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 { | |
849e5b7b DG |
254 | /* Only 3 level of verbosity (-vvv). */ |
255 | if (lttng_opt_verbose < 3) { | |
256 | lttng_opt_verbose += 1; | |
257 | } | |
cd60b05a JG |
258 | } |
259 | break; | |
260 | default: | |
261 | /* Unknown option or other error. | |
262 | * Error is printed by getopt, just return */ | |
263 | ret = -1; | |
264 | goto end; | |
265 | } | |
266 | ||
267 | /* All good. */ | |
268 | ret = 0; | |
269 | ||
270 | end: | |
271 | return ret; | |
272 | } | |
273 | ||
274 | /* | |
275 | * config_entry_handler_cb used to handle options read from a config file. | |
276 | * See config_entry_handler_cb comment in common/config/config.h for the | |
277 | * return value conventions. | |
278 | */ | |
279 | static | |
280 | int config_entry_handler(const struct config_entry *entry, void *unused) | |
281 | { | |
282 | int ret = 0, i; | |
283 | ||
284 | if (!entry || !entry->name || !entry->value) { | |
285 | ret = -EINVAL; | |
286 | goto end; | |
287 | } | |
288 | ||
289 | /* Check if the option is to be ignored */ | |
290 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
291 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
292 | goto end; | |
293 | } | |
294 | } | |
295 | ||
296 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
297 | /* Ignore if entry name is not fully matched. */ | |
298 | if (strcmp(entry->name, long_options[i].name)) { | |
299 | continue; | |
300 | } | |
301 | ||
302 | /* | |
303 | * If the option takes no argument on the command line, we have to | |
304 | * check if the value is "true". We support non-zero numeric values, | |
305 | * true, on and yes. | |
306 | */ | |
307 | if (!long_options[i].has_arg) { | |
308 | ret = config_parse_value(entry->value); | |
309 | if (ret <= 0) { | |
310 | if (ret) { | |
311 | WARN("Invalid configuration value \"%s\" for option %s", | |
312 | entry->value, entry->name); | |
313 | } | |
314 | /* False, skip boolean config option. */ | |
315 | goto end; | |
316 | } | |
317 | } | |
318 | ||
319 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
320 | goto end; | |
321 | } | |
322 | ||
323 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
324 | entry->name); | |
325 | ||
326 | end: | |
327 | return ret; | |
328 | } | |
329 | ||
330 | static | |
331 | int set_options(int argc, char **argv) | |
332 | { | |
333 | int c, ret = 0, option_index = 0; | |
334 | int orig_optopt = optopt, orig_optind = optind; | |
335 | char *default_address, *optstring; | |
336 | const char *config_path = NULL; | |
337 | ||
338 | optstring = utils_generate_optstring(long_options, | |
339 | sizeof(long_options) / sizeof(struct option)); | |
340 | if (!optstring) { | |
341 | ret = -ENOMEM; | |
342 | goto exit; | |
343 | } | |
344 | ||
345 | /* Check for the --config option */ | |
346 | ||
347 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
348 | &option_index)) != -1) { | |
349 | if (c == '?') { | |
350 | ret = -EINVAL; | |
351 | goto exit; | |
352 | } else if (c != 'f') { | |
353 | continue; | |
354 | } | |
355 | ||
356 | config_path = utils_expand_path(optarg); | |
357 | if (!config_path) { | |
358 | ERR("Failed to resolve path: %s", optarg); | |
359 | } | |
360 | } | |
361 | ||
362 | ret = config_get_section_entries(config_path, config_section_name, | |
363 | config_entry_handler, NULL); | |
364 | if (ret) { | |
365 | if (ret > 0) { | |
366 | ERR("Invalid configuration option at line %i", ret); | |
367 | ret = -1; | |
368 | } | |
369 | goto exit; | |
370 | } | |
b8aa1682 | 371 | |
cd60b05a JG |
372 | /* Reset getopt's global state */ |
373 | optopt = orig_optopt; | |
374 | optind = orig_optind; | |
b8aa1682 | 375 | while (1) { |
cd60b05a | 376 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
377 | if (c == -1) { |
378 | break; | |
379 | } | |
380 | ||
cd60b05a JG |
381 | ret = set_option(c, optarg, long_options[option_index].name); |
382 | if (ret < 0) { | |
b8aa1682 JD |
383 | goto exit; |
384 | } | |
385 | } | |
386 | ||
387 | /* assign default values */ | |
388 | if (control_uri == NULL) { | |
fa91dc52 MD |
389 | ret = asprintf(&default_address, |
390 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
391 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
392 | if (ret < 0) { |
393 | PERROR("asprintf default data address"); | |
394 | goto exit; | |
395 | } | |
396 | ||
397 | ret = uri_parse(default_address, &control_uri); | |
398 | free(default_address); | |
399 | if (ret < 0) { | |
400 | ERR("Invalid control URI specified"); | |
401 | goto exit; | |
402 | } | |
403 | } | |
404 | if (data_uri == NULL) { | |
fa91dc52 MD |
405 | ret = asprintf(&default_address, |
406 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
407 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
408 | if (ret < 0) { |
409 | PERROR("asprintf default data address"); | |
410 | goto exit; | |
411 | } | |
412 | ||
413 | ret = uri_parse(default_address, &data_uri); | |
414 | free(default_address); | |
415 | if (ret < 0) { | |
416 | ERR("Invalid data URI specified"); | |
417 | goto exit; | |
418 | } | |
419 | } | |
d3e2ba59 | 420 | if (live_uri == NULL) { |
fa91dc52 MD |
421 | ret = asprintf(&default_address, |
422 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
423 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
424 | if (ret < 0) { |
425 | PERROR("asprintf default viewer control address"); | |
426 | goto exit; | |
427 | } | |
428 | ||
429 | ret = uri_parse(default_address, &live_uri); | |
430 | free(default_address); | |
431 | if (ret < 0) { | |
432 | ERR("Invalid viewer control URI specified"); | |
433 | goto exit; | |
434 | } | |
435 | } | |
b8aa1682 JD |
436 | |
437 | exit: | |
cd60b05a | 438 | free(optstring); |
b8aa1682 JD |
439 | return ret; |
440 | } | |
441 | ||
442 | /* | |
443 | * Cleanup the daemon | |
444 | */ | |
445 | static | |
446 | void cleanup(void) | |
447 | { | |
b8aa1682 JD |
448 | DBG("Cleaning up"); |
449 | ||
095a4ae5 MD |
450 | /* free the dynamically allocated opt_output_path */ |
451 | free(opt_output_path); | |
452 | ||
a02de639 CB |
453 | /* Close thread quit pipes */ |
454 | utils_close_pipe(thread_quit_pipe); | |
455 | ||
710c1f73 DG |
456 | uri_free(control_uri); |
457 | uri_free(data_uri); | |
8d5c808e | 458 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
459 | |
460 | if (tracing_group_name_override) { | |
461 | free((void *) tracing_group_name); | |
462 | } | |
b8aa1682 JD |
463 | } |
464 | ||
465 | /* | |
466 | * Write to writable pipe used to notify a thread. | |
467 | */ | |
468 | static | |
469 | int notify_thread_pipe(int wpipe) | |
470 | { | |
6cd525e8 | 471 | ssize_t ret; |
b8aa1682 | 472 | |
6cd525e8 MD |
473 | ret = lttng_write(wpipe, "!", 1); |
474 | if (ret < 1) { | |
b8aa1682 JD |
475 | PERROR("write poll pipe"); |
476 | } | |
477 | ||
478 | return ret; | |
479 | } | |
480 | ||
65931c8b MD |
481 | static void notify_health_quit_pipe(int *pipe) |
482 | { | |
6cd525e8 | 483 | ssize_t ret; |
65931c8b | 484 | |
6cd525e8 MD |
485 | ret = lttng_write(pipe[1], "4", 1); |
486 | if (ret < 1) { | |
65931c8b MD |
487 | PERROR("write relay health quit"); |
488 | } | |
489 | } | |
490 | ||
b8aa1682 JD |
491 | /* |
492 | * Stop all threads by closing the thread quit pipe. | |
493 | */ | |
494 | static | |
495 | void stop_threads(void) | |
496 | { | |
497 | int ret; | |
498 | ||
499 | /* Stopping all threads */ | |
500 | DBG("Terminating all threads"); | |
501 | ret = notify_thread_pipe(thread_quit_pipe[1]); | |
502 | if (ret < 0) { | |
503 | ERR("write error on thread quit pipe"); | |
504 | } | |
505 | ||
65931c8b MD |
506 | notify_health_quit_pipe(health_quit_pipe); |
507 | ||
b8aa1682 | 508 | /* Dispatch thread */ |
26c9d55e | 509 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 510 | futex_nto1_wake(&relay_conn_queue.futex); |
b8aa1682 JD |
511 | } |
512 | ||
513 | /* | |
514 | * Signal handler for the daemon | |
515 | * | |
516 | * Simply stop all worker threads, leaving main() return gracefully after | |
517 | * joining all threads and calling cleanup(). | |
518 | */ | |
519 | static | |
520 | void sighandler(int sig) | |
521 | { | |
522 | switch (sig) { | |
523 | case SIGPIPE: | |
524 | DBG("SIGPIPE caught"); | |
525 | return; | |
526 | case SIGINT: | |
527 | DBG("SIGINT caught"); | |
528 | stop_threads(); | |
529 | break; | |
530 | case SIGTERM: | |
531 | DBG("SIGTERM caught"); | |
532 | stop_threads(); | |
533 | break; | |
3fd27398 MD |
534 | case SIGUSR1: |
535 | CMM_STORE_SHARED(recv_child_signal, 1); | |
536 | break; | |
b8aa1682 JD |
537 | default: |
538 | break; | |
539 | } | |
540 | } | |
541 | ||
542 | /* | |
543 | * Setup signal handler for : | |
544 | * SIGINT, SIGTERM, SIGPIPE | |
545 | */ | |
546 | static | |
547 | int set_signal_handler(void) | |
548 | { | |
549 | int ret = 0; | |
550 | struct sigaction sa; | |
551 | sigset_t sigset; | |
552 | ||
553 | if ((ret = sigemptyset(&sigset)) < 0) { | |
554 | PERROR("sigemptyset"); | |
555 | return ret; | |
556 | } | |
557 | ||
558 | sa.sa_handler = sighandler; | |
559 | sa.sa_mask = sigset; | |
560 | sa.sa_flags = 0; | |
561 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { | |
562 | PERROR("sigaction"); | |
563 | return ret; | |
564 | } | |
565 | ||
566 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
567 | PERROR("sigaction"); | |
568 | return ret; | |
569 | } | |
570 | ||
571 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
572 | PERROR("sigaction"); | |
573 | return ret; | |
574 | } | |
575 | ||
3fd27398 MD |
576 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
577 | PERROR("sigaction"); | |
578 | return ret; | |
579 | } | |
580 | ||
581 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
582 | |
583 | return ret; | |
584 | } | |
585 | ||
3fd27398 MD |
586 | void lttng_relay_notify_ready(void) |
587 | { | |
588 | /* Notify the parent of the fork() process that we are ready. */ | |
589 | if (opt_daemon || opt_background) { | |
590 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
591 | kill(child_ppid, SIGUSR1); | |
592 | } | |
593 | } | |
594 | } | |
595 | ||
b8aa1682 JD |
596 | /* |
597 | * Init thread quit pipe. | |
598 | * | |
599 | * Return -1 on error or 0 if all pipes are created. | |
600 | */ | |
601 | static | |
602 | int init_thread_quit_pipe(void) | |
603 | { | |
a02de639 | 604 | int ret; |
b8aa1682 | 605 | |
a02de639 | 606 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 607 | |
b8aa1682 JD |
608 | return ret; |
609 | } | |
610 | ||
611 | /* | |
612 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
613 | */ | |
614 | static | |
615 | int create_thread_poll_set(struct lttng_poll_event *events, int size) | |
616 | { | |
617 | int ret; | |
618 | ||
619 | if (events == NULL || size == 0) { | |
620 | ret = -1; | |
621 | goto error; | |
622 | } | |
623 | ||
624 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
625 | if (ret < 0) { | |
626 | goto error; | |
627 | } | |
628 | ||
629 | /* Add quit pipe */ | |
c7759e6a | 630 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
631 | if (ret < 0) { |
632 | goto error; | |
633 | } | |
634 | ||
635 | return 0; | |
636 | ||
637 | error: | |
638 | return ret; | |
639 | } | |
640 | ||
641 | /* | |
642 | * Check if the thread quit pipe was triggered. | |
643 | * | |
644 | * Return 1 if it was triggered else 0; | |
645 | */ | |
646 | static | |
647 | int check_thread_quit_pipe(int fd, uint32_t events) | |
648 | { | |
649 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
650 | return 1; | |
651 | } | |
652 | ||
653 | return 0; | |
654 | } | |
655 | ||
656 | /* | |
657 | * Create and init socket from uri. | |
658 | */ | |
659 | static | |
660 | struct lttcomm_sock *relay_init_sock(struct lttng_uri *uri) | |
661 | { | |
662 | int ret; | |
663 | struct lttcomm_sock *sock = NULL; | |
664 | ||
665 | sock = lttcomm_alloc_sock_from_uri(uri); | |
666 | if (sock == NULL) { | |
