Commit | Line | Data |
---|---|---|
b8aa1682 JD |
1 | /* |
2 | * Copyright (C) 2012 - Julien Desfossez <jdesfossez@efficios.com> | |
3 | * David Goulet <dgoulet@efficios.com> | |
cd60b05a | 4 | * 2013 - Jérémie Galarneau <jeremie.galarneau@efficios.com> |
b8aa1682 JD |
5 | * |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License, version 2 only, | |
8 | * as published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
11 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
12 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
13 | * more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License along | |
16 | * with this program; if not, write to the Free Software Foundation, Inc., | |
17 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
18 | */ | |
19 | ||
20 | #define _GNU_SOURCE | |
21 | #include <getopt.h> | |
22 | #include <grp.h> | |
23 | #include <limits.h> | |
24 | #include <pthread.h> | |
25 | #include <signal.h> | |
26 | #include <stdio.h> | |
27 | #include <stdlib.h> | |
28 | #include <string.h> | |
29 | #include <sys/mman.h> | |
30 | #include <sys/mount.h> | |
31 | #include <sys/resource.h> | |
32 | #include <sys/socket.h> | |
33 | #include <sys/stat.h> | |
34 | #include <sys/types.h> | |
35 | #include <sys/wait.h> | |
173af62f | 36 | #include <inttypes.h> |
b8aa1682 JD |
37 | #include <urcu/futex.h> |
38 | #include <urcu/uatomic.h> | |
39 | #include <unistd.h> | |
40 | #include <fcntl.h> | |
41 | #include <config.h> | |
42 | ||
43 | #include <lttng/lttng.h> | |
44 | #include <common/common.h> | |
45 | #include <common/compat/poll.h> | |
46 | #include <common/compat/socket.h> | |
f263b7fd | 47 | #include <common/compat/endian.h> |
b8aa1682 | 48 | #include <common/defaults.h> |
3fd27398 | 49 | #include <common/daemonize.h> |
b8aa1682 JD |
50 | #include <common/futex.h> |
51 | #include <common/sessiond-comm/sessiond-comm.h> | |
52 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
53 | #include <common/sessiond-comm/relayd.h> |
54 | #include <common/uri.h> | |
a02de639 | 55 | #include <common/utils.h> |
cd60b05a | 56 | #include <common/config/config.h> |
b8aa1682 | 57 | |
0f907de1 | 58 | #include "cmd.h" |
d3e2ba59 | 59 | #include "ctf-trace.h" |
1c20f0e2 | 60 | #include "index.h" |
0f907de1 | 61 | #include "utils.h" |
b8aa1682 | 62 | #include "lttng-relayd.h" |
d3e2ba59 | 63 | #include "live.h" |
55706a7d | 64 | #include "health-relayd.h" |
9b5e0863 | 65 | #include "testpoint.h" |
2f8f53af | 66 | #include "viewer-stream.h" |
2a174661 DG |
67 | #include "session.h" |
68 | #include "stream.h" | |
58eb9381 | 69 | #include "connection.h" |
b8aa1682 JD |
70 | |
71 | /* command line options */ | |
0f907de1 | 72 | char *opt_output_path; |
b5218ffb | 73 | static int opt_daemon, opt_background; |
3fd27398 MD |
74 | |
75 | /* | |
76 | * We need to wait for listener and live listener threads, as well as | |
77 | * health check thread, before being ready to signal readiness. | |
78 | */ | |
79 | #define NR_LTTNG_RELAY_READY 3 | |
80 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
81 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ | |
82 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
83 | ||
095a4ae5 MD |
84 | static struct lttng_uri *control_uri; |
85 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 86 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
87 | |
88 | const char *progname; | |
b8aa1682 | 89 | |
65931c8b | 90 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
91 | static int tracing_group_name_override; |
92 | ||
93 | const char * const config_section_name = "relayd"; | |
65931c8b | 94 | |
b8aa1682 JD |
95 | /* |
96 | * Quit pipe for all threads. This permits a single cancellation point | |
97 | * for all threads when receiving an event on the pipe. | |
98 | */ | |
0b242f62 | 99 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
100 | |
101 | /* | |
102 | * This pipe is used to inform the worker thread that a command is queued and | |
103 | * ready to be processed. | |
104 | */ | |
58eb9381 | 105 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 106 | |
26c9d55e | 107 | /* Shared between threads */ |
b8aa1682 JD |
108 | static int dispatch_thread_exit; |
109 | ||
110 | static pthread_t listener_thread; | |
111 | static pthread_t dispatcher_thread; | |
112 | static pthread_t worker_thread; | |
65931c8b | 113 | static pthread_t health_thread; |
b8aa1682 | 114 | |
095a4ae5 | 115 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
116 | |
117 | /* | |
118 | * Relay command queue. | |
119 | * | |
120 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
121 | * queue. | |
122 | */ | |
58eb9381 | 123 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 JD |
124 | |
125 | /* buffer allocated at startup, used to store the trace data */ | |
095a4ae5 MD |
126 | static char *data_buffer; |
127 | static unsigned int data_buffer_size; | |
b8aa1682 | 128 | |
1c20f0e2 JD |
129 | /* We need those values for the file/dir creation. */ |
130 | static uid_t relayd_uid; | |
131 | static gid_t relayd_gid; | |
132 | ||
d3e2ba59 JD |
133 | /* Global relay stream hash table. */ |
134 | struct lttng_ht *relay_streams_ht; | |
135 | ||
92c6ca54 DG |
136 | /* Global relay viewer stream hash table. */ |
137 | struct lttng_ht *viewer_streams_ht; | |
138 | ||
0a6518b0 DG |
139 | /* Global hash table that stores relay index object. */ |
140 | struct lttng_ht *indexes_ht; | |
141 | ||
55706a7d | 142 | /* Relayd health monitoring */ |
eea7556c | 143 | struct health_app *health_relayd; |
55706a7d | 144 | |
cd60b05a JG |
145 | static struct option long_options[] = { |
146 | { "control-port", 1, 0, 'C', }, | |
147 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 148 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 149 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 150 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
151 | { "group", 1, 0, 'g', }, |
152 | { "help", 0, 0, 'h', }, | |
153 | { "output", 1, 0, 'o', }, | |
154 | { "verbose", 0, 0, 'v', }, | |
155 | { "config", 1, 0, 'f' }, | |
156 | { NULL, 0, 0, 0, }, | |
157 | }; | |
158 | ||
159 | static const char *config_ignore_options[] = { "help", "config" }; | |
160 | ||
b8aa1682 JD |
161 | /* |
162 | * usage function on stderr | |
163 | */ | |
164 | static | |
165 | void usage(void) | |
166 | { | |
167 | fprintf(stderr, "Usage: %s OPTIONS\n\nOptions:\n", progname); | |
994fa64f DG |
168 | fprintf(stderr, " -h, --help Display this usage.\n"); |
169 | fprintf(stderr, " -d, --daemonize Start as a daemon.\n"); | |
b5218ffb | 170 | fprintf(stderr, " -b, --background Start as a daemon, keeping console open.\n"); |
994fa64f DG |
171 | fprintf(stderr, " -C, --control-port URL Control port listening.\n"); |
172 | fprintf(stderr, " -D, --data-port URL Data port listening.\n"); | |
8d5c808e | 173 | fprintf(stderr, " -L, --live-port URL Live view port listening.\n"); |
994fa64f DG |
174 | fprintf(stderr, " -o, --output PATH Output path for traces. Must use an absolute path.\n"); |
175 | fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n"); | |
65931c8b | 176 | fprintf(stderr, " -g, --group NAME Specify the tracing group name. (default: tracing)\n"); |
cd60b05a | 177 | fprintf(stderr, " -f --config Load daemon configuration file\n"); |
b8aa1682 JD |
178 | } |
179 | ||
cd60b05a JG |
180 | /* |
181 | * Take an option from the getopt output and set it in the right variable to be | |
182 | * used later. | |
183 | * | |
184 | * Return 0 on success else a negative value. | |
185 | */ | |
b8aa1682 | 186 | static |
cd60b05a | 187 | int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 188 | { |
cd60b05a JG |
189 | int ret; |
190 | ||
191 | switch (opt) { | |
192 | case 0: | |
193 | fprintf(stderr, "option %s", optname); | |
194 | if (arg) { | |
195 | fprintf(stderr, " with arg %s\n", arg); | |
196 | } | |
197 | break; | |
198 | case 'C': | |
199 | ret = uri_parse(arg, &control_uri); | |
200 | if (ret < 0) { | |
201 | ERR("Invalid control URI specified"); | |
202 | goto end; | |
203 | } | |
204 | if (control_uri->port == 0) { | |
205 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
206 | } | |
207 | break; | |
208 | case 'D': | |
209 | ret = uri_parse(arg, &data_uri); | |
210 | if (ret < 0) { | |
211 | ERR("Invalid data URI specified"); | |
212 | goto end; | |
213 | } | |
214 | if (data_uri->port == 0) { | |
215 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
216 | } | |
217 | break; | |
8d5c808e AM |
218 | case 'L': |
219 | ret = uri_parse(arg, &live_uri); | |
220 | if (ret < 0) { | |
221 | ERR("Invalid live URI specified"); | |
222 | goto end; | |
223 | } | |
224 | if (live_uri->port == 0) { | |
225 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
226 | } | |
227 | break; | |
cd60b05a JG |
228 | case 'd': |
229 | opt_daemon = 1; | |
230 | break; | |
b5218ffb MD |
231 | case 'b': |
232 | opt_background = 1; | |
233 | break; | |
cd60b05a JG |
234 | case 'g': |
235 | tracing_group_name = strdup(arg); | |
236 | tracing_group_name_override = 1; | |
237 | break; | |
238 | case 'h': | |
239 | usage(); | |
240 | exit(EXIT_FAILURE); | |
241 | case 'o': | |
242 | ret = asprintf(&opt_output_path, "%s", arg); | |
243 | if (ret < 0) { | |
244 | ret = -errno; | |
245 | PERROR("asprintf opt_output_path"); | |
246 | goto end; | |
247 | } | |
248 | break; | |
249 | case 'v': | |
250 | /* Verbose level can increase using multiple -v */ | |
251 | if (arg) { | |
252 | lttng_opt_verbose = config_parse_value(arg); | |
253 | } else { | |
254 | lttng_opt_verbose += 1; | |
255 | } | |
256 | break; | |
257 | default: | |
258 | /* Unknown option or other error. | |
259 | * Error is printed by getopt, just return */ | |
260 | ret = -1; | |
261 | goto end; | |
262 | } | |
263 | ||
264 | /* All good. */ | |
265 | ret = 0; | |
266 | ||
267 | end: | |
268 | return ret; | |
269 | } | |
270 | ||
271 | /* | |
272 | * config_entry_handler_cb used to handle options read from a config file. | |
273 | * See config_entry_handler_cb comment in common/config/config.h for the | |
274 | * return value conventions. | |
275 | */ | |
276 | static | |
277 | int config_entry_handler(const struct config_entry *entry, void *unused) | |
278 | { | |
279 | int ret = 0, i; | |
280 | ||
281 | if (!entry || !entry->name || !entry->value) { | |
282 | ret = -EINVAL; | |
283 | goto end; | |
284 | } | |
285 | ||
286 | /* Check if the option is to be ignored */ | |
287 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
288 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
289 | goto end; | |
290 | } | |
291 | } | |
292 | ||
293 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
294 | /* Ignore if entry name is not fully matched. */ | |
295 | if (strcmp(entry->name, long_options[i].name)) { | |
296 | continue; | |
297 | } | |
298 | ||
299 | /* | |
300 | * If the option takes no argument on the command line, we have to | |
301 | * check if the value is "true". We support non-zero numeric values, | |
302 | * true, on and yes. | |
303 | */ | |
304 | if (!long_options[i].has_arg) { | |
305 | ret = config_parse_value(entry->value); | |
306 | if (ret <= 0) { | |
307 | if (ret) { | |
308 | WARN("Invalid configuration value \"%s\" for option %s", | |
309 | entry->value, entry->name); | |
310 | } | |
311 | /* False, skip boolean config option. */ | |
312 | goto end; | |
313 | } | |
314 | } | |
315 | ||
316 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
317 | goto end; | |
318 | } | |
319 | ||
320 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
321 | entry->name); | |
322 | ||
323 | end: | |
324 | return ret; | |
325 | } | |
326 | ||
327 | static | |
328 | int set_options(int argc, char **argv) | |
329 | { | |
330 | int c, ret = 0, option_index = 0; | |
331 | int orig_optopt = optopt, orig_optind = optind; | |
332 | char *default_address, *optstring; | |
333 | const char *config_path = NULL; | |
334 | ||
335 | optstring = utils_generate_optstring(long_options, | |
336 | sizeof(long_options) / sizeof(struct option)); | |
337 | if (!optstring) { | |
338 | ret = -ENOMEM; | |
339 | goto exit; | |
340 | } | |
341 | ||
342 | /* Check for the --config option */ | |
343 | ||
344 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
345 | &option_index)) != -1) { | |
346 | if (c == '?') { | |
347 | ret = -EINVAL; | |
348 | goto exit; | |
349 | } else if (c != 'f') { | |
350 | continue; | |
351 | } | |
352 | ||
353 | config_path = utils_expand_path(optarg); | |
354 | if (!config_path) { | |
355 | ERR("Failed to resolve path: %s", optarg); | |
356 | } | |
357 | } | |
358 | ||
359 | ret = config_get_section_entries(config_path, config_section_name, | |
360 | config_entry_handler, NULL); | |
361 | if (ret) { | |
362 | if (ret > 0) { | |
363 | ERR("Invalid configuration option at line %i", ret); | |
364 | ret = -1; | |
365 | } | |
