last seen fd where | |
2530 | * here we don't really care since we gracefully | |
2531 | * continue the loop after the connection is deleted. | |
2532 | */ | |
2533 | } else { | |
2534 | /* Keep last seen port. */ | |
2535 | last_seen_data_fd = pollfd; | |
5a693d30 | 2536 | connection_put(data_conn); |
fd20dac9 | 2537 | goto restart; |
b8aa1682 JD |
2538 | } |
2539 | } | |
5a693d30 MD |
2540 | put_connection: |
2541 | connection_put(data_conn); | |
b8aa1682 | 2542 | } |
beaad64c | 2543 | last_seen_data_fd = -1; |
b8aa1682 JD |
2544 | } |
2545 | ||
f385ae0a MD |
2546 | /* Normal exit, no error */ |
2547 | ret = 0; | |
2548 | ||
095a4ae5 | 2549 | exit: |
b8aa1682 | 2550 | error: |
58eb9381 | 2551 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2552 | rcu_read_lock(); |
90e7d72f JG |
2553 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
2554 | destroy_conn, | |
58eb9381 | 2555 | sock_n.node) { |
f385ae0a | 2556 | health_code_update(); |
5a693d30 MD |
2557 | /* |
2558 | * No need to grab another ref, because we own | |
2559 | * destroy_conn. | |
2560 | */ | |
2561 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
2562 | destroy_conn); | |
b8aa1682 | 2563 | } |
94d49140 | 2564 | rcu_read_unlock(); |
5a693d30 MD |
2565 | |
2566 | lttng_poll_clean(&events); | |
7d2f7452 | 2567 | error_poll_create: |
b8aa1682 | 2568 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2569 | relay_connections_ht_error: |
58eb9381 DG |
2570 | /* Close relay conn pipes */ |
2571 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2572 | if (err) { |
2573 | DBG("Thread exited with error"); | |
2574 | } | |
b8aa1682 | 2575 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2576 | free(data_buffer); |
9b5e0863 | 2577 | error_testpoint: |
f385ae0a MD |
2578 | if (err) { |
2579 | health_error(); | |
2580 | ERR("Health error occurred in %s", __func__); | |
2581 | } | |
2582 | health_unregister(health_relayd); | |
9d1bbf21 | 2583 | rcu_unregister_thread(); |
b4aacfdc | 2584 | lttng_relay_stop_threads(); |
b8aa1682 JD |
2585 | return NULL; |
2586 | } | |
2587 | ||
2588 | /* | |
2589 | * Create the relay command pipe to wake thread_manage_apps. | |
2590 | * Closed in cleanup(). | |
2591 | */ | |
58eb9381 | 2592 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2593 | { |
a02de639 | 2594 | int ret; |
b8aa1682 | 2595 | |
58eb9381 | 2596 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2597 | |
b8aa1682 JD |
2598 | return ret; |
2599 | } | |
2600 | ||
2601 | /* | |
2602 | * main | |
2603 | */ | |
2604 | int main(int argc, char **argv) | |
2605 | { | |
178a0557 | 2606 | int ret = 0, retval = 0; |
b8aa1682 JD |
2607 | void *status; |
2608 | ||
b8aa1682 JD |
2609 | /* Parse arguments */ |
2610 | progname = argv[0]; | |
178a0557 MD |
2611 | if (set_options(argc, argv)) { |
2612 | retval = -1; | |
2613 | goto exit_options; | |
b8aa1682 JD |
2614 | } |
2615 | ||
178a0557 MD |
2616 | if (set_signal_handler()) { |
2617 | retval = -1; | |
2618 | goto exit_options; | |
b8aa1682 JD |
2619 | } |
2620 | ||
4d513a50 DG |
2621 | /* Try to create directory if -o, --output is specified. */ |
2622 | if (opt_output_path) { | |
994fa64f DG |
2623 | if (*opt_output_path != '/') { |
2624 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
2625 | retval = -1; |
2626 | goto exit_options; | |
994fa64f DG |
2627 | } |
2628 | ||
d85144f4 JG |
2629 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
2630 | -1, -1); | |
4d513a50 DG |
2631 | if (ret < 0) { |
2632 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
