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> |
7591bab1 | 5 | * 2015 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
b8aa1682 JD |
6 | * |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License, version 2 only, | |
9 | * as published by the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
14 | * more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; if not, write to the Free Software Foundation, Inc., | |
18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
19 | */ | |
20 | ||
6c1c0768 | 21 | #define _LGPL_SOURCE |
b8aa1682 JD |
22 | #include <getopt.h> |
23 | #include <grp.h> | |
24 | #include <limits.h> | |
25 | #include <pthread.h> | |
26 | #include <signal.h> | |
27 | #include <stdio.h> | |
28 | #include <stdlib.h> | |
29 | #include <string.h> | |
30 | #include <sys/mman.h> | |
31 | #include <sys/mount.h> | |
32 | #include <sys/resource.h> | |
33 | #include <sys/socket.h> | |
34 | #include <sys/stat.h> | |
35 | #include <sys/types.h> | |
36 | #include <sys/wait.h> | |
173af62f | 37 | #include <inttypes.h> |
b8aa1682 JD |
38 | #include <urcu/futex.h> |
39 | #include <urcu/uatomic.h> | |
40 | #include <unistd.h> | |
41 | #include <fcntl.h> | |
b8aa1682 JD |
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> |
e8fa9fb0 | 48 | #include <common/compat/getenv.h> |
b8aa1682 | 49 | #include <common/defaults.h> |
3fd27398 | 50 | #include <common/daemonize.h> |
b8aa1682 JD |
51 | #include <common/futex.h> |
52 | #include <common/sessiond-comm/sessiond-comm.h> | |
53 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
54 | #include <common/sessiond-comm/relayd.h> |
55 | #include <common/uri.h> | |
a02de639 | 56 | #include <common/utils.h> |
f40ef1d5 | 57 | #include <common/config/session-config.h> |
44a2fbf1 JG |
58 | #include <common/dynamic-buffer.h> |
59 | #include <common/buffer-view.h> | |
7591bab1 | 60 | #include <urcu/rculist.h> |
b8aa1682 | 61 | |
0f907de1 | 62 | #include "cmd.h" |
d3e2ba59 | 63 | #include "ctf-trace.h" |
1c20f0e2 | 64 | #include "index.h" |
0f907de1 | 65 | #include "utils.h" |
b8aa1682 | 66 | #include "lttng-relayd.h" |
d3e2ba59 | 67 | #include "live.h" |
55706a7d | 68 | #include "health-relayd.h" |
9b5e0863 | 69 | #include "testpoint.h" |
2f8f53af | 70 | #include "viewer-stream.h" |
2a174661 DG |
71 | #include "session.h" |
72 | #include "stream.h" | |
58eb9381 | 73 | #include "connection.h" |
a44ca2ca | 74 | #include "tracefile-array.h" |
2fc6b1ab | 75 | #include "tcp_keep_alive.h" |
b8aa1682 | 76 | |
44a2fbf1 JG |
77 | enum relay_connection_status { |
78 | RELAY_CONNECTION_STATUS_OK, | |
79 | /* An error occured while processing an event on the connection. */ | |
80 | RELAY_CONNECTION_STATUS_ERROR, | |
81 | /* Connection closed/shutdown cleanly. */ | |
82 | RELAY_CONNECTION_STATUS_CLOSED, | |
83 | }; | |
84 | ||
b8aa1682 | 85 | /* command line options */ |
0f907de1 | 86 | char *opt_output_path; |
b5218ffb | 87 | static int opt_daemon, opt_background; |
3fd27398 MD |
88 | |
89 | /* | |
90 | * We need to wait for listener and live listener threads, as well as | |
91 | * health check thread, before being ready to signal readiness. | |
92 | */ | |
93 | #define NR_LTTNG_RELAY_READY 3 | |
94 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
0848dba7 MD |
95 | |
96 | /* Size of receive buffer. */ | |
97 | #define RECV_DATA_BUFFER_SIZE 65536 | |
98 | ||
3fd27398 MD |
99 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ |
100 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
101 | ||
095a4ae5 MD |
102 | static struct lttng_uri *control_uri; |
103 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 104 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
105 | |
106 | const char *progname; | |
b8aa1682 | 107 | |
65931c8b | 108 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
109 | static int tracing_group_name_override; |
110 | ||
111 | const char * const config_section_name = "relayd"; | |
65931c8b | 112 | |
b8aa1682 JD |
113 | /* |
114 | * Quit pipe for all threads. This permits a single cancellation point | |
115 | * for all threads when receiving an event on the pipe. | |
116 | */ | |
0b242f62 | 117 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
118 | |
119 | /* | |
120 | * This pipe is used to inform the worker thread that a command is queued and | |
121 | * ready to be processed. | |
122 | */ | |
58eb9381 | 123 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 124 | |
26c9d55e | 125 | /* Shared between threads */ |
b8aa1682 JD |
126 | static int dispatch_thread_exit; |
127 | ||
128 | static pthread_t listener_thread; | |
129 | static pthread_t dispatcher_thread; | |
130 | static pthread_t worker_thread; | |
65931c8b | 131 | static pthread_t health_thread; |
b8aa1682 | 132 | |
7591bab1 MD |
133 | /* |
134 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
135 | * atomicity on 32-bit architectures. | |
136 | */ | |
137 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 138 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
139 | |
140 | /* | |
141 | * Relay command queue. | |
142 | * | |
143 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
144 | * queue. | |
145 | */ | |
58eb9381 | 146 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 | 147 | |
d3e2ba59 JD |
148 | /* Global relay stream hash table. */ |
149 | struct lttng_ht *relay_streams_ht; | |
150 | ||
92c6ca54 DG |
151 | /* Global relay viewer stream hash table. */ |
152 | struct lttng_ht *viewer_streams_ht; | |
153 | ||
7591bab1 MD |
154 | /* Global relay sessions hash table. */ |
155 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 156 | |
55706a7d | 157 | /* Relayd health monitoring */ |
eea7556c | 158 | struct health_app *health_relayd; |
55706a7d | 159 | |
cd60b05a JG |
160 | static struct option long_options[] = { |
161 | { "control-port", 1, 0, 'C', }, | |
162 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 163 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 164 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 165 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
166 | { "group", 1, 0, 'g', }, |
167 | { "help", 0, 0, 'h', }, | |
168 | { "output", 1, 0, 'o', }, | |
169 | { "verbose", 0, 0, 'v', }, | |
170 | { "config", 1, 0, 'f' }, | |
297af7ff | 171 | { "version", 0, 0, 'V' }, |
cd60b05a JG |
172 | { NULL, 0, 0, 0, }, |
173 | }; | |
174 | ||
297af7ff | 175 | static const char *config_ignore_options[] = { "help", "config", "version" }; |
cd60b05a | 176 | |
cd60b05a JG |
177 | /* |
178 | * Take an option from the getopt output and set it in the right variable to be | |
179 | * used later. | |
180 | * | |
181 | * Return 0 on success else a negative value. | |
182 | */ | |
7591bab1 | 183 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 184 | { |
cd60b05a JG |
185 | int ret; |
186 | ||
187 | switch (opt) { | |
188 | case 0: | |
189 | fprintf(stderr, "option %s", optname); | |
190 | if (arg) { | |
191 | fprintf(stderr, " with arg %s\n", arg); | |
192 | } | |
193 | break; | |
194 | case 'C': | |
e8fa9fb0 MD |
195 | if (lttng_is_setuid_setgid()) { |
196 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
197 | "-C, --control-port"); | |
198 | } else { | |
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 | } | |
cd60b05a JG |
207 | } |
208 | break; | |
209 | case 'D': | |
e8fa9fb0 MD |
210 | if (lttng_is_setuid_setgid()) { |
211 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
212 | "-D, -data-port"); | |
213 | } else { | |
214 | ret = uri_parse(arg, &data_uri); | |
215 | if (ret < 0) { | |
216 | ERR("Invalid data URI specified"); | |
217 | goto end; | |
218 | } | |
219 | if (data_uri->port == 0) { | |
220 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
221 | } | |
cd60b05a JG |
222 | } |
223 | break; | |
8d5c808e | 224 | case 'L': |
e8fa9fb0 MD |
225 | if (lttng_is_setuid_setgid()) { |
226 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
227 | "-L, -live-port"); | |
228 | } else { | |
229 | ret = uri_parse(arg, &live_uri); | |
230 | if (ret < 0) { | |
231 | ERR("Invalid live URI specified"); | |
232 | goto end; | |
233 | } | |
234 | if (live_uri->port == 0) { | |
235 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
236 | } | |
8d5c808e AM |
237 | } |
238 | break; | |
cd60b05a JG |
239 | case 'd': |
240 | opt_daemon = 1; | |
241 | break; | |
b5218ffb MD |
242 | case 'b': |
243 | opt_background = 1; | |
244 | break; | |
cd60b05a | 245 | case 'g': |
e8fa9fb0 MD |
246 | if (lttng_is_setuid_setgid()) { |
247 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
248 | "-g, --group"); | |
249 | } else { | |
250 | tracing_group_name = strdup(arg); | |
251 | if (tracing_group_name == NULL) { | |
252 | ret = -errno; | |
253 | PERROR("strdup"); | |
254 | goto end; | |
255 | } | |
256 | tracing_group_name_override = 1; | |
330a40bb | 257 | } |
cd60b05a JG |
258 | break; |
259 | case 'h': | |
655b5cc1 PP |
260 | ret = utils_show_man_page(8, "lttng-relayd"); |
261 | if (ret) { | |
262 | ERR("Cannot view man page lttng-relayd(8)"); | |
263 | perror("exec"); | |
264 | } | |
cd60b05a | 265 | exit(EXIT_FAILURE); |
297af7ff AW |
266 | case 'V': |
267 | fprintf(stdout, "%s\n", VERSION); | |
268 | exit(EXIT_SUCCESS); | |
cd60b05a | 269 | case 'o': |
e8fa9fb0 MD |
270 | if (lttng_is_setuid_setgid()) { |
271 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
272 | "-o, --output"); | |
273 | } else { | |
274 | ret = asprintf(&opt_output_path, "%s", arg); | |
275 | if (ret < 0) { | |
276 | ret = -errno; | |
277 | PERROR("asprintf opt_output_path"); | |
278 | goto end; | |
279 | } | |
cd60b05a JG |
280 | } |
281 | break; | |
282 | case 'v': | |
283 | /* Verbose level can increase using multiple -v */ | |
284 | if (arg) { | |
285 | lttng_opt_verbose = config_parse_value(arg); | |
286 | } else { | |
849e5b7b DG |
287 | /* Only 3 level of verbosity (-vvv). */ |
288 | if (lttng_opt_verbose < 3) { | |
289 | lttng_opt_verbose += 1; | |
290 | } | |
cd60b05a JG |
291 | } |
292 | break; | |
293 | default: | |
294 | /* Unknown option or other error. | |
295 | * Error is printed by getopt, just return */ | |
296 | ret = -1; | |
297 | goto end; | |
298 | } | |
299 | ||
300 | /* All good. */ | |
301 | ret = 0; | |
302 | ||
303 | end: | |
304 | return ret; | |
305 | } | |
306 | ||
307 | /* | |
308 | * config_entry_handler_cb used to handle options read from a config file. | |
f40ef1d5 | 309 | * See config_entry_handler_cb comment in common/config/session-config.h for the |
cd60b05a JG |
310 | * return value conventions. |
311 | */ | |
7591bab1 | 312 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
313 | { |
314 | int ret = 0, i; | |
315 | ||
316 | if (!entry || !entry->name || !entry->value) { | |
317 | ret = -EINVAL; | |
318 | goto end; | |
319 | } | |
320 | ||
321 | /* Check if the option is to be ignored */ | |
322 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
323 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
324 | goto end; | |
325 | } | |
326 | } | |
327 | ||
328 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
329 | /* Ignore if entry name is not fully matched. */ | |
330 | if (strcmp(entry->name, long_options[i].name)) { | |
331 | continue; | |
332 | } | |
333 | ||
334 | /* | |
7591bab1 MD |
335 | * If the option takes no argument on the command line, |
336 | * we have to check if the value is "true". We support | |
337 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
338 | */ |
339 | if (!long_options[i].has_arg) { | |
340 | ret = config_parse_value(entry->value); | |
341 | if (ret <= 0) { | |
342 | if (ret) { | |
343 | WARN("Invalid configuration value \"%s\" for option %s", | |
344 | entry->value, entry->name); | |
345 | } | |
346 | /* False, skip boolean config option. */ | |
347 | goto end; | |
348 | } | |
349 | } | |
350 | ||
351 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
352 | goto end; | |
353 | } | |
354 | ||
355 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
356 | entry->name); | |
357 | ||
358 | end: | |
359 | return ret; | |
360 | } | |
361 | ||
7591bab1 | 362 | static int set_options(int argc, char **argv) |
cd60b05a | 363 | { |
178a0557 | 364 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
365 | int orig_optopt = optopt, orig_optind = optind; |
366 | char *default_address, *optstring; | |
367 | const char *config_path = NULL; | |
368 | ||
369 | optstring = utils_generate_optstring(long_options, | |
370 | sizeof(long_options) / sizeof(struct option)); | |
371 | if (!optstring) { | |
178a0557 | 372 | retval = -ENOMEM; |
cd60b05a JG |
373 | goto exit; |
374 | } | |
375 | ||
376 | /* Check for the --config option */ | |
377 | ||
378 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
379 | &option_index)) != -1) { | |
380 | if (c == '?') { | |