366 | goto exit; | |
367 | } | |
b8aa1682 | 368 | |
cd60b05a JG |
369 | /* Reset getopt's global state */ |
370 | optopt = orig_optopt; | |
371 | optind = orig_optind; | |
b8aa1682 | 372 | while (1) { |
cd60b05a | 373 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
374 | if (c == -1) { |
375 | break; | |
376 | } | |
377 | ||
cd60b05a JG |
378 | ret = set_option(c, optarg, long_options[option_index].name); |
379 | if (ret < 0) { | |
b8aa1682 JD |
380 | goto exit; |
381 | } | |
382 | } | |
383 | ||
384 | /* assign default values */ | |
385 | if (control_uri == NULL) { | |
fa91dc52 MD |
386 | ret = asprintf(&default_address, |
387 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
388 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
389 | if (ret < 0) { |
390 | PERROR("asprintf default data address"); | |
391 | goto exit; | |
392 | } | |
393 | ||
394 | ret = uri_parse(default_address, &control_uri); | |
395 | free(default_address); | |
396 | if (ret < 0) { | |
397 | ERR("Invalid control URI specified"); | |
398 | goto exit; | |
399 | } | |
400 | } | |
401 | if (data_uri == NULL) { | |
fa91dc52 MD |
402 | ret = asprintf(&default_address, |
403 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
404 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
405 | if (ret < 0) { |
406 | PERROR("asprintf default data address"); | |
407 | goto exit; | |
408 | } | |
409 | ||
410 | ret = uri_parse(default_address, &data_uri); | |
411 | free(default_address); | |
412 | if (ret < 0) { | |
413 | ERR("Invalid data URI specified"); | |
414 | goto exit; | |
415 | } | |
416 | } | |
d3e2ba59 | 417 | if (live_uri == NULL) { |
fa91dc52 MD |
418 | ret = asprintf(&default_address, |
419 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
420 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
421 | if (ret < 0) { |
422 | PERROR("asprintf default viewer control address"); | |
423 | goto exit; | |
424 | } | |
425 | ||
426 | ret = uri_parse(default_address, &live_uri); | |
427 | free(default_address); | |
428 | if (ret < 0) { | |
429 | ERR("Invalid viewer control URI specified"); | |
430 | goto exit; | |
431 | } | |
432 | } | |
b8aa1682 JD |
433 | |
434 | exit: | |
cd60b05a | 435 | free(optstring); |
b8aa1682 JD |
436 | return ret; |
437 | } | |
438 | ||
439 | /* | |
440 | * Cleanup the daemon | |
441 | */ | |
442 | static | |
443 | void cleanup(void) | |
444 | { | |
b8aa1682 JD |
445 | DBG("Cleaning up"); |
446 | ||
095a4ae5 MD |
447 | /* free the dynamically allocated opt_output_path */ |
448 | free(opt_output_path); | |
449 | ||
a02de639 CB |
450 | /* Close thread quit pipes */ |
451 | utils_close_pipe(thread_quit_pipe); | |
452 | ||
710c1f73 DG |
453 | uri_free(control_uri); |
454 | uri_free(data_uri); | |
8d5c808e | 455 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
456 | |
457 | if (tracing_group_name_override) { | |
458 | free((void *) tracing_group_name); | |
459 | } | |
b8aa1682 JD |
460 | } |
461 | ||
462 | /* | |
463 | * Write to writable pipe used to notify a thread. | |
464 | */ | |
465 | static | |
466 | int notify_thread_pipe(int wpipe) | |
467 | { | |
6cd525e8 | 468 | ssize_t ret; |
b8aa1682 | 469 | |
6cd525e8 MD |
470 | ret = lttng_write(wpipe, "!", 1); |
471 | if (ret < 1) { | |
b8aa1682 JD |
472 | PERROR("write poll pipe"); |
473 | } | |
474 | ||
475 | return ret; | |
476 | } | |
477 | ||
65931c8b MD |
478 | static void notify_health_quit_pipe(int *pipe) |
479 | { | |
6cd525e8 | 480 | ssize_t ret; |
65931c8b | 481 | |
6cd525e8 MD |
482 | ret = lttng_write(pipe[1], "4", 1); |
483 | if (ret < 1) { | |
65931c8b MD |
484 | PERROR("write relay health quit"); |
485 | } | |
486 | } | |
487 | ||
b8aa1682 JD |
488 | /* |
489 | * Stop all threads by closing the thread quit pipe. | |
490 | */ | |
491 | static | |
492 | void stop_threads(void) | |
493 | { | |
494 | int ret; | |
495 | ||
496 | /* Stopping all threads */ | |
497 | DBG("Terminating all threads"); | |
498 | ret = notify_thread_pipe(thread_quit_pipe[1]); | |
499 | if (ret < 0) { | |
500 | ERR("write error on thread quit pipe"); | |
501 | } | |
502 | ||
65931c8b MD |
503 | notify_health_quit_pipe(health_quit_pipe); |
504 | ||
b8aa1682 | 505 | /* Dispatch thread */ |
26c9d55e | 506 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 507 | futex_nto1_wake(&relay_conn_queue.futex); |
b8aa1682 JD |
508 | } |
509 | ||
510 | /* | |
511 | * Signal handler for the daemon | |
512 | * | |
513 | * Simply stop all worker threads, leaving main() return gracefully after | |
514 | * joining all threads and calling cleanup(). | |
515 | */ | |
516 | static | |
517 | void sighandler(int sig) | |
518 | { | |
519 | switch (sig) { | |
520 | case SIGPIPE: | |
521 | DBG("SIGPIPE caught"); | |
522 | return; | |
523 | case SIGINT: | |
524 | DBG("SIGINT caught"); | |
525 | stop_threads(); | |
526 | break; | |
527 | case SIGTERM: | |
528 | DBG("SIGTERM caught"); | |
529 | stop_threads(); | |
530 | break; | |
3fd27398 MD |
531 | case SIGUSR1: |
532 | CMM_STORE_SHARED(recv_child_signal, 1); | |
533 | break; | |
b8aa1682 JD |
534 | default: |
535 | break; | |
536 | } | |
537 | } | |
538 | ||
539 | /* | |
540 | * Setup signal handler for : | |
541 | * SIGINT, SIGTERM, SIGPIPE | |
542 | */ | |
543 | static | |
544 | int set_signal_handler(void) | |
545 | { | |
546 | int ret = 0; | |
547 | struct sigaction sa; | |
548 | sigset_t sigset; | |
549 | ||
550 | if ((ret = sigemptyset(&sigset)) < 0) { | |
551 | PERROR("sigemptyset"); | |
552 | return ret; | |
553 | } | |
554 | ||
555 | sa.sa_handler = sighandler; | |
556 | sa.sa_mask = sigset; | |
557 | sa.sa_flags = 0; | |
558 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
559 | PERROR("sigaction"); | |
560 | return ret; | |
561 | } | |
562 | ||
563 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
564 | PERROR("sigaction"); | |
565 | return ret; | |
566 | } | |
567 | ||
568 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
569 | PERROR("sigaction"); | |
570 | return ret; | |
571 | } | |
572 | ||
3fd27398 MD |
573 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
574 | PERROR("sigaction"); | |
575 | return ret; | |
576 | } | |
577 | ||
578 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
579 | |
580 | return ret; | |
581 | } | |
582 | ||
3fd27398 MD |
583 | void lttng_relay_notify_ready(void) |
584 | { | |
585 | /* Notify the parent of the fork() process that we are ready. */ | |
586 | if (opt_daemon || opt_background) { | |
587 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
588 | kill(child_ppid, SIGUSR1); | |
589 | } | |
590 | } | |
591 | } | |
592 | ||
b8aa1682 JD |
593 | /* |
594 | * Init thread quit pipe. | |
595 | * | |
596 | * Return -1 on error or 0 if all pipes are created. | |
597 | */ | |
598 | static | |
599 | int init_thread_quit_pipe(void) | |
600 | { | |
a02de639 | 601 | int ret; |
b8aa1682 | 602 | |
a02de639 | 603 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 604 | |
b8aa1682 JD |
605 | return ret; |
606 | } | |
607 | ||
608 | /* | |
609 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
610 | */ | |
611 | static | |
612 | int create_thread_poll_set(struct lttng_poll_event *events, int size) | |
613 | { | |
614 | int ret; | |
615 | ||
616 | if (events == NULL || size == 0) { | |
617 | ret = -1; | |
618 | goto error; | |
619 | } | |
620 | ||
621 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
622 | if (ret < 0) { | |
623 | goto error; | |
624 | } | |
625 | ||
626 | /* Add quit pipe */ | |
c7759e6a | 627 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
628 | if (ret < 0) { |
629 | goto error; | |
630 | } | |
631 | ||
632 | return 0; | |
633 | ||
634 | error: | |
635 | return ret; | |
636 | } | |
637 | ||
638 | /* | |
639 | * Check if the thread quit pipe was triggered. | |
640 | * | |
641 | * Return 1 if it was triggered else 0; | |
642 | */ | |
643 | static | |
644 | int check_thread_quit_pipe(int fd, uint32_t events) | |
645 | { | |
646 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
647 | return 1; | |
648 | } | |
649 | ||
650 | return 0; | |
651 | } | |
652 | ||
653 | /* | |
654 | * Create and init socket from uri. | |
655 | */ | |
656 | static | |
657 | struct lttcomm_sock *relay_init_sock(struct lttng_uri *uri) | |
658 | { | |
659 | int ret; | |
660 | struct lttcomm_sock *sock = NULL; | |
661 | ||
662 | sock = lttcomm_alloc_sock_from_uri(uri); | |
663 | if (sock == NULL) { | |
664 | ERR("Allocating socket"); | |
665 | goto error; | |
666 | } | |
667 | ||
668 | ret = lttcomm_create_sock(sock); | |
669 | if (ret < 0) { | |
670 | goto error; | |
671 | } | |
672 | DBG("Listening on sock %d", sock->fd); | |
673 | ||
674 | ret = sock->ops->bind(sock); | |
675 | if (ret < 0) { | |
676 | goto error; | |
677 | } | |
678 | ||
679 | ret = sock->ops->listen(sock, -1); | |
680 | if (ret < 0) { | |
681 | goto error; | |
682 | ||
683 | } | |
684 | ||
685 | return sock; | |
686 | ||
687 | error: | |
688 | if (sock) { | |
689 | lttcomm_destroy_sock(sock); | |
690 | } | |
691 | return NULL; | |
692 | } | |
693 | ||
173af62f DG |
694 | /* |
695 | * Return nonzero if stream needs to be closed. | |
696 | */ | |
697 | static | |
698 | int close_stream_check(struct relay_stream *stream) | |
699 | { | |
173af62f | 700 | if (stream->close_flag && stream->prev_seq == stream->last_net_seq_num) { |
f7079f67 DG |
701 | /* |
702 | * We are about to close the stream so set the data pending flag to 1 | |
703 | * which will make the end data pending command skip the stream which | |
704 | * is now closed and ready. Note that after proceeding to a file close, | |
705 | * the written file is ready for reading. | |
706 | */ | |
707 | stream->data_pending_check_done = 1; | |
173af62f DG |
708 | return 1; |
709 | } | |
710 | return 0; | |
711 | } | |
712 | ||
2a174661 DG |
713 | static void try_close_stream(struct relay_session *session, |
714 | struct relay_stream *stream) | |
715 | { | |
716 | int ret; | |
717 | struct ctf_trace *ctf_trace; | |
718 | ||
719 | assert(session); | |
720 | assert(stream); | |
721 | ||
722 | if (!close_stream_check(stream)) { | |
723 | /* Can't close it, not ready for that. */ | |
724 | goto end; | |
725 | } | |
726 | ||
727 | ctf_trace = ctf_trace_find_by_path(session->ctf_traces_ht, | |
728 | stream->path_name); | |
729 | assert(ctf_trace); | |
730 | ||
731 | pthread_mutex_lock(&session->viewer_ready_lock); | |
732 | ctf_trace->invalid_flag = 1; | |
733 | pthread_mutex_unlock(&session->viewer_ready_lock); | |
734 | ||
735 | ret = stream_close(session, stream); | |
f94b19e6 | 736 | if (ret || session->snapshot) { |
2a174661 DG |
737 | /* Already close thus the ctf trace is being or has been destroyed. */ |
738 | goto end; | |
739 | } | |
740 | ||
741 | ctf_trace_try_destroy(session, ctf_trace); | |
742 | ||
743 | end: | |
744 | return; | |
745 | } | |
746 | ||
b8aa1682 JD |
747 | /* |
748 | * This thread manages the listening for new connections on the network | |
749 | */ | |
750 | static | |
751 | void *relay_thread_listener(void *data) | |
752 | { | |
095a4ae5 | 753 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
754 | uint32_t revents, nb_fd; |
755 | struct lttng_poll_event events; | |
756 | struct lttcomm_sock *control_sock, *data_sock; | |
757 | ||
b8aa1682 JD |
758 | DBG("[thread] Relay listener started"); |
759 | ||
55706a7d MD |
760 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
761 | ||
f385ae0a MD |
762 | health_code_update(); |
763 | ||
b8aa1682 JD |
764 | control_sock = relay_init_sock(control_uri); |
765 | if (!control_sock) { | |
095a4ae5 | 766 | goto error_sock_control; |
b8aa1682 JD |
767 | } |
768 | ||
769 | data_sock = relay_init_sock(data_uri); | |
770 | if (!data_sock) { | |
095a4ae5 | 771 | goto error_sock_relay; |
b8aa1682 JD |
772 | } |
773 | ||
774 | /* | |
775 | * Pass 3 as size here for the thread quit pipe, control and data socket. | |
776 | */ | |
777 | ret = create_thread_poll_set(&events, 3); | |
778 | if (ret < 0) { | |
779 | goto error_create_poll; | |
780 | } | |
781 | ||
782 | /* Add the control socket */ | |
783 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
784 | if (ret < 0) { | |
785 | goto error_poll_add; | |
786 | } | |
787 | ||
788 | /* Add the data socket */ | |
789 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
790 | if (ret < 0) { | |
791 | goto error_poll_add; | |
792 | } | |
793 | ||
3fd27398 MD |
794 | lttng_relay_notify_ready(); |
795 | ||
9b5e0863 MD |
796 | if (testpoint(relayd_thread_listener)) { |
797 | goto error_testpoint; | |
798 | } | |
799 | ||
b8aa1682 | 800 | while (1) { |
f385ae0a MD |
801 | health_code_update(); |
802 | ||
b8aa1682 JD |
803 | DBG("Listener accepting connections"); |
804 | ||
b8aa1682 | 805 | restart: |
f385ae0a | 806 | health_poll_entry(); |
b8aa1682 | 807 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 808 | health_poll_exit(); |