667 | ERR("Allocating socket"); | |
668 | goto error; | |
669 | } | |
670 | ||
671 | ret = lttcomm_create_sock(sock); | |
672 | if (ret < 0) { | |
673 | goto error; | |
674 | } | |
675 | DBG("Listening on sock %d", sock->fd); | |
676 | ||
677 | ret = sock->ops->bind(sock); | |
678 | if (ret < 0) { | |
679 | goto error; | |
680 | } | |
681 | ||
682 | ret = sock->ops->listen(sock, -1); | |
683 | if (ret < 0) { | |
684 | goto error; | |
685 | ||
686 | } | |
687 | ||
688 | return sock; | |
689 | ||
690 | error: | |
691 | if (sock) { | |
692 | lttcomm_destroy_sock(sock); | |
693 | } | |
694 | return NULL; | |
695 | } | |
696 | ||
173af62f DG |
697 | /* |
698 | * Return nonzero if stream needs to be closed. | |
699 | */ | |
700 | static | |
701 | int close_stream_check(struct relay_stream *stream) | |
702 | { | |
173af62f | 703 | if (stream->close_flag && stream->prev_seq == stream->last_net_seq_num) { |
f7079f67 DG |
704 | /* |
705 | * We are about to close the stream so set the data pending flag to 1 | |
706 | * which will make the end data pending command skip the stream which | |
707 | * is now closed and ready. Note that after proceeding to a file close, | |
708 | * the written file is ready for reading. | |
709 | */ | |
710 | stream->data_pending_check_done = 1; | |
173af62f DG |
711 | return 1; |
712 | } | |
713 | return 0; | |
714 | } | |
715 | ||
2a174661 DG |
716 | static void try_close_stream(struct relay_session *session, |
717 | struct relay_stream *stream) | |
718 | { | |
719 | int ret; | |
720 | struct ctf_trace *ctf_trace; | |
721 | ||
722 | assert(session); | |
723 | assert(stream); | |
724 | ||
725 | if (!close_stream_check(stream)) { | |
726 | /* Can't close it, not ready for that. */ | |
727 | goto end; | |
728 | } | |
729 | ||
730 | ctf_trace = ctf_trace_find_by_path(session->ctf_traces_ht, | |
731 | stream->path_name); | |
732 | assert(ctf_trace); | |
733 | ||
734 | pthread_mutex_lock(&session->viewer_ready_lock); | |
735 | ctf_trace->invalid_flag = 1; | |
736 | pthread_mutex_unlock(&session->viewer_ready_lock); | |
737 | ||
738 | ret = stream_close(session, stream); | |
f94b19e6 | 739 | if (ret || session->snapshot) { |
2a174661 DG |
740 | /* Already close thus the ctf trace is being or has been destroyed. */ |
741 | goto end; | |
742 | } | |
743 | ||
744 | ctf_trace_try_destroy(session, ctf_trace); | |
745 | ||
746 | end: | |
747 | return; | |
748 | } | |
749 | ||
b8aa1682 JD |
750 | /* |
751 | * This thread manages the listening for new connections on the network | |
752 | */ | |
753 | static | |
754 | void *relay_thread_listener(void *data) | |
755 | { | |
095a4ae5 | 756 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
757 | uint32_t revents, nb_fd; |
758 | struct lttng_poll_event events; | |
759 | struct lttcomm_sock *control_sock, *data_sock; | |
760 | ||
b8aa1682 JD |
761 | DBG("[thread] Relay listener started"); |
762 | ||
55706a7d MD |
763 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
764 | ||
f385ae0a MD |
765 | health_code_update(); |
766 | ||
b8aa1682 JD |
767 | control_sock = relay_init_sock(control_uri); |
768 | if (!control_sock) { | |
095a4ae5 | 769 | goto error_sock_control; |
b8aa1682 JD |
770 | } |
771 | ||
772 | data_sock = relay_init_sock(data_uri); | |
773 | if (!data_sock) { | |
095a4ae5 | 774 | goto error_sock_relay; |
b8aa1682 JD |
775 | } |
776 | ||
777 | /* | |
778 | * Pass 3 as size here for the thread quit pipe, control and data socket. | |
779 | */ | |
780 | ret = create_thread_poll_set(&events, 3); | |
781 | if (ret < 0) { | |
782 | goto error_create_poll; | |
783 | } | |
784 | ||
785 | /* Add the control socket */ | |
786 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
787 | if (ret < 0) { | |
788 | goto error_poll_add; | |
789 | } | |
790 | ||
791 | /* Add the data socket */ | |
792 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
793 | if (ret < 0) { | |
794 | goto error_poll_add; | |
795 | } | |
796 | ||
3fd27398 MD |
797 | lttng_relay_notify_ready(); |
798 | ||
9b5e0863 MD |
799 | if (testpoint(relayd_thread_listener)) { |
800 | goto error_testpoint; | |
801 | } | |
802 | ||
b8aa1682 | 803 | while (1) { |
f385ae0a MD |
804 | health_code_update(); |
805 | ||
b8aa1682 JD |
806 | DBG("Listener accepting connections"); |
807 | ||
b8aa1682 | 808 | restart: |
f385ae0a | 809 | health_poll_entry(); |
b8aa1682 | 810 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 811 | health_poll_exit(); |
b8aa1682 JD |
812 | if (ret < 0) { |
813 | /* | |
814 | * Restart interrupted system call. | |
815 | */ | |
816 | if (errno == EINTR) { | |
817 | goto restart; | |
818 | } | |
819 | goto error; | |
820 | } | |
821 | ||
0d9c5d77 DG |
822 | nb_fd = ret; |
823 | ||
b8aa1682 JD |
824 | DBG("Relay new connection received"); |
825 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
826 | health_code_update(); |
827 | ||
b8aa1682 JD |
828 | /* Fetch once the poll data */ |
829 | revents = LTTNG_POLL_GETEV(&events, i); | |
830 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
831 | ||
832 | /* Thread quit pipe has been closed. Killing thread. */ | |
833 | ret = check_thread_quit_pipe(pollfd, revents); | |
834 | if (ret) { | |
095a4ae5 MD |
835 | err = 0; |
836 | goto exit; | |
b8aa1682 JD |
837 | } |
838 | ||
839 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
840 | ERR("socket poll error"); | |
841 | goto error; | |
842 | } else if (revents & LPOLLIN) { | |
4b7f17b2 | 843 | /* |
58eb9381 DG |
844 | * Get allocated in this thread, enqueued to a global queue, |
845 | * dequeued and freed in the worker thread. | |
4b7f17b2 | 846 | */ |
58eb9381 DG |
847 | int val = 1; |
848 | struct relay_connection *new_conn; | |
4b7f17b2 | 849 | struct lttcomm_sock *newsock; |
b8aa1682 | 850 | |
58eb9381 DG |
851 | new_conn = connection_create(); |
852 | if (!new_conn) { | |
b8aa1682 JD |
853 | goto error; |
854 | } | |
855 | ||
856 | if (pollfd == data_sock->fd) { | |
58eb9381 | 857 | new_conn->type = RELAY_DATA; |
b8aa1682 | 858 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
859 | DBG("Relay data connection accepted, socket %d", |
860 | newsock->fd); | |
4b7f17b2 MD |
861 | } else { |
862 | assert(pollfd == control_sock->fd); | |
58eb9381 | 863 | new_conn->type = RELAY_CONTROL; |
b8aa1682 | 864 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
865 | DBG("Relay control connection accepted, socket %d", |
866 | newsock->fd); | |
b8aa1682 | 867 | } |
58eb9381 DG |
868 | if (!newsock) { |
869 | PERROR("accepting sock"); | |
870 | connection_free(new_conn); | |
871 | goto error; | |
872 | } | |
873 | ||
874 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
875 | sizeof(val)); | |
b8aa1682 JD |
876 | if (ret < 0) { |
877 | PERROR("setsockopt inet"); | |
4b7f17b2 | 878 | lttcomm_destroy_sock(newsock); |
58eb9381 | 879 | connection_free(new_conn); |
b8aa1682 JD |
880 | goto error; |
881 | } | |
58eb9381 DG |
882 | new_conn->sock = newsock; |
883 | ||
884 | /* Enqueue request for the dispatcher thread. */ | |
885 | cds_wfq_enqueue(&relay_conn_queue.queue, &new_conn->qnode); | |
b8aa1682 JD |
886 | |
887 | /* | |
58eb9381 DG |
888 | * Wake the dispatch queue futex. Implicit memory barrier with |
889 | * the exchange in cds_wfq_enqueue. | |
b8aa1682 | 890 | */ |
58eb9381 | 891 | futex_nto1_wake(&relay_conn_queue.futex); |
b8aa1682 JD |
892 | } |
893 | } | |
894 | } | |
895 | ||
095a4ae5 | 896 | exit: |
b8aa1682 JD |
897 | error: |
898 | error_poll_add: | |
9b5e0863 | 899 | error_testpoint: |
b8aa1682 JD |
900 | lttng_poll_clean(&events); |
901 | error_create_poll: | |
095a4ae5 MD |
902 | if (data_sock->fd >= 0) { |
903 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
904 | if (ret) { |
905 | PERROR("close"); | |
906 | } | |
b8aa1682 | 907 | } |
095a4ae5 MD |
908 | lttcomm_destroy_sock(data_sock); |
909 | error_sock_relay: | |
910 | if (control_sock->fd >= 0) { | |
911 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
912 | if (ret) { |
913 | PERROR("close"); | |
914 | } | |
b8aa1682 | 915 | } |
095a4ae5 MD |
916 | lttcomm_destroy_sock(control_sock); |
917 | error_sock_control: | |
918 | if (err) { | |
f385ae0a MD |
919 | health_error(); |
920 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 921 | } |
55706a7d | 922 | health_unregister(health_relayd); |
b8aa1682 JD |
923 | DBG("Relay listener thread cleanup complete"); |
924 | stop_threads(); | |
b8aa1682 JD |
925 | return NULL; |
926 | } | |
927 | ||
928 | /* | |
929 | * This thread manages the dispatching of the requests to worker threads | |
930 | */ | |
931 | static | |
932 | void *relay_thread_dispatcher(void *data) | |
933 | { | |
6cd525e8 MD |
934 | int err = -1; |
935 | ssize_t ret; | |
b8aa1682 | 936 | struct cds_wfq_node *node; |
58eb9381 | 937 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
938 | |
939 | DBG("[thread] Relay dispatcher started"); | |
940 | ||
55706a7d MD |
941 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
942 | ||
9b5e0863 MD |
943 | if (testpoint(relayd_thread_dispatcher)) { |
944 | goto error_testpoint; | |
945 | } | |
946 | ||
f385ae0a MD |
947 | health_code_update(); |
948 | ||
26c9d55e | 949 | while (!CMM_LOAD_SHARED(dispatch_thread_exit)) { |
f385ae0a MD |
950 | health_code_update(); |
951 | ||
b8aa1682 | 952 | /* Atomically prepare the queue futex */ |
58eb9381 | 953 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 JD |
954 | |
955 | do { | |
f385ae0a MD |
956 | health_code_update(); |
957 | ||
b8aa1682 | 958 | /* Dequeue commands */ |
58eb9381 | 959 | node = cds_wfq_dequeue_blocking(&relay_conn_queue.queue); |
b8aa1682 JD |
960 | if (node == NULL) { |
961 | DBG("Woken up but nothing in the relay command queue"); | |
962 | /* Continue thread execution */ | |
963 | break; | |
964 | } | |
58eb9381 | 965 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 966 | |
58eb9381 | 967 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
968 | |
969 | /* | |
58eb9381 DG |
970 | * Inform worker thread of the new request. This call is blocking |
971 | * so we can be assured that the data will be read at some point in | |
972 | * time or wait to the end of the world :) | |
b8aa1682 | 973 | */ |
58eb9381 DG |
974 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
975 | if (ret < 0) { | |
976 | PERROR("write connection pipe"); | |
977 | connection_destroy(new_conn); | |
b8aa1682 JD |
978 | goto error; |
979 | } | |
980 | } while (node != NULL); | |
981 | ||
982 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 983 | health_poll_entry(); |
58eb9381 | 984 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 985 | health_poll_exit(); |
b8aa1682 JD |
986 | } |
987 | ||
f385ae0a MD |
988 | /* Normal exit, no error */ |
989 | err = 0; | |
990 | ||
b8aa1682 | 991 | error: |
9b5e0863 | 992 | error_testpoint: |
f385ae0a MD |
993 | if (err) { |
994 | health_error(); | |
995 | ERR("Health error occurred in %s", __func__); | |
996 | } | |
55706a7d | 997 | health_unregister(health_relayd); |
b8aa1682 JD |
998 | DBG("Dispatch thread dying"); |
999 | stop_threads(); | |
1000 | return NULL; | |
1001 | } | |
1002 | ||
2a174661 | 1003 | static void try_close_streams(struct relay_session *session) |
d3e2ba59 | 1004 | { |
2a174661 | 1005 | struct ctf_trace *ctf_trace; |
94d49140 JD |
1006 | struct lttng_ht_iter iter; |
1007 | ||
2a174661 | 1008 | assert(session); |
94d49140 | 1009 | |
2a174661 DG |
1010 | pthread_mutex_lock(&session->viewer_ready_lock); |
1011 | rcu_read_lock(); | |
1012 | cds_lfht_for_each_entry(session->ctf_traces_ht->ht, &iter.iter, ctf_trace, | |
1013 | node.node) { | |
1014 | struct relay_stream *stream; | |
94d49140 | 1015 | |
2a174661 DG |
1016 | /* Close streams. */ |
1017 | cds_list_for_each_entry(stream, &ctf_trace->stream_list, trace_list) { | |
1018 | stream_close(session, stream); | |
94d49140 | 1019 | } |
94d49140 | 1020 | |
2a174661 DG |
1021 | ctf_trace->invalid_flag = 1; |
1022 | ctf_trace_try_destroy(session, ctf_trace); | |
