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" |
8921c6e4 | 74 | #include "tracefile-array.h" |
b8aa1682 JD |
75 | |
76 | /* command line options */ | |
0f907de1 | 77 | char *opt_output_path; |
b5218ffb | 78 | static int opt_daemon, opt_background; |
3fd27398 MD |
79 | |
80 | /* | |
81 | * We need to wait for listener and live listener threads, as well as | |
82 | * health check thread, before being ready to signal readiness. | |
83 | */ | |
84 | #define NR_LTTNG_RELAY_READY 3 | |
85 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
b95ad2e1 MD |
86 | |
87 | /* Size of receive buffer. */ | |
88 | #define RECV_DATA_BUFFER_SIZE 65536 | |
89 | ||
3fd27398 MD |
90 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ |
91 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
92 | ||
095a4ae5 MD |
93 | static struct lttng_uri *control_uri; |
94 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 95 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
96 | |
97 | const char *progname; | |
b8aa1682 | 98 | |
65931c8b | 99 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
100 | static int tracing_group_name_override; |
101 | ||
102 | const char * const config_section_name = "relayd"; | |
65931c8b | 103 | |
b8aa1682 JD |
104 | /* |
105 | * Quit pipe for all threads. This permits a single cancellation point | |
106 | * for all threads when receiving an event on the pipe. | |
107 | */ | |
0b242f62 | 108 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
109 | |
110 | /* | |
111 | * This pipe is used to inform the worker thread that a command is queued and | |
112 | * ready to be processed. | |
113 | */ | |
58eb9381 | 114 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 115 | |
26c9d55e | 116 | /* Shared between threads */ |
b8aa1682 JD |
117 | static int dispatch_thread_exit; |
118 | ||
119 | static pthread_t listener_thread; | |
120 | static pthread_t dispatcher_thread; | |
121 | static pthread_t worker_thread; | |
65931c8b | 122 | static pthread_t health_thread; |
b8aa1682 | 123 | |
5a693d30 MD |
124 | /* |
125 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
126 | * atomicity on 32-bit architectures. | |
127 | */ | |
128 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 129 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
130 | |
131 | /* | |
132 | * Relay command queue. | |
133 | * | |
134 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
135 | * queue. | |
136 | */ | |
58eb9381 | 137 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 JD |
138 | |
139 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
140 | static char *data_buffer; |
141 | static unsigned int data_buffer_size; | |
b8aa1682 | 142 | |
d3e2ba59 JD |
143 | /* Global relay stream hash table. */ |
144 | struct lttng_ht *relay_streams_ht; | |
145 | ||
92c6ca54 DG |
146 | /* Global relay viewer stream hash table. */ |
147 | struct lttng_ht *viewer_streams_ht; | |
148 | ||
5a693d30 MD |
149 | /* Global relay sessions hash table. */ |
150 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 151 | |
55706a7d | 152 | /* Relayd health monitoring */ |
eea7556c | 153 | struct health_app *health_relayd; |
55706a7d | 154 | |
cd60b05a JG |
155 | static struct option long_options[] = { |
156 | { "control-port", 1, 0, 'C', }, | |
157 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 158 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 159 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 160 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
161 | { "group", 1, 0, 'g', }, |
162 | { "help", 0, 0, 'h', }, | |
163 | { "output", 1, 0, 'o', }, | |
164 | { "verbose", 0, 0, 'v', }, | |
165 | { "config", 1, 0, 'f' }, | |
b929b5b4 | 166 | { "version", 0, 0, 'V' }, |
cd60b05a JG |
167 | { NULL, 0, 0, 0, }, |
168 | }; | |
169 | ||
b929b5b4 | 170 | static const char *config_ignore_options[] = { "help", "config", "version" }; |
cd60b05a | 171 | |
b8aa1682 JD |
172 | /* |
173 | * usage function on stderr | |
174 | */ | |
5a693d30 | 175 | static void usage(void) |
b8aa1682 JD |
176 | { |
177 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
178 | fprintf(stderr, " -h, --help Display this usage.\n"); |
179 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
b5218ffb | 180 | fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); |
994fa64f DG |
181 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); |
182 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
8d5c808e | 183 | fprintf(stderr, " -L, --live-port URL Live view port listening.\n"); |
994fa64f DG |
184 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); |
185 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 186 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 187 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
188 | } |
189 | ||
cd60b05a JG |
190 | /* |
191 | * Take an option from the getopt output and set it in the right variable to be | |
192 | * used later. | |
193 | * | |
194 | * Return 0 on success else a negative value. | |
195 | */ | |
5a693d30 | 196 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 197 | { |
cd60b05a JG |
198 | int ret; |
199 | ||
200 | switch (opt) { | |
201 | case 0: | |
202 | fprintf(stderr, "option %s", optname); | |
203 | if (arg) { | |
204 | fprintf(stderr, " with arg %s\n", arg); | |
205 | } | |
206 | break; | |
207 | case 'C': | |
e8fa9fb0 MD |
208 | if (lttng_is_setuid_setgid()) { |
209 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
210 | "-C, --control-port"); | |
211 | } else { | |
212 | ret = uri_parse(arg, &control_uri); | |
213 | if (ret < 0) { | |
214 | ERR("Invalid control URI specified"); | |
215 | goto end; | |
216 | } | |
217 | if (control_uri->port == 0) { | |
218 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
219 | } | |
cd60b05a JG |
220 | } |
221 | break; | |
222 | case 'D': | |
e8fa9fb0 MD |
223 | if (lttng_is_setuid_setgid()) { |
224 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
225 | "-D, -data-port"); | |
226 | } else { | |
227 | ret = uri_parse(arg, &data_uri); | |
228 | if (ret < 0) { | |
229 | ERR("Invalid data URI specified"); | |
230 | goto end; | |
231 | } | |
232 | if (data_uri->port == 0) { | |
233 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
234 | } | |
cd60b05a JG |
235 | } |
236 | break; | |
8d5c808e | 237 | case 'L': |
e8fa9fb0 MD |
238 | if (lttng_is_setuid_setgid()) { |
239 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
240 | "-L, -live-port"); | |
241 | } else { | |
242 | ret = uri_parse(arg, &live_uri); | |
243 | if (ret < 0) { | |
244 | ERR("Invalid live URI specified"); | |
245 | goto end; | |
246 | } | |
247 | if (live_uri->port == 0) { | |
248 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
249 | } | |
8d5c808e AM |
250 | } |
251 | break; | |
cd60b05a JG |
252 | case 'd': |
253 | opt_daemon = 1; | |
254 | break; | |
b5218ffb MD |
255 | case 'b': |
256 | opt_background = 1; | |
257 | break; | |
cd60b05a | 258 | case 'g': |
e8fa9fb0 MD |
259 | if (lttng_is_setuid_setgid()) { |
260 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
261 | "-g, --group"); | |
262 | } else { | |
263 | tracing_group_name = strdup(arg); | |
264 | if (tracing_group_name == NULL) { | |
265 | ret = -errno; | |
266 | PERROR("strdup"); | |
267 | goto end; | |
268 | } | |
269 | tracing_group_name_override = 1; | |
330a40bb | 270 | } |
cd60b05a JG |
271 | break; |
272 | case 'h': | |
273 | usage(); | |
274 | exit(EXIT_FAILURE); | |
b929b5b4 AW |
275 | case 'V': |
276 | fprintf(stdout, "%s\n", VERSION); | |
277 | exit(EXIT_SUCCESS); | |
cd60b05a | 278 | case 'o': |
e8fa9fb0 MD |
279 | if (lttng_is_setuid_setgid()) { |
280 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
281 | "-o, --output"); | |
282 | } else { | |
283 | ret = asprintf(&opt_output_path, "%s", arg); | |
284 | if (ret < 0) { | |
285 | ret = -errno; | |
286 | PERROR("asprintf opt_output_path"); | |
287 | goto end; | |
288 | } | |
cd60b05a JG |
289 | } |
290 | break; | |
291 | case 'v': | |
292 | /* Verbose level can increase using multiple -v */ | |
293 | if (arg) { | |
294 | lttng_opt_verbose = config_parse_value(arg); | |
295 | } else { | |
849e5b7b DG |
296 | /* Only 3 level of verbosity (-vvv). */ |
297 | if (lttng_opt_verbose < 3) { | |
298 | lttng_opt_verbose += 1; | |
299 | } | |
cd60b05a JG |
300 | } |
301 | break; | |
302 | default: | |
303 | /* Unknown option or other error. | |
304 | * Error is printed by getopt, just return */ | |
305 | ret = -1; | |
306 | goto end; | |
307 | } | |
308 | ||
309 | /* All good. */ | |
310 | ret = 0; | |
311 | ||
312 | end: | |
313 | return ret; | |
314 | } | |
315 | ||
316 | /* | |
317 | * config_entry_handler_cb used to handle options read from a config file. | |
318 | * See config_entry_handler_cb comment in common/config/config.h for the | |
319 | * return value conventions. | |
320 | */ | |
5a693d30 | 321 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
322 | { |
323 | int ret = 0, i; | |
324 | ||
325 | if (!entry || !entry->name || !entry->value) { | |
326 | ret = -EINVAL; | |
327 | goto end; | |
328 | } | |
329 | ||
330 | /* Check if the option is to be ignored */ | |
331 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
332 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
333 | goto end; | |
334 | } | |
335 | } | |
336 | ||
337 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
338 | /* Ignore if entry name is not fully matched. */ | |
339 | if (strcmp(entry->name, long_options[i].name)) { | |
340 | continue; | |
341 | } | |
342 | ||
343 | /* | |
5a693d30 MD |
344 | * If the option takes no argument on the command line, |
345 | * we have to check if the value is "true". We support | |
346 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
347 | */ |
348 | if (!long_options[i].has_arg) { | |
349 | ret = config_parse_value(entry->value); | |
350 | if (ret <= 0) { | |
351 | if (ret) { | |
352 | WARN("Invalid configuration value \"%s\" for option %s", | |
353 | entry->value, entry->name); | |
354 | } | |
355 | /* False, skip boolean config option. */ | |
356 | goto end; | |
357 | } | |
358 | } | |
359 | ||
360 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
361 | goto end; | |
362 | } | |
363 | ||
364 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
365 | entry->name); | |
366 | ||
367 | end: | |
368 | return ret; | |
369 | } | |
370 | ||
5a693d30 | 371 | static int set_options(int argc, char **argv) |
cd60b05a | 372 | { |
178a0557 | 373 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
374 | int orig_optopt = optopt, orig_optind = optind; |
375 | char *default_address, *optstring; | |
376 | const char *config_path = NULL; | |
377 | ||
378 | optstring = utils_generate_optstring(long_options, | |
379 | sizeof(long_options) / sizeof(struct option)); | |
380 | if (!optstring) { | |
178a0557 | 381 | retval = -ENOMEM; |
cd60b05a JG |
382 | goto exit; |
383 | } | |
384 | ||
385 | /* Check for the --config option */ | |
386 | ||
387 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
388 | &option_index)) != -1) { | |
389 | if (c == '?') { | |
178a0557 | 390 | retval = -EINVAL; |
cd60b05a JG |
391 | goto exit; |
392 | } else if (c != 'f') { | |
393 | continue; | |
394 | } | |
395 | ||
e8fa9fb0 MD |
396 | if (lttng_is_setuid_setgid()) { |
397 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
398 | "-f, --config"); | |
399 | } else { | |
400 | config_path = utils_expand_path(optarg); | |
401 | if (!config_path) { | |
402 | ERR("Failed to resolve path: %s", optarg); | |
403 | } | |
cd60b05a JG |
404 | } |
405 | } | |
406 | ||
407 | ret = config_get_section_entries(config_path, config_section_name, | |
408 | config_entry_handler, NULL); | |
409 | if (ret) { | |
410 | if (ret > 0) { | |
411 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 412 | } |
178a0557 | 413 | retval = -1; |
cd60b05a JG |
414 | goto exit; |
415 | } | |
b8aa1682 | 416 | |
cd60b05a JG |
417 | /* Reset getopt's global state */ |
418 | optopt = orig_optopt; | |
419 | optind = orig_optind; | |
b8aa1682 | 420 | while (1) { |
cd60b05a | 421 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
422 | if (c == -1) { |
423 | break; | |
424 | } | |
425 | ||
cd60b05a JG |
426 | ret = set_option(c, optarg, long_options[option_index].name); |
427 | if (ret < 0) { | |
178a0557 | 428 | retval = -1; |
b8aa1682 JD |
429 | goto exit; |
430 | } | |
431 | } | |
432 | ||
433 | /* assign default values */ | |