2633 | retval = -1; |
2634 | goto exit_options; | |
4d513a50 DG |
2635 | } |
2636 | } | |
2637 | ||
b8aa1682 | 2638 | /* Daemonize */ |
b5218ffb | 2639 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2640 | int i; |
2641 | ||
2642 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2643 | !opt_background); | |
b8aa1682 | 2644 | if (ret < 0) { |
178a0557 MD |
2645 | retval = -1; |
2646 | goto exit_options; | |
b8aa1682 | 2647 | } |
3fd27398 MD |
2648 | |
2649 | /* | |
2650 | * We are in the child. Make sure all other file | |
2651 | * descriptors are closed, in case we are called with | |
2652 | * more opened file descriptors than the standard ones. | |
2653 | */ | |
2654 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2655 | (void) close(i); | |
2656 | } | |
2657 | } | |
2658 | ||
178a0557 MD |
2659 | |
2660 | /* Initialize thread health monitoring */ | |
2661 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2662 | if (!health_relayd) { | |
2663 | PERROR("health_app_create error"); | |
2664 | retval = -1; | |
2665 | goto exit_health_app_create; | |
2666 | } | |
2667 | ||
3fd27398 | 2668 | /* Create thread quit pipe */ |
178a0557 MD |
2669 | if (init_thread_quit_pipe()) { |
2670 | retval = -1; | |
2671 | goto exit_init_data; | |
b8aa1682 JD |
2672 | } |
2673 | ||
1c20f0e2 | 2674 | /* Check if daemon is UID = 0 */ |
5a693d30 | 2675 | if (!getuid()) { |
8d5c808e | 2676 | if (control_uri->port < 1024 || data_uri->port < 1024 || live_uri->port < 1024) { |
b8aa1682 | 2677 | ERR("Need to be root to use ports < 1024"); |
178a0557 MD |
2678 | retval = -1; |
2679 | goto exit_init_data; | |
b8aa1682 JD |
2680 | } |
2681 | } | |
2682 | ||
2683 | /* Setup the thread apps communication pipe. */ | |
178a0557 MD |
2684 | if (create_relay_conn_pipe()) { |
2685 | retval = -1; | |
2686 | goto exit_init_data; | |
b8aa1682 JD |
2687 | } |
2688 | ||
2689 | /* Init relay command queue. */ | |
8bdee6e2 | 2690 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 JD |
2691 | |
2692 | /* Set up max poll set size */ | |
25b397f9 MD |
2693 | if (lttng_poll_set_max_size()) { |
2694 | retval = -1; | |
2695 | goto exit_init_data; | |
2696 | } | |
b8aa1682 | 2697 | |
554831e7 MD |
2698 | /* Initialize communication library */ |
2699 | lttcomm_init(); | |
87e45c13 | 2700 | lttcomm_inet_init(); |
554831e7 | 2701 | |
d3e2ba59 | 2702 | /* tables of sessions indexed by session ID */ |
5a693d30 MD |
2703 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2704 | if (!sessions_ht) { | |
178a0557 MD |
2705 | retval = -1; |
2706 | goto exit_init_data; | |
d3e2ba59 JD |
2707 | } |
2708 | ||
2709 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2710 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 2711 | if (!relay_streams_ht) { |
178a0557 MD |
2712 | retval = -1; |
2713 | goto exit_init_data; | |
d3e2ba59 JD |
2714 | } |
2715 | ||
2716 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2717 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2718 | if (!viewer_streams_ht) { | |
178a0557 MD |
2719 | retval = -1; |
2720 | goto exit_init_data; | |
55706a7d MD |
2721 | } |
2722 | ||
65931c8b | 2723 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
2724 | if (ret) { |
2725 | retval = -1; | |
2726 | goto exit_health_quit_pipe; | |
65931c8b MD |
2727 | } |
2728 | ||
2729 | /* Create thread to manage the client socket */ | |
2730 | ret = pthread_create(&health_thread, NULL, | |
2731 | thread_manage_health, (void *) NULL); | |
178a0557 MD |
2732 | if (ret) { |
2733 | errno = ret; | |