157df586 | 1023 | } |
2a174661 DG |
1024 | rcu_read_unlock(); |
1025 | pthread_mutex_unlock(&session->viewer_ready_lock); | |
94d49140 JD |
1026 | } |
1027 | ||
b8aa1682 | 1028 | /* |
2a174661 | 1029 | * Try to destroy a session within a connection. |
b8aa1682 | 1030 | */ |
58eb9381 | 1031 | static void destroy_session(struct relay_session *session, |
d3e2ba59 | 1032 | struct lttng_ht *sessions_ht) |
b8aa1682 | 1033 | { |
58eb9381 | 1034 | assert(session); |
2a174661 | 1035 | assert(sessions_ht); |
b8aa1682 | 1036 | |
2a174661 | 1037 | /* Indicate that this session can be destroyed from now on. */ |
58eb9381 | 1038 | session->close_flag = 1; |
b8aa1682 | 1039 | |
58eb9381 | 1040 | try_close_streams(session); |
5b6d8097 | 1041 | |
2a174661 DG |
1042 | /* |
1043 | * This will try to delete and destroy the session if no viewer is attached | |
1044 | * to it meaning the refcount is down to zero. | |
1045 | */ | |
58eb9381 | 1046 | session_try_destroy(sessions_ht, session); |
b8aa1682 JD |
1047 | } |
1048 | ||
1c20f0e2 JD |
1049 | /* |
1050 | * Copy index data from the control port to a given index object. | |
1051 | */ | |
1052 | static void copy_index_control_data(struct relay_index *index, | |
1053 | struct lttcomm_relayd_index *data) | |
1054 | { | |
1055 | assert(index); | |
1056 | assert(data); | |
1057 | ||
1058 | /* | |
1059 | * The index on disk is encoded in big endian, so we don't need to convert | |
1060 | * the data received on the network. The data_offset value is NEVER | |
1061 | * modified here and is updated by the data thread. | |
1062 | */ | |
1063 | index->index_data.packet_size = data->packet_size; | |
1064 | index->index_data.content_size = data->content_size; | |
1065 | index->index_data.timestamp_begin = data->timestamp_begin; | |
1066 | index->index_data.timestamp_end = data->timestamp_end; | |
1067 | index->index_data.events_discarded = data->events_discarded; | |
1068 | index->index_data.stream_id = data->stream_id; | |
1069 | } | |
1070 | ||
c5b6f4f0 DG |
1071 | /* |
1072 | * Handle the RELAYD_CREATE_SESSION command. | |
1073 | * | |
1074 | * On success, send back the session id or else return a negative value. | |
1075 | */ | |
1076 | static | |
1077 | int relay_create_session(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1078 | struct relay_connection *conn) |
c5b6f4f0 DG |
1079 | { |
1080 | int ret = 0, send_ret; | |
1081 | struct relay_session *session; | |
1082 | struct lttcomm_relayd_status_session reply; | |
1083 | ||
1084 | assert(recv_hdr); | |
58eb9381 | 1085 | assert(conn); |
c5b6f4f0 DG |
1086 | |
1087 | memset(&reply, 0, sizeof(reply)); | |
1088 | ||
2a174661 DG |
1089 | session = session_create(); |
1090 | if (!session) { | |
c5b6f4f0 DG |
1091 | ret = -1; |
1092 | goto error; | |
1093 | } | |
58eb9381 DG |
1094 | session->minor = conn->minor; |
1095 | session->major = conn->major; | |
1096 | conn->session_id = session->id; | |
1097 | conn->session = session; | |
c5b6f4f0 DG |
1098 | |
1099 | reply.session_id = htobe64(session->id); | |
1100 | ||
58eb9381 | 1101 | switch (conn->minor) { |
2a174661 DG |
1102 | case 1: |
1103 | case 2: | |
1104 | case 3: | |
1105 | break; | |
1106 | case 4: /* LTTng sessiond 2.4 */ | |
1107 | default: | |
58eb9381 | 1108 | ret = cmd_create_session_2_4(conn, session); |
d3e2ba59 JD |
1109 | } |
1110 | ||
58eb9381 | 1111 | lttng_ht_add_unique_u64(conn->sessions_ht, &session->session_n); |
c5b6f4f0 DG |
1112 | DBG("Created session %" PRIu64, session->id); |
1113 | ||
1114 | error: | |
1115 | if (ret < 0) { | |
1116 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1117 | } else { | |
1118 | reply.ret_code = htobe32(LTTNG_OK); | |
1119 | } | |
1120 | ||
58eb9381 | 1121 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c5b6f4f0 DG |
1122 | if (send_ret < 0) { |
1123 | ERR("Relayd sending session id"); | |
4169f5ad | 1124 | ret = send_ret; |
c5b6f4f0 DG |
1125 | } |
1126 | ||
1127 | return ret; | |
1128 | } | |
1129 | ||
a4baae1b JD |
1130 | /* |
1131 | * When we have received all the streams and the metadata for a channel, | |
1132 | * we make them visible to the viewer threads. | |
1133 | */ | |
1134 | static | |
58eb9381 | 1135 | void set_viewer_ready_flag(struct relay_connection *conn) |
a4baae1b | 1136 | { |
2a174661 | 1137 | struct relay_stream *stream, *tmp_stream; |
a4baae1b | 1138 | |
58eb9381 DG |
1139 | pthread_mutex_lock(&conn->session->viewer_ready_lock); |
1140 | cds_list_for_each_entry_safe(stream, tmp_stream, &conn->recv_head, | |
2a174661 | 1141 | recv_list) { |
a4baae1b | 1142 | stream->viewer_ready = 1; |
2a174661 | 1143 | cds_list_del(&stream->recv_list); |
a4baae1b | 1144 | } |
58eb9381 | 1145 | pthread_mutex_unlock(&conn->session->viewer_ready_lock); |
a4baae1b JD |
1146 | return; |
1147 | } | |
1148 | ||
1149 | /* | |
1150 | * Add a recv handle node to the connection recv list with the given stream | |
1151 | * handle. A new node is allocated thus must be freed when the node is deleted | |
1152 | * from the list. | |
1153 | */ | |
58eb9381 DG |
1154 | static void queue_stream(struct relay_stream *stream, |
1155 | struct relay_connection *conn) | |
a4baae1b | 1156 | { |
58eb9381 | 1157 | assert(conn); |
2a174661 | 1158 | assert(stream); |
a4baae1b | 1159 | |
58eb9381 | 1160 | cds_list_add(&stream->recv_list, &conn->recv_head); |
a4baae1b JD |
1161 | } |
1162 | ||
b8aa1682 JD |
1163 | /* |
1164 | * relay_add_stream: allocate a new stream for a session | |
1165 | */ | |
1166 | static | |
1167 | int relay_add_stream(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1168 | struct relay_connection *conn) |
b8aa1682 | 1169 | { |
2a174661 | 1170 | int ret, send_ret; |
58eb9381 | 1171 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1172 | struct relay_stream *stream = NULL; |
1173 | struct lttcomm_relayd_status_stream reply; | |
2a174661 | 1174 | struct ctf_trace *trace; |
b8aa1682 | 1175 | |
58eb9381 | 1176 | if (!session || conn->version_check_done == 0) { |
b8aa1682 JD |
1177 | ERR("Trying to add a stream before version check"); |
1178 | ret = -1; | |
1179 | goto end_no_session; | |
1180 | } | |
1181 | ||
b8aa1682 JD |
1182 | stream = zmalloc(sizeof(struct relay_stream)); |
1183 | if (stream == NULL) { | |
1184 | PERROR("relay stream zmalloc"); | |
1185 | ret = -1; | |
1186 | goto end_no_session; | |
1187 | } | |
1188 | ||
58eb9381 | 1189 | switch (conn->minor) { |
0f907de1 | 1190 | case 1: /* LTTng sessiond 2.1 */ |
58eb9381 | 1191 | ret = cmd_recv_stream_2_1(conn, stream); |
0f907de1 JD |
1192 | break; |
1193 | case 2: /* LTTng sessiond 2.2 */ | |
1194 | default: | |
58eb9381 | 1195 | ret = cmd_recv_stream_2_2(conn, stream); |
0f907de1 JD |
1196 | break; |
1197 | } | |
1198 | if (ret < 0) { | |
1199 | goto err_free_stream; | |
1200 | } | |
1201 | ||
9d1bbf21 | 1202 | rcu_read_lock(); |
b8aa1682 | 1203 | stream->stream_handle = ++last_relay_stream_id; |
173af62f | 1204 | stream->prev_seq = -1ULL; |
2a174661 | 1205 | stream->session_id = session->id; |
1c20f0e2 | 1206 | stream->index_fd = -1; |
d3e2ba59 | 1207 | stream->read_index_fd = -1; |
bb8cf449 | 1208 | stream->ctf_stream_id = -1ULL; |
2a174661 | 1209 | lttng_ht_node_init_u64(&stream->node, stream->stream_handle); |
d3e2ba59 | 1210 | pthread_mutex_init(&stream->lock, NULL); |
b8aa1682 | 1211 | |
0f907de1 | 1212 | ret = utils_mkdir_recursive(stream->path_name, S_IRWXU | S_IRWXG); |
b8aa1682 | 1213 | if (ret < 0) { |
b8aa1682 JD |
1214 | ERR("relay creating output directory"); |
1215 | goto end; | |
1216 | } | |
1217 | ||
be96a7d1 DG |
1218 | /* |
1219 | * No need to use run_as API here because whatever we receives, the relayd | |
1220 | * uses its own credentials for the stream files. | |
1221 | */ | |
0f907de1 | 1222 | ret = utils_create_stream_file(stream->path_name, stream->channel_name, |
1c20f0e2 | 1223 | stream->tracefile_size, 0, relayd_uid, relayd_gid, NULL); |
b8aa1682 | 1224 | if (ret < 0) { |
0f907de1 | 1225 | ERR("Create output file"); |
b8aa1682 JD |
1226 | goto end; |
1227 | } | |
b8aa1682 | 1228 | stream->fd = ret; |
0f907de1 JD |
1229 | if (stream->tracefile_size) { |
1230 | DBG("Tracefile %s/%s_0 created", stream->path_name, stream->channel_name); | |
1231 | } else { | |
1232 | DBG("Tracefile %s/%s created", stream->path_name, stream->channel_name); | |
1233 | } | |
b8aa1682 | 1234 | |
2a174661 DG |
1235 | trace = ctf_trace_find_by_path(session->ctf_traces_ht, stream->path_name); |
1236 | if (!trace) { | |
1237 | trace = ctf_trace_create(stream->path_name); | |
1238 | if (!trace) { | |
d3e2ba59 JD |
1239 | ret = -1; |
1240 | goto end; | |
1241 | } | |
2a174661 DG |
1242 | ctf_trace_add(session->ctf_traces_ht, trace); |
1243 | } | |
1244 | ctf_trace_get_ref(trace); | |
1245 | ||
1246 | if (!strncmp(stream->channel_name, DEFAULT_METADATA_NAME, NAME_MAX)) { | |
1247 | stream->metadata_flag = 1; | |
1248 | /* Assign quick reference to the metadata stream in the trace. */ | |
1249 | trace->metadata_stream = stream; | |
d3e2ba59 | 1250 | } |
d3e2ba59 | 1251 | |
a4baae1b | 1252 | /* |
2a174661 DG |
1253 | * Add the stream in the recv list of the connection. Once the end stream |
1254 | * message is received, this list is emptied and streams are set with the | |
1255 | * viewer ready flag. | |
a4baae1b | 1256 | */ |
58eb9381 | 1257 | queue_stream(stream, conn); |
a4baae1b | 1258 | |
2a174661 DG |
1259 | /* |
1260 | * Both in the ctf_trace object and the global stream ht since the data | |
1261 | * side of the relayd does not have the concept of session. | |
1262 | */ | |
1263 | lttng_ht_add_unique_u64(relay_streams_ht, &stream->node); | |
1264 | cds_list_add_tail(&stream->trace_list, &trace->stream_list); | |
b8aa1682 | 1265 | |
87b576ec | 1266 | session->stream_count++; |
d3e2ba59 | 1267 | |
1c20f0e2 JD |
1268 | DBG("Relay new stream added %s with ID %" PRIu64, stream->channel_name, |
1269 | stream->stream_handle); | |
b8aa1682 JD |
1270 | |
1271 | end: | |
53efb85a | 1272 | memset(&reply, 0, sizeof(reply)); |
5af40280 | 1273 | reply.handle = htobe64(stream->stream_handle); |
b8aa1682 JD |
1274 | /* send the session id to the client or a negative return code on error */ |
1275 | if (ret < 0) { | |
f73fabfd | 1276 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
5af40280 CB |
1277 | /* stream was not properly added to the ht, so free it */ |
1278 | free(stream); | |
b8aa1682 | 1279 | } else { |
f73fabfd | 1280 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1281 | } |
5af40280 | 1282 | |
58eb9381 | 1283 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1284 | sizeof(struct lttcomm_relayd_status_stream), 0); |
1285 | if (send_ret < 0) { | |
1286 | ERR("Relay sending stream id"); | |
4169f5ad | 1287 | ret = send_ret; |
b8aa1682 | 1288 | } |
9d1bbf21 | 1289 | rcu_read_unlock(); |
b8aa1682 JD |
1290 | |
1291 | end_no_session: | |
1292 | return ret; | |
0f907de1 JD |
1293 | |
1294 | err_free_stream: | |
1295 | free(stream->path_name); | |
1296 | free(stream->channel_name); | |
1297 | free(stream); | |
1298 | return ret; | |
b8aa1682 JD |
1299 | } |
1300 | ||
173af62f DG |
1301 | /* |
1302 | * relay_close_stream: close a specific stream | |
1303 | */ | |
1304 | static | |
1305 | int relay_close_stream(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1306 | struct relay_connection *conn) |
173af62f | 1307 | { |
94d49140 | 1308 | int ret, send_ret; |
58eb9381 | 1309 | struct relay_session *session = conn->session; |
173af62f DG |
1310 | struct lttcomm_relayd_close_stream stream_info; |
1311 | struct lttcomm_relayd_generic_reply reply; | |
1312 | struct relay_stream *stream; | |
173af62f DG |
1313 | |
1314 | DBG("Close stream received"); | |
1315 | ||
58eb9381 | 1316 | if (!session || conn->version_check_done == 0) { |
173af62f DG |
1317 | ERR("Trying to close a stream before version check"); |
1318 | ret = -1; | |
1319 | goto end_no_session; | |
1320 | } | |
1321 | ||
58eb9381 | 1322 | ret = conn->sock->ops->recvmsg(conn->sock, &stream_info, |