434 | if (control_uri == NULL) { | |
fa91dc52 MD |
435 | ret = asprintf(&default_address, |
436 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
437 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
438 | if (ret < 0) { |
439 | PERROR("asprintf default data address"); | |
178a0557 | 440 | retval = -1; |
b8aa1682 JD |
441 | goto exit; |
442 | } | |
443 | ||
444 | ret = uri_parse(default_address, &control_uri); | |
445 | free(default_address); | |
446 | if (ret < 0) { | |
447 | ERR("Invalid control URI specified"); | |
178a0557 | 448 | retval = -1; |
b8aa1682 JD |
449 | goto exit; |
450 | } | |
451 | } | |
452 | if (data_uri == NULL) { | |
fa91dc52 MD |
453 | ret = asprintf(&default_address, |
454 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
455 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
456 | if (ret < 0) { |
457 | PERROR("asprintf default data address"); | |
178a0557 | 458 | retval = -1; |
b8aa1682 JD |
459 | goto exit; |
460 | } | |
461 | ||
462 | ret = uri_parse(default_address, &data_uri); | |
463 | free(default_address); | |
464 | if (ret < 0) { | |
465 | ERR("Invalid data URI specified"); | |
178a0557 | 466 | retval = -1; |
b8aa1682 JD |
467 | goto exit; |
468 | } | |
469 | } | |
d3e2ba59 | 470 | if (live_uri == NULL) { |
fa91dc52 MD |
471 | ret = asprintf(&default_address, |
472 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
473 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
474 | if (ret < 0) { |
475 | PERROR("asprintf default viewer control address"); | |
178a0557 | 476 | retval = -1; |
d3e2ba59 JD |
477 | goto exit; |
478 | } | |
479 | ||
480 | ret = uri_parse(default_address, &live_uri); | |
481 | free(default_address); | |
482 | if (ret < 0) { | |
483 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 484 | retval = -1; |
d3e2ba59 JD |
485 | goto exit; |
486 | } | |
487 | } | |
b8aa1682 JD |
488 | |
489 | exit: | |
cd60b05a | 490 | free(optstring); |
178a0557 | 491 | return retval; |
b8aa1682 JD |
492 | } |
493 | ||
5a693d30 MD |
494 | static void print_global_objects(void) |
495 | { | |
496 | rcu_register_thread(); | |
497 | ||
498 | print_viewer_streams(); | |
499 | print_relay_streams(); | |
500 | print_sessions(); | |
501 | ||
502 | rcu_unregister_thread(); | |
503 | } | |
504 | ||
b8aa1682 JD |
505 | /* |
506 | * Cleanup the daemon | |
507 | */ | |
5a693d30 | 508 | static void relayd_cleanup(void) |
b8aa1682 | 509 | { |
5a693d30 MD |
510 | print_global_objects(); |
511 | ||
b8aa1682 JD |
512 | DBG("Cleaning up"); |
513 | ||
178a0557 MD |
514 | if (viewer_streams_ht) |
515 | lttng_ht_destroy(viewer_streams_ht); | |
516 | if (relay_streams_ht) | |
517 | lttng_ht_destroy(relay_streams_ht); | |
5a693d30 MD |
518 | if (sessions_ht) |
519 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 520 | |
095a4ae5 MD |
521 | /* free the dynamically allocated opt_output_path */ |
522 | free(opt_output_path); | |
523 | ||
a02de639 CB |
524 | /* Close thread quit pipes */ |
525 | utils_close_pipe(thread_quit_pipe); | |
526 | ||
710c1f73 DG |
527 | uri_free(control_uri); |
528 | uri_free(data_uri); | |
8d5c808e | 529 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
530 | |
531 | if (tracing_group_name_override) { | |
532 | free((void *) tracing_group_name); | |
533 | } | |
b8aa1682 JD |
534 | } |
535 | ||
536 | /* | |
537 | * Write to writable pipe used to notify a thread. | |
538 | */ | |
5a693d30 | 539 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 540 | { |
6cd525e8 | 541 | ssize_t ret; |
b8aa1682 | 542 | |
6cd525e8 MD |
543 | ret = lttng_write(wpipe, "!", 1); |
544 | if (ret < 1) { | |
b8aa1682 | 545 | PERROR("write poll pipe"); |
b4aacfdc | 546 | goto end; |
b8aa1682 | 547 | } |
b4aacfdc MD |
548 | ret = 0; |
549 | end: | |
b8aa1682 JD |
550 | return ret; |
551 | } | |
552 | ||
5a693d30 | 553 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 554 | { |
6cd525e8 | 555 | ssize_t ret; |
65931c8b | 556 | |
6cd525e8 MD |
557 | ret = lttng_write(pipe[1], "4", 1); |
558 | if (ret < 1) { | |
65931c8b | 559 | PERROR("write relay health quit"); |
b4aacfdc | 560 | goto end; |
65931c8b | 561 | } |
b4aacfdc MD |
562 | ret = 0; |
563 | end: | |
564 | return ret; | |
65931c8b MD |
565 | } |
566 | ||
b8aa1682 | 567 | /* |
b4aacfdc | 568 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 569 | */ |
b4aacfdc | 570 | int lttng_relay_stop_threads(void) |
b8aa1682 | 571 | { |
b4aacfdc | 572 | int retval = 0; |
b8aa1682 JD |
573 | |
574 | /* Stopping all threads */ | |
575 | DBG("Terminating all threads"); | |
b4aacfdc | 576 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 577 | ERR("write error on thread quit pipe"); |
b4aacfdc | 578 | retval = -1; |
b8aa1682 JD |
579 | } |
580 | ||
b4aacfdc MD |
581 | if (notify_health_quit_pipe(health_quit_pipe)) { |
582 | ERR("write error on health quit pipe"); | |
583 | } | |
65931c8b | 584 | |
b8aa1682 | 585 | /* Dispatch thread */ |
26c9d55e | 586 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 587 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 588 | |
b4aacfdc | 589 | if (relayd_live_stop()) { |
178a0557 | 590 | ERR("Error stopping live threads"); |
b4aacfdc | 591 | retval = -1; |
178a0557 | 592 | } |
b4aacfdc | 593 | return retval; |
b8aa1682 JD |
594 | } |
595 | ||
596 | /* | |
597 | * Signal handler for the daemon | |
598 | * | |
599 | * Simply stop all worker threads, leaving main() return gracefully after | |
600 | * joining all threads and calling cleanup(). | |
601 | */ | |
5a693d30 | 602 | static void sighandler(int sig) |
b8aa1682 JD |
603 | { |
604 | switch (sig) { | |
b8aa1682 JD |
605 | case SIGINT: |
606 | DBG("SIGINT caught"); | |
b4aacfdc MD |
607 | if (lttng_relay_stop_threads()) { |
608 | ERR("Error stopping threads"); | |
609 | } | |
b8aa1682 JD |
610 | break; |
611 | case SIGTERM: | |
612 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
613 | if (lttng_relay_stop_threads()) { |
614 | ERR("Error stopping threads"); | |
615 | } | |
b8aa1682 | 616 | break; |
3fd27398 MD |
617 | case SIGUSR1: |
618 | CMM_STORE_SHARED(recv_child_signal, 1); | |
619 | break; | |
b8aa1682 JD |
620 | default: |
621 | break; | |
622 | } | |
623 | } | |
624 | ||
625 | /* | |
626 | * Setup signal handler for : | |
627 | * SIGINT, SIGTERM, SIGPIPE | |
628 | */ | |
5a693d30 | 629 | static int set_signal_handler(void) |
b8aa1682 JD |
630 | { |
631 | int ret = 0; | |
632 | struct sigaction sa; | |
633 | sigset_t sigset; | |
634 | ||
635 | if ((ret = sigemptyset(&sigset)) < 0) { | |
636 | PERROR("sigemptyset"); | |
637 | return ret; | |
638 | } | |
639 | ||
b8aa1682 JD |
640 | sa.sa_mask = sigset; |
641 | sa.sa_flags = 0; | |
35c973f1 MD |
642 | |
643 | sa.sa_handler = sighandler; | |
b8aa1682 JD |
644 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { |
645 | PERROR("sigaction"); | |
646 | return ret; | |
647 | } | |
648 | ||
649 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
650 | PERROR("sigaction"); | |
651 | return ret; | |
652 | } | |
653 | ||
35c973f1 | 654 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
b8aa1682 JD |
655 | PERROR("sigaction"); |
656 | return ret; | |
657 | } | |
658 | ||
35c973f1 MD |
659 | sa.sa_handler = SIG_IGN; |
660 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
3fd27398 MD |
661 | PERROR("sigaction"); |
662 | return ret; | |
663 | } | |
664 | ||
665 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
666 | |
667 | return ret; | |
668 | } | |
669 | ||
3fd27398 MD |
670 | void lttng_relay_notify_ready(void) |
671 | { | |
672 | /* Notify the parent of the fork() process that we are ready. */ | |
673 | if (opt_daemon || opt_background) { | |
674 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
675 | kill(child_ppid, SIGUSR1); | |
676 | } | |
677 | } | |
678 | } | |
679 | ||
b8aa1682 JD |
680 | /* |
681 | * Init thread quit pipe. | |
682 | * | |
683 | * Return -1 on error or 0 if all pipes are created. | |
684 | */ | |
5a693d30 | 685 | static int init_thread_quit_pipe(void) |
b8aa1682 | 686 | { |
a02de639 | 687 | int ret; |
b8aa1682 | 688 | |
a02de639 | 689 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 690 | |
b8aa1682 JD |
691 | return ret; |
692 | } | |
693 | ||
694 | /* | |
695 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
696 | */ | |
5a693d30 | 697 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
698 | { |
699 | int ret; | |
700 | ||
701 | if (events == NULL || size == 0) { | |
702 | ret = -1; | |
703 | goto error; | |
704 | } | |
705 | ||
706 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
707 | if (ret < 0) { | |
708 | goto error; | |
709 | } | |
710 | ||
711 | /* Add quit pipe */ | |
c7759e6a | 712 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
713 | if (ret < 0) { |
714 | goto error; | |
715 | } | |
716 | ||
717 | return 0; | |
718 | ||
719 | error: | |
720 | return ret; | |
721 | } | |
722 | ||
723 | /* | |
724 | * Check if the thread quit pipe was triggered. | |
725 | * | |
726 | * Return 1 if it was triggered else 0; | |
727 | */ | |
5a693d30 | 728 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
729 | { |
730 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
731 | return 1; | |
732 | } | |
733 | ||
734 | return 0; | |
735 | } | |
736 | ||
737 | /* | |
738 | * Create and init socket from uri. | |
739 | */ | |
5a693d30 | 740 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
741 | { |
742 | int ret; | |
743 | struct lttcomm_sock *sock = NULL; | |
744 | ||
745 | sock = lttcomm_alloc_sock_from_uri(uri); | |
746 | if (sock == NULL) { | |
747 | ERR("Allocating socket"); | |
748 | goto error; | |
749 | } | |
750 | ||
751 | ret = lttcomm_create_sock(sock); | |
752 | if (ret < 0) { | |
753 | goto error; | |
754 | } | |
755 | DBG("Listening on sock %d", sock->fd); | |
756 | ||
757 | ret = sock->ops->bind(sock); | |
758 | if (ret < 0) { | |
759 | goto error; | |
760 | } | |
761 | ||
762 | ret = sock->ops->listen(sock, -1); | |
763 | if (ret < 0) { | |
764 | goto error; | |
765 | ||
766 | } | |
767 | ||
768 | return sock; | |
769 | ||
770 | error: | |
771 | if (sock) { | |
772 | lttcomm_destroy_sock(sock); | |
773 | } | |
774 | return NULL; | |
775 | } | |
776 | ||
777 | /* | |
778 | * This thread manages the listening for new connections on the network | |
779 | */ | |
5a693d30 | 780 | static void *relay_thread_listener(void *data) |
b8aa1682 | 781 | { |
095a4ae5 | 782 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
783 | uint32_t revents, nb_fd; |
784 | struct lttng_poll_event events; | |
785 | struct lttcomm_sock *control_sock, *data_sock; | |
786 | ||
b8aa1682 JD |
787 | DBG("[thread] Relay listener started"); |
788 | ||
55706a7d MD |
789 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
790 | ||
f385ae0a MD |
791 | health_code_update(); |
792 | ||
5a693d30 | 793 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 794 | if (!control_sock) { |
095a4ae5 | 795 | goto error_sock_control; |
b8aa1682 JD |
796 | } |
797 | ||
5a693d30 | 798 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 799 | if (!data_sock) { |
095a4ae5 | 800 | goto error_sock_relay; |
b8aa1682 JD |
801 | } |
802 | ||
803 | /* | |
5a693d30 MD |
804 | * Pass 3 as size here for the thread quit pipe, control and |
805 | * data socket. | |
b8aa1682 JD |
806 | */ |
807 | ret = create_thread_poll_set(&events, 3); | |
808 | if (ret < 0) { | |
809 | goto error_create_poll; | |
810 | } | |
811 | ||
812 | /* Add the control socket */ | |
813 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
814 | if (ret < 0) { | |
815 | goto error_poll_add; | |
816 | } | |
817 | ||
818 | /* Add the data socket */ | |
819 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
820 | if (ret < 0) { | |
821 | goto error_poll_add; | |
822 | } | |
823 | ||
3fd27398 MD |
824 | lttng_relay_notify_ready(); |
825 | ||
9b5e0863 MD |
826 | if (testpoint(relayd_thread_listener)) { |
827 | goto error_testpoint; | |
828 | } | |
829 | ||
b8aa1682 | 830 | while (1) { |
f385ae0a MD |
831 | health_code_update(); |
832 | ||
b8aa1682 JD |
833 | DBG("Listener accepting connections"); |
834 | ||
b8aa1682 | 835 | restart: |
f385ae0a | 836 | health_poll_entry(); |
b8aa1682 | 837 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 838 | health_poll_exit(); |
b8aa1682 JD |
839 | if (ret < 0) { |
840 | /* | |
841 | * Restart interrupted system call. | |
842 | */ | |
843 | if (errno == EINTR) { | |
844 | goto restart; | |
845 | } | |
846 | goto error; | |
847 | } | |
848 | ||