178a0557 | 381 | retval = -EINVAL; |
cd60b05a JG |
382 | goto exit; |
383 | } else if (c != 'f') { | |
384 | continue; | |
385 | } | |
386 | ||
e8fa9fb0 MD |
387 | if (lttng_is_setuid_setgid()) { |
388 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
389 | "-f, --config"); | |
390 | } else { | |
391 | config_path = utils_expand_path(optarg); | |
392 | if (!config_path) { | |
393 | ERR("Failed to resolve path: %s", optarg); | |
394 | } | |
cd60b05a JG |
395 | } |
396 | } | |
397 | ||
398 | ret = config_get_section_entries(config_path, config_section_name, | |
399 | config_entry_handler, NULL); | |
400 | if (ret) { | |
401 | if (ret > 0) { | |
402 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 403 | } |
178a0557 | 404 | retval = -1; |
cd60b05a JG |
405 | goto exit; |
406 | } | |
b8aa1682 | 407 | |
cd60b05a JG |
408 | /* Reset getopt's global state */ |
409 | optopt = orig_optopt; | |
410 | optind = orig_optind; | |
b8aa1682 | 411 | while (1) { |
cd60b05a | 412 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
413 | if (c == -1) { |
414 | break; | |
415 | } | |
416 | ||
cd60b05a JG |
417 | ret = set_option(c, optarg, long_options[option_index].name); |
418 | if (ret < 0) { | |
178a0557 | 419 | retval = -1; |
b8aa1682 JD |
420 | goto exit; |
421 | } | |
422 | } | |
423 | ||
424 | /* assign default values */ | |
425 | if (control_uri == NULL) { | |
fa91dc52 MD |
426 | ret = asprintf(&default_address, |
427 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
428 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
429 | if (ret < 0) { |
430 | PERROR("asprintf default data address"); | |
178a0557 | 431 | retval = -1; |
b8aa1682 JD |
432 | goto exit; |
433 | } | |
434 | ||
435 | ret = uri_parse(default_address, &control_uri); | |
436 | free(default_address); | |
437 | if (ret < 0) { | |
438 | ERR("Invalid control URI specified"); | |
178a0557 | 439 | retval = -1; |
b8aa1682 JD |
440 | goto exit; |
441 | } | |
442 | } | |
443 | if (data_uri == NULL) { | |
fa91dc52 MD |
444 | ret = asprintf(&default_address, |
445 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
446 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
447 | if (ret < 0) { |
448 | PERROR("asprintf default data address"); | |
178a0557 | 449 | retval = -1; |
b8aa1682 JD |
450 | goto exit; |
451 | } | |
452 | ||
453 | ret = uri_parse(default_address, &data_uri); | |
454 | free(default_address); | |
455 | if (ret < 0) { | |
456 | ERR("Invalid data URI specified"); | |
178a0557 | 457 | retval = -1; |
b8aa1682 JD |
458 | goto exit; |
459 | } | |
460 | } | |
d3e2ba59 | 461 | if (live_uri == NULL) { |
fa91dc52 MD |
462 | ret = asprintf(&default_address, |
463 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
464 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
465 | if (ret < 0) { |
466 | PERROR("asprintf default viewer control address"); | |
178a0557 | 467 | retval = -1; |
d3e2ba59 JD |
468 | goto exit; |
469 | } | |
470 | ||
471 | ret = uri_parse(default_address, &live_uri); | |
472 | free(default_address); | |
473 | if (ret < 0) { | |
474 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 475 | retval = -1; |
d3e2ba59 JD |
476 | goto exit; |
477 | } | |
478 | } | |
b8aa1682 JD |
479 | |
480 | exit: | |
cd60b05a | 481 | free(optstring); |
178a0557 | 482 | return retval; |
b8aa1682 JD |
483 | } |
484 | ||
7591bab1 MD |
485 | static void print_global_objects(void) |
486 | { | |
487 | rcu_register_thread(); | |
488 | ||
489 | print_viewer_streams(); | |
490 | print_relay_streams(); | |
491 | print_sessions(); | |
492 | ||
493 | rcu_unregister_thread(); | |
494 | } | |
495 | ||
b8aa1682 JD |
496 | /* |
497 | * Cleanup the daemon | |
498 | */ | |
7591bab1 | 499 | static void relayd_cleanup(void) |
b8aa1682 | 500 | { |
7591bab1 MD |
501 | print_global_objects(); |
502 | ||
b8aa1682 JD |
503 | DBG("Cleaning up"); |
504 | ||
178a0557 MD |
505 | if (viewer_streams_ht) |
506 | lttng_ht_destroy(viewer_streams_ht); | |
507 | if (relay_streams_ht) | |
508 | lttng_ht_destroy(relay_streams_ht); | |
7591bab1 MD |
509 | if (sessions_ht) |
510 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 511 | |
095a4ae5 MD |
512 | /* free the dynamically allocated opt_output_path */ |
513 | free(opt_output_path); | |
514 | ||
a02de639 CB |
515 | /* Close thread quit pipes */ |
516 | utils_close_pipe(thread_quit_pipe); | |
517 | ||
710c1f73 DG |
518 | uri_free(control_uri); |
519 | uri_free(data_uri); | |
8d5c808e | 520 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
521 | |
522 | if (tracing_group_name_override) { | |
523 | free((void *) tracing_group_name); | |
524 | } | |
b8aa1682 JD |
525 | } |
526 | ||
527 | /* | |
528 | * Write to writable pipe used to notify a thread. | |
529 | */ | |
7591bab1 | 530 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 531 | { |
6cd525e8 | 532 | ssize_t ret; |
b8aa1682 | 533 | |
6cd525e8 MD |
534 | ret = lttng_write(wpipe, "!", 1); |
535 | if (ret < 1) { | |
b8aa1682 | 536 | PERROR("write poll pipe"); |
b4aacfdc | 537 | goto end; |
b8aa1682 | 538 | } |
b4aacfdc MD |
539 | ret = 0; |
540 | end: | |
b8aa1682 JD |
541 | return ret; |
542 | } | |
543 | ||
7591bab1 | 544 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 545 | { |
6cd525e8 | 546 | ssize_t ret; |
65931c8b | 547 | |
6cd525e8 MD |
548 | ret = lttng_write(pipe[1], "4", 1); |
549 | if (ret < 1) { | |
65931c8b | 550 | PERROR("write relay health quit"); |
b4aacfdc | 551 | goto end; |
65931c8b | 552 | } |
b4aacfdc MD |
553 | ret = 0; |
554 | end: | |
555 | return ret; | |
65931c8b MD |
556 | } |
557 | ||
b8aa1682 | 558 | /* |
b4aacfdc | 559 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 560 | */ |
b4aacfdc | 561 | int lttng_relay_stop_threads(void) |
b8aa1682 | 562 | { |
b4aacfdc | 563 | int retval = 0; |
b8aa1682 JD |
564 | |
565 | /* Stopping all threads */ | |
566 | DBG("Terminating all threads"); | |
b4aacfdc | 567 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 568 | ERR("write error on thread quit pipe"); |
b4aacfdc | 569 | retval = -1; |
b8aa1682 JD |
570 | } |
571 | ||
b4aacfdc MD |
572 | if (notify_health_quit_pipe(health_quit_pipe)) { |
573 | ERR("write error on health quit pipe"); | |
574 | } | |
65931c8b | 575 | |
b8aa1682 | 576 | /* Dispatch thread */ |
26c9d55e | 577 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 578 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 579 | |
b4aacfdc | 580 | if (relayd_live_stop()) { |
178a0557 | 581 | ERR("Error stopping live threads"); |
b4aacfdc | 582 | retval = -1; |
178a0557 | 583 | } |
b4aacfdc | 584 | return retval; |
b8aa1682 JD |
585 | } |
586 | ||
587 | /* | |
588 | * Signal handler for the daemon | |
589 | * | |
590 | * Simply stop all worker threads, leaving main() return gracefully after | |
591 | * joining all threads and calling cleanup(). | |
592 | */ | |
7591bab1 | 593 | static void sighandler(int sig) |
b8aa1682 JD |
594 | { |
595 | switch (sig) { | |
b8aa1682 JD |
596 | case SIGINT: |
597 | DBG("SIGINT caught"); | |
b4aacfdc MD |
598 | if (lttng_relay_stop_threads()) { |
599 | ERR("Error stopping threads"); | |
600 | } | |
b8aa1682 JD |
601 | break; |
602 | case SIGTERM: | |
603 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
604 | if (lttng_relay_stop_threads()) { |
605 | ERR("Error stopping threads"); | |
606 | } | |
b8aa1682 | 607 | break; |
3fd27398 MD |
608 | case SIGUSR1: |
609 | CMM_STORE_SHARED(recv_child_signal, 1); | |
610 | break; | |
b8aa1682 JD |
611 | default: |
612 | break; | |
613 | } | |
614 | } | |
615 | ||
616 | /* | |
617 | * Setup signal handler for : | |
618 | * SIGINT, SIGTERM, SIGPIPE | |
619 | */ | |
7591bab1 | 620 | static int set_signal_handler(void) |
b8aa1682 JD |
621 | { |
622 | int ret = 0; | |
623 | struct sigaction sa; | |
624 | sigset_t sigset; | |
625 | ||
626 | if ((ret = sigemptyset(&sigset)) < 0) { | |
627 | PERROR("sigemptyset"); | |
628 | return ret; | |
629 | } | |
630 | ||
b8aa1682 JD |
631 | sa.sa_mask = sigset; |
632 | sa.sa_flags = 0; | |
0072e5e2 MD |
633 | |
634 | sa.sa_handler = sighandler; | |
b8aa1682 JD |
635 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { |
636 | PERROR("sigaction"); | |
637 | return ret; | |
638 | } | |
639 | ||
640 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
641 | PERROR("sigaction"); | |
642 | return ret; | |
643 | } | |
644 | ||
0072e5e2 | 645 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
b8aa1682 JD |
646 | PERROR("sigaction"); |
647 | return ret; | |
648 | } | |
649 | ||
0072e5e2 MD |
650 | sa.sa_handler = SIG_IGN; |
651 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
3fd27398 MD |
652 | PERROR("sigaction"); |
653 | return ret; | |
654 | } | |
655 | ||
656 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
657 | |
658 | return ret; | |
659 | } | |
660 | ||
3fd27398 MD |
661 | void lttng_relay_notify_ready(void) |
662 | { | |
663 | /* Notify the parent of the fork() process that we are ready. */ | |
664 | if (opt_daemon || opt_background) { | |
665 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
666 | kill(child_ppid, SIGUSR1); | |
667 | } | |
668 | } | |
669 | } | |
670 | ||
b8aa1682 JD |
671 | /* |
672 | * Init thread quit pipe. | |
673 | * | |
674 | * Return -1 on error or 0 if all pipes are created. | |
675 | */ | |
7591bab1 | 676 | static int init_thread_quit_pipe(void) |
b8aa1682 | 677 | { |
a02de639 | 678 | int ret; |
b8aa1682 | 679 | |
a02de639 | 680 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 681 | |
b8aa1682 JD |
682 | return ret; |
683 | } | |
684 | ||
685 | /* | |
686 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
687 | */ | |
7591bab1 | 688 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
689 | { |
690 | int ret; | |
691 | ||
692 | if (events == NULL || size == 0) { | |
693 | ret = -1; | |
694 | goto error; | |
695 | } | |
696 | ||
697 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
698 | if (ret < 0) { | |
699 | goto error; | |
700 | } | |
701 | ||
702 | /* Add quit pipe */ | |
c7759e6a | 703 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
704 | if (ret < 0) { |
705 | goto error; | |
706 | } | |
707 | ||
708 | return 0; | |
709 | ||
710 | error: | |
711 | return ret; | |
712 | } | |
713 | ||
714 | /* | |
715 | * Check if the thread quit pipe was triggered. | |
716 | * | |
717 | * Return 1 if it was triggered else 0; | |
718 | */ | |
7591bab1 | 719 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
720 | { |
721 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
722 | return 1; | |
723 | } | |
724 | ||
725 | return 0; | |
726 | } | |
727 | ||
728 | /* | |
729 | * Create and init socket from uri. | |
730 | */ | |
7591bab1 | 731 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
732 | { |
733 | int ret; | |
734 | struct lttcomm_sock *sock = NULL; | |
735 | ||
736 | sock = lttcomm_alloc_sock_from_uri(uri); | |
737 | if (sock == NULL) { | |
738 | ERR("Allocating socket"); | |
739 | goto error; | |
740 | } | |
741 | ||
742 | ret = lttcomm_create_sock(sock); | |
743 | if (ret < 0) { | |
744 | goto error; | |
745 | } | |
746 | DBG("Listening on sock %d", sock->fd); | |
747 | ||
748 | ret = sock->ops->bind(sock); | |
749 | if (ret < 0) { | |
750 | goto error; | |
751 | } | |
752 | ||
753 | ret = sock->ops->listen(sock, -1); | |
754 | if (ret < 0) { | |
755 | goto error; | |
756 | ||
757 | } | |
758 | ||
759 | return sock; | |
760 | ||
761 | error: | |
762 | if (sock) { | |
763 | lttcomm_destroy_sock(sock); | |
764 | } | |
765 | return NULL; | |
766 | } | |
767 | ||
768 | /* | |
769 | * This thread manages the listening for new connections on the network | |
770 | */ | |
7591bab1 | 771 | static void *relay_thread_listener(void *data) |
b8aa1682 | 772 | { |
095a4ae5 | 773 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
774 | uint32_t revents, nb_fd; |
775 | struct lttng_poll_event events; | |
776 | struct lttcomm_sock *control_sock, *data_sock; | |
777 | ||
b8aa1682 JD |
778 | DBG("[thread] Relay listener started"); |
779 | ||
55706a7d MD |
780 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
781 | ||
f385ae0a MD |
782 | health_code_update(); |
783 | ||
7591bab1 | 784 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 785 | if (!control_sock) { |
095a4ae5 | 786 | goto error_sock_control; |
b8aa1682 JD |
787 | } |
788 | ||
7591bab1 | 789 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 790 | if (!data_sock) { |
095a4ae5 | 791 | goto error_sock_relay; |
b8aa1682 JD |
792 | } |
793 | ||
794 | /* | |
7591bab1 MD |
795 | * Pass 3 as size here for the thread quit pipe, control and |
796 | * data socket. | |
b8aa1682 JD |
797 | */ |
798 | ret = create_thread_poll_set(&events, 3); | |
799 | if (ret < 0) { | |
800 | goto error_create_poll; | |
801 | } | |
802 | ||
803 | /* Add the control socket */ | |