7c5aef62 | 1323 | sizeof(struct lttcomm_relayd_close_stream), 0); |
173af62f | 1324 | if (ret < sizeof(struct lttcomm_relayd_close_stream)) { |
a6cd2b97 DG |
1325 | if (ret == 0) { |
1326 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1327 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1328 | } else { |
1329 | ERR("Relay didn't receive valid add_stream struct size : %d", ret); | |
1330 | } | |
173af62f DG |
1331 | ret = -1; |
1332 | goto end_no_session; | |
1333 | } | |
1334 | ||
1335 | rcu_read_lock(); | |
2a174661 DG |
1336 | stream = stream_find_by_id(relay_streams_ht, |
1337 | be64toh(stream_info.stream_id)); | |
173af62f DG |
1338 | if (!stream) { |
1339 | ret = -1; | |
1340 | goto end_unlock; | |
1341 | } | |
1342 | ||
8e2583a4 | 1343 | stream->last_net_seq_num = be64toh(stream_info.last_net_seq_num); |
173af62f | 1344 | stream->close_flag = 1; |
87b576ec JD |
1345 | session->stream_count--; |
1346 | assert(session->stream_count >= 0); | |
173af62f | 1347 | |
2a174661 DG |
1348 | /* Check if we can close it or else the data will do it. */ |
1349 | try_close_stream(session, stream); | |
173af62f DG |
1350 | |
1351 | end_unlock: | |
1352 | rcu_read_unlock(); | |
1353 | ||
53efb85a | 1354 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1355 | if (ret < 0) { |
f73fabfd | 1356 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1357 | } else { |
f73fabfd | 1358 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1359 | } |
58eb9381 | 1360 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f DG |
1361 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1362 | if (send_ret < 0) { | |
1363 | ERR("Relay sending stream id"); | |
4169f5ad | 1364 | ret = send_ret; |
173af62f DG |
1365 | } |
1366 | ||
1367 | end_no_session: | |
1368 | return ret; | |
1369 | } | |
1370 | ||
b8aa1682 JD |
1371 | /* |
1372 | * relay_unknown_command: send -1 if received unknown command | |
1373 | */ | |
1374 | static | |
58eb9381 | 1375 | void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1376 | { |
1377 | struct lttcomm_relayd_generic_reply reply; | |
1378 | int ret; | |
1379 | ||
53efb85a | 1380 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1381 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
58eb9381 | 1382 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1383 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1384 | if (ret < 0) { | |
1385 | ERR("Relay sending unknown command"); | |
1386 | } | |
1387 | } | |
1388 | ||
1389 | /* | |
1390 | * relay_start: send an acknowledgment to the client to tell if we are | |
1391 | * ready to receive data. We are ready if a session is established. | |
1392 | */ | |
1393 | static | |
1394 | int relay_start(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1395 | struct relay_connection *conn) |
b8aa1682 | 1396 | { |
f73fabfd | 1397 | int ret = htobe32(LTTNG_OK); |
b8aa1682 | 1398 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1399 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1400 | |
1401 | if (!session) { | |
1402 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1403 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1404 | } |
1405 | ||
53efb85a | 1406 | memset(&reply, 0, sizeof(reply)); |
b8aa1682 | 1407 | reply.ret_code = ret; |
58eb9381 | 1408 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 JD |
1409 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
1410 | if (ret < 0) { | |
1411 | ERR("Relay sending start ack"); | |
1412 | } | |
1413 | ||
1414 | return ret; | |
1415 | } | |
1416 | ||
1d4dfdef DG |
1417 | /* |
1418 | * Append padding to the file pointed by the file descriptor fd. | |
1419 | */ | |
1420 | static int write_padding_to_file(int fd, uint32_t size) | |
1421 | { | |
6cd525e8 | 1422 | ssize_t ret = 0; |
1d4dfdef DG |
1423 | char *zeros; |
1424 | ||
1425 | if (size == 0) { | |
1426 | goto end; | |
1427 | } | |
1428 | ||
1429 | zeros = zmalloc(size); | |
1430 | if (zeros == NULL) { | |
1431 | PERROR("zmalloc zeros for padding"); | |
1432 | ret = -1; | |
1433 | goto end; | |
1434 | } | |
1435 | ||
6cd525e8 MD |
1436 | ret = lttng_write(fd, zeros, size); |
1437 | if (ret < size) { | |
1d4dfdef DG |
1438 | PERROR("write padding to file"); |
1439 | } | |
1440 | ||
e986c7a1 DG |
1441 | free(zeros); |
1442 | ||
1d4dfdef DG |
1443 | end: |
1444 | return ret; | |
1445 | } | |
1446 | ||
b8aa1682 JD |
1447 | /* |
1448 | * relay_recv_metadata: receive the metada for the session. | |
1449 | */ | |
1450 | static | |
1451 | int relay_recv_metadata(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1452 | struct relay_connection *conn) |
b8aa1682 | 1453 | { |
f73fabfd | 1454 | int ret = htobe32(LTTNG_OK); |
6cd525e8 | 1455 | ssize_t size_ret; |
58eb9381 | 1456 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1457 | struct lttcomm_relayd_metadata_payload *metadata_struct; |
1458 | struct relay_stream *metadata_stream; | |
1459 | uint64_t data_size, payload_size; | |
2a174661 | 1460 | struct ctf_trace *ctf_trace; |
b8aa1682 JD |
1461 | |
1462 | if (!session) { | |
1463 | ERR("Metadata sent before version check"); | |
1464 | ret = -1; | |
1465 | goto end; | |
1466 | } | |
1467 | ||
f6416125 MD |
1468 | data_size = payload_size = be64toh(recv_hdr->data_size); |
1469 | if (data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { | |
1470 | ERR("Incorrect data size"); | |
1471 | ret = -1; | |
1472 | goto end; | |
1473 | } | |
1474 | payload_size -= sizeof(struct lttcomm_relayd_metadata_payload); | |
1475 | ||
b8aa1682 | 1476 | if (data_buffer_size < data_size) { |
d7b3776f | 1477 | /* In case the realloc fails, we can free the memory */ |
c617c0c6 MD |
1478 | char *tmp_data_ptr; |
1479 | ||
1480 | tmp_data_ptr = realloc(data_buffer, data_size); | |
1481 | if (!tmp_data_ptr) { | |
b8aa1682 | 1482 | ERR("Allocating data buffer"); |
c617c0c6 | 1483 | free(data_buffer); |
b8aa1682 JD |
1484 | ret = -1; |
1485 | goto end; | |
1486 | } | |
c617c0c6 | 1487 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
1488 | data_buffer_size = data_size; |
1489 | } | |
1490 | memset(data_buffer, 0, data_size); | |
77c7c900 | 1491 | DBG2("Relay receiving metadata, waiting for %" PRIu64 " bytes", data_size); |
58eb9381 | 1492 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 1493 | if (ret < 0 || ret != data_size) { |
a6cd2b97 DG |
1494 | if (ret == 0) { |
1495 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1496 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1497 | } else { |
1498 | ERR("Relay didn't receive the whole metadata"); | |
1499 | } | |
b8aa1682 | 1500 | ret = -1; |
b8aa1682 JD |
1501 | goto end; |
1502 | } | |
1503 | metadata_struct = (struct lttcomm_relayd_metadata_payload *) data_buffer; | |
9d1bbf21 MD |
1504 | |
1505 | rcu_read_lock(); | |
2a174661 | 1506 | metadata_stream = stream_find_by_id(relay_streams_ht, |
d3e2ba59 | 1507 | be64toh(metadata_struct->stream_id)); |
b8aa1682 JD |
1508 | if (!metadata_stream) { |
1509 | ret = -1; | |
9d1bbf21 | 1510 | goto end_unlock; |
b8aa1682 JD |
1511 | } |
1512 | ||
6cd525e8 MD |
1513 | size_ret = lttng_write(metadata_stream->fd, metadata_struct->payload, |
1514 | payload_size); | |
1515 | if (size_ret < payload_size) { | |
b8aa1682 JD |
1516 | ERR("Relay error writing metadata on file"); |
1517 | ret = -1; | |
9d1bbf21 | 1518 | goto end_unlock; |
b8aa1682 | 1519 | } |
1d4dfdef DG |
1520 | |
1521 | ret = write_padding_to_file(metadata_stream->fd, | |
1522 | be32toh(metadata_struct->padding_size)); | |
1523 | if (ret < 0) { | |
1524 | goto end_unlock; | |
1525 | } | |
2a174661 DG |
1526 | |
1527 | ctf_trace = ctf_trace_find_by_path(session->ctf_traces_ht, | |
1528 | metadata_stream->path_name); | |
1529 | assert(ctf_trace); | |
1530 | ctf_trace->metadata_received += | |
d3e2ba59 | 1531 | payload_size + be32toh(metadata_struct->padding_size); |
1d4dfdef | 1532 | |
b8aa1682 JD |
1533 | DBG2("Relay metadata written"); |
1534 | ||
9d1bbf21 | 1535 | end_unlock: |
6e3c5836 | 1536 | rcu_read_unlock(); |
b8aa1682 JD |
1537 | end: |
1538 | return ret; | |
1539 | } | |
1540 | ||
1541 | /* | |
1542 | * relay_send_version: send relayd version number | |
1543 | */ | |
1544 | static | |
1545 | int relay_send_version(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1546 | struct relay_connection *conn) |
b8aa1682 | 1547 | { |
7f51dcba | 1548 | int ret; |
092b6259 | 1549 | struct lttcomm_relayd_version reply, msg; |
b8aa1682 | 1550 | |
58eb9381 | 1551 | assert(conn); |
c5b6f4f0 | 1552 | |
58eb9381 | 1553 | conn->version_check_done = 1; |
b8aa1682 | 1554 | |
092b6259 | 1555 | /* Get version from the other side. */ |
58eb9381 | 1556 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
092b6259 | 1557 | if (ret < 0 || ret != sizeof(msg)) { |
a6cd2b97 DG |
1558 | if (ret == 0) { |
1559 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1560 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1561 | } else { |
1562 | ERR("Relay failed to receive the version values."); | |
1563 | } | |
092b6259 | 1564 | ret = -1; |
092b6259 DG |
1565 | goto end; |
1566 | } | |
1567 | ||
53efb85a | 1568 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1569 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1570 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1571 | |
1572 | /* Major versions must be the same */ | |
1573 | if (reply.major != be32toh(msg.major)) { | |
6151a90f JD |
1574 | DBG("Incompatible major versions (%u vs %u), deleting session", |
1575 | reply.major, be32toh(msg.major)); | |
58eb9381 | 1576 | destroy_session(conn->session, conn->sessions_ht); |
d4519fa3 JD |
1577 | ret = 0; |
1578 | goto end; | |
1579 | } | |
1580 | ||
58eb9381 | 1581 | conn->major = reply.major; |
0f907de1 JD |
1582 | /* We adapt to the lowest compatible version */ |
1583 | if (reply.minor <= be32toh(msg.minor)) { | |
58eb9381 | 1584 | conn->minor = reply.minor; |
0f907de1 | 1585 | } else { |
58eb9381 | 1586 | conn->minor = be32toh(msg.minor); |
0f907de1 JD |
1587 | } |
1588 | ||
6151a90f JD |
1589 | reply.major = htobe32(reply.major); |
1590 | reply.minor = htobe32(reply.minor); | |
58eb9381 | 1591 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
6151a90f JD |
1592 | sizeof(struct lttcomm_relayd_version), 0); |
1593 | if (ret < 0) { | |
1594 | ERR("Relay sending version"); | |
1595 | } | |
1596 | ||
58eb9381 DG |
1597 | DBG("Version check done using protocol %u.%u", conn->major, |
1598 | conn->minor); | |
b8aa1682 JD |
1599 | |
1600 | end: | |
1601 | return ret; | |
1602 | } | |
1603 | ||
c8f59ee5 | 1604 | /* |
6d805429 | 1605 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 DG |
1606 | */ |
1607 | static | |
6d805429 | 1608 | int relay_data_pending(struct lttcomm_relayd_hdr *recv_hdr, |
58eb9381 | 1609 | struct relay_connection *conn) |
c8f59ee5 | 1610 | { |
58eb9381 | 1611 | struct relay_session *session = conn->session; |
6d805429 | 1612 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1613 | struct lttcomm_relayd_generic_reply reply; |
1614 | struct relay_stream *stream; | |
1615 | int ret; | |
c8f59ee5 DG |
1616 | uint64_t last_net_seq_num, stream_id; |
1617 | ||
6d805429 | 1618 | DBG("Data pending command received"); |
c8f59ee5 | 1619 | |
58eb9381 | 1620 | if (!session || conn->version_check_done == 0) { |
c8f59ee5 DG |
1621 | ERR("Trying to check for data before version check"); |
1622 | ret = -1; | |
1623 | goto end_no_session; | |
1624 | } | |
1625 | ||
58eb9381 | 1626 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
c8f59ee5 | 1627 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1628 | if (ret == 0) { |
1629 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1630 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1631 | } else { |
1632 | ERR("Relay didn't receive valid data_pending struct size : %d", | |
1633 | ret); | |
1634 | } | |
c8f59ee5 DG |
1635 | ret = -1; |
1636 | goto end_no_session; | |
1637 | } | |
1638 | ||