0d9c5d77 DG |
849 | nb_fd = ret; |
850 | ||
b8aa1682 JD |
851 | DBG("Relay new connection received"); |
852 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
853 | health_code_update(); |
854 | ||
b8aa1682 JD |
855 | /* Fetch once the poll data */ |
856 | revents = LTTNG_POLL_GETEV(&events, i); | |
857 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
858 | ||
fd20dac9 | 859 | if (!revents) { |
5a693d30 MD |
860 | /* |
861 | * No activity for this FD (poll | |
862 | * implementation). | |
863 | */ | |
fd20dac9 MD |
864 | continue; |
865 | } | |
866 | ||
b8aa1682 JD |
867 | /* Thread quit pipe has been closed. Killing thread. */ |
868 | ret = check_thread_quit_pipe(pollfd, revents); | |
869 | if (ret) { | |
095a4ae5 MD |
870 | err = 0; |
871 | goto exit; | |
b8aa1682 JD |
872 | } |
873 | ||
92d6debb | 874 | if (revents & LPOLLIN) { |
4b7f17b2 | 875 | /* |
5a693d30 MD |
876 | * A new connection is requested, therefore a |
877 | * sessiond/consumerd connection is allocated in | |
878 | * this thread, enqueued to a global queue and | |
879 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 880 | */ |
58eb9381 DG |
881 | int val = 1; |
882 | struct relay_connection *new_conn; | |
4b7f17b2 | 883 | struct lttcomm_sock *newsock; |
5a693d30 | 884 | enum connection_type type; |
b8aa1682 JD |
885 | |
886 | if (pollfd == data_sock->fd) { | |
5a693d30 | 887 | type = RELAY_DATA; |
b8aa1682 | 888 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
889 | DBG("Relay data connection accepted, socket %d", |
890 | newsock->fd); | |
4b7f17b2 MD |
891 | } else { |
892 | assert(pollfd == control_sock->fd); | |
5a693d30 | 893 | type = RELAY_CONTROL; |
b8aa1682 | 894 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
895 | DBG("Relay control connection accepted, socket %d", |
896 | newsock->fd); | |
b8aa1682 | 897 | } |
58eb9381 DG |
898 | if (!newsock) { |
899 | PERROR("accepting sock"); | |
58eb9381 DG |
900 | goto error; |
901 | } | |
902 | ||
903 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
904 | sizeof(val)); | |
b8aa1682 JD |
905 | if (ret < 0) { |
906 | PERROR("setsockopt inet"); | |
4b7f17b2 | 907 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
908 | goto error; |
909 | } | |
5a693d30 MD |
910 | new_conn = connection_create(newsock, type); |
911 | if (!new_conn) { | |
912 | lttcomm_destroy_sock(newsock); | |
913 | goto error; | |
914 | } | |
58eb9381 DG |
915 | |
916 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
917 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
918 | &new_conn->qnode); | |
b8aa1682 JD |
919 | |
920 | /* | |
5a693d30 MD |
921 | * Wake the dispatch queue futex. |
922 | * Implicit memory barrier with the | |
923 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 924 | */ |
58eb9381 | 925 | futex_nto1_wake(&relay_conn_queue.futex); |
92d6debb MD |
926 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
927 | ERR("socket poll error"); | |
928 | goto error; | |
929 | } else { | |
930 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
931 | goto error; | |
b8aa1682 JD |
932 | } |
933 | } | |
934 | } | |
935 | ||
095a4ae5 | 936 | exit: |
b8aa1682 JD |
937 | error: |
938 | error_poll_add: | |
9b5e0863 | 939 | error_testpoint: |
b8aa1682 JD |
940 | lttng_poll_clean(&events); |
941 | error_create_poll: | |
095a4ae5 MD |
942 | if (data_sock->fd >= 0) { |
943 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
944 | if (ret) { |
945 | PERROR("close"); | |
946 | } | |
b8aa1682 | 947 | } |
095a4ae5 MD |
948 | lttcomm_destroy_sock(data_sock); |
949 | error_sock_relay: | |
950 | if (control_sock->fd >= 0) { | |
951 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
952 | if (ret) { |
953 | PERROR("close"); | |
954 | } | |
b8aa1682 | 955 | } |
095a4ae5 MD |
956 | lttcomm_destroy_sock(control_sock); |
957 | error_sock_control: | |
958 | if (err) { | |
f385ae0a MD |
959 | health_error(); |
960 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 961 | } |
55706a7d | 962 | health_unregister(health_relayd); |
b8aa1682 | 963 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 964 | lttng_relay_stop_threads(); |
b8aa1682 JD |
965 | return NULL; |
966 | } | |
967 | ||
968 | /* | |
969 | * This thread manages the dispatching of the requests to worker threads | |
970 | */ | |
5a693d30 | 971 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 972 | { |
6cd525e8 MD |
973 | int err = -1; |
974 | ssize_t ret; | |
8bdee6e2 | 975 | struct cds_wfcq_node *node; |
58eb9381 | 976 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
977 | |
978 | DBG("[thread] Relay dispatcher started"); | |
979 | ||
55706a7d MD |
980 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
981 | ||
9b5e0863 MD |
982 | if (testpoint(relayd_thread_dispatcher)) { |
983 | goto error_testpoint; | |
984 | } | |
985 | ||
f385ae0a MD |
986 | health_code_update(); |
987 | ||
26c9d55e | 988 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
989 | health_code_update(); |
990 | ||
b8aa1682 | 991 | /* Atomically prepare the queue futex */ |
58eb9381 | 992 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
993 | |
994 | do { | |
f385ae0a MD |
995 | health_code_update(); |
996 | ||
b8aa1682 | 997 | /* Dequeue commands */ |
8bdee6e2 SM |
998 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
999 | &relay_conn_queue.tail); | |
b8aa1682 JD |
1000 | if (node == NULL) { |
1001 | DBG("Woken up but nothing in the relay command queue"); | |
1002 | /* Continue thread execution */ | |
1003 | break; | |
1004 | } | |
58eb9381 | 1005 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 1006 | |
58eb9381 | 1007 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1008 | |
1009 | /* | |
5a693d30 MD |
1010 | * Inform worker thread of the new request. This |
1011 | * call is blocking so we can be assured that | |
1012 | * the data will be read at some point in time | |
1013 | * or wait to the end of the world :) | |
b8aa1682 | 1014 | */ |
58eb9381 DG |
1015 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1016 | if (ret < 0) { | |
1017 | PERROR("write connection pipe"); | |
5a693d30 | 1018 | connection_put(new_conn); |
b8aa1682 JD |
1019 | goto error; |
1020 | } | |
1021 | } while (node != NULL); | |
1022 | ||
1023 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1024 | health_poll_entry(); |
58eb9381 | 1025 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1026 | health_poll_exit(); |
b8aa1682 JD |
1027 | } |
1028 | ||
f385ae0a MD |
1029 | /* Normal exit, no error */ |
1030 | err = 0; | |
1031 | ||
b8aa1682 | 1032 | error: |
9b5e0863 | 1033 | error_testpoint: |
f385ae0a MD |
1034 | if (err) { |
1035 | health_error(); | |
1036 | ERR("Health error occurred in %s", __func__); | |
1037 | } | |
55706a7d | 1038 | health_unregister(health_relayd); |
b8aa1682 | 1039 | DBG("Dispatch thread dying"); |
b4aacfdc | 1040 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1041 | return NULL; |
1042 | } | |
1043 | ||
b8aa1682 | 1044 | /* |
5a693d30 | 1045 | * Set index data from the control port to a given index object. |
b8aa1682 | 1046 | */ |
5a693d30 | 1047 | static int set_index_control_data(struct relay_index *index, |
1c20f0e2 JD |
1048 | struct lttcomm_relayd_index *data) |
1049 | { | |
5a693d30 | 1050 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1051 | |
1052 | /* | |
5a693d30 MD |
1053 | * The index on disk is encoded in big endian, so we don't need |
1054 | * to convert the data received on the network. The data_offset | |
1055 | * value is NEVER modified here and is updated by the data | |
1056 | * thread. | |
1c20f0e2 | 1057 | */ |
5a693d30 MD |
1058 | index_data.packet_size = data->packet_size; |
1059 | index_data.content_size = data->content_size; | |
1060 | index_data.timestamp_begin = data->timestamp_begin; | |
1061 | index_data.timestamp_end = data->timestamp_end; | |
1062 | index_data.events_discarded = data->events_discarded; | |
1063 | index_data.stream_id = data->stream_id; | |
1064 | return relay_index_set_data(index, &index_data); | |
1c20f0e2 JD |
1065 | } |
1066 | ||
c5b6f4f0 DG |
1067 | /* |
1068 | * Handle the RELAYD_CREATE_SESSION command. | |
1069 | * | |
1070 | * On success, send back the session id or else return a negative value. | |
1071 | */ | |
5a693d30 | 1072 | static int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1073 | struct relay_connection *conn) |
c5b6f4f0 DG |
1074 | { |
1075 | int ret = 0, send_ret; | |
1076 | struct relay_session *session; | |
1077 | struct lttcomm_relayd_status_session reply; | |
5a693d30 MD |
1078 | char session_name[NAME_MAX]; |
1079 | char hostname[HOST_NAME_MAX]; | |
1080 | uint32_t live_timer = 0; | |
1081 | bool snapshot = false; | |
c5b6f4f0 | 1082 | |
5a693d30 MD |
1083 | memset(session_name, 0, NAME_MAX); |
1084 | memset(hostname, 0, HOST_NAME_MAX); | |
c5b6f4f0 DG |
1085 | |
1086 | memset(&reply, 0, sizeof(reply)); | |
1087 | ||
58eb9381 | 1088 | switch (conn->minor) { |
2a174661 DG |
1089 | case 1: |
1090 | case 2: | |
1091 | case 3: | |
1092 | break; | |
1093 | case 4: /* LTTng sessiond 2.4 */ | |
1094 | default: | |
5a693d30 MD |
1095 | ret = cmd_create_session_2_4(conn, session_name, |
1096 | hostname, &live_timer, &snapshot); | |
1097 | } | |
1098 | if (ret < 0) { | |
1099 | goto send_reply; | |
d3e2ba59 JD |
1100 | } |
1101 | ||
5a693d30 MD |
1102 | session = session_create(session_name, hostname, live_timer, |
1103 | snapshot, conn->major, conn->minor); | |
1104 | if (!session) { | |
1105 | ret = -1; | |
1106 | goto send_reply; | |
1107 | } | |
1108 | assert(!conn->session); | |
1109 | conn->session = session; | |
c5b6f4f0 DG |
1110 | DBG("Created session %" PRIu64, session->id); |
1111 | ||
5a693d30 MD |
1112 | reply.session_id = htobe64(session->id); |
1113 | ||
1114 | send_reply: | |
c5b6f4f0 DG |
1115 | if (ret < 0) { |
1116 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1117 | } else { | |
1118 | reply.ret_code = htobe32(LTTNG_OK); | |
1119 | } | |
1120 | ||
58eb9381 | 1121 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1122 | if (send_ret < 0) { |
1123 | ERR("Relayd sending session id"); | |
4169f5ad | 1124 | ret = send_ret; |
c5b6f4f0 DG |
1125 | } |
1126 | ||
1127 | return ret; | |
1128 | } | |
1129 | ||
a4baae1b JD |
1130 | /* |
1131 | * When we have received all the streams and the metadata for a channel, | |
1132 | * we make them visible to the viewer threads. | |
1133 | */ | |
5a693d30 | 1134 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1135 | { |
5a693d30 MD |
1136 | struct relay_stream *stream; |
1137 | struct relay_session *session = conn->session; | |
a4baae1b | 1138 | |
5a693d30 MD |
1139 | /* |
1140 | * We publish all streams belonging to a session atomically wrt | |
1141 | * session lock. | |
1142 | */ | |
1143 | pthread_mutex_lock(&session->lock); | |
1144 | rcu_read_lock(); | |
1145 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1146 | recv_node) { | |
1147 | stream_publish(stream); | |
a4baae1b | 1148 | } |
5a693d30 | 1149 | rcu_read_unlock(); |
a4baae1b | 1150 | |
5a693d30 MD |
1151 | /* |
1152 | * Inform the viewer that there are new streams in the session. | |
1153 | */ | |
1154 | if (session->viewer_attached) { | |
1155 | uatomic_set(&session->new_streams, 1); | |
1156 | } | |
1157 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1158 | } |
1159 | ||
b8aa1682 JD |
1160 | /* |
1161 | * relay_add_stream: allocate a new stream for a session | |
1162 | */ | |
5a693d30 | 1163 | static int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1164 | struct relay_connection *conn) |
b8aa1682 | 1165 | { |
5a693d30 MD |
1166 | int ret; |
1167 | ssize_t send_ret; | |
58eb9381 | 1168 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1169 | struct relay_stream *stream = NULL; |
1170 | struct lttcomm_relayd_status_stream reply; | |
80ec61a6 | 1171 | struct ctf_trace *trace = NULL; |
5a693d30 MD |
1172 | uint64_t stream_handle = -1ULL; |
1173 | char *path_name = NULL, *channel_name = NULL; | |
1174 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1175 | |
58eb9381 | 1176 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1177 | ERR("Trying to add a stream before version check"); |
1178 | ret = -1; | |
1179 | goto end_no_session; | |
1180 | } | |
1181 | ||
5a693d30 MD |
1182 | switch (session->minor) { |
1183 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
1184 | ret = cmd_recv_stream_2_1(conn, &path_name, | |
1185 | &channel_name); | |
0f907de1 | 1186 | break; |