804 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
805 | if (ret < 0) { | |
806 | goto error_poll_add; | |
807 | } | |
808 | ||
809 | /* Add the data socket */ | |
810 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
811 | if (ret < 0) { | |
812 | goto error_poll_add; | |
813 | } | |
814 | ||
3fd27398 MD |
815 | lttng_relay_notify_ready(); |
816 | ||
9b5e0863 MD |
817 | if (testpoint(relayd_thread_listener)) { |
818 | goto error_testpoint; | |
819 | } | |
820 | ||
b8aa1682 | 821 | while (1) { |
f385ae0a MD |
822 | health_code_update(); |
823 | ||
b8aa1682 JD |
824 | DBG("Listener accepting connections"); |
825 | ||
b8aa1682 | 826 | restart: |
f385ae0a | 827 | health_poll_entry(); |
b8aa1682 | 828 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 829 | health_poll_exit(); |
b8aa1682 JD |
830 | if (ret < 0) { |
831 | /* | |
832 | * Restart interrupted system call. | |
833 | */ | |
834 | if (errno == EINTR) { | |
835 | goto restart; | |
836 | } | |
837 | goto error; | |
838 | } | |
839 | ||
0d9c5d77 DG |
840 | nb_fd = ret; |
841 | ||
b8aa1682 JD |
842 | DBG("Relay new connection received"); |
843 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
844 | health_code_update(); |
845 | ||
b8aa1682 JD |
846 | /* Fetch once the poll data */ |
847 | revents = LTTNG_POLL_GETEV(&events, i); | |
848 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
849 | ||
fd20dac9 | 850 | if (!revents) { |
7591bab1 MD |
851 | /* |
852 | * No activity for this FD (poll | |
853 | * implementation). | |
854 | */ | |
fd20dac9 MD |
855 | continue; |
856 | } | |
857 | ||
b8aa1682 JD |
858 | /* Thread quit pipe has been closed. Killing thread. */ |
859 | ret = check_thread_quit_pipe(pollfd, revents); | |
860 | if (ret) { | |
095a4ae5 MD |
861 | err = 0; |
862 | goto exit; | |
b8aa1682 JD |
863 | } |
864 | ||
03e43155 | 865 | if (revents & LPOLLIN) { |
4b7f17b2 | 866 | /* |
7591bab1 MD |
867 | * A new connection is requested, therefore a |
868 | * sessiond/consumerd connection is allocated in | |
869 | * this thread, enqueued to a global queue and | |
870 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 871 | */ |
58eb9381 DG |
872 | int val = 1; |
873 | struct relay_connection *new_conn; | |
4b7f17b2 | 874 | struct lttcomm_sock *newsock; |
7591bab1 | 875 | enum connection_type type; |
b8aa1682 JD |
876 | |
877 | if (pollfd == data_sock->fd) { | |
7591bab1 | 878 | type = RELAY_DATA; |
b8aa1682 | 879 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
880 | DBG("Relay data connection accepted, socket %d", |
881 | newsock->fd); | |
4b7f17b2 MD |
882 | } else { |
883 | assert(pollfd == control_sock->fd); | |
7591bab1 | 884 | type = RELAY_CONTROL; |
b8aa1682 | 885 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
886 | DBG("Relay control connection accepted, socket %d", |
887 | newsock->fd); | |
b8aa1682 | 888 | } |
58eb9381 DG |
889 | if (!newsock) { |
890 | PERROR("accepting sock"); | |
58eb9381 DG |
891 | goto error; |
892 | } | |
893 | ||
894 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
895 | sizeof(val)); | |
b8aa1682 JD |
896 | if (ret < 0) { |
897 | PERROR("setsockopt inet"); | |
4b7f17b2 | 898 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
899 | goto error; |
900 | } | |
2fc6b1ab JR |
901 | |
902 | ret = socket_apply_keep_alive_config(newsock->fd); | |
903 | if (ret < 0) { | |
904 | ERR("Failed to apply TCP keep-alive configuration on socket (%i)", | |
905 | newsock->fd); | |
906 | lttcomm_destroy_sock(newsock); | |
907 | goto error; | |
908 | } | |
909 | ||
7591bab1 MD |
910 | new_conn = connection_create(newsock, type); |
911 | if (!new_conn) { | |
912 | lttcomm_destroy_sock(newsock); | |
913 | goto error; | |
914 | } | |
58eb9381 DG |
915 | |
916 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
917 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
918 | &new_conn->qnode); | |
b8aa1682 JD |
919 | |
920 | /* | |
7591bab1 MD |
921 | * Wake the dispatch queue futex. |
922 | * Implicit memory barrier with the | |
923 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 924 | */ |
58eb9381 | 925 | futex_nto1_wake(&relay_conn_queue.futex); |
03e43155 MD |
926 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
927 | ERR("socket poll error"); | |
928 | goto error; | |
929 | } else { | |
930 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
931 | goto error; | |
b8aa1682 JD |
932 | } |
933 | } | |
934 | } | |
935 | ||
095a4ae5 | 936 | exit: |
b8aa1682 JD |
937 | error: |
938 | error_poll_add: | |
9b5e0863 | 939 | error_testpoint: |
b8aa1682 JD |
940 | lttng_poll_clean(&events); |
941 | error_create_poll: | |
095a4ae5 MD |
942 | if (data_sock->fd >= 0) { |
943 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
944 | if (ret) { |
945 | PERROR("close"); | |
946 | } | |
b8aa1682 | 947 | } |
095a4ae5 MD |
948 | lttcomm_destroy_sock(data_sock); |
949 | error_sock_relay: | |
950 | if (control_sock->fd >= 0) { | |
951 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
952 | if (ret) { |
953 | PERROR("close"); | |
954 | } | |
b8aa1682 | 955 | } |
095a4ae5 MD |
956 | lttcomm_destroy_sock(control_sock); |
957 | error_sock_control: | |
958 | if (err) { | |
f385ae0a MD |
959 | health_error(); |
960 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 961 | } |
55706a7d | 962 | health_unregister(health_relayd); |
b8aa1682 | 963 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 964 | lttng_relay_stop_threads(); |
b8aa1682 JD |
965 | return NULL; |
966 | } | |
967 | ||
968 | /* | |
969 | * This thread manages the dispatching of the requests to worker threads | |
970 | */ | |
7591bab1 | 971 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 972 | { |
6cd525e8 MD |
973 | int err = -1; |
974 | ssize_t ret; | |
8bdee6e2 | 975 | struct cds_wfcq_node *node; |
58eb9381 | 976 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
977 | |
978 | DBG("[thread] Relay dispatcher started"); | |
979 | ||
55706a7d MD |
980 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
981 | ||
9b5e0863 MD |
982 | if (testpoint(relayd_thread_dispatcher)) { |
983 | goto error_testpoint; | |
984 | } | |
985 | ||
f385ae0a MD |
986 | health_code_update(); |
987 | ||
68924dc1 | 988 | for (;;) { |
f385ae0a MD |
989 | health_code_update(); |
990 | ||
b8aa1682 | 991 | /* Atomically prepare the queue futex */ |
58eb9381 | 992 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 | 993 | |
68924dc1 MD |
994 | if (CMM_LOAD_SHARED(dispatch_thread_exit)) { |
995 | break; | |
996 | } | |
997 | ||
b8aa1682 | 998 | do { |
f385ae0a MD |
999 | health_code_update(); |
1000 | ||
b8aa1682 | 1001 | /* Dequeue commands */ |
8bdee6e2 SM |
1002 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
1003 | &relay_conn_queue.tail); | |
b8aa1682 JD |
1004 | if (node == NULL) { |
1005 | DBG("Woken up but nothing in the relay command queue"); | |
1006 | /* Continue thread execution */ | |
1007 | break; | |
1008 | } | |
58eb9381 | 1009 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 1010 | |
58eb9381 | 1011 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1012 | |
1013 | /* | |
7591bab1 MD |
1014 | * Inform worker thread of the new request. This |
1015 | * call is blocking so we can be assured that | |
1016 | * the data will be read at some point in time | |
1017 | * or wait to the end of the world :) | |
b8aa1682 | 1018 | */ |
58eb9381 DG |
1019 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1020 | if (ret < 0) { | |
1021 | PERROR("write connection pipe"); | |
7591bab1 | 1022 | connection_put(new_conn); |
b8aa1682 JD |
1023 | goto error; |
1024 | } | |
1025 | } while (node != NULL); | |
1026 | ||
1027 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1028 | health_poll_entry(); |
58eb9381 | 1029 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1030 | health_poll_exit(); |
b8aa1682 JD |
1031 | } |
1032 | ||
f385ae0a MD |
1033 | /* Normal exit, no error */ |
1034 | err = 0; | |
1035 | ||
b8aa1682 | 1036 | error: |
9b5e0863 | 1037 | error_testpoint: |
f385ae0a MD |
1038 | if (err) { |
1039 | health_error(); | |
1040 | ERR("Health error occurred in %s", __func__); | |
1041 | } | |
55706a7d | 1042 | health_unregister(health_relayd); |
b8aa1682 | 1043 | DBG("Dispatch thread dying"); |
b4aacfdc | 1044 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1045 | return NULL; |
1046 | } | |
1047 | ||
b8aa1682 | 1048 | /* |
7591bab1 | 1049 | * Set index data from the control port to a given index object. |
b8aa1682 | 1050 | */ |
7591bab1 | 1051 | static int set_index_control_data(struct relay_index *index, |
234cd636 JD |
1052 | struct lttcomm_relayd_index *data, |
1053 | struct relay_connection *conn) | |
1c20f0e2 | 1054 | { |
7591bab1 | 1055 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1056 | |
1057 | /* | |
44a2fbf1 | 1058 | * The index on disk is encoded in big endian. |
1c20f0e2 | 1059 | */ |
44a2fbf1 JG |
1060 | index_data.packet_size = htobe64(data->packet_size); |
1061 | index_data.content_size = htobe64(data->content_size); | |
1062 | index_data.timestamp_begin = htobe64(data->timestamp_begin); | |
1063 | index_data.timestamp_end = htobe64(data->timestamp_end); | |
1064 | index_data.events_discarded = htobe64(data->events_discarded); | |
1065 | index_data.stream_id = htobe64(data->stream_id); | |
234cd636 JD |
1066 | |
1067 | if (conn->minor >= 8) { | |
44a2fbf1 JG |
1068 | index->index_data.stream_instance_id = htobe64(data->stream_instance_id); |
1069 | index->index_data.packet_seq_num = htobe64(data->packet_seq_num); | |
234cd636 JD |
1070 | } |
1071 | ||
7591bab1 | 1072 | return relay_index_set_data(index, &index_data); |
1c20f0e2 JD |
1073 | } |
1074 | ||
c5b6f4f0 DG |
1075 | /* |
1076 | * Handle the RELAYD_CREATE_SESSION command. | |
1077 | * | |
1078 | * On success, send back the session id or else return a negative value. | |
1079 | */ | |
44a2fbf1 JG |
1080 | static int relay_create_session(const struct lttcomm_relayd_hdr *recv_hdr, |
1081 | struct relay_connection *conn, | |
1082 | const struct lttng_buffer_view *payload) | |
c5b6f4f0 | 1083 | { |
44a2fbf1 JG |
1084 | int ret = 0; |
1085 | ssize_t send_ret; | |
c5b6f4f0 DG |
1086 | struct relay_session *session; |
1087 | struct lttcomm_relayd_status_session reply; | |
36d2e35d | 1088 | char session_name[LTTNG_NAME_MAX]; |
9cf3eb20 | 1089 | char hostname[LTTNG_HOST_NAME_MAX]; |
7591bab1 MD |
1090 | uint32_t live_timer = 0; |
1091 | bool snapshot = false; | |
c5b6f4f0 | 1092 | |
36d2e35d | 1093 | memset(session_name, 0, LTTNG_NAME_MAX); |
9cf3eb20 | 1094 | memset(hostname, 0, LTTNG_HOST_NAME_MAX); |
c5b6f4f0 DG |
1095 | |
1096 | memset(&reply, 0, sizeof(reply)); | |
1097 | ||
58eb9381 | 1098 | switch (conn->minor) { |
2a174661 DG |
1099 | case 1: |
1100 | case 2: | |
1101 | case 3: | |
1102 | break; | |
1103 | case 4: /* LTTng sessiond 2.4 */ | |
1104 | default: | |
44a2fbf1 | 1105 | ret = cmd_create_session_2_4(payload, session_name, |
7591bab1 MD |
1106 | hostname, &live_timer, &snapshot); |
1107 | } | |
1108 | if (ret < 0) { | |
1109 | goto send_reply; | |
d3e2ba59 JD |
1110 | } |
1111 | ||
7591bab1 MD |
1112 | session = session_create(session_name, hostname, live_timer, |
1113 | snapshot, conn->major, conn->minor); | |
1114 | if (!session) { | |
1115 | ret = -1; | |
1116 | goto send_reply; | |
1117 | } | |
1118 | assert(!conn->session); | |
1119 | conn->session = session; | |
c5b6f4f0 DG |
1120 | DBG("Created session %" PRIu64, session->id); |
1121 | ||
7591bab1 MD |
1122 | reply.session_id = htobe64(session->id); |
1123 | ||
1124 | send_reply: | |
c5b6f4f0 DG |
1125 | if (ret < 0) { |
1126 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1127 | } else { | |
1128 | reply.ret_code = htobe32(LTTNG_OK); | |
1129 | } | |
1130 | ||
58eb9381 | 1131 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
44a2fbf1 JG |
1132 | if (send_ret < (ssize_t) sizeof(reply)) { |
1133 | ERR("Failed to send \"create session\" command reply (ret = %zd)", | |
1134 | send_ret); | |
1135 | ret = -1; | |
c5b6f4f0 DG |
1136 | } |
1137 | ||
1138 | return ret; | |
1139 | } | |
1140 | ||
a4baae1b JD |
1141 | /* |
1142 | * When we have received all the streams and the metadata for a channel, | |
1143 | * we make them visible to the viewer threads. | |
1144 | */ | |
7591bab1 | 1145 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1146 | { |
7591bab1 MD |
1147 | struct relay_stream *stream; |
1148 | struct relay_session *session = conn->session; | |
a4baae1b | 1149 | |
7591bab1 MD |
1150 | /* |
1151 | * We publish all streams belonging to a session atomically wrt | |
1152 | * session lock. | |
1153 | */ | |
1154 | pthread_mutex_lock(&session->lock); | |
1155 | rcu_read_lock(); | |
1156 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1157 | recv_node) { | |