65931c8b | 2734 | PERROR("pthread_create health"); |
178a0557 MD |
2735 | retval = -1; |
2736 | goto exit_health_thread; | |
65931c8b MD |
2737 | } |
2738 | ||
b8aa1682 JD |
2739 | /* Setup the dispatcher thread */ |
2740 | ret = pthread_create(&dispatcher_thread, NULL, | |
2741 | relay_thread_dispatcher, (void *) NULL); | |
178a0557 MD |
2742 | if (ret) { |
2743 | errno = ret; | |
b8aa1682 | 2744 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
2745 | retval = -1; |
2746 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
2747 | } |
2748 | ||
2749 | /* Setup the worker thread */ | |
2750 | ret = pthread_create(&worker_thread, NULL, | |
5a693d30 | 2751 | relay_thread_worker, NULL); |
178a0557 MD |
2752 | if (ret) { |
2753 | errno = ret; | |
b8aa1682 | 2754 | PERROR("pthread_create worker"); |
178a0557 MD |
2755 | retval = -1; |
2756 | goto exit_worker_thread; | |
b8aa1682 JD |
2757 | } |
2758 | ||
2759 | /* Setup the listener thread */ | |
2760 | ret = pthread_create(&listener_thread, NULL, | |
2761 | relay_thread_listener, (void *) NULL); | |
178a0557 MD |
2762 | if (ret) { |
2763 | errno = ret; | |
b8aa1682 | 2764 | PERROR("pthread_create listener"); |
178a0557 MD |
2765 | retval = -1; |
2766 | goto exit_listener_thread; | |
b8aa1682 JD |
2767 | } |
2768 | ||
5a693d30 | 2769 | ret = relayd_live_create(live_uri); |
178a0557 | 2770 | if (ret) { |
d3e2ba59 | 2771 | ERR("Starting live viewer threads"); |
178a0557 | 2772 | retval = -1; |
50138f51 | 2773 | goto exit_live; |
d3e2ba59 JD |
2774 | } |
2775 | ||
178a0557 MD |
2776 | /* |
2777 | * This is where we start awaiting program completion (e.g. through | |
2778 | * signal that asks threads to teardown). | |
2779 | */ | |
2780 | ||
2781 | ret = relayd_live_join(); | |
2782 | if (ret) { | |
2783 | retval = -1; | |
2784 | } | |
50138f51 | 2785 | exit_live: |
178a0557 | 2786 | |
b8aa1682 | 2787 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
2788 | if (ret) { |
2789 | errno = ret; | |
2790 | PERROR("pthread_join listener_thread"); | |
2791 | retval = -1; | |
b8aa1682 JD |
2792 | } |
2793 | ||
178a0557 | 2794 | exit_listener_thread: |
b8aa1682 | 2795 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
2796 | if (ret) { |
2797 | errno = ret; | |
2798 | PERROR("pthread_join worker_thread"); | |
2799 | retval = -1; | |
b8aa1682 JD |
2800 | } |
2801 | ||
178a0557 | 2802 | exit_worker_thread: |
b8aa1682 | 2803 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
2804 | if (ret) { |
2805 | errno = ret; | |
2806 | PERROR("pthread_join dispatcher_thread"); | |
2807 | retval = -1; | |
b8aa1682 | 2808 | } |
178a0557 | 2809 | exit_dispatcher_thread: |
42415026 | 2810 | |
65931c8b | 2811 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
2812 | if (ret) { |
2813 | errno = ret; | |
2814 | PERROR("pthread_join health_thread"); | |
2815 | retval = -1; | |
65931c8b | 2816 | } |
178a0557 | 2817 | exit_health_thread: |
65931c8b | 2818 | |
65931c8b | 2819 | utils_close_pipe(health_quit_pipe); |
178a0557 | 2820 | exit_health_quit_pipe: |
65931c8b | 2821 | |
178a0557 | 2822 | exit_init_data: |
55706a7d | 2823 | health_app_destroy(health_relayd); |
55706a7d | 2824 | exit_health_app_create: |
178a0557 | 2825 | exit_options: |
5a693d30 MD |
2826 | relayd_cleanup(); |
2827 | ||
2828 | /* Ensure all prior call_rcu are done. */ | |
2829 | rcu_barrier(); | |
d3e2ba59 | 2830 | |
178a0557 | 2831 | if (!retval) { |
b8aa1682 | 2832 | exit(EXIT_SUCCESS); |
178a0557 MD |
2833 | } else { |
2834 | exit(EXIT_FAILURE); | |
b8aa1682 | 2835 | } |
b8aa1682 | 2836 | } |