1639 | stream_id = be64toh(msg.stream_id); | |
1640 | last_net_seq_num = be64toh(msg.last_net_seq_num); | |
1641 | ||
1642 | rcu_read_lock(); | |
2a174661 | 1643 | stream = stream_find_by_id(relay_streams_ht, stream_id); |
de91f48a | 1644 | if (stream == NULL) { |
c8f59ee5 DG |
1645 | ret = -1; |
1646 | goto end_unlock; | |
1647 | } | |
1648 | ||
6d805429 | 1649 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
c8f59ee5 DG |
1650 | " and last_seq %" PRIu64, stream_id, stream->prev_seq, |
1651 | last_net_seq_num); | |
1652 | ||
33832e64 | 1653 | /* Avoid wrapping issue */ |
39df6d9f | 1654 | if (((int64_t) (stream->prev_seq - last_net_seq_num)) >= 0) { |
6d805429 | 1655 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1656 | ret = 0; |
6d805429 DG |
1657 | } else { |
1658 | /* Data still being streamed thus pending */ | |
1659 | ret = 1; | |
c8f59ee5 DG |
1660 | } |
1661 | ||
f7079f67 DG |
1662 | /* Pending check is now done. */ |
1663 | stream->data_pending_check_done = 1; | |
1664 | ||
c8f59ee5 DG |
1665 | end_unlock: |
1666 | rcu_read_unlock(); | |
1667 | ||
53efb85a | 1668 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1669 | reply.ret_code = htobe32(ret); |
58eb9381 | 1670 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1671 | if (ret < 0) { |
6d805429 | 1672 | ERR("Relay data pending ret code failed"); |
c8f59ee5 DG |
1673 | } |
1674 | ||
1675 | end_no_session: | |
1676 | return ret; | |
1677 | } | |
1678 | ||
1679 | /* | |
1680 | * Wait for the control socket to reach a quiescent state. | |
1681 | * | |
1682 | * Note that for now, when receiving this command from the session daemon, this | |
1683 | * means that every subsequent commands or data received on the control socket | |
1684 | * has been handled. So, this is why we simply return OK here. | |
1685 | */ | |
1686 | static | |
1687 | int relay_quiescent_control(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1688 | struct relay_connection *conn) |
c8f59ee5 DG |
1689 | { |
1690 | int ret; | |
ad7051c0 DG |
1691 | uint64_t stream_id; |
1692 | struct relay_stream *stream; | |
1693 | struct lttng_ht_iter iter; | |
1694 | struct lttcomm_relayd_quiescent_control msg; | |
c8f59ee5 DG |
1695 | struct lttcomm_relayd_generic_reply reply; |
1696 | ||
1697 | DBG("Checking quiescent state on control socket"); | |
1698 | ||
58eb9381 | 1699 | if (!conn->session || conn->version_check_done == 0) { |
ad7051c0 DG |
1700 | ERR("Trying to check for data before version check"); |
1701 | ret = -1; | |
1702 | goto end_no_session; | |
1703 | } | |
1704 | ||
58eb9381 | 1705 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
ad7051c0 | 1706 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1707 | if (ret == 0) { |
1708 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1709 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1710 | } else { |
1711 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1712 | ret); | |
1713 | } | |
ad7051c0 DG |
1714 | ret = -1; |
1715 | goto end_no_session; | |
1716 | } | |
1717 | ||
1718 | stream_id = be64toh(msg.stream_id); | |
1719 | ||
1720 | rcu_read_lock(); | |
d3e2ba59 | 1721 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1722 | node.node) { |
ad7051c0 DG |
1723 | if (stream->stream_handle == stream_id) { |
1724 | stream->data_pending_check_done = 1; | |
1725 | DBG("Relay quiescent control pending flag set to %" PRIu64, | |
1726 | stream_id); | |
1727 | break; | |
1728 | } | |
1729 | } | |
1730 | rcu_read_unlock(); | |
1731 | ||
53efb85a | 1732 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1733 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 1734 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c8f59ee5 | 1735 | if (ret < 0) { |
6d805429 | 1736 | ERR("Relay data quiescent control ret code failed"); |
c8f59ee5 DG |
1737 | } |
1738 | ||
ad7051c0 | 1739 | end_no_session: |
c8f59ee5 DG |
1740 | return ret; |
1741 | } | |
1742 | ||
f7079f67 DG |
1743 | /* |
1744 | * Initialize a data pending command. This means that a client is about to ask | |
1745 | * for data pending for each stream he/she holds. Simply iterate over all | |
1746 | * streams of a session and set the data_pending_check_done flag. | |
1747 | * | |
1748 | * This command returns to the client a LTTNG_OK code. | |
1749 | */ | |
1750 | static | |
1751 | int relay_begin_data_pending(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1752 | struct relay_connection *conn) |
f7079f67 DG |
1753 | { |
1754 | int ret; | |
1755 | struct lttng_ht_iter iter; | |
1756 | struct lttcomm_relayd_begin_data_pending msg; | |
1757 | struct lttcomm_relayd_generic_reply reply; | |
1758 | struct relay_stream *stream; | |
1759 | uint64_t session_id; | |
1760 | ||
1761 | assert(recv_hdr); | |
58eb9381 | 1762 | assert(conn); |
f7079f67 DG |
1763 | |
1764 | DBG("Init streams for data pending"); | |
1765 | ||
58eb9381 | 1766 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1767 | ERR("Trying to check for data before version check"); |
1768 | ret = -1; | |
1769 | goto end_no_session; | |
1770 | } | |
1771 | ||
58eb9381 | 1772 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1773 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1774 | if (ret == 0) { |
1775 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1776 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1777 | } else { |
1778 | ERR("Relay didn't receive valid begin data_pending struct size: %d", | |
1779 | ret); | |
1780 | } | |
f7079f67 DG |
1781 | ret = -1; |
1782 | goto end_no_session; | |
1783 | } | |
1784 | ||
1785 | session_id = be64toh(msg.session_id); | |
1786 | ||
1787 | /* | |
1788 | * Iterate over all streams to set the begin data pending flag. For now, the | |
1789 | * streams are indexed by stream handle so we have to iterate over all | |
1790 | * streams to find the one associated with the right session_id. | |
1791 | */ | |
1792 | rcu_read_lock(); | |
d3e2ba59 | 1793 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 DG |
1794 | node.node) { |
1795 | if (stream->session_id == session_id) { | |
f7079f67 DG |
1796 | stream->data_pending_check_done = 0; |
1797 | DBG("Set begin data pending flag to stream %" PRIu64, | |
1798 | stream->stream_handle); | |
1799 | } | |
1800 | } | |
1801 | rcu_read_unlock(); | |
1802 | ||
53efb85a | 1803 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1804 | /* All good, send back reply. */ |
1805 | reply.ret_code = htobe32(LTTNG_OK); | |
1806 | ||
58eb9381 | 1807 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1808 | if (ret < 0) { |
1809 | ERR("Relay begin data pending send reply failed"); | |
1810 | } | |
1811 | ||
1812 | end_no_session: | |
1813 | return ret; | |
1814 | } | |
1815 | ||
1816 | /* | |
1817 | * End data pending command. This will check, for a given session id, if each | |
1818 | * stream associated with it has its data_pending_check_done flag set. If not, | |
1819 | * this means that the client lost track of the stream but the data is still | |
1820 | * being streamed on our side. In this case, we inform the client that data is | |
1821 | * inflight. | |
1822 | * | |
1823 | * Return to the client if there is data in flight or not with a ret_code. | |
1824 | */ | |
1825 | static | |
1826 | int relay_end_data_pending(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1827 | struct relay_connection *conn) |
f7079f67 DG |
1828 | { |
1829 | int ret; | |
1830 | struct lttng_ht_iter iter; | |
1831 | struct lttcomm_relayd_end_data_pending msg; | |
1832 | struct lttcomm_relayd_generic_reply reply; | |
1833 | struct relay_stream *stream; | |
1834 | uint64_t session_id; | |
1835 | uint32_t is_data_inflight = 0; | |
1836 | ||
1837 | assert(recv_hdr); | |
58eb9381 | 1838 | assert(conn); |
f7079f67 DG |
1839 | |
1840 | DBG("End data pending command"); | |
1841 | ||
58eb9381 | 1842 | if (!conn->session || conn->version_check_done == 0) { |
f7079f67 DG |
1843 | ERR("Trying to check for data before version check"); |
1844 | ret = -1; | |
1845 | goto end_no_session; | |
1846 | } | |
1847 | ||
58eb9381 | 1848 | ret = conn->sock->ops->recvmsg(conn->sock, &msg, sizeof(msg), 0); |
f7079f67 | 1849 | if (ret < sizeof(msg)) { |
a6cd2b97 DG |
1850 | if (ret == 0) { |
1851 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1852 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 DG |
1853 | } else { |
1854 | ERR("Relay didn't receive valid end data_pending struct size: %d", | |
1855 | ret); | |
1856 | } | |
f7079f67 DG |
1857 | ret = -1; |
1858 | goto end_no_session; | |
1859 | } | |
1860 | ||
1861 | session_id = be64toh(msg.session_id); | |
1862 | ||
1863 | /* Iterate over all streams to see if the begin data pending flag is set. */ | |
1864 | rcu_read_lock(); | |
d3e2ba59 | 1865 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 DG |
1866 | node.node) { |
1867 | if (stream->session_id == session_id && | |
f94b19e6 | 1868 | !stream->data_pending_check_done && !stream->terminated_flag) { |
f7079f67 DG |
1869 | is_data_inflight = 1; |
1870 | DBG("Data is still in flight for stream %" PRIu64, | |
1871 | stream->stream_handle); | |
1872 | break; | |
1873 | } | |
1874 | } | |
1875 | rcu_read_unlock(); | |
1876 | ||
53efb85a | 1877 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1878 | /* All good, send back reply. */ |
1879 | reply.ret_code = htobe32(is_data_inflight); | |
1880 | ||
58eb9381 | 1881 | ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
f7079f67 DG |
1882 | if (ret < 0) { |
1883 | ERR("Relay end data pending send reply failed"); | |
1884 | } | |
1885 | ||
1886 | end_no_session: | |
1887 | return ret; | |
1888 | } | |
1889 | ||
1c20f0e2 JD |
1890 | /* |
1891 | * Receive an index for a specific stream. | |
1892 | * | |
1893 | * Return 0 on success else a negative value. | |
1894 | */ | |
1895 | static | |
1896 | int relay_recv_index(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 1897 | struct relay_connection *conn) |
1c20f0e2 JD |
1898 | { |
1899 | int ret, send_ret, index_created = 0; | |
58eb9381 | 1900 | struct relay_session *session = conn->session; |
1c20f0e2 JD |
1901 | struct lttcomm_relayd_index index_info; |
1902 | struct relay_index *index, *wr_index = NULL; | |
1903 | struct lttcomm_relayd_generic_reply reply; | |
1904 | struct relay_stream *stream; | |
1905 | uint64_t net_seq_num; | |
1906 | ||
58eb9381 | 1907 | assert(conn); |
1c20f0e2 JD |
1908 | |
1909 | DBG("Relay receiving index"); | |
1910 | ||
58eb9381 | 1911 | if (!session || conn->version_check_done == 0) { |
1c20f0e2 JD |
1912 | ERR("Trying to close a stream before version check"); |
1913 | ret = -1; | |
1914 | goto end_no_session; | |
1915 | } | |
1916 | ||
58eb9381 | 1917 | ret = conn->sock->ops->recvmsg(conn->sock, &index_info, |
1c20f0e2 JD |
1918 | sizeof(index_info), 0); |
1919 | if (ret < sizeof(index_info)) { | |
1920 | if (ret == 0) { | |
1921 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 1922 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
1c20f0e2 JD |
1923 | } else { |
1924 | ERR("Relay didn't receive valid index struct size : %d", ret); | |
1925 | } | |
1926 | ret = -1; | |
1927 | goto end_no_session; | |
1928 | } | |
1929 | ||
1930 | net_seq_num = be64toh(index_info.net_seq_num); | |
1931 | ||
1932 | rcu_read_lock(); | |
2a174661 DG |
1933 | stream = stream_find_by_id(relay_streams_ht, |
1934 | be64toh(index_info.relay_stream_id)); | |
1c20f0e2 JD |
1935 | if (!stream) { |
1936 | ret = -1; | |
1937 | goto end_rcu_unlock; | |
1938 | } | |
1939 | ||
d3e2ba59 JD |
1940 | /* Live beacon handling */ |
1941 | if (index_info.packet_size == 0) { | |
1942 | DBG("Received live beacon for stream %" PRIu64, stream->stream_handle); | |
1943 | ||
1944 | /* | |
6f6e4c37 JD |
1945 | * Only flag a stream inactive when it has already received data |
1946 | * and no indexes are in flight. | |
d3e2ba59 | 1947 | */ |
6f6e4c37 | 1948 | if (stream->total_index_received > 0 && stream->indexes_in_flight == 0) { |