1158 | stream_publish(stream); | |
a4baae1b | 1159 | } |
7591bab1 | 1160 | rcu_read_unlock(); |
a4baae1b | 1161 | |
7591bab1 MD |
1162 | /* |
1163 | * Inform the viewer that there are new streams in the session. | |
1164 | */ | |
1165 | if (session->viewer_attached) { | |
1166 | uatomic_set(&session->new_streams, 1); | |
1167 | } | |
1168 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1169 | } |
1170 | ||
b8aa1682 JD |
1171 | /* |
1172 | * relay_add_stream: allocate a new stream for a session | |
1173 | */ | |
44a2fbf1 JG |
1174 | static int relay_add_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1175 | struct relay_connection *conn, | |
1176 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1177 | { |
7591bab1 MD |
1178 | int ret; |
1179 | ssize_t send_ret; | |
58eb9381 | 1180 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1181 | struct relay_stream *stream = NULL; |
1182 | struct lttcomm_relayd_status_stream reply; | |
4030a636 | 1183 | struct ctf_trace *trace = NULL; |
7591bab1 MD |
1184 | uint64_t stream_handle = -1ULL; |
1185 | char *path_name = NULL, *channel_name = NULL; | |
1186 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1187 | |
44a2fbf1 | 1188 | if (!session || !conn->version_check_done) { |
b8aa1682 JD |
1189 | ERR("Trying to add a stream before version check"); |
1190 | ret = -1; | |
1191 | goto end_no_session; | |
1192 | } | |
1193 | ||
7591bab1 MD |
1194 | switch (session->minor) { |
1195 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
44a2fbf1 | 1196 | ret = cmd_recv_stream_2_1(payload, &path_name, |
7591bab1 | 1197 | &channel_name); |
0f907de1 | 1198 | break; |
7591bab1 | 1199 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1200 | default: |
44a2fbf1 | 1201 | ret = cmd_recv_stream_2_2(payload, &path_name, |
7591bab1 | 1202 | &channel_name, &tracefile_size, &tracefile_count); |
0f907de1 JD |
1203 | break; |
1204 | } | |
1205 | if (ret < 0) { | |
7591bab1 | 1206 | goto send_reply; |
b8aa1682 JD |
1207 | } |
1208 | ||
7591bab1 | 1209 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1210 | if (!trace) { |
7591bab1 | 1211 | goto send_reply; |
2a174661 | 1212 | } |
7591bab1 | 1213 | /* This stream here has one reference on the trace. */ |
2a174661 | 1214 | |
7591bab1 MD |
1215 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1216 | stream_handle = ++last_relay_stream_id; | |
1217 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1218 | |
7591bab1 MD |
1219 | /* We pass ownership of path_name and channel_name. */ |
1220 | stream = stream_create(trace, stream_handle, path_name, | |
1221 | channel_name, tracefile_size, tracefile_count); | |
1222 | path_name = NULL; | |
1223 | channel_name = NULL; | |
a4baae1b | 1224 | |
2a174661 | 1225 | /* |
7591bab1 MD |
1226 | * Streams are the owners of their trace. Reference to trace is |
1227 | * kept within stream_create(). | |
2a174661 | 1228 | */ |
7591bab1 | 1229 | ctf_trace_put(trace); |
d3e2ba59 | 1230 | |
7591bab1 | 1231 | send_reply: |
53efb85a | 1232 | memset(&reply, 0, sizeof(reply)); |
7591bab1 MD |
1233 | reply.handle = htobe64(stream_handle); |
1234 | if (!stream) { | |
f73fabfd | 1235 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1236 | } else { |
f73fabfd | 1237 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1238 | } |
5af40280 | 1239 | |
58eb9381 | 1240 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 | 1241 | sizeof(struct lttcomm_relayd_status_stream), 0); |
44a2fbf1 JG |
1242 | if (send_ret < (ssize_t) sizeof(reply)) { |
1243 | ERR("Failed to send \"add stream\" command reply (ret = %zd)", | |
1244 | send_ret); | |
1245 | ret = -1; | |
b8aa1682 JD |
1246 | } |
1247 | ||
1248 | end_no_session: | |
7591bab1 MD |
1249 | free(path_name); |
1250 | free(channel_name); | |
0f907de1 | 1251 | return ret; |
b8aa1682 JD |
1252 | } |
1253 | ||
173af62f DG |
1254 | /* |
1255 | * relay_close_stream: close a specific stream | |
1256 | */ | |
44a2fbf1 JG |
1257 | static int relay_close_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1258 | struct relay_connection *conn, | |
1259 | const struct lttng_buffer_view *payload) | |
173af62f | 1260 | { |
44a2fbf1 JG |
1261 | int ret; |
1262 | ssize_t send_ret; | |
58eb9381 | 1263 | struct relay_session *session = conn->session; |
173af62f DG |
1264 | struct lttcomm_relayd_close_stream stream_info; |
1265 | struct lttcomm_relayd_generic_reply reply; | |
1266 | struct relay_stream *stream; | |
173af62f DG |
1267 | |
1268 | DBG("Close stream received"); | |
1269 | ||
44a2fbf1 | 1270 | if (!session || !conn->version_check_done) { |
173af62f DG |
1271 | ERR("Trying to close a stream before version check"); |
1272 | ret = -1; | |
1273 | goto end_no_session; | |
1274 | } | |
1275 | ||
44a2fbf1 JG |
1276 | if (payload->size < sizeof(stream_info)) { |
1277 | ERR("Unexpected payload size in \"relay_close_stream\": expected >= %zu bytes, got %zu bytes", | |
1278 | sizeof(stream_info), payload->size); | |
173af62f DG |
1279 | ret = -1; |
1280 | goto end_no_session; | |
1281 | } | |
44a2fbf1 JG |
1282 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1283 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1284 | stream_info.last_net_seq_num = be64toh(stream_info.last_net_seq_num); | |
173af62f | 1285 | |
44a2fbf1 | 1286 | stream = stream_get_by_id(stream_info.stream_id); |
173af62f DG |
1287 | if (!stream) { |
1288 | ret = -1; | |
7591bab1 | 1289 | goto end; |
173af62f | 1290 | } |
77f7bd85 MD |
1291 | |
1292 | /* | |
1293 | * Set last_net_seq_num before the close flag. Required by data | |
1294 | * pending check. | |
1295 | */ | |
7591bab1 | 1296 | pthread_mutex_lock(&stream->lock); |
44a2fbf1 | 1297 | stream->last_net_seq_num = stream_info.last_net_seq_num; |
77f7bd85 MD |
1298 | pthread_mutex_unlock(&stream->lock); |
1299 | ||
bda7c7b9 JG |
1300 | /* |
1301 | * This is one of the conditions which may trigger a stream close | |
1302 | * with the others being: | |
1303 | * 1) A close command is received for a stream | |
1304 | * 2) The control connection owning the stream is closed | |
1305 | * 3) We have received all of the stream's data _after_ a close | |
1306 | * request. | |
1307 | */ | |
1308 | try_stream_close(stream); | |
7591bab1 MD |
1309 | if (stream->is_metadata) { |
1310 | struct relay_viewer_stream *vstream; | |
173af62f | 1311 | |
7591bab1 MD |
1312 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1313 | if (vstream) { | |
1314 | if (vstream->metadata_sent == stream->metadata_received) { | |
1315 | /* | |
1316 | * Since all the metadata has been sent to the | |
1317 | * viewer and that we have a request to close | |
1318 | * its stream, we can safely teardown the | |
1319 | * corresponding metadata viewer stream. | |
1320 | */ | |
1321 | viewer_stream_put(vstream); | |
1322 | } | |
1323 | /* Put local reference. */ | |
1324 | viewer_stream_put(vstream); | |
1325 | } | |
1326 | } | |
7591bab1 | 1327 | stream_put(stream); |
44a2fbf1 | 1328 | ret = 0; |
173af62f | 1329 | |
7591bab1 | 1330 | end: |
53efb85a | 1331 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1332 | if (ret < 0) { |
f73fabfd | 1333 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1334 | } else { |
f73fabfd | 1335 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1336 | } |
58eb9381 | 1337 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f | 1338 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
44a2fbf1 JG |
1339 | if (send_ret < (ssize_t) sizeof(reply)) { |
1340 | ERR("Failed to send \"close stream\" command reply (ret = %zd)", | |
1341 | send_ret); | |
1342 | ret = -1; | |
173af62f DG |
1343 | } |
1344 | ||
1345 | end_no_session: | |
1346 | return ret; | |
1347 | } | |
1348 | ||
93ec662e JD |
1349 | /* |
1350 | * relay_reset_metadata: reset a metadata stream | |
1351 | */ | |
1352 | static | |
44a2fbf1 JG |
1353 | int relay_reset_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1354 | struct relay_connection *conn, | |
1355 | const struct lttng_buffer_view *payload) | |
93ec662e | 1356 | { |
44a2fbf1 JG |
1357 | int ret; |
1358 | ssize_t send_ret; | |
93ec662e JD |
1359 | struct relay_session *session = conn->session; |
1360 | struct lttcomm_relayd_reset_metadata stream_info; | |
1361 | struct lttcomm_relayd_generic_reply reply; | |
1362 | struct relay_stream *stream; | |
1363 | ||
1364 | DBG("Reset metadata received"); | |
1365 | ||
44a2fbf1 | 1366 | if (!session || !conn->version_check_done) { |
93ec662e JD |
1367 | ERR("Trying to reset a metadata stream before version check"); |
1368 | ret = -1; | |
1369 | goto end_no_session; | |
1370 | } | |
1371 | ||
44a2fbf1 JG |
1372 | if (payload->size < sizeof(stream_info)) { |
1373 | ERR("Unexpected payload size in \"relay_reset_metadata\": expected >= %zu bytes, got %zu bytes", | |
1374 | sizeof(stream_info), payload->size); | |
93ec662e JD |
1375 | ret = -1; |
1376 | goto end_no_session; | |
1377 | } | |
44a2fbf1 JG |
1378 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1379 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1380 | stream_info.version = be64toh(stream_info.version); | |
1381 | ||
1382 | DBG("Update metadata to version %" PRIu64, stream_info.version); | |
93ec662e JD |
1383 | |
1384 | /* Unsupported for live sessions for now. */ | |
1385 | if (session->live_timer != 0) { | |
1386 | ret = -1; | |
1387 | goto end; | |
1388 | } | |
1389 | ||
44a2fbf1 | 1390 | stream = stream_get_by_id(stream_info.stream_id); |
93ec662e JD |
1391 | if (!stream) { |
1392 | ret = -1; | |
1393 | goto end; | |
1394 | } | |
1395 | pthread_mutex_lock(&stream->lock); | |
1396 | if (!stream->is_metadata) { | |
1397 | ret = -1; | |
1398 | goto end_unlock; | |
1399 | } | |
1400 | ||
1401 | ret = utils_rotate_stream_file(stream->path_name, stream->channel_name, | |
1402 | 0, 0, -1, -1, stream->stream_fd->fd, NULL, | |
1403 | &stream->stream_fd->fd); | |
1404 | if (ret < 0) { | |
1405 | ERR("Failed to rotate metadata file %s of channel %s", | |
1406 | stream->path_name, stream->channel_name); | |
1407 | goto end_unlock; | |
1408 | } | |
1409 | ||
1410 | end_unlock: | |
1411 | pthread_mutex_unlock(&stream->lock); | |
1412 | stream_put(stream); | |
1413 | ||
1414 | end: | |
1415 | memset(&reply, 0, sizeof(reply)); | |
1416 | if (ret < 0) { | |
1417 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1418 | } else { | |
1419 | reply.ret_code = htobe32(LTTNG_OK); | |
1420 | } | |
1421 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1422 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
44a2fbf1 JG |
1423 | if (send_ret < (ssize_t) sizeof(reply)) { |
1424 | ERR("Failed to send \"reset metadata\" command reply (ret = %zd)", | |
1425 | send_ret); | |
1426 | ret = -1; | |
93ec662e JD |
1427 | } |
1428 | ||
1429 | end_no_session: | |
1430 | return ret; | |
1431 | } | |
1432 | ||
b8aa1682 JD |
1433 | /* |
1434 | * relay_unknown_command: send -1 if received unknown command | |
1435 | */ | |
7591bab1 | 1436 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1437 | { |
1438 | struct lttcomm_relayd_generic_reply reply; | |
44a2fbf1 | 1439 | ssize_t send_ret; |
b8aa1682 | 1440 | |
53efb85a | 1441 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1442 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
44a2fbf1 JG |
1443 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1444 | if (send_ret < sizeof(reply)) { | |
1445 | ERR("Failed to send \"unknown command\" command reply (ret = %zd)", send_ret); | |
b8aa1682 JD |
1446 | } |
1447 | } | |
1448 | ||
1449 | /* | |
1450 | * relay_start: send an acknowledgment to the client to tell if we are | |
1451 | * ready to receive data. We are ready if a session is established. | |
1452 | */ | |
44a2fbf1 JG |
1453 | static int relay_start(const struct lttcomm_relayd_hdr *recv_hdr, |
1454 | struct relay_connection *conn, | |
1455 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1456 | { |
44a2fbf1 JG |
1457 | int ret = 0; |
1458 | ssize_t send_ret; | |
b8aa1682 | 1459 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1460 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1461 | |
1462 | if (!session) { | |
1463 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1464 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1465 | } |
1466 | ||
53efb85a | 1467 | memset(&reply, 0, sizeof(reply)); |
44a2fbf1 JG |
1468 | reply.ret_code = htobe32(LTTNG_OK); |
1469 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1470 | sizeof(reply), 0); | |
1471 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1472 | ERR("Failed to send \"relay_start\" command reply (ret = %zd)", | |
1473 | send_ret); | |
1474 | ret = -1; | |
b8aa1682 JD |
1475 | } |
1476 | ||
1477 | return ret; | |
1478 | } | |
1479 | ||
1d4dfdef DG |
1480 | /* |
1481 | * Append padding to the file pointed by the file descriptor fd. | |