b8aa1682 JD |
809 | if (ret < 0) { |
810 | /* | |
811 | * Restart interrupted system call. | |
812 | */ | |
813 | if (errno == EINTR) { | |
814 | goto restart; | |
815 | } | |
816 | goto error; | |
817 | } | |
818 | ||
0d9c5d77 DG |
819 | nb_fd = ret; |
820 | ||
b8aa1682 JD |
821 | DBG("Relay new connection received"); |
822 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
823 | health_code_update(); |
824 | ||
b8aa1682 JD |
825 | /* Fetch once the poll data */ |
826 | revents = LTTNG_POLL_GETEV(&events, i); | |
827 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
828 | ||
829 | /* Thread quit pipe has been closed. Killing thread. */ | |
830 | ret = check_thread_quit_pipe(pollfd, revents); | |
831 | if (ret) { | |
095a4ae5 MD |
832 | err = 0; |
833 | goto exit; | |
b8aa1682 JD |
834 | } |
835 | ||
836 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
837 | ERR("socket poll error"); | |
838 | goto error; | |
839 | } else if (revents & LPOLLIN) { | |
4b7f17b2 | 840 | /* |
58eb9381 DG |
841 | * Get allocated in this thread, enqueued to a global queue, |
842 | * dequeued and freed in the worker thread. | |
4b7f17b2 | 843 | */ |
58eb9381 DG |
844 | int val = 1; |
845 | struct relay_connection *new_conn; | |
4b7f17b2 | 846 | struct lttcomm_sock *newsock; |
b8aa1682 | 847 | |
58eb9381 DG |
848 | new_conn = connection_create(); |
849 | if (!new_conn) { | |
b8aa1682 JD |
850 | goto error; |
851 | } | |
852 | ||
853 | if (pollfd == data_sock->fd) { | |
58eb9381 | 854 | new_conn->type = RELAY_DATA; |
b8aa1682 | 855 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
856 | DBG("Relay data connection accepted, socket %d", |
857 | newsock->fd); | |
4b7f17b2 MD |
858 | } else { |
859 | assert(pollfd == control_sock->fd); | |
58eb9381 | 860 | new_conn->type = RELAY_CONTROL; |
b8aa1682 | 861 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
862 | DBG("Relay control connection accepted, socket %d", |
863 | newsock->fd); | |
b8aa1682 | 864 | } |
58eb9381 DG |
865 | if (!newsock) { |
866 | PERROR("accepting sock"); | |
867 | connection_free(new_conn); | |
868 | goto error; | |
869 | } | |
870 | ||
871 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
872 | sizeof(val)); | |
b8aa1682 JD |
873 | if (ret < 0) { |
874 | PERROR("setsockopt inet"); | |
4b7f17b2 | 875 | lttcomm_destroy_sock(newsock); |
58eb9381 | 876 | connection_free(new_conn); |
b8aa1682 JD |
877 | goto error; |
878 | } | |
58eb9381 DG |
879 | new_conn->sock = newsock; |
880 | ||
881 | /* Enqueue request for the dispatcher thread. */ | |
882 | cds_wfq_enqueue(&relay_conn_queue.queue, &new_conn->qnode); | |
b8aa1682 JD |
883 | |
884 | /* | |
58eb9381 DG |
885 | * Wake the dispatch queue futex. Implicit memory barrier with |
886 | * the exchange in cds_wfq_enqueue. | |
b8aa1682 | 887 | */ |
58eb9381 | 888 | futex_nto1_wake(&relay_conn_queue.futex); |
b8aa1682 JD |
889 | } |
890 | } | |
891 | } | |
892 | ||
095a4ae5 | 893 | exit: |
b8aa1682 JD |
894 | error: |
895 | error_poll_add: | |
9b5e0863 | 896 | error_testpoint: |
b8aa1682 JD |
897 | lttng_poll_clean(&events); |
898 | error_create_poll: | |
095a4ae5 MD |
899 | if (data_sock->fd >= 0) { |
900 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
901 | if (ret) { |
902 | PERROR("close"); | |
903 | } | |
b8aa1682 | 904 | } |
095a4ae5 MD |
905 | lttcomm_destroy_sock(data_sock); |
906 | error_sock_relay: | |
907 | if (control_sock->fd >= 0) { | |
908 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
909 | if (ret) { |
910 | PERROR("close"); | |
911 | } | |
b8aa1682 | 912 | } |
095a4ae5 MD |
913 | lttcomm_destroy_sock(control_sock); |
914 | error_sock_control: | |
915 | if (err) { | |
f385ae0a MD |
916 | health_error(); |
917 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 918 | } |
55706a7d | 919 | health_unregister(health_relayd); |
b8aa1682 JD |
920 | DBG("Relay listener thread cleanup complete"); |
921 | stop_threads(); | |
b8aa1682 JD |
922 | return NULL; |
923 | } | |
924 | ||
925 | /* | |
926 | * This thread manages the dispatching of the requests to worker threads | |
927 | */ | |
928 | static | |
929 | void *relay_thread_dispatcher(void *data) | |
930 | { | |
6cd525e8 MD |
931 | int err = -1; |
932 | ssize_t ret; | |
b8aa1682 | 933 | struct cds_wfq_node *node; |
58eb9381 | 934 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
935 | |
936 | DBG("[thread] Relay dispatcher started"); | |
937 | ||
55706a7d MD |
938 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
939 | ||
9b5e0863 MD |
940 | if (testpoint(relayd_thread_dispatcher)) { |
941 | goto error_testpoint; | |
942 | } | |
943 | ||
f385ae0a MD |
944 | health_code_update(); |
945 | ||
26c9d55e | 946 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
947 | health_code_update(); |
948 | ||
b8aa1682 | 949 | /* Atomically prepare the queue futex */ |
58eb9381 | 950 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
951 | |
952 | do { | |
f385ae0a MD |
953 | health_code_update(); |
954 | ||
b8aa1682 | 955 | /* Dequeue commands */ |
58eb9381 | 956 | node = cds_wfq_dequeue_blocking(&relay_conn_queue.queue); |
b8aa1682 JD |
957 | if (node == NULL) { |
958 | DBG("Woken up but nothing in the relay command queue"); | |
959 | /* Continue thread execution */ | |
960 | break; | |
961 | } | |
58eb9381 | 962 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 963 | |
58eb9381 | 964 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
965 | |
966 | /* | |
58eb9381 DG |
967 | * Inform worker thread of the new request. This call is blocking |
968 | * so we can be assured that the data will be read at some point in | |
969 | * time or wait to the end of the world :) | |
b8aa1682 | 970 | */ |
58eb9381 DG |
971 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
972 | if (ret < 0) { | |
973 | PERROR("write connection pipe"); | |
974 | connection_destroy(new_conn); | |
b8aa1682 JD |
975 | goto error; |
976 | } | |
977 | } while (node != NULL); | |
978 | ||
979 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 980 | health_poll_entry(); |
58eb9381 | 981 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 982 | health_poll_exit(); |
b8aa1682 JD |
983 | } |
984 | ||
f385ae0a MD |
985 | /* Normal exit, no error */ |
986 | err = 0; | |
987 | ||
b8aa1682 | 988 | error: |
9b5e0863 | 989 | error_testpoint: |
f385ae0a MD |
990 | if (err) { |
991 | health_error(); | |
992 | ERR("Health error occurred in %s", __func__); | |
993 | } | |
55706a7d | 994 | health_unregister(health_relayd); |
b8aa1682 JD |
995 | DBG("Dispatch thread dying"); |
996 | stop_threads(); | |
997 | return NULL; | |
998 | } | |
999 | ||
2a174661 | 1000 | static void try_close_streams(struct relay_session *session) |
d3e2ba59 | 1001 | { |
2a174661 | 1002 | struct ctf_trace *ctf_trace; |
94d49140 JD |
1003 | struct lttng_ht_iter iter; |
1004 | ||
2a174661 | 1005 | assert(session); |
94d49140 | 1006 | |
2a174661 DG |
1007 | pthread_mutex_lock(&session->viewer_ready_lock); |
1008 | rcu_read_lock(); | |
1009 | cds_lfht_for_each_entry(session->ctf_traces_ht->ht, &iter.iter, ctf_trace, | |
1010 | node.node) { | |
1011 | struct relay_stream *stream; | |
94d49140 | 1012 | |
2a174661 DG |
1013 | /* Close streams. */ |
1014 | cds_list_for_each_entry(stream, &ctf_trace->stream_list, trace_list) { | |
1015 | stream_close(session, stream); | |
94d49140 | 1016 | } |
94d49140 | 1017 | |
2a174661 DG |
1018 | ctf_trace->invalid_flag = 1; |
1019 | ctf_trace_try_destroy(session, ctf_trace); | |
157df586 | 1020 | } |
2a174661 DG |
1021 | rcu_read_unlock(); |
1022 | pthread_mutex_unlock(&session->viewer_ready_lock); | |
94d49140 JD |
1023 | } |
1024 | ||
b8aa1682 | 1025 | /* |
2a174661 | 1026 | * Try to destroy a session within a connection. |
b8aa1682 | 1027 | */ |
58eb9381 | 1028 | static void destroy_session(struct relay_session *session, |
d3e2ba59 | 1029 | struct lttng_ht *sessions_ht) |
b8aa1682 | 1030 | { |
58eb9381 | 1031 | assert(session); |
2a174661 | 1032 | assert(sessions_ht); |
b8aa1682 | 1033 | |
2a174661 | 1034 | /* Indicate that this session can be destroyed from now on. */ |
58eb9381 | 1035 | session->close_flag = 1; |
b8aa1682 | 1036 | |
58eb9381 | 1037 | try_close_streams(session); |
5b6d8097 | 1038 | |
2a174661 DG |
1039 | /* |
1040 | * This will try to delete and destroy the session if no viewer is attached | |
1041 | * to it meaning the refcount is down to zero. | |
1042 | */ | |
58eb9381 | 1043 | session_try_destroy(sessions_ht, session); |
b8aa1682 JD |
1044 | } |
1045 | ||
1c20f0e2 JD |
1046 | /* |
1047 | * Copy index data from the control port to a given index object. | |
1048 | */ | |
1049 | static void copy_index_control_data(struct relay_index *index, | |
1050 | struct lttcomm_relayd_index *data) | |
1051 | { | |
1052 | assert(index); | |
1053 | assert(data); | |
1054 | ||
1055 | /* | |
1056 | * The index on disk is encoded in big endian, so we don't need to convert | |
1057 | * the data received on the network. The data_offset value is NEVER | |
1058 | * modified here and is updated by the data thread. | |
1059 | */ | |
1060 | index->index_data.packet_size = data->packet_size; | |
1061 | index->index_data.content_size = data->content_size; | |
1062 | index->index_data.timestamp_begin = data->timestamp_begin; | |
1063 | index->index_data.timestamp_end = data->timestamp_end; | |
1064 | index->index_data.events_discarded = data->events_discarded; | |
1065 | index->index_data.stream_id = data->stream_id; | |
1066 | } | |
1067 | ||
c5b6f4f0 DG |
1068 | /* |
1069 | * Handle the RELAYD_CREATE_SESSION command. | |
1070 | * | |
1071 | * On success, send back the session id or else return a negative value. | |
1072 | */ | |
1073 | static | |
1074 | int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1075 | struct relay_connection *conn) |
c5b6f4f0 DG |
1076 | { |
1077 | int ret = 0, send_ret; | |
1078 | struct relay_session *session; | |
1079 | struct lttcomm_relayd_status_session reply; | |
1080 | ||
1081 | assert(recv_hdr); | |
58eb9381 | 1082 | assert(conn); |
c5b6f4f0 DG |
1083 | |
1084 | memset(&reply, 0, sizeof(reply)); | |
1085 | ||
2a174661 DG |
1086 | session = session_create(); |
1087 | if (!session) { | |
c5b6f4f0 DG |
1088 | ret = -1; |
1089 | goto error; | |
1090 | } | |
58eb9381 DG |
1091 | session->minor = conn->minor; |
1092 | session->major = conn->major; | |
1093 | conn->session_id = session->id; | |
1094 | conn->session = session; | |
c5b6f4f0 DG |
1095 | |
1096 | reply.session_id = htobe64(session->id); | |
1097 | ||
58eb9381 | 1098 | switch (conn->minor) { |
2a174661 DG |
1099 | case 1: |
1100 | case 2: | |
1101 | case 3: | |
1102 | break; | |
1103 | case 4: /* LTTng sessiond 2.4 */ | |
1104 | default: | |
58eb9381 | 1105 | ret = cmd_create_session_2_4(conn, session); |
d3e2ba59 JD |
1106 | } |
1107 | ||
58eb9381 | 1108 | lttng_ht_add_unique_u64(conn->sessions_ht, &session->session_n); |
c5b6f4f0 DG |
1109 | DBG("Created session %" PRIu64, session->id); |
1110 | ||
1111 | error: | |
1112 | if (ret < 0) { | |
1113 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1114 | } else { | |
1115 | reply.ret_code = htobe32(LTTNG_OK); | |
1116 | } | |
1117 | ||
58eb9381 | 1118 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1119 | if (send_ret < 0) { |
1120 | ERR("Relayd sending session id"); | |
4169f5ad | 1121 | ret = send_ret; |
c5b6f4f0 DG |
1122 | } |
1123 | ||
1124 | return ret; | |
1125 | } | |
1126 | ||
a4baae1b JD |
1127 | /* |
1128 | * When we have received all the streams and the metadata for a channel, | |
1129 | * we make them visible to the viewer threads. | |
1130 | */ | |
1131 | static | |
58eb9381 | 1132 | void set_viewer_ready_flag(struct relay_connection *conn) |
a4baae1b | 1133 | { |
2a174661 | 1134 | struct relay_stream *stream, *tmp_stream; |
a4baae1b | 1135 | |
58eb9381 DG |
1136 | pthread_mutex_lock(&conn->session->viewer_ready_lock); |
1137 | cds_list_for_each_entry_safe(stream, tmp_stream, &conn->recv_head, | |
2a174661 | 1138 | recv_list) { |
a4baae1b | 1139 | stream->viewer_ready = 1; |
2a174661 | 1140 | cds_list_del(&stream->recv_list); |
a4baae1b | 1141 | } |
58eb9381 | 1142 | pthread_mutex_unlock(&conn->session->viewer_ready_lock); |
a4baae1b JD |
1143 | return; |
1144 | } | |
1145 | ||
1146 | /* | |
1147 | * Add a recv handle node to the connection recv list with the given stream | |
1148 | * handle. A new node is allocated thus must be freed when the node is deleted | |
1149 | * from the list. | |
1150 | */ | |
58eb9381 DG |
1151 | static void queue_stream(struct relay_stream *stream, |
1152 | struct relay_connection *conn) | |
a4baae1b | 1153 | { |
58eb9381 | 1154 | assert(conn); |
2a174661 | 1155 | assert(stream); |
a4baae1b | 1156 | |
58eb9381 | 1157 | cds_list_add(&stream->recv_list, &conn->recv_head); |
a4baae1b JD |
1158 | } |
1159 | ||
b8aa1682 JD |
1160 | /* |