d3e2ba59 JD |
1949 | stream->beacon_ts_end = be64toh(index_info.timestamp_end); |
1950 | } | |
1951 | ret = 0; | |
1952 | goto end_rcu_unlock; | |
1953 | } else { | |
1954 | stream->beacon_ts_end = -1ULL; | |
1955 | } | |
1956 | ||
0a6518b0 | 1957 | index = relay_index_find(stream->stream_handle, net_seq_num); |
1c20f0e2 JD |
1958 | if (!index) { |
1959 | /* A successful creation will add the object to the HT. */ | |
1960 | index = relay_index_create(stream->stream_handle, net_seq_num); | |
1961 | if (!index) { | |
1962 | goto end_rcu_unlock; | |
1963 | } | |
1964 | index_created = 1; | |
6f6e4c37 | 1965 | stream->indexes_in_flight++; |
1c20f0e2 JD |
1966 | } |
1967 | ||
1968 | copy_index_control_data(index, &index_info); | |
bb8cf449 JD |
1969 | if (stream->ctf_stream_id == -1ULL) { |
1970 | stream->ctf_stream_id = be64toh(index_info.stream_id); | |
1971 | } | |
1c20f0e2 JD |
1972 | |
1973 | if (index_created) { | |
1974 | /* | |
1975 | * Try to add the relay index object to the hash table. If an object | |
1976 | * already exist, destroy back the index created, set the data in this | |
1977 | * object and write it on disk. | |
1978 | */ | |
0a6518b0 | 1979 | relay_index_add(index, &wr_index); |
1c20f0e2 JD |
1980 | if (wr_index) { |
1981 | copy_index_control_data(wr_index, &index_info); | |
1982 | free(index); | |
1983 | } | |
1984 | } else { | |
1985 | /* The index already exists so write it on disk. */ | |
1986 | wr_index = index; | |
1987 | } | |
1988 | ||
1989 | /* Do we have a writable ready index to write on disk. */ | |
1990 | if (wr_index) { | |
0a6518b0 | 1991 | ret = relay_index_write(wr_index->fd, wr_index); |
1c20f0e2 JD |
1992 | if (ret < 0) { |
1993 | goto end_rcu_unlock; | |
1994 | } | |
d3e2ba59 | 1995 | stream->total_index_received++; |
6f6e4c37 JD |
1996 | stream->indexes_in_flight--; |
1997 | assert(stream->indexes_in_flight >= 0); | |
1c20f0e2 JD |
1998 | } |
1999 | ||
2000 | end_rcu_unlock: | |
2001 | rcu_read_unlock(); | |
2002 | ||
53efb85a | 2003 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
2004 | if (ret < 0) { |
2005 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
2006 | } else { | |
2007 | reply.ret_code = htobe32(LTTNG_OK); | |
2008 | } | |
58eb9381 | 2009 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1c20f0e2 JD |
2010 | if (send_ret < 0) { |
2011 | ERR("Relay sending close index id reply"); | |
2012 | ret = send_ret; | |
2013 | } | |
2014 | ||
2015 | end_no_session: | |
2016 | return ret; | |
2017 | } | |
2018 | ||
a4baae1b JD |
2019 | /* |
2020 | * Receive the streams_sent message. | |
2021 | * | |
2022 | * Return 0 on success else a negative value. | |
2023 | */ | |
2024 | static | |
2025 | int relay_streams_sent(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 2026 | struct relay_connection *conn) |
a4baae1b JD |
2027 | { |
2028 | int ret, send_ret; | |
2029 | struct lttcomm_relayd_generic_reply reply; | |
2030 | ||
58eb9381 | 2031 | assert(conn); |
a4baae1b JD |
2032 | |
2033 | DBG("Relay receiving streams_sent"); | |
2034 | ||
58eb9381 | 2035 | if (!conn->session || conn->version_check_done == 0) { |
a4baae1b JD |
2036 | ERR("Trying to close a stream before version check"); |
2037 | ret = -1; | |
2038 | goto end_no_session; | |
2039 | } | |
2040 | ||
2041 | /* | |
2042 | * Flag every pending stream in the connection recv list that they are | |
2043 | * ready to be used by the viewer. | |
2044 | */ | |
58eb9381 | 2045 | set_viewer_ready_flag(conn); |
a4baae1b | 2046 | |
4a9daf17 JD |
2047 | /* |
2048 | * Inform the viewer that there are new streams in the session. | |
2049 | */ | |
c5141fe5 JD |
2050 | if (conn->session->viewer_refcount) { |
2051 | uatomic_set(&conn->session->new_streams, 1); | |
2052 | } | |
4a9daf17 | 2053 | |
53efb85a | 2054 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2055 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2056 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
a4baae1b JD |
2057 | if (send_ret < 0) { |
2058 | ERR("Relay sending sent_stream reply"); | |
2059 | ret = send_ret; | |
2060 | } else { | |
2061 | /* Success. */ | |
2062 | ret = 0; | |
2063 | } | |
2064 | ||
2065 | end_no_session: | |
2066 | return ret; | |
2067 | } | |
2068 | ||
b8aa1682 | 2069 | /* |
d3e2ba59 | 2070 | * Process the commands received on the control socket |
b8aa1682 JD |
2071 | */ |
2072 | static | |
2073 | int relay_process_control(struct lttcomm_relayd_hdr *recv_hdr, | |
58eb9381 | 2074 | struct relay_connection *conn) |
b8aa1682 JD |
2075 | { |
2076 | int ret = 0; | |
2077 | ||
2078 | switch (be32toh(recv_hdr->cmd)) { | |
b8aa1682 | 2079 | case RELAYD_CREATE_SESSION: |
58eb9381 | 2080 | ret = relay_create_session(recv_hdr, conn); |
b8aa1682 | 2081 | break; |
b8aa1682 | 2082 | case RELAYD_ADD_STREAM: |
58eb9381 | 2083 | ret = relay_add_stream(recv_hdr, conn); |
b8aa1682 JD |
2084 | break; |
2085 | case RELAYD_START_DATA: | |
58eb9381 | 2086 | ret = relay_start(recv_hdr, conn); |
b8aa1682 JD |
2087 | break; |
2088 | case RELAYD_SEND_METADATA: | |
58eb9381 | 2089 | ret = relay_recv_metadata(recv_hdr, conn); |
b8aa1682 JD |
2090 | break; |
2091 | case RELAYD_VERSION: | |
58eb9381 | 2092 | ret = relay_send_version(recv_hdr, conn); |
b8aa1682 | 2093 | break; |
173af62f | 2094 | case RELAYD_CLOSE_STREAM: |
58eb9381 | 2095 | ret = relay_close_stream(recv_hdr, conn); |
173af62f | 2096 | break; |
6d805429 | 2097 | case RELAYD_DATA_PENDING: |
58eb9381 | 2098 | ret = relay_data_pending(recv_hdr, conn); |
c8f59ee5 DG |
2099 | break; |
2100 | case RELAYD_QUIESCENT_CONTROL: | |
58eb9381 | 2101 | ret = relay_quiescent_control(recv_hdr, conn); |
c8f59ee5 | 2102 | break; |
f7079f67 | 2103 | case RELAYD_BEGIN_DATA_PENDING: |
58eb9381 | 2104 | ret = relay_begin_data_pending(recv_hdr, conn); |
f7079f67 DG |
2105 | break; |
2106 | case RELAYD_END_DATA_PENDING: | |
58eb9381 | 2107 | ret = relay_end_data_pending(recv_hdr, conn); |
f7079f67 | 2108 | break; |
1c20f0e2 | 2109 | case RELAYD_SEND_INDEX: |
58eb9381 | 2110 | ret = relay_recv_index(recv_hdr, conn); |
1c20f0e2 | 2111 | break; |
a4baae1b | 2112 | case RELAYD_STREAMS_SENT: |
58eb9381 | 2113 | ret = relay_streams_sent(recv_hdr, conn); |
a4baae1b | 2114 | break; |
b8aa1682 JD |
2115 | case RELAYD_UPDATE_SYNC_INFO: |
2116 | default: | |
2117 | ERR("Received unknown command (%u)", be32toh(recv_hdr->cmd)); | |
58eb9381 | 2118 | relay_unknown_command(conn); |
b8aa1682 JD |
2119 | ret = -1; |
2120 | goto end; | |
2121 | } | |
2122 | ||
2123 | end: | |
2124 | return ret; | |
2125 | } | |
2126 | ||
7d2f7452 DG |
2127 | /* |
2128 | * Handle index for a data stream. | |
2129 | * | |
2130 | * RCU read side lock MUST be acquired. | |
2131 | * | |
2132 | * Return 0 on success else a negative value. | |
2133 | */ | |
2134 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
2135 | int rotate_index) | |
2136 | { | |
2137 | int ret = 0, index_created = 0; | |
2138 | uint64_t stream_id, data_offset; | |
2139 | struct relay_index *index, *wr_index = NULL; | |
2140 | ||
2141 | assert(stream); | |
2142 | ||
2143 | stream_id = stream->stream_handle; | |
2144 | /* Get data offset because we are about to update the index. */ | |
2145 | data_offset = htobe64(stream->tracefile_size_current); | |
2146 | ||
2147 | /* | |
2148 | * Lookup for an existing index for that stream id/sequence number. If on | |
2149 | * exists, the control thread already received the data for it thus we need | |
2150 | * to write it on disk. | |
2151 | */ | |
2152 | index = relay_index_find(stream_id, net_seq_num); | |
2153 | if (!index) { | |
2154 | /* A successful creation will add the object to the HT. */ | |
2155 | index = relay_index_create(stream_id, net_seq_num); | |
2156 | if (!index) { | |
2157 | ret = -1; | |
2158 | goto error; | |
2159 | } | |
2160 | index_created = 1; | |
6f6e4c37 | 2161 | stream->indexes_in_flight++; |
7d2f7452 DG |
2162 | } |
2163 | ||
2164 | if (rotate_index || stream->index_fd < 0) { | |
2165 | index->to_close_fd = stream->index_fd; | |
2166 | ret = index_create_file(stream->path_name, stream->channel_name, | |
2167 | relayd_uid, relayd_gid, stream->tracefile_size, | |
2168 | stream->tracefile_count_current); | |
2169 | if (ret < 0) { | |
2170 | /* This will close the stream's index fd if one. */ | |
2171 | relay_index_free_safe(index); | |
2172 | goto error; | |
2173 | } | |
2174 | stream->index_fd = ret; | |
2175 | } | |
2176 | index->fd = stream->index_fd; | |
2177 | index->index_data.offset = data_offset; | |
2178 | ||
2179 | if (index_created) { | |
2180 | /* | |
2181 | * Try to add the relay index object to the hash table. If an object | |
2182 | * already exist, destroy back the index created and set the data. | |
2183 | */ | |
2184 | relay_index_add(index, &wr_index); | |
2185 | if (wr_index) { | |
2186 | /* Copy back data from the created index. */ | |
2187 | wr_index->fd = index->fd; | |
2188 | wr_index->to_close_fd = index->to_close_fd; | |
2189 | wr_index->index_data.offset = data_offset; | |
2190 | free(index); | |
2191 | } | |
2192 | } else { | |
2193 | /* The index already exists so write it on disk. */ | |
2194 | wr_index = index; | |
2195 | } | |
2196 | ||
2197 | /* Do we have a writable ready index to write on disk. */ | |
2198 | if (wr_index) { | |
2199 | ret = relay_index_write(wr_index->fd, wr_index); | |
2200 | if (ret < 0) { | |
2201 | goto error; | |
2202 | } | |
2203 | stream->total_index_received++; | |
6f6e4c37 JD |
2204 | stream->indexes_in_flight--; |
2205 | assert(stream->indexes_in_flight >= 0); | |
7d2f7452 DG |
2206 | } |
2207 | ||
2208 | error: | |
2209 | return ret; | |
2210 | } | |
2211 | ||
b8aa1682 JD |
2212 | /* |
2213 | * relay_process_data: Process the data received on the data socket | |
2214 | */ | |
2215 | static | |
58eb9381 | 2216 | int relay_process_data(struct relay_connection *conn) |
b8aa1682 | 2217 | { |
7d2f7452 | 2218 | int ret = 0, rotate_index = 0; |
6cd525e8 | 2219 | ssize_t size_ret; |
b8aa1682 JD |
2220 | struct relay_stream *stream; |
2221 | struct lttcomm_relayd_data_hdr data_hdr; | |
7d2f7452 | 2222 | uint64_t stream_id; |
173af62f | 2223 | uint64_t net_seq_num; |
b8aa1682 | 2224 | uint32_t data_size; |
2a174661 | 2225 | struct relay_session *session; |
b8aa1682 | 2226 | |
58eb9381 DG |
2227 | assert(conn); |
2228 | ||
2229 | ret = conn->sock->ops->recvmsg(conn->sock, &data_hdr, | |
7c5aef62 | 2230 | sizeof(struct lttcomm_relayd_data_hdr), 0); |
b8aa1682 | 2231 | if (ret <= 0) { |
a6cd2b97 DG |
2232 | if (ret == 0) { |
2233 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2234 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2235 | } else { |
58eb9381 | 2236 | ERR("Unable to receive data header on sock %d", conn->sock->fd); |
a6cd2b97 | 2237 | } |
b8aa1682 JD |
2238 | ret = -1; |
2239 | goto end; | |
2240 | } | |
2241 | ||
2242 | stream_id = be64toh(data_hdr.stream_id); | |
9d1bbf21 MD |
2243 | |
2244 | rcu_read_lock(); | |
2a174661 | 2245 | stream = stream_find_by_id(relay_streams_ht, stream_id); |
b8aa1682 JD |
2246 | if (!stream) { |
2247 | ret = -1; | |
1c20f0e2 | 2248 | goto end_rcu_unlock; |
b8aa1682 JD |
2249 | } |
2250 | ||
58eb9381 | 2251 | session = session_find_by_id(conn->sessions_ht, stream->session_id); |
2a174661 DG |
2252 | assert(session); |
2253 | ||
b8aa1682 JD |
2254 | data_size = be32toh(data_hdr.data_size); |
2255 | if (data_buffer_size < data_size) { | |
c617c0c6 MD |
2256 | char *tmp_data_ptr; |
2257 | ||
2258 | tmp_data_ptr = realloc(data_buffer, data_size); | |
2259 | if (!tmp_data_ptr) { | |
b8aa1682 | 2260 | ERR("Allocating data buffer"); |
c617c0c6 | 2261 | free(data_buffer); |
b8aa1682 | 2262 | ret = -1; |
1c20f0e2 | 2263 | goto end_rcu_unlock; |
b8aa1682 | 2264 | } |
c617c0c6 | 2265 | data_buffer = tmp_data_ptr; |
b8aa1682 JD |
2266 | data_buffer_size = data_size; |
2267 | } | |
2268 | memset(data_buffer, 0, data_size); | |
2269 | ||
173af62f DG |