5a693d30 | 1187 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1188 | default: |
5a693d30 MD |
1189 | ret = cmd_recv_stream_2_2(conn, &path_name, |
1190 | &channel_name, &tracefile_size, &tracefile_count); | |
0f907de1 JD |
1191 | break; |
1192 | } | |
1193 | if (ret < 0) { | |
5a693d30 | 1194 | goto send_reply; |
b8aa1682 JD |
1195 | } |
1196 | ||
5a693d30 | 1197 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1198 | if (!trace) { |
5a693d30 | 1199 | goto send_reply; |
2a174661 | 1200 | } |
5a693d30 | 1201 | /* This stream here has one reference on the trace. */ |
2a174661 | 1202 | |
5a693d30 MD |
1203 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1204 | stream_handle = ++last_relay_stream_id; | |
1205 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1206 | |
5a693d30 MD |
1207 | /* We pass ownership of path_name and channel_name. */ |
1208 | stream = stream_create(trace, stream_handle, path_name, | |
1209 | channel_name, tracefile_size, tracefile_count); | |
1210 | path_name = NULL; | |
1211 | channel_name = NULL; | |
a4baae1b | 1212 | |
2a174661 | 1213 | /* |
5a693d30 MD |
1214 | * Streams are the owners of their trace. Reference to trace is |
1215 | * kept within stream_create(). | |
2a174661 | 1216 | */ |
5a693d30 | 1217 | ctf_trace_put(trace); |
d3e2ba59 | 1218 | |
5a693d30 | 1219 | send_reply: |
53efb85a | 1220 | memset(&reply, 0, sizeof(reply)); |
5a693d30 MD |
1221 | reply.handle = htobe64(stream_handle); |
1222 | if (!stream) { | |
f73fabfd | 1223 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1224 | } else { |
f73fabfd | 1225 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1226 | } |
5af40280 | 1227 | |
58eb9381 | 1228 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1229 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1230 | if (send_ret < 0) { | |
1231 | ERR("Relay sending stream id"); | |
5a693d30 | 1232 | ret = (int) send_ret; |
b8aa1682 JD |
1233 | } |
1234 | ||
1235 | end_no_session: | |
5a693d30 MD |
1236 | free(path_name); |
1237 | free(channel_name); | |
0f907de1 | 1238 | return ret; |
b8aa1682 JD |
1239 | } |
1240 | ||
173af62f DG |
1241 | /* |
1242 | * relay_close_stream: close a specific stream | |
1243 | */ | |
5a693d30 | 1244 | static int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1245 | struct relay_connection *conn) |
173af62f | 1246 | { |
94d49140 | 1247 | int ret, send_ret; |
58eb9381 | 1248 | struct relay_session *session = conn->session; |
173af62f DG |
1249 | struct lttcomm_relayd_close_stream stream_info; |
1250 | struct lttcomm_relayd_generic_reply reply; | |
1251 | struct relay_stream *stream; | |
173af62f DG |
1252 | |
1253 | DBG("Close stream received"); | |
1254 | ||
58eb9381 | 1255 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1256 | ERR("Trying to close a stream before version check"); |
1257 | ret = -1; | |
1258 | goto end_no_session; | |
1259 | } | |
1260 | ||
58eb9381 | 1261 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1262 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1263 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1264 | if (ret == 0) { |
1265 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1266 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1267 | } else { |
1268 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1269 | } | |
173af62f DG |
1270 | ret = -1; |
1271 | goto end_no_session; | |
1272 | } | |
1273 | ||
5a693d30 | 1274 | stream = stream_get_by_id(be64toh(stream_info.stream_id)); |
173af62f DG |
1275 | if (!stream) { |
1276 | ret = -1; | |
5a693d30 | 1277 | goto end; |
173af62f | 1278 | } |
a12886a5 MD |
1279 | |
1280 | /* | |
1281 | * Set last_net_seq_num before the close flag. Required by data | |
1282 | * pending check. | |
1283 | */ | |
5a693d30 | 1284 | pthread_mutex_lock(&stream->lock); |
8e2583a4 | 1285 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
a12886a5 MD |
1286 | pthread_mutex_unlock(&stream->lock); |
1287 | ||
a08107cc JG |
1288 | /* |
1289 | * This is one of the conditions which may trigger a stream close | |
1290 | * with the others being: | |
1291 | * 1) A close command is received for a stream | |
1292 | * 2) The control connection owning the stream is closed | |
1293 | * 3) We have received all of the stream's data _after_ a close | |
1294 | * request. | |
1295 | */ | |
1296 | try_stream_close(stream); | |
5a693d30 MD |
1297 | if (stream->is_metadata) { |
1298 | struct relay_viewer_stream *vstream; | |
173af62f | 1299 | |
5a693d30 MD |
1300 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1301 | if (vstream) { | |
1302 | if (vstream->metadata_sent == stream->metadata_received) { | |
1303 | /* | |
1304 | * Since all the metadata has been sent to the | |
1305 | * viewer and that we have a request to close | |
1306 | * its stream, we can safely teardown the | |
1307 | * corresponding metadata viewer stream. | |
1308 | */ | |
1309 | viewer_stream_put(vstream); | |
1310 | } | |
1311 | /* Put local reference. */ | |
1312 | viewer_stream_put(vstream); | |
1313 | } | |
1314 | } | |
5a693d30 | 1315 | stream_put(stream); |
173af62f | 1316 | |
5a693d30 | 1317 | end: |
53efb85a | 1318 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1319 | if (ret < 0) { |
f73fabfd | 1320 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1321 | } else { |
f73fabfd | 1322 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1323 | } |
58eb9381 | 1324 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1325 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1326 | if (send_ret < 0) { | |
1327 | ERR("Relay sending stream id"); | |
4169f5ad | 1328 | ret = send_ret; |
173af62f DG |
1329 | } |
1330 | ||
1331 | end_no_session: | |
1332 | return ret; | |
1333 | } | |
1334 | ||
b8aa1682 JD |
1335 | /* |
1336 | * relay_unknown_command: send -1 if received unknown command | |
1337 | */ | |
5a693d30 | 1338 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1339 | { |
1340 | struct lttcomm_relayd_generic_reply reply; | |
1341 | int ret; | |
1342 | ||
53efb85a | 1343 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1344 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
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 unknown command"); | |
1349 | } | |
1350 | } | |
1351 | ||
1352 | /* | |
1353 | * relay_start: send an acknowledgment to the client to tell if we are | |
1354 | * ready to receive data. We are ready if a session is established. | |
1355 | */ | |
5a693d30 | 1356 | static int relay_start(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1357 | struct relay_connection *conn) |
b8aa1682 | 1358 | { |
f73fabfd | 1359 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1360 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1361 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1362 | |
1363 | if (!session) { | |
1364 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1365 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1366 | } |
1367 | ||
53efb85a | 1368 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1369 | reply.ret_code = ret; |
58eb9381 | 1370 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1371 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1372 | if (ret < 0) { | |
1373 | ERR("Relay sending start ack"); | |
1374 | } | |
1375 | ||
1376 | return ret; | |
1377 | } | |
1378 | ||
1d4dfdef DG |
1379 | /* |
1380 | * Append padding to the file pointed by the file descriptor fd. | |
1381 | */ | |
1382 | static int write_padding_to_file(int fd, uint32_t size) | |
1383 | { | |
6cd525e8 | 1384 | ssize_t ret = 0; |
1d4dfdef DG |
1385 | char *zeros; |
1386 | ||
1387 | if (size == 0) { | |
1388 | goto end; | |
1389 | } | |
1390 | ||
1391 | zeros = zmalloc(size); | |
1392 | if (zeros == NULL) { | |
1393 | PERROR("zmalloc zeros for padding"); | |
1394 | ret = -1; | |
1395 | goto end; | |
1396 | } | |
1397 | ||
6cd525e8 MD |
1398 | ret = lttng_write(fd, zeros, size); |
1399 | if (ret < size) { | |
1d4dfdef DG |
1400 | PERROR("write padding to file"); |
1401 | } | |
1402 | ||
e986c7a1 DG |
1403 | free(zeros); |
1404 | ||
1d4dfdef DG |
1405 | end: |
1406 | return ret; | |
1407 | } | |
1408 | ||
b8aa1682 | 1409 | /* |
5a693d30 | 1410 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1411 | */ |
5a693d30 | 1412 | static int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1413 | struct relay_connection *conn) |
b8aa1682 | 1414 | { |
7fcdafc9 | 1415 | int ret = 0; |
6cd525e8 | 1416 | ssize_t size_ret; |
58eb9381 | 1417 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1418 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1419 | struct relay_stream *metadata_stream; | |
1420 | uint64_t data_size, payload_size; | |
1421 | ||
1422 | if (!session) { | |
1423 | ERR("Metadata sent before version check"); | |
1424 | ret = -1; | |
1425 | goto end; | |
1426 | } | |
1427 | ||
f6416125 MD |
1428 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1429 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1430 | ERR("Incorrect data size"); | |
1431 | ret = -1; | |
1432 | goto end; | |
1433 | } | |
1434 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1435 | ||
b8aa1682 | 1436 | if (data_buffer_size < data_size) { |
d7b3776f | 1437 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1438 | char *tmp_data_ptr; |
1439 | ||
1440 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1441 | if (!tmp_data_ptr) { | |
b8aa1682 | 1442 | ERR("Allocating data buffer"); |
c617c0c6 | 1443 | free(data_buffer); |
b8aa1682 JD |
1444 | ret = -1; |
1445 | goto end; | |
1446 | } | |
c617c0c6 | 1447 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1448 | data_buffer_size = data_size; |
1449 | } | |
1450 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1451 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
7fcdafc9 JG |
1452 | size_ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
1453 | if (size_ret < 0 || size_ret != data_size) { | |
1454 | if (size_ret == 0) { | |
a6cd2b97 | 1455 | /* Orderly shutdown. Not necessary to print an error. */ |
58eb9381 | 1456 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1457 | } else { |
1458 | ERR("Relay didn't receive the whole metadata"); | |
1459 | } | |
b8aa1682 | 1460 | ret = -1; |
b8aa1682 JD |
1461 | goto end; |
1462 | } | |
1463 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 | 1464 | |
5a693d30 | 1465 | metadata_stream = stream_get_by_id(be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1466 | if (!metadata_stream) { |
1467 | ret = -1; | |
5a693d30 | 1468 | goto end; |
b8aa1682 JD |
1469 | } |
1470 | ||
5a693d30 MD |
1471 | pthread_mutex_lock(&metadata_stream->lock); |
1472 | ||
1473 | size_ret = lttng_write(metadata_stream->stream_fd->fd, metadata_struct->payload, | |
6cd525e8 MD |
1474 | payload_size); |
1475 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1476 | ERR("Relay error writing metadata on file"); |
1477 | ret = -1; | |
5a693d30 | 1478 | goto end_put; |
b8aa1682 | 1479 | } |
1d4dfdef | 1480 | |
7fcdafc9 | 1481 | size_ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
1d4dfdef | 1482 | be32toh(metadata_struct->padding_size)); |
7fcdafc9 | 1483 | if (size_ret < 0) { |
5a693d30 | 1484 | goto end_put; |
1d4dfdef | 1485 | } |
2a174661 | 1486 | |
5a693d30 | 1487 | metadata_stream->metadata_received += |
d3e2ba59 | 1488 | payload_size + be32toh(metadata_struct->padding_size); |
5a693d30 MD |
1489 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1490 | metadata_stream->metadata_received); | |
1d4dfdef | 1491 | |
5a693d30 MD |
1492 | end_put: |
1493 | pthread_mutex_unlock(&metadata_stream->lock); | |
1494 | stream_put(metadata_stream); | |
b8aa1682 JD |
1495 | end: |
1496 | return ret; | |
1497 | } | |
1498 | ||
1499 | /* | |
1500 | * relay_send_version: send relayd version number | |
1501 | */ | |
5a693d30 | 1502 | static int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1503 | struct relay_connection *conn) |
b8aa1682 | 1504 | { |
7f51dcba | 1505 | int ret; |
092b6259 | 1506 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1507 | |
58eb9381 | 1508 | conn->version_check_done = 1; |
b8aa1682 | 1509 | |
092b6259 | 1510 | /* Get version from the other side. */ |
58eb9381 | 1511 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1512 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1513 | if (ret == 0) { |
1514 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1515 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1516 | } else { |