1161 | * relay_add_stream: allocate a new stream for a session | |
1162 | */ | |
1163 | static | |
1164 | int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1165 | struct relay_connection *conn) |
b8aa1682 | 1166 | { |
2a174661 | 1167 | int ret, 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; | |
2a174661 | 1171 | struct ctf_trace *trace; |
b8aa1682 | 1172 | |
58eb9381 | 1173 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1174 | ERR("Trying to add a stream before version check"); |
1175 | ret = -1; | |
1176 | goto end_no_session; | |
1177 | } | |
1178 | ||
b8aa1682 JD |
1179 | stream = zmalloc(sizeof(struct relay_stream)); |
1180 | if (stream == NULL) { | |
1181 | PERROR("relay stream zmalloc"); | |
1182 | ret = -1; | |
1183 | goto end_no_session; | |
1184 | } | |
1185 | ||
58eb9381 | 1186 | switch (conn->minor) { |
0f907de1 | 1187 | case 1: /* LTTng sessiond 2.1 */ |
58eb9381 | 1188 | ret = cmd_recv_stream_2_1(conn, stream); |
0f907de1 JD |
1189 | break; |
1190 | case 2: /* LTTng sessiond 2.2 */ | |
1191 | default: | |
58eb9381 | 1192 | ret = cmd_recv_stream_2_2(conn, stream); |
0f907de1 JD |
1193 | break; |
1194 | } | |
1195 | if (ret < 0) { | |
1196 | goto err_free_stream; | |
1197 | } | |
1198 | ||
9d1bbf21 | 1199 | rcu_read_lock(); |
b8aa1682 | 1200 | stream->stream_handle = ++last_relay_stream_id; |
173af62f | 1201 | stream->prev_seq = -1ULL; |
2a174661 | 1202 | stream->session_id = session->id; |
1c20f0e2 | 1203 | stream->index_fd = -1; |
d3e2ba59 | 1204 | stream->read_index_fd = -1; |
2a174661 | 1205 | lttng_ht_node_init_u64(&stream->node, stream->stream_handle); |
d3e2ba59 | 1206 | pthread_mutex_init(&stream->lock, NULL); |
b8aa1682 | 1207 | |
0f907de1 | 1208 | ret = utils_mkdir_recursive(stream->path_name, S_IRWXU | S_IRWXG); |
b8aa1682 | 1209 | if (ret < 0) { |
b8aa1682 JD |
1210 | ERR("relay creating output directory"); |
1211 | goto end; | |
1212 | } | |
1213 | ||
be96a7d1 DG |
1214 | /* |
1215 | * No need to use run_as API here because whatever we receives, the relayd | |
1216 | * uses its own credentials for the stream files. | |
1217 | */ | |
0f907de1 | 1218 | ret = utils_create_stream_file(stream->path_name, stream->channel_name, |
1c20f0e2 | 1219 | stream->tracefile_size, 0, relayd_uid, relayd_gid, NULL); |
b8aa1682 | 1220 | if (ret < 0) { |
0f907de1 | 1221 | ERR("Create output file"); |
b8aa1682 JD |
1222 | goto end; |
1223 | } | |
b8aa1682 | 1224 | stream->fd = ret; |
0f907de1 JD |
1225 | if (stream->tracefile_size) { |
1226 | DBG("Tracefile %s/%s_0 created", stream->path_name, stream->channel_name); | |
1227 | } else { | |
1228 | DBG("Tracefile %s/%s created", stream->path_name, stream->channel_name); | |
1229 | } | |
b8aa1682 | 1230 | |
2a174661 DG |
1231 | trace = ctf_trace_find_by_path(session->ctf_traces_ht, stream->path_name); |
1232 | if (!trace) { | |
1233 | trace = ctf_trace_create(stream->path_name); | |
1234 | if (!trace) { | |
d3e2ba59 JD |
1235 | ret = -1; |
1236 | goto end; | |
1237 | } | |
2a174661 DG |
1238 | ctf_trace_add(session->ctf_traces_ht, trace); |
1239 | } | |
1240 | ctf_trace_get_ref(trace); | |
1241 | ||
1242 | if (!strncmp(stream->channel_name, DEFAULT_METADATA_NAME, NAME_MAX)) { | |
1243 | stream->metadata_flag = 1; | |
1244 | /* Assign quick reference to the metadata stream in the trace. */ | |
1245 | trace->metadata_stream = stream; | |
d3e2ba59 | 1246 | } |
d3e2ba59 | 1247 | |
a4baae1b | 1248 | /* |
2a174661 DG |
1249 | * Add the stream in the recv list of the connection. Once the end stream |
1250 | * message is received, this list is emptied and streams are set with the | |
1251 | * viewer ready flag. | |
a4baae1b | 1252 | */ |
58eb9381 | 1253 | queue_stream(stream, conn); |
a4baae1b | 1254 | |
2a174661 DG |
1255 | /* |
1256 | * Both in the ctf_trace object and the global stream ht since the data | |
1257 | * side of the relayd does not have the concept of session. | |
1258 | */ | |
1259 | lttng_ht_add_unique_u64(relay_streams_ht, &stream->node); | |
1260 | cds_list_add_tail(&stream->trace_list, &trace->stream_list); | |
b8aa1682 | 1261 | |
87b576ec | 1262 | session->stream_count++; |
d3e2ba59 | 1263 | |
1c20f0e2 JD |
1264 | DBG("Relay new stream added %s with ID %" PRIu64, stream->channel_name, |
1265 | stream->stream_handle); | |
b8aa1682 JD |
1266 | |
1267 | end: | |
53efb85a | 1268 | memset(&reply, 0, sizeof(reply)); |
5af40280 | 1269 | reply.handle = htobe64(stream->stream_handle); |
b8aa1682 JD |
1270 | /* send the session id to the client or a negative return code on error */ |
1271 | if (ret < 0) { | |
f73fabfd | 1272 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
5af40280 CB |
1273 | /* stream was not properly added to the ht, so free it */ |
1274 | free(stream); | |
b8aa1682 | 1275 | } else { |
f73fabfd | 1276 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1277 | } |
5af40280 | 1278 | |
58eb9381 | 1279 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1280 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1281 | if (send_ret < 0) { | |
1282 | ERR("Relay sending stream id"); | |
4169f5ad | 1283 | ret = send_ret; |
b8aa1682 | 1284 | } |
9d1bbf21 | 1285 | rcu_read_unlock(); |
b8aa1682 JD |
1286 | |
1287 | end_no_session: | |
1288 | return ret; | |
0f907de1 JD |
1289 | |
1290 | err_free_stream: | |
1291 | free(stream->path_name); | |
1292 | free(stream->channel_name); | |
1293 | free(stream); | |
1294 | return ret; | |
b8aa1682 JD |
1295 | } |
1296 | ||
173af62f DG |
1297 | /* |
1298 | * relay_close_stream: close a specific stream | |
1299 | */ | |
1300 | static | |
1301 | int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1302 | struct relay_connection *conn) |
173af62f | 1303 | { |
94d49140 | 1304 | int ret, send_ret; |
58eb9381 | 1305 | struct relay_session *session = conn->session; |
173af62f DG |
1306 | struct lttcomm_relayd_close_stream stream_info; |
1307 | struct lttcomm_relayd_generic_reply reply; | |
1308 | struct relay_stream *stream; | |
173af62f DG |
1309 | |
1310 | DBG("Close stream received"); | |
1311 | ||
58eb9381 | 1312 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1313 | ERR("Trying to close a stream before version check"); |
1314 | ret = -1; | |
1315 | goto end_no_session; | |
1316 | } | |
1317 | ||
58eb9381 | 1318 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1319 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1320 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1321 | if (ret == 0) { |
1322 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1323 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1324 | } else { |
1325 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1326 | } | |
173af62f DG |
1327 | ret = -1; |
1328 | goto end_no_session; | |
1329 | } | |
1330 | ||
1331 | rcu_read_lock(); | |
2a174661 DG |
1332 | stream = stream_find_by_id(relay_streams_ht, |
1333 | be64toh(stream_info.stream_id)); | |
173af62f DG |
1334 | if (!stream) { |
1335 | ret = -1; | |
1336 | goto end_unlock; | |
1337 | } | |
1338 | ||
8e2583a4 | 1339 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
173af62f | 1340 | stream->close_flag = 1; |
87b576ec JD |
1341 | session->stream_count--; |
1342 | assert(session->stream_count >= 0); | |
173af62f | 1343 | |
2a174661 DG |
1344 | /* Check if we can close it or else the data will do it. */ |
1345 | try_close_stream(session, stream); | |
173af62f DG |
1346 | |
1347 | end_unlock: | |
1348 | rcu_read_unlock(); | |
1349 | ||
53efb85a | 1350 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1351 | if (ret < 0) { |
f73fabfd | 1352 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1353 | } else { |
f73fabfd | 1354 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1355 | } |
58eb9381 | 1356 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1357 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1358 | if (send_ret < 0) { | |
1359 | ERR("Relay sending stream id"); | |
4169f5ad | 1360 | ret = send_ret; |
173af62f DG |
1361 | } |
1362 | ||
1363 | end_no_session: | |
1364 | return ret; | |
1365 | } | |
1366 | ||
b8aa1682 JD |
1367 | /* |
1368 | * relay_unknown_command: send -1 if received unknown command | |
1369 | */ | |
1370 | static | |
58eb9381 | 1371 | void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1372 | { |
1373 | struct lttcomm_relayd_generic_reply reply; | |
1374 | int ret; | |
1375 | ||
53efb85a | 1376 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1377 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
58eb9381 | 1378 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1379 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1380 | if (ret < 0) { | |
1381 | ERR("Relay sending unknown command"); | |
1382 | } | |
1383 | } | |
1384 | ||
1385 | /* | |
1386 | * relay_start: send an acknowledgment to the client to tell if we are | |
1387 | * ready to receive data. We are ready if a session is established. | |
1388 | */ | |
1389 | static | |
1390 | int relay_start(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1391 | struct relay_connection *conn) |
b8aa1682 | 1392 | { |
f73fabfd | 1393 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1394 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1395 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1396 | |
1397 | if (!session) { | |
1398 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1399 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1400 | } |
1401 | ||
53efb85a | 1402 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1403 | reply.ret_code = ret; |
58eb9381 | 1404 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1405 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1406 | if (ret < 0) { | |
1407 | ERR("Relay sending start ack"); | |
1408 | } | |
1409 | ||
1410 | return ret; | |
1411 | } | |
1412 | ||
1d4dfdef DG |
1413 | /* |
1414 | * Append padding to the file pointed by the file descriptor fd. | |
1415 | */ | |
1416 | static int write_padding_to_file(int fd, uint32_t size) | |
1417 | { | |
6cd525e8 | 1418 | ssize_t ret = 0; |
1d4dfdef DG |
1419 | char *zeros; |
1420 | ||
1421 | if (size == 0) { | |
1422 | goto end; | |
1423 | } | |
1424 | ||
1425 | zeros = zmalloc(size); | |
1426 | if (zeros == NULL) { | |
1427 | PERROR("zmalloc zeros for padding"); | |
1428 | ret = -1; | |
1429 | goto end; | |
1430 | } | |
1431 | ||
6cd525e8 MD |
1432 | ret = lttng_write(fd, zeros, size); |
1433 | if (ret < size) { | |
1d4dfdef DG |
1434 | PERROR("write padding to file"); |
1435 | } | |
1436 | ||
e986c7a1 DG |
1437 | free(zeros); |
1438 | ||
1d4dfdef DG |
1439 | end: |
1440 | return ret; | |
1441 | } | |
1442 | ||
b8aa1682 JD |
1443 | /* |
1444 | * relay_recv_metadata: receive the metada for the session. | |
1445 | */ | |
1446 | static | |
1447 | int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1448 | struct relay_connection *conn) |
b8aa1682 | 1449 | { |
f73fabfd | 1450 | int ret = htobe32(LTTNG_OK); |
6cd525e8 | 1451 | ssize_t size_ret; |
58eb9381 | 1452 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1453 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1454 | struct relay_stream *metadata_stream; | |
1455 | uint64_t data_size, payload_size; | |
2a174661 | 1456 | struct ctf_trace *ctf_trace; |
b8aa1682 JD |
1457 | |
1458 | if (!session) { | |
1459 | ERR("Metadata sent before version check"); | |
1460 | ret = -1; | |
1461 | goto end; | |
1462 | } | |
1463 | ||
f6416125 MD |
1464 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1465 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1466 | ERR("Incorrect data size"); | |
1467 | ret = -1; | |
1468 | goto end; | |
1469 | } | |
1470 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1471 | ||
b8aa1682 | 1472 | if (data_buffer_size < data_size) { |
d7b3776f | 1473 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1474 | char *tmp_data_ptr; |
1475 | ||
1476 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1477 | if (!tmp_data_ptr) { | |
b8aa1682 | 1478 | ERR("Allocating data buffer"); |
c617c0c6 | 1479 | free(data_buffer); |
b8aa1682 JD |
1480 | ret = -1; |
1481 | goto end; | |
1482 | } | |
c617c0c6 | 1483 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1484 | data_buffer_size = data_size; |
1485 | } | |
1486 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1487 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
58eb9381 | 1488 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 1489 | if (ret < 0 || ret != data_size) { |
a6cd2b97 DG |
1490 | if (ret == 0) { |
1491 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1492 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1493 | } else { |
1494 | ERR("Relay didn't receive the whole metadata"); | |
1495 | } | |
b8aa1682 | 1496 | ret = -1; |
b8aa1682 JD |
1497 | goto end; |
1498 | } | |
1499 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 MD |
1500 | |
1501 | rcu_read_lock(); | |