2270 | net_seq_num = be64toh(data_hdr.net_seq_num); |
2271 | ||
77c7c900 | 2272 | DBG3("Receiving data of size %u for stream id %" PRIu64 " seqnum %" PRIu64, |
173af62f | 2273 | data_size, stream_id, net_seq_num); |
58eb9381 | 2274 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, data_size, 0); |
b8aa1682 | 2275 | if (ret <= 0) { |
a6cd2b97 DG |
2276 | if (ret == 0) { |
2277 | /* Orderly shutdown. Not necessary to print an error. */ | |
58eb9381 | 2278 | DBG("Socket %d did an orderly shutdown", conn->sock->fd); |
a6cd2b97 | 2279 | } |
b8aa1682 | 2280 | ret = -1; |
1c20f0e2 | 2281 | goto end_rcu_unlock; |
b8aa1682 JD |
2282 | } |
2283 | ||
1c20f0e2 | 2284 | /* Check if a rotation is needed. */ |
0f907de1 JD |
2285 | if (stream->tracefile_size > 0 && |
2286 | (stream->tracefile_size_current + data_size) > | |
2287 | stream->tracefile_size) { | |
6b6b9a5a JD |
2288 | struct relay_viewer_stream *vstream; |
2289 | uint64_t new_id; | |
2290 | ||
2291 | new_id = (stream->tracefile_count_current + 1) % | |
2292 | stream->tracefile_count; | |
2293 | /* | |
2294 | * When we wrap-around back to 0, we start overwriting old | |
2295 | * trace data. | |
2296 | */ | |
2297 | if (!stream->tracefile_overwrite && new_id == 0) { | |
2298 | stream->tracefile_overwrite = 1; | |
2299 | } | |
2300 | pthread_mutex_lock(&stream->viewer_stream_rotation_lock); | |
2301 | if (stream->tracefile_overwrite) { | |
2302 | stream->oldest_tracefile_id = | |
2303 | (stream->oldest_tracefile_id + 1) % | |
2304 | stream->tracefile_count; | |
2305 | } | |
2f8f53af | 2306 | vstream = viewer_stream_find_by_id(stream->stream_handle); |
6b6b9a5a JD |
2307 | if (vstream) { |
2308 | /* | |
2309 | * The viewer is reading a file about to be | |
2310 | * overwritten. Close the FDs it is | |
2311 | * currently using and let it handle the fault. | |
2312 | */ | |
2313 | if (vstream->tracefile_count_current == new_id) { | |
cef0f7d5 | 2314 | pthread_mutex_lock(&vstream->overwrite_lock); |
6b6b9a5a | 2315 | vstream->abort_flag = 1; |
cef0f7d5 | 2316 | pthread_mutex_unlock(&vstream->overwrite_lock); |
6b6b9a5a JD |
2317 | DBG("Streaming side setting abort_flag on stream %s_%lu\n", |
2318 | stream->channel_name, new_id); | |
2319 | } else if (vstream->tracefile_count_current == | |
2320 | stream->tracefile_count_current) { | |
2321 | /* | |
2322 | * The reader and writer were in the | |
2323 | * same trace file, inform the viewer | |
2324 | * that no new index will ever be added | |
2325 | * to this file. | |
2326 | */ | |
2327 | vstream->close_write_flag = 1; | |
2328 | } | |
2329 | } | |
1c20f0e2 JD |
2330 | ret = utils_rotate_stream_file(stream->path_name, stream->channel_name, |
2331 | stream->tracefile_size, stream->tracefile_count, | |
2332 | relayd_uid, relayd_gid, stream->fd, | |
2333 | &(stream->tracefile_count_current), &stream->fd); | |
cef0f7d5 | 2334 | stream->total_index_received = 0; |
6b6b9a5a | 2335 | pthread_mutex_unlock(&stream->viewer_stream_rotation_lock); |
0f907de1 | 2336 | if (ret < 0) { |
1c20f0e2 JD |
2337 | ERR("Rotating stream output file"); |
2338 | goto end_rcu_unlock; | |
0f907de1 | 2339 | } |
a6976990 DG |
2340 | /* Reset current size because we just perform a stream rotation. */ |
2341 | stream->tracefile_size_current = 0; | |
1c20f0e2 JD |
2342 | rotate_index = 1; |
2343 | } | |
2344 | ||
1c20f0e2 | 2345 | /* |
7d2f7452 DG |
2346 | * Index are handled in protocol version 2.4 and above. Also, snapshot and |
2347 | * index are NOT supported. | |
1c20f0e2 | 2348 | */ |
2a174661 | 2349 | if (session->minor >= 4 && !session->snapshot) { |
7d2f7452 | 2350 | ret = handle_index_data(stream, net_seq_num, rotate_index); |
1c20f0e2 | 2351 | if (ret < 0) { |
1c20f0e2 JD |
2352 | goto end_rcu_unlock; |
2353 | } | |
1c20f0e2 JD |
2354 | } |
2355 | ||
7d2f7452 | 2356 | /* Write data to stream output fd. */ |
6cd525e8 MD |
2357 | size_ret = lttng_write(stream->fd, data_buffer, data_size); |
2358 | if (size_ret < data_size) { | |
b8aa1682 JD |
2359 | ERR("Relay error writing data to file"); |
2360 | ret = -1; | |
1c20f0e2 | 2361 | goto end_rcu_unlock; |
b8aa1682 | 2362 | } |
1d4dfdef | 2363 | |
5ab7344e JD |
2364 | DBG2("Relay wrote %d bytes to tracefile for stream id %" PRIu64, |
2365 | ret, stream->stream_handle); | |
2366 | ||
1d4dfdef DG |
2367 | ret = write_padding_to_file(stream->fd, be32toh(data_hdr.padding_size)); |
2368 | if (ret < 0) { | |
1c20f0e2 | 2369 | goto end_rcu_unlock; |
1d4dfdef | 2370 | } |
1c20f0e2 | 2371 | stream->tracefile_size_current += data_size + be32toh(data_hdr.padding_size); |
1d4dfdef | 2372 | |
173af62f DG |
2373 | stream->prev_seq = net_seq_num; |
2374 | ||
2a174661 | 2375 | try_close_stream(session, stream); |
173af62f | 2376 | |
1c20f0e2 | 2377 | end_rcu_unlock: |
9d1bbf21 | 2378 | rcu_read_unlock(); |
b8aa1682 JD |
2379 | end: |
2380 | return ret; | |
2381 | } | |
2382 | ||
2383 | static | |
58eb9381 | 2384 | void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2385 | { |
2386 | int ret; | |
2387 | ||
58eb9381 DG |
2388 | assert(events); |
2389 | ||
2390 | (void) lttng_poll_del(events, pollfd); | |
b8aa1682 JD |
2391 | |
2392 | ret = close(pollfd); | |
2393 | if (ret < 0) { | |
2394 | ERR("Closing pollfd %d", pollfd); | |
2395 | } | |
2396 | } | |
2397 | ||
58eb9381 DG |
2398 | static void destroy_connection(struct lttng_ht *relay_connections_ht, |
2399 | struct relay_connection *conn) | |
9d1bbf21 | 2400 | { |
58eb9381 DG |
2401 | assert(relay_connections_ht); |
2402 | assert(conn); | |
2a174661 | 2403 | |
58eb9381 | 2404 | connection_delete(relay_connections_ht, conn); |
2a174661 | 2405 | |
58eb9381 | 2406 | /* For the control socket, we try to destroy the session. */ |
fd6d7293 | 2407 | if (conn->type == RELAY_CONTROL && conn->session) { |
58eb9381 | 2408 | destroy_session(conn->session, conn->sessions_ht); |
9d1bbf21 | 2409 | } |
5b6d8097 | 2410 | |
58eb9381 | 2411 | connection_destroy(conn); |
b8aa1682 JD |
2412 | } |
2413 | ||
2414 | /* | |
2415 | * This thread does the actual work | |
2416 | */ | |
2417 | static | |
2418 | void *relay_thread_worker(void *data) | |
2419 | { | |
beaad64c DG |
2420 | int ret, err = -1, last_seen_data_fd = -1; |
2421 | uint32_t nb_fd; | |
58eb9381 | 2422 | struct relay_connection *conn; |
b8aa1682 JD |
2423 | struct lttng_poll_event events; |
2424 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2425 | struct lttng_ht_iter iter; |
b8aa1682 | 2426 | struct lttcomm_relayd_hdr recv_hdr; |
d3e2ba59 JD |
2427 | struct relay_local_data *relay_ctx = (struct relay_local_data *) data; |
2428 | struct lttng_ht *sessions_ht = relay_ctx->sessions_ht; | |
b8aa1682 JD |
2429 | |
2430 | DBG("[thread] Relay worker started"); | |
2431 | ||
9d1bbf21 MD |
2432 | rcu_register_thread(); |
2433 | ||
55706a7d MD |
2434 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2435 | ||
9b5e0863 MD |
2436 | if (testpoint(relayd_thread_worker)) { |
2437 | goto error_testpoint; | |
2438 | } | |
2439 | ||
f385ae0a MD |
2440 | health_code_update(); |
2441 | ||
b8aa1682 JD |
2442 | /* table of connections indexed on socket */ |
2443 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2444 | if (!relay_connections_ht) { |
2445 | goto relay_connections_ht_error; | |
2446 | } | |
b8aa1682 | 2447 | |
1c20f0e2 JD |
2448 | /* Tables of received indexes indexed by index handle and net_seq_num. */ |
2449 | indexes_ht = lttng_ht_new(0, LTTNG_HT_TYPE_TWO_U64); | |
2450 | if (!indexes_ht) { | |
2451 | goto indexes_ht_error; | |
2452 | } | |
2453 | ||
b8aa1682 JD |
2454 | ret = create_thread_poll_set(&events, 2); |
2455 | if (ret < 0) { | |
2456 | goto error_poll_create; | |
2457 | } | |
2458 | ||
58eb9381 | 2459 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2460 | if (ret < 0) { |
2461 | goto error; | |
2462 | } | |
2463 | ||
beaad64c | 2464 | restart: |
b8aa1682 | 2465 | while (1) { |
beaad64c DG |
2466 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2467 | ||
f385ae0a MD |
2468 | health_code_update(); |
2469 | ||
b8aa1682 | 2470 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2471 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2472 | health_poll_entry(); |
b8aa1682 | 2473 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2474 | health_poll_exit(); |
b8aa1682 JD |
2475 | if (ret < 0) { |
2476 | /* | |
2477 | * Restart interrupted system call. | |
2478 | */ | |
2479 | if (errno == EINTR) { | |
2480 | goto restart; | |
2481 | } | |
2482 | goto error; | |
2483 | } | |
2484 | ||
0d9c5d77 DG |
2485 | nb_fd = ret; |
2486 | ||
beaad64c DG |
2487 | /* |
2488 | * Process control. The control connection is prioritised so we don't | |
2489 | * starve it with high throughout put tracing data on the data | |
2490 | * connection. | |
2491 | */ | |
b8aa1682 JD |
2492 | for (i = 0; i < nb_fd; i++) { |
2493 | /* Fetch once the poll data */ | |
beaad64c DG |
2494 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2495 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2496 | |
f385ae0a MD |
2497 | health_code_update(); |
2498 | ||
b8aa1682 JD |
2499 | /* Thread quit pipe has been closed. Killing thread. */ |
2500 | ret = check_thread_quit_pipe(pollfd, revents); | |
2501 | if (ret) { | |
095a4ae5 MD |
2502 | err = 0; |
2503 | goto exit; | |
b8aa1682 JD |
2504 | } |
2505 | ||
58eb9381 DG |
2506 | /* Inspect the relay conn pipe for new connection */ |
2507 | if (pollfd == relay_conn_pipe[0]) { | |
b8aa1682 | 2508 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
58eb9381 | 2509 | ERR("Relay connection pipe error"); |
b8aa1682 JD |
2510 | goto error; |
2511 | } else if (revents & LPOLLIN) { | |
58eb9381 | 2512 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2513 | if (ret < 0) { |
2514 | goto error; | |
2515 | } | |
58eb9381 DG |
2516 | conn->sessions_ht = sessions_ht; |
2517 | connection_init(conn); | |
2518 | lttng_poll_add(&events, conn->sock->fd, | |
2519 | LPOLLIN | LPOLLRDHUP); | |
2520 | rcu_read_lock(); | |
2521 | lttng_ht_add_unique_ulong(relay_connections_ht, | |
2522 | &conn->sock_n); | |
9d1bbf21 | 2523 | rcu_read_unlock(); |
58eb9381 | 2524 | DBG("Connection socket %d added", conn->sock->fd); |
b8aa1682 | 2525 | } |
58eb9381 DG |
2526 | } else { |
2527 | rcu_read_lock(); | |
2528 | conn = connection_find_by_sock(relay_connections_ht, pollfd); | |
2529 | /* If not found, there is a synchronization issue. */ | |
2530 | assert(conn); | |
2531 | ||
2532 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
2533 | cleanup_connection_pollfd(&events, pollfd); | |
2534 | destroy_connection(relay_connections_ht, conn); | |
beaad64c DG |
2535 | if (last_seen_data_fd == pollfd) { |
2536 | last_seen_data_fd = last_notdel_data_fd; | |
2537 | } | |
b8aa1682 | 2538 | } else if (revents & LPOLLIN) { |
58eb9381 DG |
2539 | if (conn->type == RELAY_CONTROL) { |
2540 | ret = conn->sock->ops->recvmsg(conn->sock, &recv_hdr, | |
2541 | sizeof(recv_hdr), 0); | |
b8aa1682 | 2542 | if (ret <= 0) { |
58eb9381 DG |
2543 | /* Connection closed */ |
2544 | cleanup_connection_pollfd(&events, pollfd); | |
2545 | destroy_connection(relay_connections_ht, conn); | |
b8aa1682 JD |
2546 | DBG("Control connection closed with %d", pollfd); |
2547 | } else { | |
58eb9381 | 2548 | ret = relay_process_control(&recv_hdr, conn); |
b8aa1682 | 2549 | if (ret < 0) { |
beaad64c | 2550 | /* Clear the session on error. */ |
58eb9381 DG |
2551 | cleanup_connection_pollfd(&events, pollfd); |
2552 | destroy_connection(relay_connections_ht, conn); | |
b8aa1682 JD |
2553 | DBG("Connection closed with %d", pollfd); |
2554 | } | |
beaad64c | 2555 | seen_control = 1; |
b8aa1682 | 2556 | } |
beaad64c DG |
2557 | } else { |
2558 | /* | |
2559 | * Flag the last seen data fd not deleted. It will be | |
2560 | * used as the last seen fd if any fd gets deleted in | |
2561 | * this first loop. | |
2562 | */ | |
2563 | last_notdel_data_fd = pollfd; | |
2564 | } | |
58eb9381 DG |
2565 | } else { |
2566 | ERR("Unknown poll events %u for sock %d", revents, pollfd); | |