1517 | ERR("Relay failed to receive the version values."); | |
1518 | } | |
092b6259 | 1519 | ret = -1; |
092b6259 DG |
1520 | goto end; |
1521 | } | |
1522 | ||
53efb85a | 1523 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1524 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1525 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1526 | |
1527 | /* Major versions must be the same */ | |
1528 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1529 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1530 | reply.major, be32toh(msg.major)); | |
5a693d30 | 1531 | connection_put(conn); |
d4519fa3 JD |
1532 | ret = 0; |
1533 | goto end; | |
1534 | } | |
1535 | ||
58eb9381 | 1536 | conn->major = reply.major; |
0f907de1 JD |
1537 | /* We adapt to the lowest compatible version */ |
1538 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1539 | conn->minor = reply.minor; |
0f907de1 | 1540 | } else { |
58eb9381 | 1541 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1542 | } |
1543 | ||
6151a90f JD |
1544 | reply.major = htobe32(reply.major); |
1545 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1546 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1547 | sizeof(struct lttcomm_relayd_version), 0); |
1548 | if (ret < 0) { | |
1549 | ERR("Relay sending version"); | |
1550 | } | |
1551 | ||
58eb9381 DG |
1552 | DBG("Version check done using protocol %u.%u", conn->major, |
1553 | conn->minor); | |
b8aa1682 JD |
1554 | |
1555 | end: | |
1556 | return ret; | |
1557 | } | |
1558 | ||
c8f59ee5 | 1559 | /* |
6d805429 | 1560 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1561 | */ |
5a693d30 | 1562 | static int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1563 | struct relay_connection *conn) |
c8f59ee5 | 1564 | { |
58eb9381 | 1565 | struct relay_session *session = conn->session; |
6d805429 | 1566 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1567 | struct lttcomm_relayd_generic_reply reply; |
1568 | struct relay_stream *stream; | |
1569 | int ret; | |
c8f59ee5 DG |
1570 | uint64_t last_net_seq_num, stream_id; |
1571 | ||
6d805429 | 1572 | DBG("Data pending command received"); |
c8f59ee5 | 1573 | |
58eb9381 | 1574 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1575 | ERR("Trying to check for data before version check"); |
1576 | ret = -1; | |
1577 | goto end_no_session; | |
1578 | } | |
1579 | ||
58eb9381 | 1580 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1581 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1582 | if (ret == 0) { |
1583 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1584 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1585 | } else { |
1586 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1587 | ret); | |
1588 | } | |
c8f59ee5 DG |
1589 | ret = -1; |
1590 | goto end_no_session; | |
1591 | } | |
1592 | ||
1593 | stream_id = be64toh(msg.stream_id); | |
1594 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1595 | ||
5a693d30 | 1596 | stream = stream_get_by_id(stream_id); |
de91f48a | 1597 | if (stream == NULL) { |
c8f59ee5 | 1598 | ret = -1; |
5a693d30 | 1599 | goto end; |
c8f59ee5 DG |
1600 | } |
1601 | ||
5a693d30 MD |
1602 | pthread_mutex_lock(&stream->lock); |
1603 | ||
6d805429 | 1604 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1605 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1606 | last_net_seq_num); | |
1607 | ||
33832e64 | 1608 | /* Avoid wrapping issue */ |
39df6d9f | 1609 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1610 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1611 | ret = 0; |
6d805429 DG |
1612 | } else { |
1613 | /* Data still being streamed thus pending */ | |
1614 | ret = 1; | |
c8f59ee5 DG |
1615 | } |
1616 | ||
5a693d30 MD |
1617 | stream->data_pending_check_done = true; |
1618 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1619 | |
5a693d30 MD |
1620 | stream_put(stream); |
1621 | end: | |
c8f59ee5 | 1622 | |
53efb85a | 1623 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1624 | reply.ret_code = htobe32(ret); |
58eb9381 | 1625 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1626 | if (ret < 0) { |
6d805429 | 1627 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1628 | } |
1629 | ||
1630 | end_no_session: | |
1631 | return ret; | |
1632 | } | |
1633 | ||
1634 | /* | |
1635 | * Wait for the control socket to reach a quiescent state. | |
1636 | * | |
5a693d30 MD |
1637 | * Note that for now, when receiving this command from the session |
1638 | * daemon, this means that every subsequent commands or data received on | |
1639 | * the control socket has been handled. So, this is why we simply return | |
1640 | * OK here. | |
c8f59ee5 | 1641 | */ |
5a693d30 | 1642 | static int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1643 | struct relay_connection *conn) |
c8f59ee5 DG |
1644 | { |
1645 | int ret; | |
ad7051c0 DG |
1646 | uint64_t stream_id; |
1647 | struct relay_stream *stream; | |
ad7051c0 | 1648 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1649 | struct lttcomm_relayd_generic_reply reply; |
1650 | ||
1651 | DBG("Checking quiescent state on control socket"); | |
1652 | ||
58eb9381 | 1653 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1654 | ERR("Trying to check for data before version check"); |
1655 | ret = -1; | |
1656 | goto end_no_session; | |
1657 | } | |
1658 | ||
58eb9381 | 1659 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1660 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1661 | if (ret == 0) { |
1662 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1663 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1664 | } else { |
1665 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1666 | ret); | |
1667 | } | |
ad7051c0 DG |
1668 | ret = -1; |
1669 | goto end_no_session; | |
1670 | } | |
1671 | ||
1672 | stream_id = be64toh(msg.stream_id); | |
5a693d30 MD |
1673 | stream = stream_get_by_id(stream_id); |
1674 | if (!stream) { | |
1675 | goto reply; | |
1676 | } | |
1677 | pthread_mutex_lock(&stream->lock); | |
1678 | stream->data_pending_check_done = true; | |
1679 | pthread_mutex_unlock(&stream->lock); | |
1680 | DBG("Relay quiescent control pending flag set to %" PRIu64, stream_id); | |
1681 | stream_put(stream); | |
1682 | reply: | |
53efb85a | 1683 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1684 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1685 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1686 | if (ret < 0) { |
6d805429 | 1687 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1688 | } |
1689 | ||
ad7051c0 | 1690 | end_no_session: |
c8f59ee5 DG |
1691 | return ret; |
1692 | } | |
1693 | ||
f7079f67 | 1694 | /* |
5a693d30 MD |
1695 | * Initialize a data pending command. This means that a consumer is about |
1696 | * to ask for data pending for each stream it holds. Simply iterate over | |
1697 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1698 | * |
1699 | * This command returns to the client a LTTNG_OK code. | |
1700 | */ | |
5a693d30 | 1701 | static int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1702 | struct relay_connection *conn) |
f7079f67 DG |
1703 | { |
1704 | int ret; | |
1705 | struct lttng_ht_iter iter; | |
1706 | struct lttcomm_relayd_begin_data_pending msg; | |
1707 | struct lttcomm_relayd_generic_reply reply; | |
1708 | struct relay_stream *stream; | |
1709 | uint64_t session_id; | |
1710 | ||
1711 | assert(recv_hdr); | |
58eb9381 | 1712 | assert(conn); |
f7079f67 DG |
1713 | |
1714 | DBG("Init streams for data pending"); | |
1715 | ||
58eb9381 | 1716 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1717 | ERR("Trying to check for data before version check"); |
1718 | ret = -1; | |
1719 | goto end_no_session; | |
1720 | } | |
1721 | ||
58eb9381 | 1722 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1723 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1724 | if (ret == 0) { |
1725 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1726 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1727 | } else { |
1728 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1729 | ret); | |
1730 | } | |
f7079f67 DG |
1731 | ret = -1; |
1732 | goto end_no_session; | |
1733 | } | |
1734 | ||
1735 | session_id = be64toh(msg.session_id); | |
1736 | ||
1737 | /* | |
5a693d30 MD |
1738 | * Iterate over all streams to set the begin data pending flag. |
1739 | * For now, the streams are indexed by stream handle so we have | |
1740 | * to iterate over all streams to find the one associated with | |
1741 | * the right session_id. | |
f7079f67 DG |
1742 | */ |
1743 | rcu_read_lock(); | |
d3e2ba59 | 1744 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1745 | node.node) { |
5a693d30 MD |
1746 | if (!stream_get(stream)) { |
1747 | continue; | |
1748 | } | |
1749 | if (stream->trace->session->id == session_id) { | |
1750 | pthread_mutex_lock(&stream->lock); | |
1751 | stream->data_pending_check_done = false; | |
1752 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1753 | DBG("Set begin data pending flag to stream %" PRIu64, |
1754 | stream->stream_handle); | |
1755 | } | |
5a693d30 | 1756 | stream_put(stream); |
f7079f67 DG |
1757 | } |
1758 | rcu_read_unlock(); | |
1759 | ||
53efb85a | 1760 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1761 | /* All good, send back reply. */ |
1762 | reply.ret_code = htobe32(LTTNG_OK); | |
1763 | ||
58eb9381 | 1764 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1765 | if (ret < 0) { |
1766 | ERR("Relay begin data pending send reply failed"); | |
1767 | } | |
1768 | ||
1769 | end_no_session: | |
1770 | return ret; | |
1771 | } | |
1772 | ||
1773 | /* | |
5a693d30 MD |
1774 | * End data pending command. This will check, for a given session id, if |
1775 | * each stream associated with it has its data_pending_check_done flag | |
1776 | * set. If not, this means that the client lost track of the stream but | |
1777 | * the data is still being streamed on our side. In this case, we inform | |
1778 | * the client that data is in flight. | |
f7079f67 DG |
1779 | * |
1780 | * Return to the client if there is data in flight or not with a ret_code. | |
1781 | */ | |
5a693d30 | 1782 | static int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1783 | struct relay_connection *conn) |
f7079f67 DG |
1784 | { |
1785 | int ret; | |
1786 | struct lttng_ht_iter iter; | |
1787 | struct lttcomm_relayd_end_data_pending msg; | |
1788 | struct lttcomm_relayd_generic_reply reply; | |
1789 | struct relay_stream *stream; | |
1790 | uint64_t session_id; | |
1791 | uint32_t is_data_inflight = 0; | |
1792 | ||
f7079f67 DG |
1793 | DBG("End data pending command"); |
1794 | ||
58eb9381 | 1795 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1796 | ERR("Trying to check for data before version check"); |
1797 | ret = -1; | |
1798 | goto end_no_session; | |
1799 | } | |
1800 | ||
58eb9381 | 1801 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1802 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1803 | if (ret == 0) { |
1804 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1805 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1806 | } else { |
1807 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1808 | ret); | |
1809 | } | |
f7079f67 DG |
1810 | ret = -1; |
1811 | goto end_no_session; | |
1812 | } | |
1813 | ||
1814 | session_id = be64toh(msg.session_id); | |
1815 | ||
5a693d30 MD |
1816 | /* |
1817 | * Iterate over all streams to see if the begin data pending | |
1818 | * flag is set. | |
1819 | */ | |
f7079f67 | 1820 | rcu_read_lock(); |
d3e2ba59 | 1821 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1822 | node.node) { |
5a693d30 MD |
1823 | if (!stream_get(stream)) { |
1824 | continue; | |
1825 | } | |
1826 | if (stream->trace->session->id != session_id) { | |
1827 | stream_put(stream); | |
1828 | continue; | |
1829 | } | |
1830 | pthread_mutex_lock(&stream->lock); | |
1831 | if (!stream->data_pending_check_done) { | |
1832 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1833 | is_data_inflight = 1; | |
1834 | DBG("Data is still in flight for stream %" PRIu64, | |
1835 | stream->stream_handle); | |
1836 | pthread_mutex_unlock(&stream->lock); | |
1837 | stream_put(stream); | |
1838 | break; | |
1839 | } | |
f7079f67 | 1840 | } |
5a693d30 MD |
1841 | pthread_mutex_unlock(&stream->lock); |
1842 | stream_put(stream); | |
f7079f67 DG |
1843 | } |