2a174661 | 1502 | metadata_stream = stream_find_by_id(relay_streams_ht, |
d3e2ba59 | 1503 | be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1504 | if (!metadata_stream) { |
1505 | ret = -1; | |
9d1bbf21 | 1506 | goto end_unlock; |
b8aa1682 JD |
1507 | } |
1508 | ||
6cd525e8 MD |
1509 | size_ret = lttng_write(metadata_stream->fd, metadata_struct->payload, |
1510 | payload_size); | |
1511 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1512 | ERR("Relay error writing metadata on file"); |
1513 | ret = -1; | |
9d1bbf21 | 1514 | goto end_unlock; |
b8aa1682 | 1515 | } |
1d4dfdef DG |
1516 | |
1517 | ret = write_padding_to_file(metadata_stream->fd, | |
1518 | be32toh(metadata_struct->padding_size)); | |
1519 | if (ret < 0) { | |
1520 | goto end_unlock; | |
1521 | } | |
2a174661 DG |
1522 | |
1523 | ctf_trace = ctf_trace_find_by_path(session->ctf_traces_ht, | |
1524 | metadata_stream->path_name); | |
1525 | assert(ctf_trace); | |
1526 | ctf_trace->metadata_received += | |
d3e2ba59 | 1527 | payload_size + be32toh(metadata_struct->padding_size); |
1d4dfdef | 1528 | |
b8aa1682 JD |
1529 | DBG2("Relay metadata written"); |
1530 | ||
9d1bbf21 | 1531 | end_unlock: |
6e3c5836 | 1532 | rcu_read_unlock(); |
b8aa1682 JD |
1533 | end: |
1534 | return ret; | |
1535 | } | |
1536 | ||
1537 | /* | |
1538 | * relay_send_version: send relayd version number | |
1539 | */ | |
1540 | static | |
1541 | int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1542 | struct relay_connection *conn) |
b8aa1682 | 1543 | { |
7f51dcba | 1544 | int ret; |
092b6259 | 1545 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1546 | |
58eb9381 | 1547 | assert(conn); |
c5b6f4f0 | 1548 | |
58eb9381 | 1549 | conn->version_check_done = 1; |
b8aa1682 | 1550 | |
092b6259 | 1551 | /* Get version from the other side. */ |
58eb9381 | 1552 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1553 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1554 | if (ret == 0) { |
1555 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1556 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1557 | } else { |
1558 | ERR("Relay failed to receive the version values."); | |
1559 | } | |
092b6259 | 1560 | ret = -1; |
092b6259 DG |
1561 | goto end; |
1562 | } | |
1563 | ||
53efb85a | 1564 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1565 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1566 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1567 | |
1568 | /* Major versions must be the same */ | |
1569 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1570 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1571 | reply.major, be32toh(msg.major)); | |
58eb9381 | 1572 | destroy_session(conn->session, conn->sessions_ht); |
d4519fa3 JD |
1573 | ret = 0; |
1574 | goto end; | |
1575 | } | |
1576 | ||
58eb9381 | 1577 | conn->major = reply.major; |
0f907de1 JD |
1578 | /* We adapt to the lowest compatible version */ |
1579 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1580 | conn->minor = reply.minor; |
0f907de1 | 1581 | } else { |
58eb9381 | 1582 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1583 | } |
1584 | ||
6151a90f JD |
1585 | reply.major = htobe32(reply.major); |
1586 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1587 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1588 | sizeof(struct lttcomm_relayd_version), 0); |
1589 | if (ret < 0) { | |
1590 | ERR("Relay sending version"); | |
1591 | } | |
1592 | ||
58eb9381 DG |
1593 | DBG("Version check done using protocol %u.%u", conn->major, |
1594 | conn->minor); | |
b8aa1682 JD |
1595 | |
1596 | end: | |
1597 | return ret; | |
1598 | } | |
1599 | ||
c8f59ee5 | 1600 | /* |
6d805429 | 1601 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 DG |
1602 | */ |
1603 | static | |
6d805429 | 1604 | int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1605 | struct relay_connection *conn) |
c8f59ee5 | 1606 | { |
58eb9381 | 1607 | struct relay_session *session = conn->session; |
6d805429 | 1608 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1609 | struct lttcomm_relayd_generic_reply reply; |
1610 | struct relay_stream *stream; | |
1611 | int ret; | |
c8f59ee5 DG |
1612 | uint64_t last_net_seq_num, stream_id; |
1613 | ||
6d805429 | 1614 | DBG("Data pending command received"); |
c8f59ee5 | 1615 | |
58eb9381 | 1616 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1617 | ERR("Trying to check for data before version check"); |
1618 | ret = -1; | |
1619 | goto end_no_session; | |
1620 | } | |
1621 | ||
58eb9381 | 1622 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1623 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1624 | if (ret == 0) { |
1625 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1626 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1627 | } else { |
1628 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1629 | ret); | |
1630 | } | |
c8f59ee5 DG |
1631 | ret = -1; |
1632 | goto end_no_session; | |
1633 | } | |
1634 | ||
1635 | stream_id = be64toh(msg.stream_id); | |
1636 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1637 | ||
1638 | rcu_read_lock(); | |
2a174661 | 1639 | stream = stream_find_by_id(relay_streams_ht, stream_id); |
de91f48a | 1640 | if (stream == NULL) { |
c8f59ee5 DG |
1641 | ret = -1; |
1642 | goto end_unlock; | |
1643 | } | |
1644 | ||
6d805429 | 1645 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1646 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1647 | last_net_seq_num); | |
1648 | ||
33832e64 | 1649 | /* Avoid wrapping issue */ |
39df6d9f | 1650 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1651 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1652 | ret = 0; |
6d805429 DG |
1653 | } else { |
1654 | /* Data still being streamed thus pending */ | |
1655 | ret = 1; | |
c8f59ee5 DG |
1656 | } |
1657 | ||
f7079f67 DG |
1658 | /* Pending check is now done. */ |
1659 | stream->data_pending_check_done = 1; | |
1660 | ||
c8f59ee5 DG |
1661 | end_unlock: |
1662 | rcu_read_unlock(); | |
1663 | ||
53efb85a | 1664 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1665 | reply.ret_code = htobe32(ret); |
58eb9381 | 1666 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1667 | if (ret < 0) { |
6d805429 | 1668 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1669 | } |
1670 | ||
1671 | end_no_session: | |
1672 | return ret; | |
1673 | } | |
1674 | ||
1675 | /* | |
1676 | * Wait for the control socket to reach a quiescent state. | |
1677 | * | |
1678 | * Note that for now, when receiving this command from the session daemon, this | |
1679 | * means that every subsequent commands or data received on the control socket | |
1680 | * has been handled. So, this is why we simply return OK here. | |
1681 | */ | |
1682 | static | |
1683 | int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1684 | struct relay_connection *conn) |
c8f59ee5 DG |
1685 | { |
1686 | int ret; | |
ad7051c0 DG |
1687 | uint64_t stream_id; |
1688 | struct relay_stream *stream; | |
1689 | struct lttng_ht_iter iter; | |
1690 | struct lttcomm_relayd_quiescent_control msg; | |
c8f59ee5 DG |
1691 | struct lttcomm_relayd_generic_reply reply; |
1692 | ||
1693 | DBG("Checking quiescent state on control socket"); | |
1694 | ||
58eb9381 | 1695 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1696 | ERR("Trying to check for data before version check"); |
1697 | ret = -1; | |
1698 | goto end_no_session; | |
1699 | } | |
1700 | ||
58eb9381 | 1701 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1702 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1703 | if (ret == 0) { |
1704 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1705 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1706 | } else { |
1707 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1708 | ret); | |
1709 | } | |
ad7051c0 DG |
1710 | ret = -1; |
1711 | goto end_no_session; | |
1712 | } | |
1713 | ||
1714 | stream_id = be64toh(msg.stream_id); | |
1715 | ||
1716 | rcu_read_lock(); | |
d3e2ba59 | 1717 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1718 | node.node) { |
ad7051c0 DG |
1719 | if (stream->stream_handle == stream_id) { |
1720 | stream->data_pending_check_done = 1; | |
1721 | DBG("Relay quiescent control pending flag set to %" PRIu64, | |
1722 | stream_id); | |
1723 | break; | |
1724 | } | |
1725 | } | |
1726 | rcu_read_unlock(); | |
1727 | ||
53efb85a | 1728 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1729 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1730 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1731 | if (ret < 0) { |
6d805429 | 1732 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1733 | } |
1734 | ||
ad7051c0 | 1735 | end_no_session: |
c8f59ee5 DG |
1736 | return ret; |
1737 | } | |
1738 | ||
f7079f67 DG |
1739 | /* |
1740 | * Initialize a data pending command. This means that a client is about to ask | |
1741 | * for data pending for each stream he/she holds. Simply iterate over all | |
1742 | * streams of a session and set the data_pending_check_done flag. | |
1743 | * | |
1744 | * This command returns to the client a LTTNG_OK code. | |
1745 | */ | |
1746 | static | |
1747 | int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1748 | struct relay_connection *conn) |
f7079f67 DG |
1749 | { |
1750 | int ret; | |
1751 | struct lttng_ht_iter iter; | |
1752 | struct lttcomm_relayd_begin_data_pending msg; | |
1753 | struct lttcomm_relayd_generic_reply reply; | |
1754 | struct relay_stream *stream; | |
1755 | uint64_t session_id; | |
1756 | ||
1757 | assert(recv_hdr); | |
58eb9381 | 1758 | assert(conn); |
f7079f67 DG |
1759 | |
1760 | DBG("Init streams for data pending"); | |
1761 | ||
58eb9381 | 1762 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1763 | ERR("Trying to check for data before version check"); |
1764 | ret = -1; | |
1765 | goto end_no_session; | |
1766 | } | |
1767 | ||
58eb9381 | 1768 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1769 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1770 | if (ret == 0) { |
1771 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1772 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1773 | } else { |
1774 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1775 | ret); | |
1776 | } | |
f7079f67 DG |
1777 | ret = -1; |
1778 | goto end_no_session; | |
1779 | } | |
1780 | ||
1781 | session_id = be64toh(msg.session_id); | |
1782 | ||
1783 | /* | |
1784 | * Iterate over all streams to set the begin data pending flag. For now, the | |
1785 | * streams are indexed by stream handle so we have to iterate over all | |
1786 | * streams to find the one associated with the right session_id. | |
1787 | */ | |
1788 | rcu_read_lock(); | |
d3e2ba59 | 1789 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 DG |
1790 | node.node) { |
1791 | if (stream->session_id == session_id) { | |
f7079f67 DG |
1792 | stream->data_pending_check_done = 0; |
1793 | DBG("Set begin data pending flag to stream %" PRIu64, | |
1794 | stream->stream_handle); | |
1795 | } | |
1796 | } | |
1797 | rcu_read_unlock(); | |
1798 | ||
53efb85a | 1799 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1800 | /* All good, send back reply. */ |
1801 | reply.ret_code = htobe32(LTTNG_OK); | |
1802 | ||
58eb9381 | 1803 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1804 | if (ret < 0) { |
1805 | ERR("Relay begin data pending send reply failed"); | |
1806 | } | |
1807 | ||
1808 | end_no_session: | |
1809 | return ret; | |
1810 | } | |
1811 | ||
1812 | /* | |
1813 | * End data pending command. This will check, for a given session id, if each | |
1814 | * stream associated with it has its data_pending_check_done flag set. If not, | |
1815 | * this means that the client lost track of the stream but the data is still | |
1816 | * being streamed on our side. In this case, we inform the client that data is | |
1817 | * inflight. | |
1818 | * | |
1819 | * Return to the client if there is data in flight or not with a ret_code. | |
1820 | */ | |
1821 | static | |
1822 | int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1823 | struct relay_connection *conn) |
f7079f67 DG |
1824 | { |
1825 | int ret; | |
1826 | struct lttng_ht_iter iter; | |
1827 | struct lttcomm_relayd_end_data_pending msg; | |
1828 | struct lttcomm_relayd_generic_reply reply; | |
1829 | struct relay_stream *stream; | |
1830 | uint64_t session_id; | |
1831 | uint32_t is_data_inflight = 0; | |
1832 | ||
1833 | assert(recv_hdr); | |
58eb9381 | 1834 | assert(conn); |
f7079f67 DG |
1835 | |
1836 | DBG("End data pending command"); | |
1837 | ||
58eb9381 | 1838 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1839 | ERR("Trying to check for data before version check"); |
1840 | ret = -1; | |
1841 | goto end_no_session; | |
1842 | } | |
1843 | ||