1482 | */ | |
1483 | static int write_padding_to_file(int fd, uint32_t size) | |
1484 | { | |
6cd525e8 | 1485 | ssize_t ret = 0; |
1d4dfdef DG |
1486 | char *zeros; |
1487 | ||
1488 | if (size == 0) { | |
1489 | goto end; | |
1490 | } | |
1491 | ||
1492 | zeros = zmalloc(size); | |
1493 | if (zeros == NULL) { | |
1494 | PERROR("zmalloc zeros for padding"); | |
1495 | ret = -1; | |
1496 | goto end; | |
1497 | } | |
1498 | ||
6cd525e8 MD |
1499 | ret = lttng_write(fd, zeros, size); |
1500 | if (ret < size) { | |
1d4dfdef DG |
1501 | PERROR("write padding to file"); |
1502 | } | |
1503 | ||
e986c7a1 DG |
1504 | free(zeros); |
1505 | ||
1d4dfdef DG |
1506 | end: |
1507 | return ret; | |
1508 | } | |
1509 | ||
b8aa1682 | 1510 | /* |
7591bab1 | 1511 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1512 | */ |
44a2fbf1 JG |
1513 | static int relay_recv_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1514 | struct relay_connection *conn, | |
1515 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1516 | { |
32d1569c | 1517 | int ret = 0; |
6cd525e8 | 1518 | ssize_t size_ret; |
58eb9381 | 1519 | struct relay_session *session = conn->session; |
44a2fbf1 | 1520 | struct lttcomm_relayd_metadata_payload metadata_payload_header; |
b8aa1682 | 1521 | struct relay_stream *metadata_stream; |
44a2fbf1 | 1522 | uint64_t metadata_payload_size; |
b8aa1682 JD |
1523 | |
1524 | if (!session) { | |
1525 | ERR("Metadata sent before version check"); | |
1526 | ret = -1; | |
1527 | goto end; | |
1528 | } | |
1529 | ||
44a2fbf1 | 1530 | if (recv_hdr->data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { |
f6416125 MD |
1531 | ERR("Incorrect data size"); |
1532 | ret = -1; | |
1533 | goto end; | |
1534 | } | |
44a2fbf1 JG |
1535 | metadata_payload_size = recv_hdr->data_size - |
1536 | sizeof(struct lttcomm_relayd_metadata_payload); | |
f6416125 | 1537 | |
44a2fbf1 JG |
1538 | memcpy(&metadata_payload_header, payload->data, |
1539 | sizeof(metadata_payload_header)); | |
1540 | metadata_payload_header.stream_id = be64toh( | |
1541 | metadata_payload_header.stream_id); | |
1542 | metadata_payload_header.padding_size = be32toh( | |
1543 | metadata_payload_header.padding_size); | |
c617c0c6 | 1544 | |
44a2fbf1 | 1545 | metadata_stream = stream_get_by_id(metadata_payload_header.stream_id); |
b8aa1682 JD |
1546 | if (!metadata_stream) { |
1547 | ret = -1; | |
7591bab1 | 1548 | goto end; |
b8aa1682 JD |
1549 | } |
1550 | ||
7591bab1 MD |
1551 | pthread_mutex_lock(&metadata_stream->lock); |
1552 | ||
44a2fbf1 JG |
1553 | size_ret = lttng_write(metadata_stream->stream_fd->fd, |
1554 | payload->data + sizeof(metadata_payload_header), | |
1555 | metadata_payload_size); | |
1556 | if (size_ret < metadata_payload_size) { | |
b8aa1682 JD |
1557 | ERR("Relay error writing metadata on file"); |
1558 | ret = -1; | |
7591bab1 | 1559 | goto end_put; |
b8aa1682 | 1560 | } |
1d4dfdef | 1561 | |
32d1569c | 1562 | size_ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
44a2fbf1 JG |
1563 | metadata_payload_header.padding_size); |
1564 | if (size_ret < (int64_t) metadata_payload_header.padding_size) { | |
1565 | ret = -1; | |
7591bab1 | 1566 | goto end_put; |
1d4dfdef | 1567 | } |
2a174661 | 1568 | |
7591bab1 | 1569 | metadata_stream->metadata_received += |
44a2fbf1 | 1570 | metadata_payload_size + metadata_payload_header.padding_size; |
7591bab1 MD |
1571 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1572 | metadata_stream->metadata_received); | |
1d4dfdef | 1573 | |
7591bab1 MD |
1574 | end_put: |
1575 | pthread_mutex_unlock(&metadata_stream->lock); | |
1576 | stream_put(metadata_stream); | |
b8aa1682 JD |
1577 | end: |
1578 | return ret; | |
1579 | } | |
1580 | ||
1581 | /* | |
1582 | * relay_send_version: send relayd version number | |
1583 | */ | |
44a2fbf1 JG |
1584 | static int relay_send_version(const struct lttcomm_relayd_hdr *recv_hdr, |
1585 | struct relay_connection *conn, | |
1586 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1587 | { |
7f51dcba | 1588 | int ret; |
44a2fbf1 | 1589 | ssize_t send_ret; |
092b6259 | 1590 | struct lttcomm_relayd_version reply, msg; |
40096142 | 1591 | bool compatible = true; |
b8aa1682 | 1592 | |
44a2fbf1 | 1593 | conn->version_check_done = true; |
b8aa1682 | 1594 | |
092b6259 | 1595 | /* Get version from the other side. */ |
44a2fbf1 JG |
1596 | if (payload->size < sizeof(msg)) { |
1597 | ERR("Unexpected payload size in \"relay_send_version\": expected >= %zu bytes, got %zu bytes", | |
1598 | sizeof(msg), payload->size); | |
092b6259 | 1599 | ret = -1; |
092b6259 DG |
1600 | goto end; |
1601 | } | |
1602 | ||
44a2fbf1 JG |
1603 | memcpy(&msg, payload->data, sizeof(msg)); |
1604 | msg.major = be32toh(msg.major); | |
1605 | msg.minor = be32toh(msg.minor); | |
1606 | ||
53efb85a | 1607 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1608 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1609 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1610 | |
1611 | /* Major versions must be the same */ | |
44a2fbf1 | 1612 | if (reply.major != msg.major) { |
6151a90f | 1613 | DBG("Incompatible major versions (%u vs %u), deleting session", |
44a2fbf1 | 1614 | reply.major, msg.major); |
40096142 | 1615 | compatible = false; |
d4519fa3 JD |
1616 | } |
1617 | ||
58eb9381 | 1618 | conn->major = reply.major; |
0f907de1 | 1619 | /* We adapt to the lowest compatible version */ |
44a2fbf1 | 1620 | if (reply.minor <= msg.minor) { |
58eb9381 | 1621 | conn->minor = reply.minor; |
0f907de1 | 1622 | } else { |
44a2fbf1 | 1623 | conn->minor = msg.minor; |
0f907de1 JD |
1624 | } |
1625 | ||
6151a90f JD |
1626 | reply.major = htobe32(reply.major); |
1627 | reply.minor = htobe32(reply.minor); | |
44a2fbf1 JG |
1628 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
1629 | sizeof(reply), 0); | |
1630 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1631 | ERR("Failed to send \"send version\" command reply (ret = %zd)", | |
1632 | send_ret); | |
1633 | ret = -1; | |
1634 | goto end; | |
1635 | } else { | |
1636 | ret = 0; | |
6151a90f JD |
1637 | } |
1638 | ||
40096142 JD |
1639 | if (!compatible) { |
1640 | ret = -1; | |
1641 | goto end; | |
1642 | } | |
1643 | ||
58eb9381 DG |
1644 | DBG("Version check done using protocol %u.%u", conn->major, |
1645 | conn->minor); | |
b8aa1682 JD |
1646 | |
1647 | end: | |
1648 | return ret; | |
1649 | } | |
1650 | ||
c8f59ee5 | 1651 | /* |
6d805429 | 1652 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1653 | */ |
44a2fbf1 JG |
1654 | static int relay_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1655 | struct relay_connection *conn, | |
1656 | const struct lttng_buffer_view *payload) | |
c8f59ee5 | 1657 | { |
58eb9381 | 1658 | struct relay_session *session = conn->session; |
6d805429 | 1659 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1660 | struct lttcomm_relayd_generic_reply reply; |
1661 | struct relay_stream *stream; | |
44a2fbf1 | 1662 | ssize_t send_ret; |
c8f59ee5 | 1663 | int ret; |
c8f59ee5 | 1664 | |
6d805429 | 1665 | DBG("Data pending command received"); |
c8f59ee5 | 1666 | |
44a2fbf1 | 1667 | if (!session || !conn->version_check_done) { |
c8f59ee5 DG |
1668 | ERR("Trying to check for data before version check"); |
1669 | ret = -1; | |
1670 | goto end_no_session; | |
1671 | } | |
1672 | ||
44a2fbf1 JG |
1673 | if (payload->size < sizeof(msg)) { |
1674 | ERR("Unexpected payload size in \"relay_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1675 | sizeof(msg), payload->size); | |
c8f59ee5 DG |
1676 | ret = -1; |
1677 | goto end_no_session; | |
1678 | } | |
44a2fbf1 JG |
1679 | memcpy(&msg, payload->data, sizeof(msg)); |
1680 | msg.stream_id = be64toh(msg.stream_id); | |
1681 | msg.last_net_seq_num = be64toh(msg.last_net_seq_num); | |
c8f59ee5 | 1682 | |
44a2fbf1 | 1683 | stream = stream_get_by_id(msg.stream_id); |
de91f48a | 1684 | if (stream == NULL) { |
c8f59ee5 | 1685 | ret = -1; |
7591bab1 | 1686 | goto end; |
c8f59ee5 DG |
1687 | } |
1688 | ||
7591bab1 MD |
1689 | pthread_mutex_lock(&stream->lock); |
1690 | ||
6d805429 | 1691 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
44a2fbf1 JG |
1692 | " and last_seq %" PRIu64, msg.stream_id, |
1693 | stream->prev_seq, msg.last_net_seq_num); | |
c8f59ee5 | 1694 | |
33832e64 | 1695 | /* Avoid wrapping issue */ |
44a2fbf1 | 1696 | if (((int64_t) (stream->prev_seq - msg.last_net_seq_num)) >= 0) { |
6d805429 | 1697 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1698 | ret = 0; |
6d805429 DG |
1699 | } else { |
1700 | /* Data still being streamed thus pending */ | |
1701 | ret = 1; | |
c8f59ee5 DG |
1702 | } |
1703 | ||
7591bab1 MD |
1704 | stream->data_pending_check_done = true; |
1705 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1706 | |
7591bab1 MD |
1707 | stream_put(stream); |
1708 | end: | |
c8f59ee5 | 1709 | |
53efb85a | 1710 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1711 | reply.ret_code = htobe32(ret); |
44a2fbf1 JG |
1712 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1713 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1714 | ERR("Failed to send \"data pending\" command reply (ret = %zd)", | |
1715 | send_ret); | |
1716 | ret = -1; | |
c8f59ee5 DG |
1717 | } |
1718 | ||
1719 | end_no_session: | |
1720 | return ret; | |
1721 | } | |
1722 | ||
1723 | /* | |
1724 | * Wait for the control socket to reach a quiescent state. | |
1725 | * | |
7591bab1 MD |
1726 | * Note that for now, when receiving this command from the session |
1727 | * daemon, this means that every subsequent commands or data received on | |
1728 | * the control socket has been handled. So, this is why we simply return | |
1729 | * OK here. | |
c8f59ee5 | 1730 | */ |
44a2fbf1 JG |
1731 | static int relay_quiescent_control(const struct lttcomm_relayd_hdr *recv_hdr, |
1732 | struct relay_connection *conn, | |
1733 | const struct lttng_buffer_view *payload) | |
c8f59ee5 DG |
1734 | { |
1735 | int ret; | |
44a2fbf1 | 1736 | ssize_t send_ret; |
ad7051c0 | 1737 | struct relay_stream *stream; |
ad7051c0 | 1738 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1739 | struct lttcomm_relayd_generic_reply reply; |
1740 | ||
1741 | DBG("Checking quiescent state on control socket"); | |
1742 | ||
44a2fbf1 | 1743 | if (!conn->session || !conn->version_check_done) { |
ad7051c0 DG |
1744 | ERR("Trying to check for data before version check"); |
1745 | ret = -1; | |
1746 | goto end_no_session; | |
1747 | } | |
1748 | ||
44a2fbf1 JG |
1749 | if (payload->size < sizeof(msg)) { |
1750 | ERR("Unexpected payload size in \"relay_quiescent_control\": expected >= %zu bytes, got %zu bytes", | |
1751 | sizeof(msg), payload->size); | |
ad7051c0 DG |
1752 | ret = -1; |
1753 | goto end_no_session; | |
1754 | } | |
44a2fbf1 JG |
1755 | memcpy(&msg, payload->data, sizeof(msg)); |
1756 | msg.stream_id = be64toh(msg.stream_id); | |
ad7051c0 | 1757 | |
44a2fbf1 | 1758 | stream = stream_get_by_id(msg.stream_id); |
7591bab1 MD |
1759 | if (!stream) { |
1760 | goto reply; | |
1761 | } | |
1762 | pthread_mutex_lock(&stream->lock); | |
1763 | stream->data_pending_check_done = true; | |
1764 | pthread_mutex_unlock(&stream->lock); | |
44a2fbf1 JG |
1765 | |
1766 | DBG("Relay quiescent control pending flag set to %" PRIu64, msg.stream_id); | |
7591bab1 MD |
1767 | stream_put(stream); |
1768 | reply: | |
53efb85a | 1769 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1770 | reply.ret_code = htobe32(LTTNG_OK); |
44a2fbf1 JG |
1771 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1772 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1773 | ERR("Failed to send \"quiescent control\" command reply (ret = %zd)", | |
1774 | send_ret); | |
1775 | ret = -1; | |
1776 | } else { | |
1777 | ret = 0; | |
c8f59ee5 DG |
1778 | } |
1779 | ||
ad7051c0 | 1780 | end_no_session: |
c8f59ee5 DG |
1781 | return ret; |
1782 | } | |
1783 | ||
f7079f67 | 1784 | /* |
7591bab1 MD |
1785 | * Initialize a data pending command. This means that a consumer is about |
1786 | * to ask for data pending for each stream it holds. Simply iterate over | |
1787 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1788 | * |
1789 | * This command returns to the client a LTTNG_OK code. | |
1790 | */ | |
44a2fbf1 JG |
1791 | static int relay_begin_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1792 | struct relay_connection *conn, | |
1793 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
1794 | { |
1795 | int ret; | |
44a2fbf1 | 1796 | ssize_t send_ret; |
f7079f67 DG |
1797 | struct lttng_ht_iter iter; |
1798 | struct lttcomm_relayd_begin_data_pending msg; | |
1799 | struct lttcomm_relayd_generic_reply reply; | |
1800 | struct relay_stream *stream; | |
f7079f67 DG |
1801 | |
1802 | assert(recv_hdr); | |