1844 | rcu_read_unlock(); | |
1845 | ||
53efb85a | 1846 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1847 | /* All good, send back reply. */ |
1848 | reply.ret_code = htobe32(is_data_inflight); | |
1849 | ||
58eb9381 | 1850 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1851 | if (ret < 0) { |
1852 | ERR("Relay end data pending send reply failed"); | |
1853 | } | |
1854 | ||
1855 | end_no_session: | |
1856 | return ret; | |
1857 | } | |
1858 | ||
1c20f0e2 JD |
1859 | /* |
1860 | * Receive an index for a specific stream. | |
1861 | * | |
1862 | * Return 0 on success else a negative value. | |
1863 | */ | |
5a693d30 | 1864 | static int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1865 | struct relay_connection *conn) |
1c20f0e2 | 1866 | { |
5a693d30 | 1867 | int ret, send_ret; |
58eb9381 | 1868 | struct relay_session *session = conn->session; |
1c20f0e2 | 1869 | struct lttcomm_relayd_index index_info; |
5a693d30 | 1870 | struct relay_index *index; |
1c20f0e2 JD |
1871 | struct lttcomm_relayd_generic_reply reply; |
1872 | struct relay_stream *stream; | |
1873 | uint64_t net_seq_num; | |
1874 | ||
58eb9381 | 1875 | assert(conn); |
1c20f0e2 JD |
1876 | |
1877 | DBG("Relay receiving index"); | |
1878 | ||
58eb9381 | 1879 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1880 | ERR("Trying to close a stream before version check"); |
1881 | ret = -1; | |
1882 | goto end_no_session; | |
1883 | } | |
1884 | ||
58eb9381 | 1885 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1886 | sizeof(index_info), 0); |
1887 | if (ret < sizeof(index_info)) { | |
1888 | if (ret == 0) { | |
1889 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1890 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1891 | } else { |
1892 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1893 | } | |
1894 | ret = -1; | |
1895 | goto end_no_session; | |
1896 | } | |
1897 | ||
1898 | net_seq_num = be64toh(index_info.net_seq_num); | |
1899 | ||
5a693d30 | 1900 | stream = stream_get_by_id(be64toh(index_info.relay_stream_id)); |
1c20f0e2 | 1901 | if (!stream) { |
5a693d30 | 1902 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 1903 | ret = -1; |
5a693d30 | 1904 | goto end; |
1c20f0e2 | 1905 | } |
5a693d30 | 1906 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 1907 | |
d3e2ba59 JD |
1908 | /* Live beacon handling */ |
1909 | if (index_info.packet_size == 0) { | |
5a693d30 MD |
1910 | DBG("Received live beacon for stream %" PRIu64, |
1911 | stream->stream_handle); | |
d3e2ba59 JD |
1912 | |
1913 | /* | |
5a693d30 MD |
1914 | * Only flag a stream inactive when it has already |
1915 | * received data and no indexes are in flight. | |
d3e2ba59 | 1916 | */ |
8921c6e4 | 1917 | if (stream->index_received_seqcount > 0 |
5a693d30 MD |
1918 | && stream->indexes_in_flight == 0) { |
1919 | stream->beacon_ts_end = | |
1920 | be64toh(index_info.timestamp_end); | |
d3e2ba59 JD |
1921 | } |
1922 | ret = 0; | |
5a693d30 | 1923 | goto end_stream_put; |
d3e2ba59 JD |
1924 | } else { |
1925 | stream->beacon_ts_end = -1ULL; | |
1926 | } | |
1927 | ||
528f2ffa JD |
1928 | if (stream->ctf_stream_id == -1ULL) { |
1929 | stream->ctf_stream_id = be64toh(index_info.stream_id); | |
1930 | } | |
5a693d30 MD |
1931 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
1932 | if (!index) { | |
1933 | ret = -1; | |
1934 | ERR("relay_index_get_by_id_or_create index NULL"); | |
1935 | goto end_stream_put; | |
1c20f0e2 | 1936 | } |
5a693d30 MD |
1937 | if (set_index_control_data(index, &index_info)) { |
1938 | ERR("set_index_control_data error"); | |
1939 | relay_index_put(index); | |
1940 | ret = -1; | |
1941 | goto end_stream_put; | |
1942 | } | |
1943 | ret = relay_index_try_flush(index); | |
1944 | if (ret == 0) { | |
8921c6e4 MD |
1945 | tracefile_array_commit_seq(stream->tfa); |
1946 | stream->index_received_seqcount++; | |
5a693d30 MD |
1947 | } else if (ret > 0) { |
1948 | /* no flush. */ | |
1949 | ret = 0; | |
1950 | } else { | |
1951 | ERR("relay_index_try_flush error %d", ret); | |
1952 | relay_index_put(index); | |
1953 | ret = -1; | |
1c20f0e2 JD |
1954 | } |
1955 | ||
5a693d30 MD |
1956 | end_stream_put: |
1957 | pthread_mutex_unlock(&stream->lock); | |
1958 | stream_put(stream); | |
1959 | ||
1960 | end: | |
1c20f0e2 | 1961 | |
53efb85a | 1962 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
1963 | if (ret < 0) { |
1964 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1965 | } else { | |
1966 | reply.ret_code = htobe32(LTTNG_OK); | |
1967 | } | |
58eb9381 | 1968 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
1969 | if (send_ret < 0) { |
1970 | ERR("Relay sending close index id reply"); | |
1971 | ret = send_ret; | |
1972 | } | |
1973 | ||
1974 | end_no_session: | |
1975 | return ret; | |
1976 | } | |
1977 | ||
a4baae1b JD |
1978 | /* |
1979 | * Receive the streams_sent message. | |
1980 | * | |
1981 | * Return 0 on success else a negative value. | |
1982 | */ | |
5a693d30 | 1983 | static int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1984 | struct relay_connection *conn) |
a4baae1b JD |
1985 | { |
1986 | int ret, send_ret; | |
1987 | struct lttcomm_relayd_generic_reply reply; | |
1988 | ||
58eb9381 | 1989 | assert(conn); |
a4baae1b JD |
1990 | |
1991 | DBG("Relay receiving streams_sent"); | |
1992 | ||
58eb9381 | 1993 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
1994 | ERR("Trying to close a stream before version check"); |
1995 | ret = -1; | |
1996 | goto end_no_session; | |
1997 | } | |
1998 | ||
1999 | /* | |
5a693d30 MD |
2000 | * Publish every pending stream in the connection recv list which are |
2001 | * now ready to be used by the viewer. | |
4a9daf17 | 2002 | */ |
5a693d30 | 2003 | publish_connection_local_streams(conn); |
4a9daf17 | 2004 | |
53efb85a | 2005 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2006 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2007 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
2008 | if (send_ret < 0) { |
2009 | ERR("Relay sending sent_stream reply"); | |
2010 | ret = send_ret; | |
2011 | } else { | |
2012 | /* Success. */ | |
2013 | ret = 0; | |
2014 | } | |
2015 | ||
2016 | end_no_session: | |
2017 | return ret; | |
2018 | } | |
2019 | ||
b8aa1682 | 2020 | /* |
d3e2ba59 | 2021 | * Process the commands received on the control socket |
b8aa1682 | 2022 | */ |
5a693d30 | 2023 | static int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 2024 | struct relay_connection *conn) |
b8aa1682 JD |
2025 | { |
2026 | int ret = 0; | |
2027 | ||
2028 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 2029 | case RELAYD_CREATE_SESSION: |
58eb9381 | 2030 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 2031 | break; |
b8aa1682 | 2032 | case RELAYD_ADD_STREAM: |
58eb9381 | 2033 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
2034 | break; |
2035 | case RELAYD_START_DATA: | |
58eb9381 | 2036 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
2037 | break; |
2038 | case RELAYD_SEND_METADATA: | |
58eb9381 | 2039 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
2040 | break; |
2041 | case RELAYD_VERSION: | |
58eb9381 | 2042 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 2043 | break; |
173af62f | 2044 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 2045 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 2046 | break; |
6d805429 | 2047 | case RELAYD_DATA_PENDING: |
58eb9381 | 2048 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
2049 | break; |
2050 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 2051 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 2052 | break; |
f7079f67 | 2053 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 2054 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
2055 | break; |
2056 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2057 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2058 | break; |
1c20f0e2 | 2059 | case RELAYD_SEND_INDEX: |
58eb9381 | 2060 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2061 | break; |
a4baae1b | 2062 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2063 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2064 | break; |
b8aa1682 JD |
2065 | case RELAYD_UPDATE_SYNC_INFO: |
2066 | default: | |
2067 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2068 | relay_unknown_command(conn); |
b8aa1682 JD |
2069 | ret = -1; |
2070 | goto end; | |
2071 | } | |
2072 | ||
2073 | end: | |
2074 | return ret; | |
2075 | } | |
2076 | ||
7d2f7452 DG |
2077 | /* |
2078 | * Handle index for a data stream. | |
2079 | * | |
5a693d30 | 2080 | * Called with the stream lock held. |
7d2f7452 DG |
2081 | * |
2082 | * Return 0 on success else a negative value. | |
2083 | */ | |
2084 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2085 | int rotate_index) | |
2086 | { | |
5a693d30 MD |
2087 | int ret = 0; |
2088 | uint64_t data_offset; | |
2089 | struct relay_index *index; | |
7d2f7452 | 2090 | |
7d2f7452 DG |
2091 | /* Get data offset because we are about to update the index. */ |
2092 | data_offset = htobe64(stream->tracefile_size_current); | |
2093 | ||
de948e0e MD |
2094 | DBG("handle_index_data: stream %" PRIu64 " net_seq_num %" PRIu64 " data offset %" PRIu64, |
2095 | stream->stream_handle, net_seq_num, stream->tracefile_size_current); | |
5a693d30 | 2096 | |
7d2f7452 | 2097 | /* |
5a693d30 MD |
2098 | * Lookup for an existing index for that stream id/sequence |
2099 | * number. If it exists, the control thread has already received the | |
2100 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2101 | */ |
5a693d30 | 2102 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2103 | if (!index) { |
5a693d30 MD |
2104 | ret = -1; |
2105 | goto end; | |
7d2f7452 DG |
2106 | } |
2107 | ||
5a693d30 MD |
2108 | if (rotate_index || !stream->index_fd) { |
2109 | int fd; | |
2110 | ||
2111 | /* Put ref on previous index_fd. */ | |
2112 | if (stream->index_fd) { | |
2113 | stream_fd_put(stream->index_fd); | |
2114 | stream->index_fd = NULL; | |
7d2f7452 | 2115 | } |
7d2f7452 | 2116 | |
5a693d30 MD |
2117 | fd = index_create_file(stream->path_name, stream->channel_name, |
2118 | -1, -1, stream->tracefile_size, | |
8921c6e4 | 2119 | tracefile_array_get_file_index_head(stream->tfa)); |
5a693d30 MD |
2120 | if (fd < 0) { |
2121 | ret = -1; | |
2122 | /* Put self-ref for this index due to error. */ | |
2123 | relay_index_put(index); | |
2124 | goto end; | |
2125 | } | |
2126 | stream->index_fd = stream_fd_create(fd); | |
2127 | if (!stream->index_fd) { | |
2128 | ret = -1; | |
2129 | if (close(fd)) { | |
2130 | PERROR("Error closing FD %d", fd); | |
2131 | } | |
2132 | /* Put self-ref for this index due to error. */ | |
2133 | relay_index_put(index); | |
2134 | /* Will put the local ref. */ | |
2135 | goto end; | |
7d2f7452 | 2136 | } |
7d2f7452 DG |
2137 | } |
2138 | ||
5a693d30 MD |
2139 | if (relay_index_set_fd(index, stream->index_fd, data_offset)) { |
2140 | ret = -1; | |
2141 | /* Put self-ref for this index due to error. */ | |
2142 | relay_index_put(index); | |
2143 | goto end; | |
7d2f7452 DG |
2144 | } |
2145 | ||
5a693d30 MD |
2146 | ret = relay_index_try_flush(index); |
2147 | if (ret == 0) { | |
8921c6e4 MD |
2148 | tracefile_array_commit_seq(stream->tfa); |
2149 | stream->index_received_seqcount++; | |
5a693d30 MD |
2150 | } else if (ret > 0) { |
2151 | /* No flush. */ | |
2152 | ret = 0; | |
2153 | } else { | |
2154 | /* Put self-ref for this index due to error. */ | |
2155 | relay_index_put(index); | |
2156 | ret = -1; | |
2157 | } | |
2158 | end: | |
7d2f7452 DG |
2159 | return ret; |
2160 | } | |
2161 | ||
b8aa1682 JD |
2162 | /* |
2163 | * relay_process_data: Process the data received on the data socket | |
2164 | */ | |
5a693d30 | 2165 | static int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2166 | { |
7d2f7452 | 2167 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2168 | ssize_t size_ret; |
b8aa1682 JD |
2169 | struct relay_stream *stream; |
2170 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2171 | uint64_t stream_id; |
173af62f | 2172 | uint64_t net_seq_num; |
b8aa1682 | 2173 | uint32_t data_size; |
2a174661 | 2174 | struct relay_session *session; |
a08107cc | 2175 | bool new_stream = false, close_requested = false; |
b95ad2e1 MD |
2176 | size_t chunk_size = RECV_DATA_BUFFER_SIZE; |
2177 | size_t recv_off = 0; | |
2178 | char data_buffer[chunk_size]; | |
b8aa1682 | 2179 | |