58eb9381 | 1844 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1845 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1846 | if (ret == 0) { |
1847 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1848 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1849 | } else { |
1850 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1851 | ret); | |
1852 | } | |
f7079f67 DG |
1853 | ret = -1; |
1854 | goto end_no_session; | |
1855 | } | |
1856 | ||
1857 | session_id = be64toh(msg.session_id); | |
1858 | ||
1859 | /* Iterate over all streams to see if the begin data pending flag is set. */ | |
1860 | rcu_read_lock(); | |
d3e2ba59 | 1861 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 DG |
1862 | node.node) { |
1863 | if (stream->session_id == session_id && | |
f94b19e6 | 1864 | !stream->data_pending_check_done && !stream->terminated_flag) { |
f7079f67 DG |
1865 | is_data_inflight = 1; |
1866 | DBG("Data is still in flight for stream %" PRIu64, | |
1867 | stream->stream_handle); | |
1868 | break; | |
1869 | } | |
1870 | } | |
1871 | rcu_read_unlock(); | |
1872 | ||
53efb85a | 1873 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1874 | /* All good, send back reply. */ |
1875 | reply.ret_code = htobe32(is_data_inflight); | |
1876 | ||
58eb9381 | 1877 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1878 | if (ret < 0) { |
1879 | ERR("Relay end data pending send reply failed"); | |
1880 | } | |
1881 | ||
1882 | end_no_session: | |
1883 | return ret; | |
1884 | } | |
1885 | ||
1c20f0e2 JD |
1886 | /* |
1887 | * Receive an index for a specific stream. | |
1888 | * | |
1889 | * Return 0 on success else a negative value. | |
1890 | */ | |
1891 | static | |
1892 | int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1893 | struct relay_connection *conn) |
1c20f0e2 JD |
1894 | { |
1895 | int ret, send_ret, index_created = 0; | |
58eb9381 | 1896 | struct relay_session *session = conn->session; |
1c20f0e2 JD |
1897 | struct lttcomm_relayd_index index_info; |
1898 | struct relay_index *index, *wr_index = NULL; | |
1899 | struct lttcomm_relayd_generic_reply reply; | |
1900 | struct relay_stream *stream; | |
1901 | uint64_t net_seq_num; | |
1902 | ||
58eb9381 | 1903 | assert(conn); |
1c20f0e2 JD |
1904 | |
1905 | DBG("Relay receiving index"); | |
1906 | ||
58eb9381 | 1907 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1908 | ERR("Trying to close a stream before version check"); |
1909 | ret = -1; | |
1910 | goto end_no_session; | |
1911 | } | |
1912 | ||
58eb9381 | 1913 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1914 | sizeof(index_info), 0); |
1915 | if (ret < sizeof(index_info)) { | |
1916 | if (ret == 0) { | |
1917 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1918 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1919 | } else { |
1920 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1921 | } | |
1922 | ret = -1; | |
1923 | goto end_no_session; | |
1924 | } | |
1925 | ||
1926 | net_seq_num = be64toh(index_info.net_seq_num); | |
1927 | ||
1928 | rcu_read_lock(); | |
2a174661 DG |
1929 | stream = stream_find_by_id(relay_streams_ht, |
1930 | be64toh(index_info.relay_stream_id)); | |
1c20f0e2 JD |
1931 | if (!stream) { |
1932 | ret = -1; | |
1933 | goto end_rcu_unlock; | |
1934 | } | |
1935 | ||
d3e2ba59 JD |
1936 | /* Live beacon handling */ |
1937 | if (index_info.packet_size == 0) { | |
1938 | DBG("Received live beacon for stream %" PRIu64, stream->stream_handle); | |
1939 | ||
1940 | /* | |
1941 | * Only flag a stream inactive when it has already received data. | |
1942 | */ | |
1943 | if (stream->total_index_received > 0) { | |
1944 | stream->beacon_ts_end = be64toh(index_info.timestamp_end); | |
1945 | } | |
1946 | ret = 0; | |
1947 | goto end_rcu_unlock; | |
1948 | } else { | |
1949 | stream->beacon_ts_end = -1ULL; | |
1950 | } | |
1951 | ||
0a6518b0 | 1952 | index = relay_index_find(stream->stream_handle, net_seq_num); |
1c20f0e2 JD |
1953 | if (!index) { |
1954 | /* A successful creation will add the object to the HT. */ | |
1955 | index = relay_index_create(stream->stream_handle, net_seq_num); | |
1956 | if (!index) { | |
1957 | goto end_rcu_unlock; | |
1958 | } | |
1959 | index_created = 1; | |
1960 | } | |
1961 | ||
1962 | copy_index_control_data(index, &index_info); | |
1963 | ||
1964 | if (index_created) { | |
1965 | /* | |
1966 | * Try to add the relay index object to the hash table. If an object | |
1967 | * already exist, destroy back the index created, set the data in this | |
1968 | * object and write it on disk. | |
1969 | */ | |
0a6518b0 | 1970 | relay_index_add(index, &wr_index); |
1c20f0e2 JD |
1971 | if (wr_index) { |
1972 | copy_index_control_data(wr_index, &index_info); | |
1973 | free(index); | |
1974 | } | |
1975 | } else { | |
1976 | /* The index already exists so write it on disk. */ | |
1977 | wr_index = index; | |
1978 | } | |
1979 | ||
1980 | /* Do we have a writable ready index to write on disk. */ | |
1981 | if (wr_index) { | |
0a6518b0 | 1982 | ret = relay_index_write(wr_index->fd, wr_index); |
1c20f0e2 JD |
1983 | if (ret < 0) { |
1984 | goto end_rcu_unlock; | |
1985 | } | |
d3e2ba59 | 1986 | stream->total_index_received++; |
1c20f0e2 JD |
1987 | } |
1988 | ||
1989 | end_rcu_unlock: | |
1990 | rcu_read_unlock(); | |
1991 | ||
53efb85a | 1992 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
1993 | if (ret < 0) { |
1994 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1995 | } else { | |
1996 | reply.ret_code = htobe32(LTTNG_OK); | |
1997 | } | |
58eb9381 | 1998 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
1999 | if (send_ret < 0) { |
2000 | ERR("Relay sending close index id reply"); | |
2001 | ret = send_ret; | |
2002 | } | |
2003 | ||
2004 | end_no_session: | |
2005 | return ret; | |
2006 | } | |
2007 | ||
a4baae1b JD |
2008 | /* |
2009 | * Receive the streams_sent message. | |
2010 | * | |
2011 | * Return 0 on success else a negative value. | |
2012 | */ | |
2013 | static | |
2014 | int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 2015 | struct relay_connection *conn) |
a4baae1b JD |
2016 | { |
2017 | int ret, send_ret; | |
2018 | struct lttcomm_relayd_generic_reply reply; | |
2019 | ||
58eb9381 | 2020 | assert(conn); |
a4baae1b JD |
2021 | |
2022 | DBG("Relay receiving streams_sent"); | |
2023 | ||
58eb9381 | 2024 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
2025 | ERR("Trying to close a stream before version check"); |
2026 | ret = -1; | |
2027 | goto end_no_session; | |
2028 | } | |
2029 | ||
2030 | /* | |
2031 | * Flag every pending stream in the connection recv list that they are | |
2032 | * ready to be used by the viewer. | |
2033 | */ | |
58eb9381 | 2034 | set_viewer_ready_flag(conn); |
a4baae1b | 2035 | |
4a9daf17 JD |
2036 | /* |
2037 | * Inform the viewer that there are new streams in the session. | |
2038 | */ | |
c5141fe5 JD |
2039 | if (conn->session->viewer_refcount) { |
2040 | uatomic_set(&conn->session->new_streams, 1); | |
2041 | } | |
4a9daf17 | 2042 | |
53efb85a | 2043 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2044 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2045 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
2046 | if (send_ret < 0) { |
2047 | ERR("Relay sending sent_stream reply"); | |
2048 | ret = send_ret; | |
2049 | } else { | |
2050 | /* Success. */ | |
2051 | ret = 0; | |
2052 | } | |
2053 | ||
2054 | end_no_session: | |
2055 | return ret; | |
2056 | } | |
2057 | ||
b8aa1682 | 2058 | /* |
d3e2ba59 | 2059 | * Process the commands received on the control socket |
b8aa1682 JD |
2060 | */ |
2061 | static | |
2062 | int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 2063 | struct relay_connection *conn) |
b8aa1682 JD |
2064 | { |
2065 | int ret = 0; | |
2066 | ||
2067 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 2068 | case RELAYD_CREATE_SESSION: |
58eb9381 | 2069 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 2070 | break; |
b8aa1682 | 2071 | case RELAYD_ADD_STREAM: |
58eb9381 | 2072 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
2073 | break; |
2074 | case RELAYD_START_DATA: | |
58eb9381 | 2075 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
2076 | break; |
2077 | case RELAYD_SEND_METADATA: | |
58eb9381 | 2078 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
2079 | break; |
2080 | case RELAYD_VERSION: | |
58eb9381 | 2081 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 2082 | break; |
173af62f | 2083 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 2084 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 2085 | break; |
6d805429 | 2086 | case RELAYD_DATA_PENDING: |
58eb9381 | 2087 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
2088 | break; |
2089 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 2090 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 2091 | break; |
f7079f67 | 2092 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 2093 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
2094 | break; |
2095 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2096 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2097 | break; |
1c20f0e2 | 2098 | case RELAYD_SEND_INDEX: |
58eb9381 | 2099 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2100 | break; |
a4baae1b | 2101 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2102 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2103 | break; |
b8aa1682 JD |
2104 | case RELAYD_UPDATE_SYNC_INFO: |
2105 | default: | |
2106 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2107 | relay_unknown_command(conn); |
b8aa1682 JD |
2108 | ret = -1; |
2109 | goto end; | |
2110 | } | |
2111 | ||
2112 | end: | |
2113 | return ret; | |
2114 | } | |
2115 | ||
7d2f7452 DG |
2116 | /* |
2117 | * Handle index for a data stream. | |
2118 | * | |
2119 | * RCU read side lock MUST be acquired. | |
2120 | * | |
2121 | * Return 0 on success else a negative value. | |
2122 | */ | |
2123 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2124 | int rotate_index) | |
2125 | { | |
2126 | int ret = 0, index_created = 0; | |
2127 | uint64_t stream_id, data_offset; | |
2128 | struct relay_index *index, *wr_index = NULL; | |
2129 | ||
2130 | assert(stream); | |
2131 | ||
2132 | stream_id = stream->stream_handle; | |
2133 | /* Get data offset because we are about to update the index. */ | |
2134 | data_offset = htobe64(stream->tracefile_size_current); | |
2135 | ||
2136 | /* | |
2137 | * Lookup for an existing index for that stream id/sequence number. If on | |
2138 | * exists, the control thread already received the data for it thus we need | |
2139 | * to write it on disk. | |
2140 | */ | |
2141 | index = relay_index_find(stream_id, net_seq_num); | |
2142 | if (!index) { | |
2143 | /* A successful creation will add the object to the HT. */ | |
2144 | index = relay_index_create(stream_id, net_seq_num); | |
2145 | if (!index) { | |
2146 | ret = -1; | |
2147 | goto error; | |
2148 | } | |
2149 | index_created = 1; | |
2150 | } | |
2151 | ||
2152 | if (rotate_index || stream->index_fd < 0) { | |
2153 | index->to_close_fd = stream->index_fd; | |
2154 | ret = index_create_file(stream->path_name, stream->channel_name, | |
2155 | relayd_uid, relayd_gid, stream->tracefile_size, | |
2156 | stream->tracefile_count_current); | |
2157 | if (ret < 0) { | |
2158 | /* This will close the stream's index fd if one. */ | |
2159 | relay_index_free_safe(index); | |
2160 | goto error; | |
2161 | } | |
2162 | stream->index_fd = ret; | |
2163 | } | |
2164 | index->fd = stream->index_fd; | |
2165 | index->index_data.offset = data_offset; | |
2166 | ||
2167 | if (index_created) { | |
2168 | /* | |
2169 | * Try to add the relay index object to the hash table. If an object | |
2170 | * already exist, destroy back the index created and set the data. | |
2171 | */ | |
2172 | relay_index_add(index, &wr_index); | |
2173 | if (wr_index) { | |
2174 | /* Copy back data from the created index. */ | |
2175 | wr_index->fd = index->fd; | |
2176 | wr_index->to_close_fd = index->to_close_fd; | |
2177 | wr_index->index_data.offset = data_offset; | |
2178 | free(index); | |
2179 | } | |
2180 | } else { | |
2181 | /* The index already exists so write it on disk. */ | |
2182 | wr_index = index; | |
2183 | } | |
2184 | ||
2185 | /* Do we have a writable ready index to write on disk. */ | |
2186 | if (wr_index) { | |
2187 | ret = relay_index_write(wr_index->fd, wr_index); | |
2188 | if (ret < 0) { | |
2189 | goto error; | |
2190 | } | |
2191 | stream->total_index_received++; | |
2192 | } | |
2193 | ||
2194 | error: | |
2195 | return ret; | |
2196 | } | |
2197 | ||
b8aa1682 JD |
2198 | /* |
2199 | * relay_process_data: Process the data received on the data socket | |
2200 | */ | |
2201 | static | |
58eb9381 | 2202 | int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2203 | { |