58eb9381 | 1803 | assert(conn); |
f7079f67 DG |
1804 | |
1805 | DBG("Init streams for data pending"); | |
1806 | ||
44a2fbf1 | 1807 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
1808 | ERR("Trying to check for data before version check"); |
1809 | ret = -1; | |
1810 | goto end_no_session; | |
1811 | } | |
1812 | ||
44a2fbf1 JG |
1813 | if (payload->size < sizeof(msg)) { |
1814 | ERR("Unexpected payload size in \"relay_begin_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1815 | sizeof(msg), payload->size); | |
f7079f67 DG |
1816 | ret = -1; |
1817 | goto end_no_session; | |
1818 | } | |
44a2fbf1 JG |
1819 | memcpy(&msg, payload->data, sizeof(msg)); |
1820 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 DG |
1821 | |
1822 | /* | |
7591bab1 MD |
1823 | * Iterate over all streams to set the begin data pending flag. |
1824 | * For now, the streams are indexed by stream handle so we have | |
1825 | * to iterate over all streams to find the one associated with | |
1826 | * the right session_id. | |
f7079f67 DG |
1827 | */ |
1828 | rcu_read_lock(); | |
d3e2ba59 | 1829 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1830 | node.node) { |
7591bab1 MD |
1831 | if (!stream_get(stream)) { |
1832 | continue; | |
1833 | } | |
44a2fbf1 | 1834 | if (stream->trace->session->id == msg.session_id) { |
7591bab1 MD |
1835 | pthread_mutex_lock(&stream->lock); |
1836 | stream->data_pending_check_done = false; | |
1837 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1838 | DBG("Set begin data pending flag to stream %" PRIu64, |
1839 | stream->stream_handle); | |
1840 | } | |
7591bab1 | 1841 | stream_put(stream); |
f7079f67 DG |
1842 | } |
1843 | rcu_read_unlock(); | |
1844 | ||
53efb85a | 1845 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1846 | /* All good, send back reply. */ |
1847 | reply.ret_code = htobe32(LTTNG_OK); | |
1848 | ||
44a2fbf1 JG |
1849 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1850 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1851 | ERR("Failed to send \"begin data pending\" command reply (ret = %zd)", | |
1852 | send_ret); | |
1853 | ret = -1; | |
1854 | } else { | |
1855 | ret = 0; | |
f7079f67 DG |
1856 | } |
1857 | ||
1858 | end_no_session: | |
1859 | return ret; | |
1860 | } | |
1861 | ||
1862 | /* | |
7591bab1 MD |
1863 | * End data pending command. This will check, for a given session id, if |
1864 | * each stream associated with it has its data_pending_check_done flag | |
1865 | * set. If not, this means that the client lost track of the stream but | |
1866 | * the data is still being streamed on our side. In this case, we inform | |
1867 | * the client that data is in flight. | |
f7079f67 DG |
1868 | * |
1869 | * Return to the client if there is data in flight or not with a ret_code. | |
1870 | */ | |
44a2fbf1 JG |
1871 | static int relay_end_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1872 | struct relay_connection *conn, | |
1873 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
1874 | { |
1875 | int ret; | |
44a2fbf1 | 1876 | ssize_t send_ret; |
f7079f67 DG |
1877 | struct lttng_ht_iter iter; |
1878 | struct lttcomm_relayd_end_data_pending msg; | |
1879 | struct lttcomm_relayd_generic_reply reply; | |
1880 | struct relay_stream *stream; | |
f7079f67 DG |
1881 | uint32_t is_data_inflight = 0; |
1882 | ||
f7079f67 DG |
1883 | DBG("End data pending command"); |
1884 | ||
44a2fbf1 | 1885 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
1886 | ERR("Trying to check for data before version check"); |
1887 | ret = -1; | |
1888 | goto end_no_session; | |
1889 | } | |
1890 | ||
44a2fbf1 JG |
1891 | if (payload->size < sizeof(msg)) { |
1892 | ERR("Unexpected payload size in \"relay_end_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1893 | sizeof(msg), payload->size); | |
f7079f67 DG |
1894 | ret = -1; |
1895 | goto end_no_session; | |
1896 | } | |
44a2fbf1 JG |
1897 | memcpy(&msg, payload->data, sizeof(msg)); |
1898 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 | 1899 | |
7591bab1 MD |
1900 | /* |
1901 | * Iterate over all streams to see if the begin data pending | |
1902 | * flag is set. | |
1903 | */ | |
f7079f67 | 1904 | rcu_read_lock(); |
d3e2ba59 | 1905 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1906 | node.node) { |
7591bab1 MD |
1907 | if (!stream_get(stream)) { |
1908 | continue; | |
1909 | } | |
44a2fbf1 | 1910 | if (stream->trace->session->id != msg.session_id) { |
7591bab1 MD |
1911 | stream_put(stream); |
1912 | continue; | |
1913 | } | |
1914 | pthread_mutex_lock(&stream->lock); | |
1915 | if (!stream->data_pending_check_done) { | |
1916 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1917 | is_data_inflight = 1; | |
1918 | DBG("Data is still in flight for stream %" PRIu64, | |
1919 | stream->stream_handle); | |
1920 | pthread_mutex_unlock(&stream->lock); | |
1921 | stream_put(stream); | |
1922 | break; | |
1923 | } | |
f7079f67 | 1924 | } |
7591bab1 MD |
1925 | pthread_mutex_unlock(&stream->lock); |
1926 | stream_put(stream); | |
f7079f67 DG |
1927 | } |
1928 | rcu_read_unlock(); | |
1929 | ||
53efb85a | 1930 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1931 | /* All good, send back reply. */ |
1932 | reply.ret_code = htobe32(is_data_inflight); | |
1933 | ||
44a2fbf1 JG |
1934 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1935 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1936 | ERR("Failed to send \"end data pending\" command reply (ret = %zd)", | |
1937 | send_ret); | |
1938 | ret = -1; | |
1939 | } else { | |
1940 | ret = 0; | |
f7079f67 DG |
1941 | } |
1942 | ||
1943 | end_no_session: | |
1944 | return ret; | |
1945 | } | |
1946 | ||
1c20f0e2 JD |
1947 | /* |
1948 | * Receive an index for a specific stream. | |
1949 | * | |
1950 | * Return 0 on success else a negative value. | |
1951 | */ | |
44a2fbf1 JG |
1952 | static int relay_recv_index(const struct lttcomm_relayd_hdr *recv_hdr, |
1953 | struct relay_connection *conn, | |
1954 | const struct lttng_buffer_view *payload) | |
1c20f0e2 | 1955 | { |
44a2fbf1 JG |
1956 | int ret; |
1957 | ssize_t send_ret; | |
58eb9381 | 1958 | struct relay_session *session = conn->session; |
1c20f0e2 | 1959 | struct lttcomm_relayd_index index_info; |
7591bab1 | 1960 | struct relay_index *index; |
1c20f0e2 JD |
1961 | struct lttcomm_relayd_generic_reply reply; |
1962 | struct relay_stream *stream; | |
e0547b83 | 1963 | size_t msg_len; |
1c20f0e2 | 1964 | |
58eb9381 | 1965 | assert(conn); |
1c20f0e2 JD |
1966 | |
1967 | DBG("Relay receiving index"); | |
1968 | ||
44a2fbf1 | 1969 | if (!session || !conn->version_check_done) { |
1c20f0e2 JD |
1970 | ERR("Trying to close a stream before version check"); |
1971 | ret = -1; | |
1972 | goto end_no_session; | |
1973 | } | |
1974 | ||
e0547b83 MD |
1975 | msg_len = lttcomm_relayd_index_len( |
1976 | lttng_to_index_major(conn->major, conn->minor), | |
1977 | lttng_to_index_minor(conn->major, conn->minor)); | |
44a2fbf1 JG |
1978 | if (payload->size < msg_len) { |
1979 | ERR("Unexpected payload size in \"relay_recv_index\": expected >= %zu bytes, got %zu bytes", | |
1980 | msg_len, payload->size); | |
1c20f0e2 JD |
1981 | ret = -1; |
1982 | goto end_no_session; | |
1983 | } | |
44a2fbf1 JG |
1984 | memcpy(&index_info, payload->data, msg_len); |
1985 | index_info.relay_stream_id = be64toh(index_info.relay_stream_id); | |
1986 | index_info.net_seq_num = be64toh(index_info.net_seq_num); | |
1987 | index_info.packet_size = be64toh(index_info.packet_size); | |
1988 | index_info.content_size = be64toh(index_info.content_size); | |
1989 | index_info.timestamp_begin = be64toh(index_info.timestamp_begin); | |
1990 | index_info.timestamp_end = be64toh(index_info.timestamp_end); | |
1991 | index_info.events_discarded = be64toh(index_info.events_discarded); | |
1992 | index_info.stream_id = be64toh(index_info.stream_id); | |
1c20f0e2 | 1993 | |
44a2fbf1 JG |
1994 | if (conn->minor >= 8) { |
1995 | index_info.stream_instance_id = | |
1996 | be64toh(index_info.stream_instance_id); | |
1997 | index_info.packet_seq_num = be64toh(index_info.packet_seq_num); | |
1998 | } | |
1c20f0e2 | 1999 | |
44a2fbf1 | 2000 | stream = stream_get_by_id(index_info.relay_stream_id); |
1c20f0e2 | 2001 | if (!stream) { |
7591bab1 | 2002 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 2003 | ret = -1; |
7591bab1 | 2004 | goto end; |
1c20f0e2 | 2005 | } |
7591bab1 | 2006 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 2007 | |
d3e2ba59 JD |
2008 | /* Live beacon handling */ |
2009 | if (index_info.packet_size == 0) { | |
7591bab1 MD |
2010 | DBG("Received live beacon for stream %" PRIu64, |
2011 | stream->stream_handle); | |
d3e2ba59 JD |
2012 | |
2013 | /* | |
7591bab1 MD |
2014 | * Only flag a stream inactive when it has already |
2015 | * received data and no indexes are in flight. | |
d3e2ba59 | 2016 | */ |
a44ca2ca | 2017 | if (stream->index_received_seqcount > 0 |
7591bab1 | 2018 | && stream->indexes_in_flight == 0) { |
44a2fbf1 | 2019 | stream->beacon_ts_end = index_info.timestamp_end; |
d3e2ba59 JD |
2020 | } |
2021 | ret = 0; | |
7591bab1 | 2022 | goto end_stream_put; |
d3e2ba59 JD |
2023 | } else { |
2024 | stream->beacon_ts_end = -1ULL; | |
2025 | } | |
2026 | ||
528f2ffa | 2027 | if (stream->ctf_stream_id == -1ULL) { |
44a2fbf1 | 2028 | stream->ctf_stream_id = index_info.stream_id; |
528f2ffa | 2029 | } |
44a2fbf1 | 2030 | index = relay_index_get_by_id_or_create(stream, index_info.net_seq_num); |
7591bab1 MD |
2031 | if (!index) { |
2032 | ret = -1; | |
2033 | ERR("relay_index_get_by_id_or_create index NULL"); | |
2034 | goto end_stream_put; | |
1c20f0e2 | 2035 | } |
234cd636 | 2036 | if (set_index_control_data(index, &index_info, conn)) { |
7591bab1 MD |
2037 | ERR("set_index_control_data error"); |
2038 | relay_index_put(index); | |
2039 | ret = -1; | |
2040 | goto end_stream_put; | |
2041 | } | |
2042 | ret = relay_index_try_flush(index); | |
2043 | if (ret == 0) { | |
a44ca2ca MD |
2044 | tracefile_array_commit_seq(stream->tfa); |
2045 | stream->index_received_seqcount++; | |
7591bab1 MD |
2046 | } else if (ret > 0) { |
2047 | /* no flush. */ | |
2048 | ret = 0; | |
2049 | } else { | |
2050 | ERR("relay_index_try_flush error %d", ret); | |
2051 | relay_index_put(index); | |
2052 | ret = -1; | |
1c20f0e2 JD |
2053 | } |
2054 | ||
7591bab1 MD |
2055 | end_stream_put: |
2056 | pthread_mutex_unlock(&stream->lock); | |
2057 | stream_put(stream); | |
2058 | ||
2059 | end: | |
1c20f0e2 | 2060 | |
53efb85a | 2061 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
2062 | if (ret < 0) { |
2063 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
2064 | } else { | |
2065 | reply.ret_code = htobe32(LTTNG_OK); | |
2066 | } | |
58eb9381 | 2067 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
44a2fbf1 JG |
2068 | if (send_ret < (ssize_t) sizeof(reply)) { |
2069 | ERR("Failed to send \"recv index\" command reply (ret = %zd)", send_ret); | |
2070 | ret = -1; | |
1c20f0e2 JD |
2071 | } |
2072 | ||
2073 | end_no_session: | |
2074 | return ret; | |
2075 | } | |
2076 | ||
a4baae1b JD |
2077 | /* |
2078 | * Receive the streams_sent message. | |
2079 | * | |
2080 | * Return 0 on success else a negative value. | |
2081 | */ | |
44a2fbf1 JG |
2082 | static int relay_streams_sent(const struct lttcomm_relayd_hdr *recv_hdr, |
2083 | struct relay_connection *conn, | |
2084 | const struct lttng_buffer_view *payload) | |
a4baae1b | 2085 | { |
44a2fbf1 JG |
2086 | int ret; |
2087 | ssize_t send_ret; | |
a4baae1b JD |
2088 | struct lttcomm_relayd_generic_reply reply; |
2089 | ||
58eb9381 | 2090 | assert(conn); |
a4baae1b JD |
2091 | |
2092 | DBG("Relay receiving streams_sent"); | |
2093 | ||
44a2fbf1 | 2094 | if (!conn->session || !conn->version_check_done) { |
a4baae1b JD |
2095 | ERR("Trying to close a stream before version check"); |
2096 | ret = -1; | |
2097 | goto end_no_session; | |
2098 | } | |
2099 | ||
2100 | /* | |
7591bab1 MD |
2101 | * Publish every pending stream in the connection recv list which are |
2102 | * now ready to be used by the viewer. | |
4a9daf17 | 2103 | */ |
7591bab1 | 2104 | publish_connection_local_streams(conn); |
4a9daf17 | 2105 | |
53efb85a | 2106 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2107 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2108 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
44a2fbf1 JG |
2109 | if (send_ret < (ssize_t) sizeof(reply)) { |
2110 | ERR("Failed to send \"streams sent\" command reply (ret = %zd)", | |
2111 | send_ret); | |
2112 | ret = -1; | |
a4baae1b JD |
2113 | } else { |
2114 | /* Success. */ | |
2115 | ret = 0; | |
2116 | } | |
2117 | ||
2118 | end_no_session: | |
2119 | return ret; | |
2120 | } | |
2121 | ||
44a2fbf1 JG |
2122 | #define DBG_CMD(cmd_name, conn) \ |