beaad64c DG |
2567 | } |
2568 | rcu_read_unlock(); | |
2569 | } | |
2570 | } | |
2571 | ||
2572 | /* | |
2573 | * The last loop handled a control request, go back to poll to make | |
2574 | * sure we prioritise the control socket. | |
2575 | */ | |
2576 | if (seen_control) { | |
2577 | continue; | |
2578 | } | |
2579 | ||
2580 | if (last_seen_data_fd >= 0) { | |
2581 | for (i = 0; i < nb_fd; i++) { | |
2582 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2583 | |
2584 | health_code_update(); | |
2585 | ||
beaad64c DG |
2586 | if (last_seen_data_fd == pollfd) { |
2587 | idx = i; | |
2588 | break; | |
2589 | } | |
2590 | } | |
2591 | } | |
2592 | ||
2593 | /* Process data connection. */ | |
2594 | for (i = idx + 1; i < nb_fd; i++) { | |
2595 | /* Fetch the poll data. */ | |
2596 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
2597 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
2598 | ||
f385ae0a MD |
2599 | health_code_update(); |
2600 | ||
beaad64c | 2601 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 2602 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
2603 | continue; |
2604 | } | |
2605 | ||
2606 | if (revents) { | |
2607 | rcu_read_lock(); | |
58eb9381 DG |
2608 | conn = connection_find_by_sock(relay_connections_ht, pollfd); |
2609 | if (!conn) { | |
beaad64c DG |
2610 | /* Skip it. Might be removed before. */ |
2611 | rcu_read_unlock(); | |
2612 | continue; | |
2613 | } | |
beaad64c DG |
2614 | |
2615 | if (revents & LPOLLIN) { | |
58eb9381 | 2616 | if (conn->type != RELAY_DATA) { |
beaad64c DG |
2617 | continue; |
2618 | } | |
2619 | ||
58eb9381 DG |
2620 | ret = relay_process_data(conn); |
2621 | /* Connection closed */ | |
beaad64c | 2622 | if (ret < 0) { |
58eb9381 DG |
2623 | cleanup_connection_pollfd(&events, pollfd); |
2624 | destroy_connection(relay_connections_ht, conn); | |
beaad64c DG |
2625 | DBG("Data connection closed with %d", pollfd); |
2626 | /* | |
2627 | * Every goto restart call sets the last seen fd where | |
2628 | * here we don't really care since we gracefully | |
2629 | * continue the loop after the connection is deleted. | |
2630 | */ | |
2631 | } else { | |
2632 | /* Keep last seen port. */ | |
2633 | last_seen_data_fd = pollfd; | |
2634 | rcu_read_unlock(); | |
2635 | goto restart; | |
b8aa1682 JD |
2636 | } |
2637 | } | |
9d1bbf21 | 2638 | rcu_read_unlock(); |
b8aa1682 JD |
2639 | } |
2640 | } | |
beaad64c | 2641 | last_seen_data_fd = -1; |
b8aa1682 JD |
2642 | } |
2643 | ||
f385ae0a MD |
2644 | /* Normal exit, no error */ |
2645 | ret = 0; | |
2646 | ||
095a4ae5 | 2647 | exit: |
b8aa1682 JD |
2648 | error: |
2649 | lttng_poll_clean(&events); | |
2650 | ||
58eb9381 | 2651 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 2652 | rcu_read_lock(); |
58eb9381 DG |
2653 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, conn, |
2654 | sock_n.node) { | |
f385ae0a | 2655 | health_code_update(); |
58eb9381 | 2656 | destroy_connection(relay_connections_ht, conn); |
b8aa1682 | 2657 | } |
94d49140 | 2658 | rcu_read_unlock(); |
7d2f7452 DG |
2659 | error_poll_create: |
2660 | lttng_ht_destroy(indexes_ht); | |
1c20f0e2 | 2661 | indexes_ht_error: |
b8aa1682 | 2662 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 2663 | relay_connections_ht_error: |
58eb9381 DG |
2664 | /* Close relay conn pipes */ |
2665 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
2666 | if (err) { |
2667 | DBG("Thread exited with error"); | |
2668 | } | |
b8aa1682 | 2669 | DBG("Worker thread cleanup complete"); |
095a4ae5 | 2670 | free(data_buffer); |
9b5e0863 | 2671 | error_testpoint: |
f385ae0a MD |
2672 | if (err) { |
2673 | health_error(); | |
2674 | ERR("Health error occurred in %s", __func__); | |
2675 | } | |
2676 | health_unregister(health_relayd); | |
9d1bbf21 | 2677 | rcu_unregister_thread(); |
f385ae0a | 2678 | stop_threads(); |
b8aa1682 JD |
2679 | return NULL; |
2680 | } | |
2681 | ||
2682 | /* | |
2683 | * Create the relay command pipe to wake thread_manage_apps. | |
2684 | * Closed in cleanup(). | |
2685 | */ | |
58eb9381 | 2686 | static int create_relay_conn_pipe(void) |
b8aa1682 | 2687 | { |
a02de639 | 2688 | int ret; |
b8aa1682 | 2689 | |
58eb9381 | 2690 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 2691 | |
b8aa1682 JD |
2692 | return ret; |
2693 | } | |
2694 | ||
2695 | /* | |
2696 | * main | |
2697 | */ | |
2698 | int main(int argc, char **argv) | |
2699 | { | |
2700 | int ret = 0; | |
2701 | void *status; | |
d3e2ba59 | 2702 | struct relay_local_data *relay_ctx; |
b8aa1682 | 2703 | |
b8aa1682 JD |
2704 | /* Parse arguments */ |
2705 | progname = argv[0]; | |
cd60b05a | 2706 | if ((ret = set_options(argc, argv)) < 0) { |
a02de639 | 2707 | goto exit; |
b8aa1682 JD |
2708 | } |
2709 | ||
2710 | if ((ret = set_signal_handler()) < 0) { | |
2711 | goto exit; | |
2712 | } | |
2713 | ||
4d513a50 DG |
2714 | /* Try to create directory if -o, --output is specified. */ |
2715 | if (opt_output_path) { | |
994fa64f DG |
2716 | if (*opt_output_path != '/') { |
2717 | ERR("Please specify an absolute path for -o, --output PATH"); | |
2718 | goto exit; | |
2719 | } | |
2720 | ||
4d513a50 DG |
2721 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG); |
2722 | if (ret < 0) { | |
2723 | ERR("Unable to create %s", opt_output_path); | |
2724 | goto exit; | |
2725 | } | |
2726 | } | |
2727 | ||
b8aa1682 | 2728 | /* Daemonize */ |
b5218ffb | 2729 | if (opt_daemon || opt_background) { |
3fd27398 MD |
2730 | int i; |
2731 | ||
2732 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
2733 | !opt_background); | |
b8aa1682 | 2734 | if (ret < 0) { |
a02de639 | 2735 | goto exit; |
b8aa1682 | 2736 | } |
3fd27398 MD |
2737 | |
2738 | /* | |
2739 | * We are in the child. Make sure all other file | |
2740 | * descriptors are closed, in case we are called with | |
2741 | * more opened file descriptors than the standard ones. | |
2742 | */ | |
2743 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
2744 | (void) close(i); | |
2745 | } | |
2746 | } | |
2747 | ||
2748 | /* Create thread quit pipe */ | |
2749 | if ((ret = init_thread_quit_pipe()) < 0) { | |
2750 | goto error; | |
b8aa1682 JD |
2751 | } |
2752 | ||
1c20f0e2 JD |
2753 | /* We need those values for the file/dir creation. */ |
2754 | relayd_uid = getuid(); | |
2755 | relayd_gid = getgid(); | |
b8aa1682 | 2756 | |
1c20f0e2 JD |
2757 | /* Check if daemon is UID = 0 */ |
2758 | if (relayd_uid == 0) { | |
8d5c808e | 2759 | if (control_uri->port < 1024 || data_uri->port < 1024 || live_uri->port < 1024) { |
b8aa1682 JD |
2760 | ERR("Need to be root to use ports < 1024"); |
2761 | ret = -1; | |
a02de639 | 2762 | goto exit; |
b8aa1682 JD |
2763 | } |
2764 | } | |
2765 | ||
2766 | /* Setup the thread apps communication pipe. */ | |
58eb9381 | 2767 | if ((ret = create_relay_conn_pipe()) < 0) { |
b8aa1682 JD |
2768 | goto exit; |
2769 | } | |
2770 | ||
2771 | /* Init relay command queue. */ | |
58eb9381 | 2772 | cds_wfq_init(&relay_conn_queue.queue); |
b8aa1682 JD |
2773 | |
2774 | /* Set up max poll set size */ | |
2775 | lttng_poll_set_max_size(); | |
2776 | ||
554831e7 MD |
2777 | /* Initialize communication library */ |
2778 | lttcomm_init(); | |
87e45c13 | 2779 | lttcomm_inet_init(); |
554831e7 | 2780 | |
d3e2ba59 JD |
2781 | relay_ctx = zmalloc(sizeof(struct relay_local_data)); |
2782 | if (!relay_ctx) { | |
2783 | PERROR("relay_ctx"); | |
2784 | goto exit; | |
2785 | } | |
2786 | ||
2787 | /* tables of sessions indexed by session ID */ | |
2a174661 | 2788 | relay_ctx->sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 JD |
2789 | if (!relay_ctx->sessions_ht) { |
2790 | goto exit_relay_ctx_sessions; | |
2791 | } | |
2792 | ||
2793 | /* tables of streams indexed by stream ID */ | |
2a174661 | 2794 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 JD |
2795 | if (!relay_streams_ht) { |
2796 | goto exit_relay_ctx_streams; | |
2797 | } | |
2798 | ||
2799 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
2800 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
2801 | if (!viewer_streams_ht) { | |
d3e2ba59 JD |
2802 | goto exit_relay_ctx_viewer_streams; |
2803 | } | |
2804 | ||
55706a7d MD |
2805 | /* Initialize thread health monitoring */ |
2806 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
2807 | if (!health_relayd) { | |
2808 | PERROR("health_app_create error"); | |
2809 | goto exit_health_app_create; | |
2810 | } | |
2811 | ||
65931c8b MD |
2812 | ret = utils_create_pipe(health_quit_pipe); |
2813 | if (ret < 0) { | |
2814 | goto error_health_pipe; | |
2815 | } | |
2816 | ||
2817 | /* Create thread to manage the client socket */ | |
2818 | ret = pthread_create(&health_thread, NULL, | |
2819 | thread_manage_health, (void *) NULL); | |
2820 | if (ret != 0) { | |
2821 | PERROR("pthread_create health"); | |
2822 | goto health_error; | |
2823 | } | |
2824 | ||
b8aa1682 JD |
2825 | /* Setup the dispatcher thread */ |
2826 | ret = pthread_create(&dispatcher_thread, NULL, | |
2827 | relay_thread_dispatcher, (void *) NULL); | |
2828 | if (ret != 0) { | |
2829 | PERROR("pthread_create dispatcher"); | |
2830 | goto exit_dispatcher; | |
2831 | } | |
2832 | ||
2833 | /* Setup the worker thread */ | |
2834 | ret = pthread_create(&worker_thread, NULL, | |
d3e2ba59 | 2835 | relay_thread_worker, (void *) relay_ctx); |
b8aa1682 JD |
2836 | if (ret != 0) { |
2837 | PERROR("pthread_create worker"); | |
2838 | goto exit_worker; | |
2839 | } | |
2840 | ||
2841 | /* Setup the listener thread */ | |
2842 | ret = pthread_create(&listener_thread, NULL, | |
2843 | relay_thread_listener, (void *) NULL); | |
2844 | if (ret != 0) { | |
2845 | PERROR("pthread_create listener"); | |
2846 | goto exit_listener; | |
2847 | } | |
2848 | ||
0b242f62 | 2849 | ret = live_start_threads(live_uri, relay_ctx); |
d3e2ba59 JD |
2850 | if (ret != 0) { |
2851 | ERR("Starting live viewer threads"); | |
50138f51 | 2852 | goto exit_live; |
d3e2ba59 JD |
2853 | } |
2854 | ||
50138f51 | 2855 | exit_live: |
b8aa1682 JD |
2856 | ret = pthread_join(listener_thread, &status); |
2857 | if (ret != 0) { | |
2858 | PERROR("pthread_join"); | |
2859 | goto error; /* join error, exit without cleanup */ | |
2860 | } | |
2861 | ||
50138f51 | 2862 | exit_listener: |
b8aa1682 JD |
2863 | ret = pthread_join(worker_thread, &status); |
2864 | if (ret != 0) { | |
2865 | PERROR("pthread_join"); | |
2866 | goto error; /* join error, exit without cleanup */ | |
2867 | } | |
2868 | ||
50138f51 | 2869 | exit_worker: |
b8aa1682 JD |
2870 | ret = pthread_join(dispatcher_thread, &status); |
2871 | if (ret != 0) { | |
2872 | PERROR("pthread_join"); | |
2873 | goto error; /* join error, exit without cleanup */ | |
2874 | } | |
42415026 | 2875 | |
50138f51 | 2876 | exit_dispatcher: |
65931c8b MD |
2877 | ret = pthread_join(health_thread, &status); |
2878 | if (ret != 0) { | |
2879 | PERROR("pthread_join health thread"); | |
2880 | goto error; /* join error, exit without cleanup */ | |
2881 | } | |
2882 | ||
bd11e201 MD |
2883 | /* |
2884 | * Stop live threads only after joining other threads. | |
2885 | */ | |
2886 | live_stop_threads(); | |
2887 | ||
65931c8b MD |
2888 | health_error: |
2889 | utils_close_pipe(health_quit_pipe); | |
2890 | ||
2891 | error_health_pipe: | |
55706a7d MD |
2892 | health_app_destroy(health_relayd); |
2893 | ||
2894 | exit_health_app_create: | |
92c6ca54 | 2895 | lttng_ht_destroy(viewer_streams_ht); |
d3e2ba59 JD |
2896 | |
2897 | exit_relay_ctx_viewer_streams: | |
2898 | lttng_ht_destroy(relay_streams_ht); | |
2899 | ||
2900 | exit_relay_ctx_streams: | |
2901 | lttng_ht_destroy(relay_ctx->sessions_ht); | |
2902 | ||
2903 | exit_relay_ctx_sessions: | |
2904 | free(relay_ctx); | |
b8aa1682 JD |
2905 | |
2906 | exit: | |
2907 | cleanup(); | |
2908 | if (!ret) { | |
2909 | exit(EXIT_SUCCESS); | |
2910 | } | |
a02de639 | 2911 | |
b8aa1682 JD |
2912 | error: |
2913 | exit(EXIT_FAILURE); | |
2914 | } |