7d2f7452 | 2204 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2205 | ssize_t size_ret; |
b8aa1682 JD |
2206 | struct relay_stream *stream; |
2207 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2208 | uint64_t stream_id; |
173af62f | 2209 | uint64_t net_seq_num; |
b8aa1682 | 2210 | uint32_t data_size; |
2a174661 | 2211 | struct relay_session *session; |
b8aa1682 | 2212 | |
58eb9381 DG |
2213 | assert(conn); |
2214 | ||
2215 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, | |
7c5aef62 | 2216 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2217 | if (ret <= 0) { |
a6cd2b97 DG |
2218 | if (ret == 0) { |
2219 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2220 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2221 | } else { |
58eb9381 | 2222 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2223 | } |
b8aa1682 JD |
2224 | ret = -1; |
2225 | goto end; | |
2226 | } | |
2227 | ||
2228 | stream_id = be64toh(data_hdr.stream_id); | |
9d1bbf21 MD |
2229 | |
2230 | rcu_read_lock(); | |
2a174661 | 2231 | stream = stream_find_by_id(relay_streams_ht, stream_id); |
b8aa1682 JD |
2232 | if (!stream) { |
2233 | ret = -1; | |
1c20f0e2 | 2234 | goto end_rcu_unlock; |
b8aa1682 JD |
2235 | } |
2236 | ||
58eb9381 | 2237 | session = session_find_by_id(conn->sessions_ht, stream->session_id); |
2a174661 DG |
2238 | assert(session); |
2239 | ||
b8aa1682 JD |
2240 | data_size = be32toh(data_hdr.data_size); |
2241 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2242 | char *tmp_data_ptr; |
2243 | ||
2244 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2245 | if (!tmp_data_ptr) { | |
b8aa1682 | 2246 | ERR("Allocating data buffer"); |
c617c0c6 | 2247 | free(data_buffer); |
b8aa1682 | 2248 | ret = -1; |
1c20f0e2 | 2249 | goto end_rcu_unlock; |
b8aa1682 | 2250 | } |
c617c0c6 | 2251 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2252 | data_buffer_size = data_size; |
2253 | } | |
2254 | memset(data_buffer, 0, data_size); | |
2255 | ||
173af62f DG |
2256 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2257 | ||
77c7c900 | 2258 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2259 | data_size, stream_id, net_seq_num); |
58eb9381 | 2260 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 2261 | if (ret <= 0) { |
a6cd2b97 DG |
2262 | if (ret == 0) { |
2263 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2264 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2265 | } |
b8aa1682 | 2266 | ret = -1; |
1c20f0e2 | 2267 | goto end_rcu_unlock; |
b8aa1682 JD |
2268 | } |
2269 | ||
1c20f0e2 | 2270 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2271 | if (stream->tracefile_size > 0 && |
2272 | (stream->tracefile_size_current + data_size) > | |
2273 | stream->tracefile_size) { | |
6b6b9a5a JD |
2274 | struct relay_viewer_stream *vstream; |
2275 | uint64_t new_id; | |
2276 | ||
2277 | new_id = (stream->tracefile_count_current + 1) % | |
2278 | stream->tracefile_count; | |
2279 | /* | |
2280 | * When we wrap-around back to 0, we start overwriting old | |
2281 | * trace data. | |
2282 | */ | |
2283 | if (!stream->tracefile_overwrite && new_id == 0) { | |
2284 | stream->tracefile_overwrite = 1; | |
2285 | } | |
2286 | pthread_mutex_lock(&stream->viewer_stream_rotation_lock); | |
2287 | if (stream->tracefile_overwrite) { | |
2288 | stream->oldest_tracefile_id = | |
2289 | (stream->oldest_tracefile_id + 1) % | |
2290 | stream->tracefile_count; | |
2291 | } | |
2f8f53af | 2292 | vstream = viewer_stream_find_by_id(stream->stream_handle); |
6b6b9a5a JD |
2293 | if (vstream) { |
2294 | /* | |
2295 | * The viewer is reading a file about to be | |
2296 | * overwritten. Close the FDs it is | |
2297 | * currently using and let it handle the fault. | |
2298 | */ | |
2299 | if (vstream->tracefile_count_current == new_id) { | |
cef0f7d5 | 2300 | pthread_mutex_lock(&vstream->overwrite_lock); |
6b6b9a5a | 2301 | vstream->abort_flag = 1; |
cef0f7d5 | 2302 | pthread_mutex_unlock(&vstream->overwrite_lock); |
6b6b9a5a JD |
2303 | DBG("Streaming side setting abort_flag on stream %s_%lu\n", |
2304 | stream->channel_name, new_id); | |
2305 | } else if (vstream->tracefile_count_current == | |
2306 | stream->tracefile_count_current) { | |
2307 | /* | |
2308 | * The reader and writer were in the | |
2309 | * same trace file, inform the viewer | |
2310 | * that no new index will ever be added | |
2311 | * to this file. | |
2312 | */ | |
2313 | vstream->close_write_flag = 1; | |
2314 | } | |
2315 | } | |
1c20f0e2 JD |
2316 | ret = utils_rotate_stream_file(stream->path_name, stream->channel_name, |
2317 | stream->tracefile_size, stream->tracefile_count, | |
2318 | relayd_uid, relayd_gid, stream->fd, | |
2319 | &(stream->tracefile_count_current), &stream->fd); | |
cef0f7d5 | 2320 | stream->total_index_received = 0; |
6b6b9a5a | 2321 | pthread_mutex_unlock(&stream->viewer_stream_rotation_lock); |
0f907de1 | 2322 | if (ret < 0) { |
1c20f0e2 JD |
2323 | ERR("Rotating stream output file"); |
2324 | goto end_rcu_unlock; | |
0f907de1 | 2325 | } |
a6976990 DG |
2326 | /* Reset current size because we just perform a stream rotation. */ |
2327 | stream->tracefile_size_current = 0; | |
1c20f0e2 JD |
2328 | rotate_index = 1; |
2329 | } | |
2330 | ||
1c20f0e2 | 2331 | /* |
7d2f7452 DG |
2332 | * Index are handled in protocol version 2.4 and above. Also, snapshot and |
2333 | * index are NOT supported. | |
1c20f0e2 | 2334 | */ |
2a174661 | 2335 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2336 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2337 | if (ret < 0) { |
1c20f0e2 JD |
2338 | goto end_rcu_unlock; |
2339 | } | |
1c20f0e2 JD |
2340 | } |
2341 | ||
7d2f7452 | 2342 | /* Write data to stream output fd. */ |
6cd525e8 MD |
2343 | size_ret = lttng_write(stream->fd, data_buffer, data_size); |
2344 | if (size_ret < data_size) { | |
b8aa1682 JD |
2345 | ERR("Relay error writing data to file"); |
2346 | ret = -1; | |
1c20f0e2 | 2347 | goto end_rcu_unlock; |
b8aa1682 | 2348 | } |
1d4dfdef | 2349 | |
5ab7344e JD |
2350 | DBG2("Relay wrote %d bytes to tracefile for stream id %" PRIu64, |
2351 | ret, stream->stream_handle); | |
2352 | ||
1d4dfdef DG |
2353 | ret = write_padding_to_file(stream->fd, be32toh(data_hdr.padding_size)); |
2354 | if (ret < 0) { | |
1c20f0e2 | 2355 | goto end_rcu_unlock; |
1d4dfdef | 2356 | } |
1c20f0e2 | 2357 | stream->tracefile_size_current += data_size + be32toh(data_hdr.padding_size); |
1d4dfdef | 2358 | |
173af62f DG |
2359 | stream->prev_seq = net_seq_num; |
2360 | ||
2a174661 | 2361 | try_close_stream(session, stream); |
173af62f | 2362 | |
1c20f0e2 | 2363 | end_rcu_unlock: |
9d1bbf21 | 2364 | rcu_read_unlock(); |
b8aa1682 JD |
2365 | end: |
2366 | return ret; | |
2367 | } | |
2368 | ||
2369 | static | |
58eb9381 | 2370 | void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2371 | { |
2372 | int ret; | |
2373 | ||
58eb9381 DG |
2374 | assert(events); |
2375 | ||
2376 | (void) lttng_poll_del(events, pollfd); | |
b8aa1682 JD |
2377 | |
2378 | ret = close(pollfd); | |
2379 | if (ret < 0) { | |
2380 | ERR("Closing pollfd %d", pollfd); | |
2381 | } | |
2382 | } | |
2383 | ||
58eb9381 DG |
2384 | static void destroy_connection(struct lttng_ht *relay_connections_ht, |
2385 | struct relay_connection *conn) | |
9d1bbf21 | 2386 | { |
58eb9381 DG |
2387 | assert(relay_connections_ht); |
2388 | assert(conn); | |
2a174661 | 2389 | |
58eb9381 | 2390 | connection_delete(relay_connections_ht, conn); |
2a174661 | 2391 | |
58eb9381 | 2392 | /* For the control socket, we try to destroy the session. */ |
fd6d7293 | 2393 | if (conn->type == RELAY_CONTROL && conn->session) { |
58eb9381 | 2394 | destroy_session(conn->session, conn->sessions_ht); |
9d1bbf21 | 2395 | } |
5b6d8097 | 2396 | |
58eb9381 | 2397 | connection_destroy(conn); |
b8aa1682 JD |
2398 | } |
2399 | ||
2400 | /* | |
2401 | * This thread does the actual work | |
2402 | */ | |
2403 | static | |
2404 | void *relay_thread_worker(void *data) | |
2405 | { | |
beaad64c DG |
2406 | int ret, err = -1, last_seen_data_fd = -1; |
2407 | uint32_t nb_fd; | |
58eb9381 | 2408 | struct relay_connection *conn; |
b8aa1682 JD |
2409 | struct lttng_poll_event events; |
2410 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2411 | struct lttng_ht_iter iter; |
b8aa1682 | 2412 | struct lttcomm_relayd_hdr recv_hdr; |
d3e2ba59 JD |
2413 | struct relay_local_data *relay_ctx = (struct relay_local_data *) data; |
2414 | struct lttng_ht *sessions_ht = relay_ctx->sessions_ht; | |
b8aa1682 JD |
2415 | |
2416 | DBG("[thread] Relay worker started"); | |
2417 | ||
9d1bbf21 MD |
2418 | rcu_register_thread(); |
2419 | ||
55706a7d MD |
2420 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2421 | ||
9b5e0863 MD |
2422 | if (testpoint(relayd_thread_worker)) { |
2423 | goto error_testpoint; | |
2424 | } | |
2425 | ||
f385ae0a MD |
2426 | health_code_update(); |
2427 | ||
b8aa1682 JD |
2428 | /* table of connections indexed on socket */ |
2429 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2430 | if (!relay_connections_ht) { |
2431 | goto relay_connections_ht_error; | |
2432 | } | |
b8aa1682 | 2433 | |
1c20f0e2 JD |
2434 | /* Tables of received indexes indexed by index handle and net_seq_num. */ |
2435 | indexes_ht = lttng_ht_new(0, LTTNG_HT_TYPE_TWO_U64); | |
2436 | if (!indexes_ht) { | |
2437 | goto indexes_ht_error; | |
2438 | } | |
2439 | ||
b8aa1682 JD |
2440 | ret = create_thread_poll_set(&events, 2); |
2441 | if (ret < 0) { | |
2442 | goto error_poll_create; | |
2443 | } | |
2444 | ||
58eb9381 | 2445 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2446 | if (ret < 0) { |
2447 | goto error; | |
2448 | } | |
2449 | ||
beaad64c | 2450 | restart: |
b8aa1682 | 2451 | while (1) { |
beaad64c DG |
2452 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2453 | ||
f385ae0a MD |
2454 | health_code_update(); |
2455 | ||
b8aa1682 | 2456 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2457 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2458 | health_poll_entry(); |
b8aa1682 | 2459 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2460 | health_poll_exit(); |
b8aa1682 JD |
2461 | if (ret < 0) { |
2462 | /* | |
2463 | * Restart interrupted system call. | |
2464 | */ | |
2465 | if (errno == EINTR) { | |
2466 | goto restart; | |
2467 | } | |
2468 | goto error; | |
2469 | } | |
2470 | ||
0d9c5d77 DG |
2471 | nb_fd = ret; |
2472 | ||
beaad64c DG |
2473 | /* |
2474 | * Process control. The control connection is prioritised so we don't | |
2475 | * starve it with high throughout put tracing data on the data | |
2476 | * connection. | |
2477 | */ | |
b8aa1682 JD |
2478 | for (i = 0; i < nb_fd; i++) { |
2479 | /* Fetch once the poll data */ | |
beaad64c DG |
2480 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2481 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2482 | |
f385ae0a MD |
2483 | health_code_update(); |
2484 | ||
b8aa1682 JD |
2485 | /* Thread quit pipe has been closed. Killing thread. */ |
2486 | ret = check_thread_quit_pipe(pollfd, revents); | |
2487 | if (ret) { | |
095a4ae5 MD |
2488 | err = 0; |
2489 | goto exit; | |
b8aa1682 JD |
2490 | } |
2491 | ||
58eb9381 DG |
2492 | /* Inspect the relay conn pipe for new connection */ |
2493 | if (pollfd == relay_conn_pipe[0]) { | |
b8aa1682 | 2494 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
58eb9381 | 2495 | ERR("Relay connection pipe error"); |
b8aa1682 JD |
2496 | goto error; |
2497 | } else if (revents & LPOLLIN) { | |
58eb9381 | 2498 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2499 | if (ret < 0) { |
2500 | goto error; | |
2501 | } | |
58eb9381 DG |
2502 | conn->sessions_ht = sessions_ht; |
2503 | connection_init(conn); | |
2504 | lttng_poll_add(&events, conn->sock->fd, | |
2505 | LPOLLIN | LPOLLRDHUP); | |
2506 | rcu_read_lock(); | |
2507 | lttng_ht_add_unique_ulong(relay_connections_ht, | |
2508 | &conn->sock_n); | |
9d1bbf21 | 2509 | rcu_read_unlock(); |
58eb9381 | 2510 | DBG("Connection socket %d added", conn->sock->fd); |
b8aa1682 | 2511 | } |
58eb9381 DG |
2512 | } else { |
2513 | rcu_read_lock(); | |
2514 | conn = connection_find_by_sock(relay_connections_ht, pollfd); | |
2515 | /* If not found, there is a synchronization issue. */ | |
2516 | assert(conn); | |
2517 | ||
2518 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2519 | cleanup_connection_pollfd(&events, pollfd); | |
2520 | destroy_connection(relay_connections_ht, conn); | |
beaad64c DG |
2521 | if (last_seen_data_fd == pollfd) { |
2522 | last_seen_data_fd = last_notdel_data_fd; | |
2523 | } | |
b8aa1682 | 2524 | } else if (revents & LPOLLIN) { |
58eb9381 DG |
2525 | if (conn->type == RELAY_CONTROL) { |
2526 | ret = conn->sock->ops->recvmsg(conn->sock, &recv_hdr, | |
2527 | sizeof(recv_hdr), 0); | |
b8aa1682 | 2528 | if (ret <= 0) { |
58eb9381 DG |
2529 | /* Connection closed */ |
2530 | cleanup_connection_pollfd(&events, pollfd); | |