2123 | DBG3("Processing \"%s\" command for socket %i", cmd_name, conn->sock->fd); | |
2124 | ||
2125 | static int relay_process_control_command(struct relay_connection *conn, | |
2126 | const struct lttcomm_relayd_hdr *header, | |
2127 | const struct lttng_buffer_view *payload) | |
b8aa1682 JD |
2128 | { |
2129 | int ret = 0; | |
2130 | ||
44a2fbf1 | 2131 | switch (header->cmd) { |
b8aa1682 | 2132 | case RELAYD_CREATE_SESSION: |
44a2fbf1 JG |
2133 | DBG_CMD("RELAYD_CREATE_SESSION", conn); |
2134 | ret = relay_create_session(header, conn, payload); | |
b8aa1682 | 2135 | break; |
b8aa1682 | 2136 | case RELAYD_ADD_STREAM: |
44a2fbf1 JG |
2137 | DBG_CMD("RELAYD_ADD_STREAM", conn); |
2138 | ret = relay_add_stream(header, conn, payload); | |
b8aa1682 JD |
2139 | break; |
2140 | case RELAYD_START_DATA: | |
44a2fbf1 JG |
2141 | DBG_CMD("RELAYD_START_DATA", conn); |
2142 | ret = relay_start(header, conn, payload); | |
b8aa1682 JD |
2143 | break; |
2144 | case RELAYD_SEND_METADATA: | |
44a2fbf1 JG |
2145 | DBG_CMD("RELAYD_SEND_METADATA", conn); |
2146 | ret = relay_recv_metadata(header, conn, payload); | |
b8aa1682 JD |
2147 | break; |
2148 | case RELAYD_VERSION: | |
44a2fbf1 JG |
2149 | DBG_CMD("RELAYD_VERSION", conn); |
2150 | ret = relay_send_version(header, conn, payload); | |
b8aa1682 | 2151 | break; |
173af62f | 2152 | case RELAYD_CLOSE_STREAM: |
44a2fbf1 JG |
2153 | DBG_CMD("RELAYD_CLOSE_STREAM", conn); |
2154 | ret = relay_close_stream(header, conn, payload); | |
173af62f | 2155 | break; |
6d805429 | 2156 | case RELAYD_DATA_PENDING: |
44a2fbf1 JG |
2157 | DBG_CMD("RELAYD_DATA_PENDING", conn); |
2158 | ret = relay_data_pending(header, conn, payload); | |
c8f59ee5 DG |
2159 | break; |
2160 | case RELAYD_QUIESCENT_CONTROL: | |
44a2fbf1 JG |
2161 | DBG_CMD("RELAYD_QUIESCENT_CONTROL", conn); |
2162 | ret = relay_quiescent_control(header, conn, payload); | |
c8f59ee5 | 2163 | break; |
f7079f67 | 2164 | case RELAYD_BEGIN_DATA_PENDING: |
44a2fbf1 JG |
2165 | DBG_CMD("RELAYD_BEGIN_DATA_PENDING", conn); |
2166 | ret = relay_begin_data_pending(header, conn, payload); | |
f7079f67 DG |
2167 | break; |
2168 | case RELAYD_END_DATA_PENDING: | |
44a2fbf1 JG |
2169 | DBG_CMD("RELAYD_END_DATA_PENDING", conn); |
2170 | ret = relay_end_data_pending(header, conn, payload); | |
f7079f67 | 2171 | break; |
1c20f0e2 | 2172 | case RELAYD_SEND_INDEX: |
44a2fbf1 JG |
2173 | DBG_CMD("RELAYD_SEND_INDEX", conn); |
2174 | ret = relay_recv_index(header, conn, payload); | |
1c20f0e2 | 2175 | break; |
a4baae1b | 2176 | case RELAYD_STREAMS_SENT: |
44a2fbf1 JG |
2177 | DBG_CMD("RELAYD_STREAMS_SENT", conn); |
2178 | ret = relay_streams_sent(header, conn, payload); | |
a4baae1b | 2179 | break; |
93ec662e | 2180 | case RELAYD_RESET_METADATA: |
44a2fbf1 JG |
2181 | DBG_CMD("RELAYD_RESET_METADATA", conn); |
2182 | ret = relay_reset_metadata(header, conn, payload); | |
93ec662e | 2183 | break; |
b8aa1682 JD |
2184 | case RELAYD_UPDATE_SYNC_INFO: |
2185 | default: | |
44a2fbf1 | 2186 | ERR("Received unknown command (%u)", header->cmd); |
58eb9381 | 2187 | relay_unknown_command(conn); |
b8aa1682 JD |
2188 | ret = -1; |
2189 | goto end; | |
2190 | } | |
2191 | ||
2192 | end: | |
2193 | return ret; | |
2194 | } | |
2195 | ||
44a2fbf1 JG |
2196 | static enum relay_connection_status relay_process_control_receive_payload( |
2197 | struct relay_connection *conn) | |
2198 | { | |
2199 | int ret = 0; | |
2200 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2201 | struct lttng_dynamic_buffer *reception_buffer = | |
2202 | &conn->protocol.ctrl.reception_buffer; | |
2203 | struct ctrl_connection_state_receive_payload *state = | |
2204 | &conn->protocol.ctrl.state.receive_payload; | |
2205 | struct lttng_buffer_view payload_view; | |
2206 | ||
2207 | if (state->left_to_receive == 0) { | |
2208 | /* Short-circuit for payload-less commands. */ | |
2209 | goto reception_complete; | |
2210 | } | |
2211 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2212 | reception_buffer->data + state->received, | |
2213 | state->left_to_receive, MSG_DONTWAIT); | |
2214 | if (ret < 0) { | |
2215 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2216 | PERROR("Unable to receive command payload on sock %d", | |
2217 | conn->sock->fd); | |
2218 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2219 | } | |
2220 | goto end; | |
2221 | } else if (ret == 0) { | |
2222 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
2223 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
2224 | goto end; | |
2225 | } | |
2226 | ||
2227 | assert(ret > 0); | |
2228 | assert(ret <= state->left_to_receive); | |
2229 | ||
2230 | state->left_to_receive -= ret; | |
2231 | state->received += ret; | |
2232 | ||
2233 | if (state->left_to_receive > 0) { | |
2234 | /* | |
2235 | * Can't transition to the protocol's next state, wait to | |
2236 | * receive the rest of the header. | |
2237 | */ | |
2238 | DBG3("Partial reception of control connection protocol payload (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2239 | state->received, state->left_to_receive, | |
2240 | conn->sock->fd); | |
2241 | goto end; | |
2242 | } | |
2243 | ||
2244 | reception_complete: | |
2245 | DBG("Done receiving control command payload: fd = %i, payload size = %" PRIu64 " bytes", | |
2246 | conn->sock->fd, state->received); | |
2247 | /* | |
2248 | * The payload required to process the command has been received. | |
2249 | * A view to the reception buffer is forwarded to the various | |
2250 | * commands and the state of the control is reset on success. | |
2251 | * | |
2252 | * Commands are responsible for sending their reply to the peer. | |
2253 | */ | |
2254 | payload_view = lttng_buffer_view_from_dynamic_buffer(reception_buffer, | |
2255 | 0, -1); | |
2256 | ret = relay_process_control_command(conn, | |
2257 | &state->header, &payload_view); | |
2258 | if (ret < 0) { | |
2259 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2260 | goto end; | |
2261 | } | |
2262 | ||
2263 | ret = connection_reset_protocol_state(conn); | |
2264 | if (ret) { | |
2265 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2266 | } | |
2267 | end: | |
2268 | return status; | |
2269 | } | |
2270 | ||
2271 | static enum relay_connection_status relay_process_control_receive_header( | |
2272 | struct relay_connection *conn) | |
2273 | { | |
2274 | int ret = 0; | |
2275 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2276 | struct lttcomm_relayd_hdr header; | |
2277 | struct lttng_dynamic_buffer *reception_buffer = | |
2278 | &conn->protocol.ctrl.reception_buffer; | |
2279 | struct ctrl_connection_state_receive_header *state = | |
2280 | &conn->protocol.ctrl.state.receive_header; | |
2281 | ||
2282 | assert(state->left_to_receive != 0); | |
2283 | ||
2284 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2285 | reception_buffer->data + state->received, | |
2286 | state->left_to_receive, MSG_DONTWAIT); | |
2287 | if (ret < 0) { | |
2288 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2289 | PERROR("Unable to receive control command header on sock %d", | |
2290 | conn->sock->fd); | |
2291 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2292 | } | |
2293 | goto end; | |
2294 | } else if (ret == 0) { | |
2295 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
2296 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
2297 | goto end; | |
2298 | } | |
2299 | ||
2300 | assert(ret > 0); | |
2301 | assert(ret <= state->left_to_receive); | |
2302 | ||
2303 | state->left_to_receive -= ret; | |
2304 | state->received += ret; | |
2305 | ||
2306 | if (state->left_to_receive > 0) { | |
2307 | /* | |
2308 | * Can't transition to the protocol's next state, wait to | |
2309 | * receive the rest of the header. | |
2310 | */ | |
2311 | DBG3("Partial reception of control connection protocol header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2312 | state->received, state->left_to_receive, | |
2313 | conn->sock->fd); | |
2314 | goto end; | |
2315 | } | |
2316 | ||
2317 | /* Transition to next state: receiving the command's payload. */ | |
2318 | conn->protocol.ctrl.state_id = | |
2319 | CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
2320 | memcpy(&header, reception_buffer->data, sizeof(header)); | |
2321 | header.circuit_id = be64toh(header.circuit_id); | |
2322 | header.data_size = be64toh(header.data_size); | |
2323 | header.cmd = be32toh(header.cmd); | |
2324 | header.cmd_version = be32toh(header.cmd_version); | |
2325 | memcpy(&conn->protocol.ctrl.state.receive_payload.header, | |
2326 | &header, sizeof(header)); | |
2327 | ||
2328 | DBG("Done receiving control command header: fd = %i, cmd = %" PRIu32 ", cmd_version = %" PRIu32 ", payload size = %" PRIu64 " bytes", | |
2329 | conn->sock->fd, header.cmd, header.cmd_version, | |
2330 | header.data_size); | |
2331 | ||
2332 | if (header.data_size > DEFAULT_NETWORK_RELAYD_CTRL_MAX_PAYLOAD_SIZE) { | |
2333 | ERR("Command header indicates a payload (%" PRIu64 " bytes) that exceeds the maximal payload size allowed on a control connection.", | |
2334 | header.data_size); | |
2335 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2336 | goto end; | |
2337 | } | |
2338 | ||
2339 | conn->protocol.ctrl.state.receive_payload.left_to_receive = | |
2340 | header.data_size; | |
2341 | conn->protocol.ctrl.state.receive_payload.received = 0; | |
2342 | ret = lttng_dynamic_buffer_set_size(reception_buffer, | |
2343 | header.data_size); | |
2344 | if (ret) { | |
2345 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2346 | goto end; | |
2347 | } | |
2348 | ||
2349 | if (header.data_size == 0) { | |
2350 | /* | |
2351 | * Manually invoke the next state as the poll loop | |
2352 | * will not wake-up to allow us to proceed further. | |
2353 | */ | |
2354 | status = relay_process_control_receive_payload(conn); | |
2355 | } | |
2356 | end: | |
2357 | return status; | |
2358 | } | |
2359 | ||
2360 | /* | |
2361 | * Process the commands received on the control socket | |
2362 | */ | |
2363 | static enum relay_connection_status relay_process_control( | |
2364 | struct relay_connection *conn) | |
2365 | { | |
2366 | enum relay_connection_status status; | |
2367 | ||
2368 | switch (conn->protocol.ctrl.state_id) { | |
2369 | case CTRL_CONNECTION_STATE_RECEIVE_HEADER: | |
2370 | status = relay_process_control_receive_header(conn); | |
2371 | break; | |
2372 | case CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
2373 | status = relay_process_control_receive_payload(conn); | |
2374 | break; | |
2375 | default: | |
2376 | ERR("Unknown control connection protocol state encountered."); | |
2377 | abort(); | |
2378 | } | |
2379 | ||
2380 | return status; | |
2381 | } | |
2382 | ||
7d2f7452 DG |
2383 | /* |
2384 | * Handle index for a data stream. | |
2385 | * | |
7591bab1 | 2386 | * Called with the stream lock held. |
7d2f7452 DG |
2387 | * |
2388 | * Return 0 on success else a negative value. | |
2389 | */ | |
2390 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
44a2fbf1 | 2391 | bool rotate_index) |
7d2f7452 | 2392 | { |
7591bab1 MD |
2393 | int ret = 0; |
2394 | uint64_t data_offset; | |
2395 | struct relay_index *index; | |
7d2f7452 | 2396 | |
7d2f7452 DG |
2397 | /* Get data offset because we are about to update the index. */ |
2398 | data_offset = htobe64(stream->tracefile_size_current); | |
2399 | ||
65d66b19 MD |
2400 | DBG("handle_index_data: stream %" PRIu64 " net_seq_num %" PRIu64 " data offset %" PRIu64, |
2401 | stream->stream_handle, net_seq_num, stream->tracefile_size_current); | |
7591bab1 | 2402 | |
7d2f7452 | 2403 | /* |
7591bab1 MD |
2404 | * Lookup for an existing index for that stream id/sequence |
2405 | * number. If it exists, the control thread has already received the | |
2406 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2407 | */ |
7591bab1 | 2408 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2409 | if (!index) { |
7591bab1 MD |
2410 | ret = -1; |
2411 | goto end; | |
7d2f7452 DG |
2412 | } |
2413 | ||
e0547b83 MD |
2414 | if (rotate_index || !stream->index_file) { |
2415 | uint32_t major, minor; | |
7591bab1 | 2416 | |
e0547b83 MD |
2417 | /* Put ref on previous index_file. */ |
2418 | if (stream->index_file) { | |
2419 | lttng_index_file_put(stream->index_file); | |
2420 | stream->index_file = NULL; | |
7d2f7452 | 2421 | } |
e0547b83 MD |
2422 | major = stream->trace->session->major; |
2423 | minor = stream->trace->session->minor; | |
2424 | stream->index_file = lttng_index_file_create(stream->path_name, | |
2425 | stream->channel_name, | |
7591bab1 | 2426 | -1, -1, stream->tracefile_size, |
e0547b83 MD |
2427 | tracefile_array_get_file_index_head(stream->tfa), |
2428 | lttng_to_index_major(major, minor), | |
2429 | lttng_to_index_minor(major, minor)); | |