58eb9381 | 2180 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, |
7c5aef62 | 2181 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2182 | if (ret <= 0) { |
a6cd2b97 DG |
2183 | if (ret == 0) { |
2184 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2185 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2186 | } else { |
58eb9381 | 2187 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2188 | } |
b8aa1682 JD |
2189 | ret = -1; |
2190 | goto end; | |
2191 | } | |
2192 | ||
2193 | stream_id = be64toh(data_hdr.stream_id); | |
5a693d30 | 2194 | stream = stream_get_by_id(stream_id); |
b8aa1682 | 2195 | if (!stream) { |
de948e0e | 2196 | ERR("relay_process_data: Cannot find stream %" PRIu64, stream_id); |
b8aa1682 | 2197 | ret = -1; |
5a693d30 | 2198 | goto end; |
b8aa1682 | 2199 | } |
5a693d30 | 2200 | session = stream->trace->session; |
b8aa1682 | 2201 | data_size = be32toh(data_hdr.data_size); |
b8aa1682 | 2202 | |
173af62f DG |
2203 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2204 | ||
77c7c900 | 2205 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2206 | data_size, stream_id, net_seq_num); |
b8aa1682 | 2207 | |
5a693d30 MD |
2208 | pthread_mutex_lock(&stream->lock); |
2209 | ||
1c20f0e2 | 2210 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2211 | if (stream->tracefile_size > 0 && |
2212 | (stream->tracefile_size_current + data_size) > | |
2213 | stream->tracefile_size) { | |
8921c6e4 MD |
2214 | uint64_t old_id, new_id; |
2215 | ||
2216 | old_id = tracefile_array_get_file_index_head(stream->tfa); | |
2217 | tracefile_array_file_rotate(stream->tfa); | |
2218 | ||
2219 | /* new_id is updated by utils_rotate_stream_file. */ | |
2220 | new_id = old_id; | |
6b6b9a5a | 2221 | |
5a693d30 MD |
2222 | ret = utils_rotate_stream_file(stream->path_name, |
2223 | stream->channel_name, stream->tracefile_size, | |
2224 | stream->tracefile_count, -1, | |
2225 | -1, stream->stream_fd->fd, | |
8921c6e4 | 2226 | &new_id, &stream->stream_fd->fd); |
0f907de1 | 2227 | if (ret < 0) { |
1c20f0e2 | 2228 | ERR("Rotating stream output file"); |
5a693d30 MD |
2229 | goto end_stream_unlock; |
2230 | } | |
5a693d30 MD |
2231 | /* |
2232 | * Reset current size because we just performed a stream | |
2233 | * rotation. | |
2234 | */ | |
a6976990 | 2235 | stream->tracefile_size_current = 0; |
1c20f0e2 JD |
2236 | rotate_index = 1; |
2237 | } | |
2238 | ||
1c20f0e2 | 2239 | /* |
5a693d30 MD |
2240 | * Index are handled in protocol version 2.4 and above. Also, |
2241 | * snapshot and index are NOT supported. | |
1c20f0e2 | 2242 | */ |
2a174661 | 2243 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2244 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2245 | if (ret < 0) { |
de948e0e MD |
2246 | ERR("handle_index_data: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
2247 | stream->stream_handle, net_seq_num, ret); | |
5a693d30 | 2248 | goto end_stream_unlock; |
1c20f0e2 | 2249 | } |
1c20f0e2 JD |
2250 | } |
2251 | ||
b95ad2e1 MD |
2252 | for (recv_off = 0; recv_off < data_size; recv_off += chunk_size) { |
2253 | size_t recv_size = min(data_size - recv_off, chunk_size); | |
1d4dfdef | 2254 | |
b95ad2e1 MD |
2255 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, recv_size, 0); |
2256 | if (ret <= 0) { | |
2257 | if (ret == 0) { | |
2258 | /* Orderly shutdown. Not necessary to print an error. */ | |
2259 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); | |
2260 | } else { | |
2261 | ERR("Socket %d error %d", conn->sock->fd, ret); | |
2262 | } | |
2263 | ret = -1; | |
2264 | goto end_stream_unlock; | |
2265 | } | |
2266 | ||
2267 | /* Write data to stream output fd. */ | |
2268 | size_ret = lttng_write(stream->stream_fd->fd, data_buffer, | |
2269 | recv_size); | |
2270 | if (size_ret < recv_size) { | |
2271 | ERR("Relay error writing data to file"); | |
2272 | ret = -1; | |
2273 | goto end_stream_unlock; | |
2274 | } | |
2275 | ||
2276 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, | |
2277 | size_ret, stream->stream_handle); | |
2278 | } | |
5ab7344e | 2279 | |
5a693d30 MD |
2280 | ret = write_padding_to_file(stream->stream_fd->fd, |
2281 | be32toh(data_hdr.padding_size)); | |
1d4dfdef | 2282 | if (ret < 0) { |
de948e0e MD |
2283 | ERR("write_padding_to_file: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
2284 | stream->stream_handle, net_seq_num, ret); | |
5a693d30 | 2285 | goto end_stream_unlock; |
1d4dfdef | 2286 | } |
5a693d30 MD |
2287 | stream->tracefile_size_current += |
2288 | data_size + be32toh(data_hdr.padding_size); | |
1b95be80 MD |
2289 | if (stream->prev_seq == -1ULL) { |
2290 | new_stream = true; | |
2291 | } | |
2292 | ||
173af62f DG |
2293 | stream->prev_seq = net_seq_num; |
2294 | ||
5a693d30 | 2295 | end_stream_unlock: |
a08107cc | 2296 | close_requested = stream->close_requested; |
5a693d30 | 2297 | pthread_mutex_unlock(&stream->lock); |
a08107cc JG |
2298 | if (close_requested) { |
2299 | try_stream_close(stream); | |
2300 | } | |
2301 | ||
1b95be80 MD |
2302 | if (new_stream) { |
2303 | pthread_mutex_lock(&session->lock); | |
2304 | uatomic_set(&session->new_streams, 1); | |
2305 | pthread_mutex_unlock(&session->lock); | |
2306 | } | |
5a693d30 | 2307 | stream_put(stream); |
b8aa1682 JD |
2308 | end: |
2309 | return ret; | |
2310 | } | |
2311 | ||
5a693d30 | 2312 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2313 | { |
2314 | int ret; | |
2315 | ||
58eb9381 | 2316 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2317 | |
2318 | ret = close(pollfd); | |
2319 | if (ret < 0) { | |
2320 | ERR("Closing pollfd %d", pollfd); | |
2321 | } | |
2322 | } | |
2323 | ||
5a693d30 MD |
2324 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2325 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2326 | { |
5a693d30 | 2327 | const char *type_str; |
2a174661 | 2328 | |
5a693d30 MD |
2329 | switch (conn->type) { |
2330 | case RELAY_DATA: | |
2331 | type_str = "Data"; | |
2332 | break; | |
2333 | case RELAY_CONTROL: | |
2334 | type_str = "Control"; | |
2335 | break; | |
2336 | case RELAY_VIEWER_COMMAND: | |
2337 | type_str = "Viewer Command"; | |
2338 | break; | |
2339 | case RELAY_VIEWER_NOTIFICATION: | |
2340 | type_str = "Viewer Notification"; | |
2341 | break; | |
2342 | default: | |
2343 | type_str = "Unknown"; | |
9d1bbf21 | 2344 | } |
5a693d30 MD |
2345 | cleanup_connection_pollfd(events, pollfd); |
2346 | connection_put(conn); | |
2347 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2348 | } |
2349 | ||
2350 | /* | |
2351 | * This thread does the actual work | |
2352 | */ | |
5a693d30 | 2353 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2354 | { |
beaad64c DG |
2355 | int ret, err = -1, last_seen_data_fd = -1; |
2356 | uint32_t nb_fd; | |
b8aa1682 JD |
2357 | struct lttng_poll_event events; |
2358 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2359 | struct lttng_ht_iter iter; |
b8aa1682 | 2360 | struct lttcomm_relayd_hdr recv_hdr; |
90e7d72f | 2361 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2362 | |
2363 | DBG("[thread] Relay worker started"); | |
2364 | ||
9d1bbf21 MD |
2365 | rcu_register_thread(); |
2366 | ||
55706a7d MD |
2367 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2368 | ||
9b5e0863 MD |
2369 | if (testpoint(relayd_thread_worker)) { |
2370 | goto error_testpoint; | |
2371 | } | |
2372 | ||
f385ae0a MD |
2373 | health_code_update(); |
2374 | ||
b8aa1682 JD |
2375 | /* table of connections indexed on socket */ |
2376 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2377 | if (!relay_connections_ht) { |
2378 | goto relay_connections_ht_error; | |
2379 | } | |
b8aa1682 | 2380 | |
b8aa1682 JD |
2381 | ret = create_thread_poll_set(&events, 2); |
2382 | if (ret < 0) { | |
2383 | goto error_poll_create; | |
2384 | } | |
2385 | ||
58eb9381 | 2386 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2387 | if (ret < 0) { |
2388 | goto error; | |
2389 | } | |
2390 | ||
beaad64c | 2391 | restart: |
b8aa1682 | 2392 | while (1) { |
beaad64c DG |
2393 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2394 | ||
f385ae0a MD |
2395 | health_code_update(); |
2396 | ||
b8aa1682 | 2397 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2398 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2399 | health_poll_entry(); |
b8aa1682 | 2400 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2401 | health_poll_exit(); |
b8aa1682 JD |
2402 | if (ret < 0) { |
2403 | /* | |
2404 | * Restart interrupted system call. | |
2405 | */ | |
2406 | if (errno == EINTR) { | |
2407 | goto restart; | |
2408 | } | |
2409 | goto error; | |
2410 | } | |
2411 | ||
0d9c5d77 DG |
2412 | nb_fd = ret; |
2413 | ||
beaad64c | 2414 | /* |
5a693d30 MD |
2415 | * Process control. The control connection is |
2416 | * prioritized so we don't starve it with high | |
2417 | * throughput tracing data on the data connection. | |
beaad64c | 2418 | */ |
b8aa1682 JD |
2419 | for (i = 0; i < nb_fd; i++) { |
2420 | /* Fetch once the poll data */ | |
beaad64c DG |
2421 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2422 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2423 | |
f385ae0a MD |
2424 | health_code_update(); |
2425 | ||
fd20dac9 | 2426 | if (!revents) { |
5a693d30 MD |
2427 | /* |
2428 | * No activity for this FD (poll | |
2429 | * implementation). | |
2430 | */ | |
fd20dac9 MD |
2431 | continue; |
2432 | } | |
2433 | ||
b8aa1682 JD |
2434 | /* Thread quit pipe has been closed. Killing thread. */ |
2435 | ret = check_thread_quit_pipe(pollfd, revents); | |
2436 | if (ret) { | |
095a4ae5 MD |
2437 | err = 0; |
2438 | goto exit; | |
b8aa1682 JD |
2439 | } |
2440 | ||
58eb9381 DG |
2441 | /* Inspect the relay conn pipe for new connection */ |
2442 | if (pollfd == relay_conn_pipe[0]) { | |
92d6debb | 2443 | if (revents & LPOLLIN) { |
90e7d72f JG |
2444 | struct relay_connection *conn; |
2445 | ||
58eb9381 | 2446 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2447 | if (ret < 0) { |
2448 | goto error; | |
2449 | } | |
58eb9381 DG |
2450 | lttng_poll_add(&events, conn->sock->fd, |
2451 | LPOLLIN | LPOLLRDHUP); | |
5a693d30 | 2452 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2453 | DBG("Connection socket %d added", conn->sock->fd); |
92d6debb MD |
2454 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2455 | ERR("Relay connection pipe error"); | |
2456 | goto error; | |
2457 | } else { | |
2458 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
2459 | goto error; | |
b8aa1682 | 2460 | } |
58eb9381 | 2461 | } else { |
90e7d72f JG |
2462 | struct relay_connection *ctrl_conn; |
2463 | ||
5a693d30 | 2464 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 2465 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 2466 | assert(ctrl_conn); |
58eb9381 | 2467 | |
92d6debb MD |
2468 | if (ctrl_conn->type == RELAY_DATA) { |
2469 | if (revents & LPOLLIN) { | |
beaad64c DG |
2470 | /* |
2471 | * Flag the last seen data fd not deleted. It will be | |
2472 | * used as the last seen fd if any fd gets deleted in | |
2473 | * this first loop. | |
2474 | */ | |
2475 | last_notdel_data_fd = pollfd; | |
2476 | } | |
92d6debb MD |
2477 | goto put_ctrl_connection; |
2478 | } | |
2479 | assert(ctrl_conn->type == RELAY_CONTROL); | |
2480 | ||
2481 | if (revents & LPOLLIN) { | |
2482 | ret = ctrl_conn->sock->ops->recvmsg(ctrl_conn->sock, | |
2483 | &recv_hdr, sizeof(recv_hdr), 0); | |
2484 | if (ret <= 0) { | |
2485 | /* Connection closed */ | |
2486 | relay_thread_close_connection(&events, pollfd, | |
2487 | ctrl_conn); | |
2488 | } else { | |
2489 | ret = relay_process_control(&recv_hdr, ctrl_conn); | |
2490 | if (ret < 0) { | |
2491 | /* Clear the session on error. */ | |
2492 | relay_thread_close_connection(&events, | |
2493 | pollfd, ctrl_conn); | |
2494 | } | |
2495 | seen_control = 1; | |
2496 | } | |
2497 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2498 | relay_thread_close_connection(&events, | |
2499 | pollfd, ctrl_conn); | |
2500 | if (last_seen_data_fd == pollfd) { | |
2501 | last_seen_data_fd = last_notdel_data_fd; | |
2502 | } | |
58eb9381 | 2503 | } else { |
92d6debb MD |
2504 | ERR("Unexpected poll events %u for control sock %d", |
2505 | revents, pollfd); | |
2506 | connection_put(ctrl_conn); | |
2507 | goto error; | |
beaad64c | 2508 | } |
92d6debb | 2509 | put_ctrl_connection: |
5a693d30 | 2510 | connection_put(ctrl_conn); |
beaad64c DG |
2511 | } |
2512 | } | |
2513 | ||
2514 | /* | |