2531 | destroy_connection(relay_connections_ht, conn); | |
b8aa1682 JD |
2532 | DBG("Control connection closed with %d", pollfd); |
2533 | } else { | |
58eb9381 | 2534 | ret = relay_process_control(&recv_hdr, conn); |
b8aa1682 | 2535 | if (ret < 0) { |
beaad64c | 2536 | /* Clear the session on error. */ |
58eb9381 DG |
2537 | cleanup_connection_pollfd(&events, pollfd); |
2538 | destroy_connection(relay_connections_ht, conn); | |
b8aa1682 JD |
2539 | DBG("Connection closed with %d", pollfd); |
2540 | } | |
beaad64c | 2541 | seen_control = 1; |
b8aa1682 | 2542 | } |
beaad64c DG |
2543 | } else { |
2544 | /* | |
2545 | * Flag the last seen data fd not deleted. It will be | |
2546 | * used as the last seen fd if any fd gets deleted in | |
2547 | * this first loop. | |
2548 | */ | |
2549 | last_notdel_data_fd = pollfd; | |
2550 | } | |
58eb9381 DG |
2551 | } else { |
2552 | ERR("Unknown poll events %u for sock %d", revents, pollfd); | |
beaad64c DG |
2553 | } |
2554 | rcu_read_unlock(); | |
2555 | } | |
2556 | } | |
2557 | ||
2558 | /* | |
2559 | * The last loop handled a control request, go back to poll to make | |
2560 | * sure we prioritise the control socket. | |
2561 | */ | |
2562 | if (seen_control) { | |
2563 | continue; | |
2564 | } | |
2565 | ||
2566 | if (last_seen_data_fd >= 0) { | |
2567 | for (i = 0; i < nb_fd; i++) { | |
2568 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2569 | |
2570 | health_code_update(); | |
2571 | ||
beaad64c DG |
2572 | if (last_seen_data_fd == pollfd) { |
2573 | idx = i; | |
2574 | break; | |
2575 | } | |
2576 | } | |
2577 | } | |
2578 | ||
2579 | /* Process data connection. */ | |
2580 | for (i = idx + 1; i < nb_fd; i++) { | |
2581 | /* Fetch the poll data. */ | |
2582 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2583 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
2584 | ||
f385ae0a MD |
2585 | health_code_update(); |
2586 | ||
beaad64c | 2587 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2588 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2589 | continue; |
2590 | } | |
2591 | ||
2592 | if (revents) { | |
2593 | rcu_read_lock(); | |
58eb9381 DG |
2594 | conn = connection_find_by_sock(relay_connections_ht, pollfd); |
2595 | if (!conn) { | |
beaad64c DG |
2596 | /* Skip it. Might be removed before. */ |
2597 | rcu_read_unlock(); | |
2598 | continue; | |
2599 | } | |
beaad64c DG |
2600 | |
2601 | if (revents & LPOLLIN) { | |
58eb9381 | 2602 | if (conn->type != RELAY_DATA) { |
beaad64c DG |
2603 | continue; |
2604 | } | |
2605 | ||
58eb9381 DG |
2606 | ret = relay_process_data(conn); |
2607 | /* Connection closed */ | |
beaad64c | 2608 | if (ret < 0) { |
58eb9381 DG |
2609 | cleanup_connection_pollfd(&events, pollfd); |
2610 | destroy_connection(relay_connections_ht, conn); | |
beaad64c DG |
2611 | DBG("Data connection closed with %d", pollfd); |
2612 | /* | |
2613 | * Every goto restart call sets the last seen fd where | |
2614 | * here we don't really care since we gracefully | |
2615 | * continue the loop after the connection is deleted. | |
2616 | */ | |
2617 | } else { | |
2618 | /* Keep last seen port. */ | |
2619 | last_seen_data_fd = pollfd; | |
2620 | rcu_read_unlock(); | |
2621 | goto restart; | |
b8aa1682 JD |
2622 | } |
2623 | } | |
9d1bbf21 | 2624 | rcu_read_unlock(); |
b8aa1682 JD |
2625 | } |
2626 | } | |
beaad64c | 2627 | last_seen_data_fd = -1; |
b8aa1682 JD |
2628 | } |
2629 | ||
f385ae0a MD |
2630 | /* Normal exit, no error */ |
2631 | ret = 0; | |
2632 | ||
095a4ae5 | 2633 | exit: |
b8aa1682 JD |
2634 | error: |
2635 | lttng_poll_clean(&events); | |
2636 | ||
58eb9381 | 2637 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2638 | rcu_read_lock(); |
58eb9381 DG |
2639 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, conn, |
2640 | sock_n.node) { | |
f385ae0a | 2641 | health_code_update(); |
58eb9381 | 2642 | destroy_connection(relay_connections_ht, conn); |
b8aa1682 | 2643 | } |
94d49140 | 2644 | rcu_read_unlock(); |
7d2f7452 DG |
2645 | error_poll_create: |
2646 | lttng_ht_destroy(indexes_ht); | |
1c20f0e2 | 2647 | indexes_ht_error: |
b8aa1682 | 2648 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2649 | relay_connections_ht_error: |
58eb9381 DG |
2650 | /* Close relay conn pipes */ |
2651 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2652 | if (err) { |
2653 | DBG("Thread exited with error"); | |
2654 | } | |
b8aa1682 | 2655 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2656 | free(data_buffer); |
9b5e0863 | 2657 | error_testpoint: |
f385ae0a MD |
2658 | if (err) { |
2659 | health_error(); | |
2660 | ERR("Health error occurred in %s", __func__); | |
2661 | } | |
2662 | health_unregister(health_relayd); | |
9d1bbf21 | 2663 | rcu_unregister_thread(); |
f385ae0a | 2664 | stop_threads(); |
b8aa1682 JD |
2665 | return NULL; |
2666 | } | |
2667 | ||
2668 | /* | |
2669 | * Create the relay command pipe to wake thread_manage_apps. | |
2670 | * Closed in cleanup(). | |
2671 | */ | |
58eb9381 | 2672 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2673 | { |
a02de639 | 2674 | int ret; |
b8aa1682 | 2675 | |
58eb9381 | 2676 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2677 | |
b8aa1682 JD |
2678 | return ret; |
2679 | } | |
2680 | ||
2681 | /* | |
2682 | * main | |
2683 | */ | |
2684 | int main(int argc, char **argv) | |
2685 | { | |
2686 | int ret = 0; | |
2687 | void *status; | |
d3e2ba59 | 2688 | struct relay_local_data *relay_ctx; |
b8aa1682 | 2689 | |
b8aa1682 JD |
2690 | /* Parse arguments */ |
2691 | progname = argv[0]; | |
cd60b05a | 2692 | if ((ret = set_options(argc, argv)) < 0) { |
a02de639 | 2693 | goto exit; |
b8aa1682 JD |
2694 | } |
2695 | ||
2696 | if ((ret = set_signal_handler()) < 0) { | |
2697 | goto exit; | |
2698 | } | |
2699 | ||
4d513a50 DG |
2700 | /* Try to create directory if -o, --output is specified. */ |
2701 | if (opt_output_path) { | |
994fa64f DG |
2702 | if (*opt_output_path != '/') { |
2703 | ERR("Please specify an absolute path for -o, --output PATH"); | |
2704 | goto exit; | |
2705 | } | |
2706 | ||
4d513a50 DG |
2707 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG); |
2708 | if (ret < 0) { | |
2709 | ERR("Unable to create %s", opt_output_path); | |
2710 | goto exit; | |
2711 | } | |
2712 | } | |
2713 | ||
b8aa1682 | 2714 | /* Daemonize */ |
b5218ffb | 2715 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2716 | int i; |
2717 | ||
2718 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2719 | !opt_background); | |
b8aa1682 | 2720 | if (ret < 0) { |
a02de639 | 2721 | goto exit; |
b8aa1682 | 2722 | } |
3fd27398 MD |
2723 | |
2724 | /* | |
2725 | * We are in the child. Make sure all other file | |
2726 | * descriptors are closed, in case we are called with | |
2727 | * more opened file descriptors than the standard ones. | |
2728 | */ | |
2729 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2730 | (void) close(i); | |
2731 | } | |
2732 | } | |
2733 | ||
2734 | /* Create thread quit pipe */ | |
2735 | if ((ret = init_thread_quit_pipe()) < 0) { | |
2736 | goto error; | |
b8aa1682 JD |
2737 | } |
2738 | ||
1c20f0e2 JD |
2739 | /* We need those values for the file/dir creation. */ |
2740 | relayd_uid = getuid(); | |
2741 | relayd_gid = getgid(); | |
b8aa1682 | 2742 | |
1c20f0e2 JD |
2743 | /* Check if daemon is UID = 0 */ |
2744 | if (relayd_uid == 0) { | |
8d5c808e | 2745 | if (control_uri->port < 1024 || data_uri->port < 1024 || live_uri->port < 1024) { |
b8aa1682 JD |
2746 | ERR("Need to be root to use ports < 1024"); |
2747 | ret = -1; | |
a02de639 | 2748 | goto exit; |
b8aa1682 JD |
2749 | } |
2750 | } | |
2751 | ||
2752 | /* Setup the thread apps communication pipe. */ | |
58eb9381 | 2753 | if ((ret = create_relay_conn_pipe()) < 0) { |
b8aa1682 JD |
2754 | goto exit; |
2755 | } | |
2756 | ||
2757 | /* Init relay command queue. */ | |
58eb9381 | 2758 | cds_wfq_init(&relay_conn_queue.queue); |
b8aa1682 JD |
2759 | |
2760 | /* Set up max poll set size */ | |
2761 | lttng_poll_set_max_size(); | |
2762 | ||
554831e7 MD |
2763 | /* Initialize communication library */ |
2764 | lttcomm_init(); | |
87e45c13 | 2765 | lttcomm_inet_init(); |
554831e7 | 2766 | |
d3e2ba59 JD |
2767 | relay_ctx = zmalloc(sizeof(struct relay_local_data)); |
2768 | if (!relay_ctx) { | |
2769 | PERROR("relay_ctx"); | |
2770 | goto exit; | |
2771 | } | |
2772 | ||
2773 | /* tables of sessions indexed by session ID */ | |
2a174661 | 2774 | relay_ctx->sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 JD |
2775 | if (!relay_ctx->sessions_ht) { |
2776 | goto exit_relay_ctx_sessions; | |
2777 | } | |
2778 | ||
2779 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2780 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 JD |
2781 | if (!relay_streams_ht) { |
2782 | goto exit_relay_ctx_streams; | |
2783 | } | |
2784 | ||
2785 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2786 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2787 | if (!viewer_streams_ht) { | |
d3e2ba59 JD |
2788 | goto exit_relay_ctx_viewer_streams; |
2789 | } | |
2790 | ||
55706a7d MD |
2791 | /* Initialize thread health monitoring */ |
2792 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2793 | if (!health_relayd) { | |
2794 | PERROR("health_app_create error"); | |
2795 | goto exit_health_app_create; | |
2796 | } | |
2797 | ||
65931c8b MD |
2798 | ret = utils_create_pipe(health_quit_pipe); |
2799 | if (ret < 0) { | |
2800 | goto error_health_pipe; | |
2801 | } | |
2802 | ||
2803 | /* Create thread to manage the client socket */ | |
2804 | ret = pthread_create(&health_thread, NULL, | |
2805 | thread_manage_health, (void *) NULL); | |
2806 | if (ret != 0) { | |
2807 | PERROR("pthread_create health"); | |
2808 | goto health_error; | |
2809 | } | |
2810 | ||
b8aa1682 JD |
2811 | /* Setup the dispatcher thread */ |
2812 | ret = pthread_create(&dispatcher_thread, NULL, | |
2813 | relay_thread_dispatcher, (void *) NULL); | |
2814 | if (ret != 0) { | |
2815 | PERROR("pthread_create dispatcher"); | |
2816 | goto exit_dispatcher; | |
2817 | } | |
2818 | ||
2819 | /* Setup the worker thread */ | |
2820 | ret = pthread_create(&worker_thread, NULL, | |
d3e2ba59 | 2821 | relay_thread_worker, (void *) relay_ctx); |
b8aa1682 JD |
2822 | if (ret != 0) { |
2823 | PERROR("pthread_create worker"); | |
2824 | goto exit_worker; | |
2825 | } | |
2826 | ||
2827 | /* Setup the listener thread */ | |
2828 | ret = pthread_create(&listener_thread, NULL, | |
2829 | relay_thread_listener, (void *) NULL); | |
2830 | if (ret != 0) { | |
2831 | PERROR("pthread_create listener"); | |
2832 | goto exit_listener; | |
2833 | } | |
2834 | ||
0b242f62 | 2835 | ret = live_start_threads(live_uri, relay_ctx); |
d3e2ba59 JD |
2836 | if (ret != 0) { |
2837 | ERR("Starting live viewer threads"); | |
50138f51 | 2838 | goto exit_live; |
d3e2ba59 JD |
2839 | } |
2840 | ||
50138f51 | 2841 | exit_live: |
b8aa1682 JD |
2842 | ret = pthread_join(listener_thread, &status); |
2843 | if (ret != 0) { | |
2844 | PERROR("pthread_join"); | |
2845 | goto error; /* join error, exit without cleanup */ | |
2846 | } | |
2847 | ||
50138f51 | 2848 | exit_listener: |
b8aa1682 JD |
2849 | ret = pthread_join(worker_thread, &status); |
2850 | if (ret != 0) { | |
2851 | PERROR("pthread_join"); | |
2852 | goto error; /* join error, exit without cleanup */ | |
2853 | } | |
2854 | ||
50138f51 | 2855 | exit_worker: |
b8aa1682 JD |
2856 | ret = pthread_join(dispatcher_thread, &status); |
2857 | if (ret != 0) { | |
2858 | PERROR("pthread_join"); | |
2859 | goto error; /* join error, exit without cleanup */ | |
2860 | } | |
42415026 | 2861 | |
50138f51 | 2862 | exit_dispatcher: |
65931c8b MD |
2863 | ret = pthread_join(health_thread, &status); |
2864 | if (ret != 0) { | |
2865 | PERROR("pthread_join health thread"); | |
2866 | goto error; /* join error, exit without cleanup */ | |
2867 | } | |
2868 | ||
bd11e201 MD |
2869 | /* |
2870 | * Stop live threads only after joining other threads. | |
2871 | */ | |
2872 | live_stop_threads(); | |
2873 | ||
65931c8b MD |
2874 | health_error: |
2875 | utils_close_pipe(health_quit_pipe); | |
2876 | ||
2877 | error_health_pipe: | |
55706a7d MD |
2878 | health_app_destroy(health_relayd); |
2879 | ||
2880 | exit_health_app_create: | |
92c6ca54 | 2881 | lttng_ht_destroy(viewer_streams_ht); |
d3e2ba59 JD |
2882 | |
2883 | exit_relay_ctx_viewer_streams: | |
2884 | lttng_ht_destroy(relay_streams_ht); | |
2885 | ||
2886 | exit_relay_ctx_streams: | |
2887 | lttng_ht_destroy(relay_ctx->sessions_ht); | |
2888 | ||
2889 | exit_relay_ctx_sessions: | |
2890 | free(relay_ctx); | |
b8aa1682 JD |
2891 | |
2892 | exit: | |
2893 | cleanup(); | |
2894 | if (!ret) { | |
2895 | exit(EXIT_SUCCESS); | |
2896 | } | |
a02de639 | 2897 | |
b8aa1682 JD |
2898 | error: |
2899 | exit(EXIT_FAILURE); | |
2900 | } |