2515 | * The last loop handled a control request, go back to poll to make | |
2516 | * sure we prioritise the control socket. | |
2517 | */ | |
2518 | if (seen_control) { | |
2519 | continue; | |
2520 | } | |
2521 | ||
2522 | if (last_seen_data_fd >= 0) { | |
2523 | for (i = 0; i < nb_fd; i++) { | |
2524 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2525 | |
2526 | health_code_update(); | |
2527 | ||
beaad64c DG |
2528 | if (last_seen_data_fd == pollfd) { |
2529 | idx = i; | |
2530 | break; | |
2531 | } | |
2532 | } | |
2533 | } | |
2534 | ||
2535 | /* Process data connection. */ | |
2536 | for (i = idx + 1; i < nb_fd; i++) { | |
2537 | /* Fetch the poll data. */ | |
2538 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2539 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 2540 | struct relay_connection *data_conn; |
beaad64c | 2541 | |
f385ae0a MD |
2542 | health_code_update(); |
2543 | ||
fd20dac9 MD |
2544 | if (!revents) { |
2545 | /* No activity for this FD (poll implementation). */ | |
2546 | continue; | |
2547 | } | |
2548 | ||
beaad64c | 2549 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2550 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2551 | continue; |
2552 | } | |
2553 | ||
5a693d30 | 2554 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 2555 | if (!data_conn) { |
fd20dac9 | 2556 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
2557 | continue; |
2558 | } | |
92d6debb MD |
2559 | if (data_conn->type == RELAY_CONTROL) { |
2560 | goto put_data_connection; | |
2561 | } | |
2562 | assert(data_conn->type == RELAY_DATA); | |
fd20dac9 MD |
2563 | |
2564 | if (revents & LPOLLIN) { | |
90e7d72f | 2565 | ret = relay_process_data(data_conn); |
fd20dac9 MD |
2566 | /* Connection closed */ |
2567 | if (ret < 0) { | |
5a693d30 | 2568 | relay_thread_close_connection(&events, pollfd, |
92d6debb | 2569 | data_conn); |
fd20dac9 MD |
2570 | /* |
2571 | * Every goto restart call sets the last seen fd where | |
2572 | * here we don't really care since we gracefully | |
2573 | * continue the loop after the connection is deleted. | |
2574 | */ | |
2575 | } else { | |
2576 | /* Keep last seen port. */ | |
2577 | last_seen_data_fd = pollfd; | |
5a693d30 | 2578 | connection_put(data_conn); |
fd20dac9 | 2579 | goto restart; |
b8aa1682 | 2580 | } |
92d6debb MD |
2581 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2582 | relay_thread_close_connection(&events, pollfd, | |
2583 | data_conn); | |
2584 | } else { | |
2585 | ERR("Unknown poll events %u for data sock %d", | |
2586 | revents, pollfd); | |
b8aa1682 | 2587 | } |
92d6debb | 2588 | put_data_connection: |
5a693d30 | 2589 | connection_put(data_conn); |
b8aa1682 | 2590 | } |
beaad64c | 2591 | last_seen_data_fd = -1; |
b8aa1682 JD |
2592 | } |
2593 | ||
f385ae0a MD |
2594 | /* Normal exit, no error */ |
2595 | ret = 0; | |
2596 | ||
095a4ae5 | 2597 | exit: |
b8aa1682 | 2598 | error: |
58eb9381 | 2599 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2600 | rcu_read_lock(); |
90e7d72f JG |
2601 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
2602 | destroy_conn, | |
58eb9381 | 2603 | sock_n.node) { |
f385ae0a | 2604 | health_code_update(); |
5a693d30 MD |
2605 | /* |
2606 | * No need to grab another ref, because we own | |
2607 | * destroy_conn. | |
2608 | */ | |
2609 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
2610 | destroy_conn); | |
b8aa1682 | 2611 | } |
94d49140 | 2612 | rcu_read_unlock(); |
5a693d30 MD |
2613 | |
2614 | lttng_poll_clean(&events); | |
7d2f7452 | 2615 | error_poll_create: |
b8aa1682 | 2616 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2617 | relay_connections_ht_error: |
58eb9381 DG |
2618 | /* Close relay conn pipes */ |
2619 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2620 | if (err) { |
2621 | DBG("Thread exited with error"); | |
2622 | } | |
b8aa1682 | 2623 | DBG("Worker thread cleanup complete"); |
9b5e0863 | 2624 | error_testpoint: |
f385ae0a MD |
2625 | if (err) { |
2626 | health_error(); | |
2627 | ERR("Health error occurred in %s", __func__); | |
2628 | } | |
2629 | health_unregister(health_relayd); | |
9d1bbf21 | 2630 | rcu_unregister_thread(); |
b4aacfdc | 2631 | lttng_relay_stop_threads(); |
b8aa1682 JD |
2632 | return NULL; |
2633 | } | |
2634 | ||
2635 | /* | |
2636 | * Create the relay command pipe to wake thread_manage_apps. | |
2637 | * Closed in cleanup(). | |
2638 | */ | |
58eb9381 | 2639 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2640 | { |
a02de639 | 2641 | int ret; |
b8aa1682 | 2642 | |
58eb9381 | 2643 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2644 | |
b8aa1682 JD |
2645 | return ret; |
2646 | } | |
2647 | ||
2648 | /* | |
2649 | * main | |
2650 | */ | |
2651 | int main(int argc, char **argv) | |
2652 | { | |
178a0557 | 2653 | int ret = 0, retval = 0; |
b8aa1682 JD |
2654 | void *status; |
2655 | ||
b8aa1682 JD |
2656 | /* Parse arguments */ |
2657 | progname = argv[0]; | |
178a0557 MD |
2658 | if (set_options(argc, argv)) { |
2659 | retval = -1; | |
2660 | goto exit_options; | |
b8aa1682 JD |
2661 | } |
2662 | ||
178a0557 MD |
2663 | if (set_signal_handler()) { |
2664 | retval = -1; | |
2665 | goto exit_options; | |
b8aa1682 JD |
2666 | } |
2667 | ||
4d513a50 DG |
2668 | /* Try to create directory if -o, --output is specified. */ |
2669 | if (opt_output_path) { | |
994fa64f DG |
2670 | if (*opt_output_path != '/') { |
2671 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
2672 | retval = -1; |
2673 | goto exit_options; | |
994fa64f DG |
2674 | } |
2675 | ||
d85144f4 JG |
2676 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
2677 | -1, -1); | |
4d513a50 DG |
2678 | if (ret < 0) { |
2679 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
2680 | retval = -1; |
2681 | goto exit_options; | |
4d513a50 DG |
2682 | } |
2683 | } | |
2684 | ||
b8aa1682 | 2685 | /* Daemonize */ |
b5218ffb | 2686 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2687 | int i; |
2688 | ||
2689 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2690 | !opt_background); | |
b8aa1682 | 2691 | if (ret < 0) { |
178a0557 MD |
2692 | retval = -1; |
2693 | goto exit_options; | |
b8aa1682 | 2694 | } |
3fd27398 MD |
2695 | |
2696 | /* | |
2697 | * We are in the child. Make sure all other file | |
2698 | * descriptors are closed, in case we are called with | |
2699 | * more opened file descriptors than the standard ones. | |
2700 | */ | |
2701 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2702 | (void) close(i); | |
2703 | } | |
2704 | } | |
2705 | ||
178a0557 MD |
2706 | |
2707 | /* Initialize thread health monitoring */ | |
2708 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2709 | if (!health_relayd) { | |
2710 | PERROR("health_app_create error"); | |
2711 | retval = -1; | |
2712 | goto exit_health_app_create; | |
2713 | } | |
2714 | ||
3fd27398 | 2715 | /* Create thread quit pipe */ |
178a0557 MD |
2716 | if (init_thread_quit_pipe()) { |
2717 | retval = -1; | |
2718 | goto exit_init_data; | |
b8aa1682 JD |
2719 | } |
2720 | ||
b8aa1682 | 2721 | /* Setup the thread apps communication pipe. */ |
178a0557 MD |
2722 | if (create_relay_conn_pipe()) { |
2723 | retval = -1; | |
2724 | goto exit_init_data; | |
b8aa1682 JD |
2725 | } |
2726 | ||
2727 | /* Init relay command queue. */ | |
8bdee6e2 | 2728 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 JD |
2729 | |
2730 | /* Set up max poll set size */ | |
25b397f9 MD |
2731 | if (lttng_poll_set_max_size()) { |
2732 | retval = -1; | |
2733 | goto exit_init_data; | |
2734 | } | |
b8aa1682 | 2735 | |
554831e7 MD |
2736 | /* Initialize communication library */ |
2737 | lttcomm_init(); | |
87e45c13 | 2738 | lttcomm_inet_init(); |
554831e7 | 2739 | |
d3e2ba59 | 2740 | /* tables of sessions indexed by session ID */ |
5a693d30 MD |
2741 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2742 | if (!sessions_ht) { | |
178a0557 MD |
2743 | retval = -1; |
2744 | goto exit_init_data; | |
d3e2ba59 JD |
2745 | } |
2746 | ||
2747 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2748 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 2749 | if (!relay_streams_ht) { |
178a0557 MD |
2750 | retval = -1; |
2751 | goto exit_init_data; | |
d3e2ba59 JD |
2752 | } |
2753 | ||
2754 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2755 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2756 | if (!viewer_streams_ht) { | |
178a0557 MD |
2757 | retval = -1; |
2758 | goto exit_init_data; | |
55706a7d MD |
2759 | } |
2760 | ||
65931c8b | 2761 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
2762 | if (ret) { |
2763 | retval = -1; | |
2764 | goto exit_health_quit_pipe; | |
65931c8b MD |
2765 | } |
2766 | ||
2767 | /* Create thread to manage the client socket */ | |
2768 | ret = pthread_create(&health_thread, NULL, | |
2769 | thread_manage_health, (void *) NULL); | |
178a0557 MD |
2770 | if (ret) { |
2771 | errno = ret; | |
65931c8b | 2772 | PERROR("pthread_create health"); |
178a0557 MD |
2773 | retval = -1; |
2774 | goto exit_health_thread; | |
65931c8b MD |
2775 | } |
2776 | ||
b8aa1682 JD |
2777 | /* Setup the dispatcher thread */ |
2778 | ret = pthread_create(&dispatcher_thread, NULL, | |
2779 | relay_thread_dispatcher, (void *) NULL); | |
178a0557 MD |
2780 | if (ret) { |
2781 | errno = ret; | |
b8aa1682 | 2782 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
2783 | retval = -1; |
2784 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
2785 | } |
2786 | ||
2787 | /* Setup the worker thread */ | |
2788 | ret = pthread_create(&worker_thread, NULL, | |
5a693d30 | 2789 | relay_thread_worker, NULL); |
178a0557 MD |
2790 | if (ret) { |
2791 | errno = ret; | |
b8aa1682 | 2792 | PERROR("pthread_create worker"); |
178a0557 MD |
2793 | retval = -1; |
2794 | goto exit_worker_thread; | |
b8aa1682 JD |
2795 | } |
2796 | ||
2797 | /* Setup the listener thread */ | |
2798 | ret = pthread_create(&listener_thread, NULL, | |
2799 | relay_thread_listener, (void *) NULL); | |
178a0557 MD |
2800 | if (ret) { |
2801 | errno = ret; | |
b8aa1682 | 2802 | PERROR("pthread_create listener"); |
178a0557 MD |
2803 | retval = -1; |
2804 | goto exit_listener_thread; | |
b8aa1682 JD |
2805 | } |
2806 | ||
5a693d30 | 2807 | ret = relayd_live_create(live_uri); |
178a0557 | 2808 | if (ret) { |
d3e2ba59 | 2809 | ERR("Starting live viewer threads"); |
178a0557 | 2810 | retval = -1; |
50138f51 | 2811 | goto exit_live; |
d3e2ba59 JD |
2812 | } |
2813 | ||
178a0557 MD |
2814 | /* |
2815 | * This is where we start awaiting program completion (e.g. through | |
2816 | * signal that asks threads to teardown). | |
2817 | */ | |
2818 | ||
2819 | ret = relayd_live_join(); | |
2820 | if (ret) { | |
2821 | retval = -1; | |
2822 | } | |
50138f51 | 2823 | exit_live: |
178a0557 | 2824 | |
b8aa1682 | 2825 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
2826 | if (ret) { |
2827 | errno = ret; | |
2828 | PERROR("pthread_join listener_thread"); | |
2829 | retval = -1; | |
b8aa1682 JD |
2830 | } |
2831 | ||
178a0557 | 2832 | exit_listener_thread: |
b8aa1682 | 2833 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
2834 | if (ret) { |
2835 | errno = ret; | |
2836 | PERROR("pthread_join worker_thread"); | |
2837 | retval = -1; | |
b8aa1682 JD |
2838 | } |
2839 | ||
178a0557 | 2840 | exit_worker_thread: |
b8aa1682 | 2841 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
2842 | if (ret) { |
2843 | errno = ret; | |
2844 | PERROR("pthread_join dispatcher_thread"); | |
2845 | retval = -1; | |
b8aa1682 | 2846 | } |
178a0557 | 2847 | exit_dispatcher_thread: |
42415026 | 2848 | |
65931c8b | 2849 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
2850 | if (ret) { |
2851 | errno = ret; | |
2852 | PERROR("pthread_join health_thread"); | |
2853 | retval = -1; | |
65931c8b | 2854 | } |
178a0557 | 2855 | exit_health_thread: |
65931c8b | 2856 | |
65931c8b | 2857 | utils_close_pipe(health_quit_pipe); |
178a0557 | 2858 | exit_health_quit_pipe: |
65931c8b | 2859 | |
178a0557 | 2860 | exit_init_data: |
55706a7d | 2861 | health_app_destroy(health_relayd); |
55706a7d | 2862 | exit_health_app_create: |
178a0557 | 2863 | exit_options: |
5a693d30 MD |
2864 | relayd_cleanup(); |
2865 | ||
2866 | /* Ensure all prior call_rcu are done. */ | |
2867 | rcu_barrier(); | |
d3e2ba59 | 2868 | |
178a0557 | 2869 | if (!retval) { |
b8aa1682 | 2870 | exit(EXIT_SUCCESS); |
178a0557 MD |
2871 | } else { |
2872 | exit(EXIT_FAILURE); | |
b8aa1682 | 2873 | } |
b8aa1682 | 2874 | } |