2430 | if (!stream->index_file) { | |
7591bab1 MD |
2431 | ret = -1; |
2432 | /* Put self-ref for this index due to error. */ | |
2433 | relay_index_put(index); | |
e0547b83 | 2434 | index = NULL; |
7591bab1 | 2435 | goto end; |
7d2f7452 | 2436 | } |
7d2f7452 DG |
2437 | } |
2438 | ||
e0547b83 | 2439 | if (relay_index_set_file(index, stream->index_file, data_offset)) { |
7591bab1 MD |
2440 | ret = -1; |
2441 | /* Put self-ref for this index due to error. */ | |
2442 | relay_index_put(index); | |
e0547b83 | 2443 | index = NULL; |
7591bab1 | 2444 | goto end; |
7d2f7452 DG |
2445 | } |
2446 | ||
7591bab1 MD |
2447 | ret = relay_index_try_flush(index); |
2448 | if (ret == 0) { | |
a44ca2ca MD |
2449 | tracefile_array_commit_seq(stream->tfa); |
2450 | stream->index_received_seqcount++; | |
7591bab1 MD |
2451 | } else if (ret > 0) { |
2452 | /* No flush. */ | |
2453 | ret = 0; | |
2454 | } else { | |
2455 | /* Put self-ref for this index due to error. */ | |
2456 | relay_index_put(index); | |
e0547b83 | 2457 | index = NULL; |
7591bab1 MD |
2458 | ret = -1; |
2459 | } | |
2460 | end: | |
7d2f7452 DG |
2461 | return ret; |
2462 | } | |
2463 | ||
44a2fbf1 JG |
2464 | static enum relay_connection_status relay_process_data_receive_header( |
2465 | struct relay_connection *conn) | |
b8aa1682 | 2466 | { |
44a2fbf1 JG |
2467 | int ret; |
2468 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2469 | struct data_connection_state_receive_header *state = | |
2470 | &conn->protocol.data.state.receive_header; | |
2471 | struct lttcomm_relayd_data_hdr header; | |
b8aa1682 | 2472 | struct relay_stream *stream; |
b8aa1682 | 2473 | |
44a2fbf1 JG |
2474 | assert(state->left_to_receive != 0); |
2475 | ||
2476 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2477 | state->header_reception_buffer + state->received, | |
2478 | state->left_to_receive, MSG_DONTWAIT); | |
2479 | if (ret < 0) { | |
2480 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2481 | PERROR("Unable to receive data header on sock %d", conn->sock->fd); | |
2482 | status = RELAY_CONNECTION_STATUS_ERROR; | |
a6cd2b97 | 2483 | } |
44a2fbf1 JG |
2484 | goto end; |
2485 | } else if (ret == 0) { | |
2486 | /* Orderly shutdown. Not necessary to print an error. */ | |
2487 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
2488 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
b8aa1682 JD |
2489 | goto end; |
2490 | } | |
2491 | ||
44a2fbf1 JG |
2492 | assert(ret > 0); |
2493 | assert(ret <= state->left_to_receive); | |
2494 | ||
2495 | state->left_to_receive -= ret; | |
2496 | state->received += ret; | |
2497 | ||
2498 | if (state->left_to_receive > 0) { | |
2499 | /* | |
2500 | * Can't transition to the protocol's next state, wait to | |
2501 | * receive the rest of the header. | |
2502 | */ | |
2503 | DBG3("Partial reception of data connection header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2504 | state->received, state->left_to_receive, | |
2505 | conn->sock->fd); | |
2506 | ret = 0; | |
7591bab1 | 2507 | goto end; |
b8aa1682 | 2508 | } |
b8aa1682 | 2509 | |
44a2fbf1 JG |
2510 | /* Transition to next state: receiving the payload. */ |
2511 | conn->protocol.data.state_id = DATA_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
173af62f | 2512 | |
44a2fbf1 JG |
2513 | memcpy(&header, state->header_reception_buffer, sizeof(header)); |
2514 | header.circuit_id = be64toh(header.circuit_id); | |
2515 | header.stream_id = be64toh(header.stream_id); | |
2516 | header.data_size = be32toh(header.data_size); | |
2517 | header.net_seq_num = be64toh(header.net_seq_num); | |
2518 | header.padding_size = be32toh(header.padding_size); | |
2519 | memcpy(&conn->protocol.data.state.receive_payload.header, &header, sizeof(header)); | |
2520 | ||
2521 | conn->protocol.data.state.receive_payload.left_to_receive = | |
2522 | header.data_size; | |
2523 | conn->protocol.data.state.receive_payload.received = 0; | |
2524 | conn->protocol.data.state.receive_payload.rotate_index = false; | |
2525 | ||
2526 | DBG("Received data connection header on fd %i: circuit_id = %" PRIu64 ", stream_id = %" PRIu64 ", data_size = %" PRIu32 ", net_seq_num = %" PRIu64 ", padding_size = %" PRIu32, | |
2527 | conn->sock->fd, header.circuit_id, | |
2528 | header.stream_id, header.data_size, | |
2529 | header.net_seq_num, header.padding_size); | |
2530 | ||
2531 | stream = stream_get_by_id(header.stream_id); | |
2532 | if (!stream) { | |
2533 | DBG("relay_process_data_receive_payload: Cannot find stream %" PRIu64, | |
2534 | header.stream_id); | |
2535 | /* Protocol error. */ | |
2536 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2537 | goto end; | |
2538 | } | |
b8aa1682 | 2539 | |
7591bab1 MD |
2540 | pthread_mutex_lock(&stream->lock); |
2541 | ||
1c20f0e2 | 2542 | /* Check if a rotation is needed. */ |
0f907de1 | 2543 | if (stream->tracefile_size > 0 && |
44a2fbf1 | 2544 | (stream->tracefile_size_current + header.data_size) > |
0f907de1 | 2545 | stream->tracefile_size) { |
a44ca2ca MD |
2546 | uint64_t old_id, new_id; |
2547 | ||
2548 | old_id = tracefile_array_get_file_index_head(stream->tfa); | |
2549 | tracefile_array_file_rotate(stream->tfa); | |
2550 | ||
2551 | /* new_id is updated by utils_rotate_stream_file. */ | |
2552 | new_id = old_id; | |
6b6b9a5a | 2553 | |
7591bab1 MD |
2554 | ret = utils_rotate_stream_file(stream->path_name, |
2555 | stream->channel_name, stream->tracefile_size, | |
2556 | stream->tracefile_count, -1, | |
2557 | -1, stream->stream_fd->fd, | |
a44ca2ca | 2558 | &new_id, &stream->stream_fd->fd); |
0f907de1 | 2559 | if (ret < 0) { |
44a2fbf1 JG |
2560 | ERR("Failed to rotate stream output file"); |
2561 | status = RELAY_CONNECTION_STATUS_ERROR; | |
7591bab1 MD |
2562 | goto end_stream_unlock; |
2563 | } | |
44a2fbf1 | 2564 | |
7591bab1 MD |
2565 | /* |
2566 | * Reset current size because we just performed a stream | |
2567 | * rotation. | |
2568 | */ | |
a6976990 | 2569 | stream->tracefile_size_current = 0; |
44a2fbf1 | 2570 | conn->protocol.data.state.receive_payload.rotate_index = true; |
1c20f0e2 JD |
2571 | } |
2572 | ||
44a2fbf1 JG |
2573 | ret = 0; |
2574 | end_stream_unlock: | |
2575 | pthread_mutex_unlock(&stream->lock); | |
2576 | stream_put(stream); | |
2577 | end: | |
2578 | return status; | |
2579 | } | |
2580 | ||
2581 | static enum relay_connection_status relay_process_data_receive_payload( | |
2582 | struct relay_connection *conn) | |
2583 | { | |
2584 | int ret; | |
2585 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2586 | struct relay_stream *stream; | |
2587 | struct data_connection_state_receive_payload *state = | |
2588 | &conn->protocol.data.state.receive_payload; | |
2589 | const size_t chunk_size = RECV_DATA_BUFFER_SIZE; | |
2590 | char data_buffer[chunk_size]; | |
2591 | bool partial_recv = false; | |
2592 | bool new_stream = false, close_requested = false; | |
2593 | uint64_t left_to_receive = state->left_to_receive; | |
2594 | struct relay_session *session; | |
2595 | ||
2596 | DBG3("Receiving data for stream id %" PRIu64 " seqnum %" PRIu64 ", %" PRIu64" bytes received, %" PRIu64 " bytes left to receive", | |
2597 | state->header.stream_id, state->header.net_seq_num, | |
2598 | state->received, left_to_receive); | |
2599 | ||
2600 | stream = stream_get_by_id(state->header.stream_id); | |
2601 | if (!stream) { | |
2602 | /* Protocol error. */ | |
2603 | ERR("relay_process_data_receive_payload: cannot find stream %" PRIu64, | |
2604 | state->header.stream_id); | |
2605 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2606 | goto end; | |
2607 | } | |
2608 | ||
2609 | pthread_mutex_lock(&stream->lock); | |
2610 | session = stream->trace->session; | |
2611 | if (!conn->session) { | |
2612 | ret = connection_set_session(conn, session); | |
2613 | if (ret) { | |
2614 | status = RELAY_CONNECTION_STATUS_ERROR; | |
7591bab1 | 2615 | goto end_stream_unlock; |
1c20f0e2 | 2616 | } |
1c20f0e2 JD |
2617 | } |
2618 | ||
44a2fbf1 JG |
2619 | /* |
2620 | * The size of the "chunk" received on any iteration is bounded by: | |
2621 | * - the data left to receive, | |
2622 | * - the data immediately available on the socket, | |
2623 | * - the on-stack data buffer | |
2624 | */ | |
2625 | while (left_to_receive > 0 && !partial_recv) { | |
2626 | ssize_t write_ret; | |
2627 | size_t recv_size = min(left_to_receive, chunk_size); | |
1d4dfdef | 2628 | |
44a2fbf1 JG |
2629 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, |
2630 | recv_size, MSG_DONTWAIT); | |
2631 | if (ret < 0) { | |
2632 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2633 | PERROR("Socket %d error", conn->sock->fd); | |
2634 | status = RELAY_CONNECTION_STATUS_ERROR; | |
0848dba7 | 2635 | } |
0848dba7 | 2636 | goto end_stream_unlock; |
44a2fbf1 JG |
2637 | } else if (ret == 0) { |
2638 | /* No more data ready to be consumed on socket. */ | |
2639 | DBG3("No more data ready for consumption on data socket of stream id %" PRIu64, | |
2640 | state->header.stream_id); | |
2641 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
2642 | break; | |
2643 | } else if (ret < (int) recv_size) { | |
2644 | /* | |
2645 | * All the data available on the socket has been | |
2646 | * consumed. | |
2647 | */ | |
2648 | partial_recv = true; | |
0848dba7 MD |
2649 | } |
2650 | ||
44a2fbf1 JG |
2651 | recv_size = ret; |
2652 | ||
0848dba7 | 2653 | /* Write data to stream output fd. */ |
44a2fbf1 | 2654 | write_ret = lttng_write(stream->stream_fd->fd, data_buffer, |
0848dba7 | 2655 | recv_size); |
44a2fbf1 | 2656 | if (write_ret < (ssize_t) recv_size) { |
0848dba7 | 2657 | ERR("Relay error writing data to file"); |
44a2fbf1 | 2658 | status = RELAY_CONNECTION_STATUS_ERROR; |
0848dba7 MD |
2659 | goto end_stream_unlock; |
2660 | } | |
2661 | ||
44a2fbf1 JG |
2662 | left_to_receive -= recv_size; |
2663 | state->received += recv_size; | |
2664 | state->left_to_receive = left_to_receive; | |
2665 | ||
0848dba7 | 2666 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, |
44a2fbf1 JG |
2667 | write_ret, stream->stream_handle); |
2668 | } | |
2669 | ||
2670 | if (state->left_to_receive > 0) { | |
2671 | /* | |
2672 | * Did not receive all the data expected, wait for more data to | |
2673 | * become available on the socket. | |
2674 | */ | |
2675 | DBG3("Partial receive on data connection of stream id %" PRIu64 ", %" PRIu64 " bytes received, %" PRIu64 " bytes left to receive", | |
2676 | state->header.stream_id, state->received, | |
2677 | state->left_to_receive); | |
2678 | goto end_stream_unlock; | |
0848dba7 | 2679 | } |
5ab7344e | 2680 | |
7591bab1 | 2681 | ret = write_padding_to_file(stream->stream_fd->fd, |
44a2fbf1 JG |
2682 | state->header.padding_size); |
2683 | if ((int64_t) ret < (int64_t) state->header.padding_size) { | |
65d66b19 | 2684 | ERR("write_padding_to_file: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
44a2fbf1 JG |
2685 | stream->stream_handle, |
2686 | state->header.net_seq_num, ret); | |
2687 | status = RELAY_CONNECTION_STATUS_ERROR; | |
7591bab1 | 2688 | goto end_stream_unlock; |
1d4dfdef | 2689 | } |
44a2fbf1 JG |
2690 | |
2691 | ||
2692 | if (session->minor >= 4 && !session->snapshot) { | |
2693 | ret = handle_index_data(stream, state->header.net_seq_num, | |
2694 | state->rotate_index); | |
2695 | if (ret < 0) { | |
2696 | ERR("handle_index_data: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", | |
2697 | stream->stream_handle, | |
2698 | state->header.net_seq_num, ret); | |
2699 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2700 | goto end_stream_unlock; | |
2701 | } | |
2702 | } | |
2703 | ||
2704 | stream->tracefile_size_current += state->header.data_size + | |
2705 | state->header.padding_size; | |
2706 | ||
c0bae11d MD |
2707 | if (stream->prev_seq == -1ULL) { |
2708 | new_stream = true; | |
2709 | } | |
2710 | ||
44a2fbf1 JG |
2711 | stream->prev_seq = state->header.net_seq_num; |
2712 | ||
2713 | /* | |
2714 | * Resetting the protocol state (to RECEIVE_HEADER) will trash the | |
2715 | * contents of *state which are aliased (union) to the same location as | |
2716 | * the new state. Don't use it beyond this point. | |
2717 | */ | |
2718 | connection_reset_protocol_state(conn); | |
2719 | state = NULL; | |
173af62f | 2720 | |
7591bab1 | 2721 | end_stream_unlock: |
bda7c7b9 | 2722 | close_requested = stream->close_requested; |
7591bab1 | 2723 | pthread_mutex_unlock(&stream->lock); |
44a2fbf1 | 2724 | if (close_requested && left_to_receive == 0) { |
bda7c7b9 JG |
2725 | try_stream_close(stream); |
2726 | } | |
2727 | ||
c0bae11d MD |
2728 | if (new_stream) { |
2729 | pthread_mutex_lock(&session->lock); | |
2730 | uatomic_set(&session->new_streams, 1); | |
2731 | pthread_mutex_unlock(&session->lock); | |
2732 | } | |
44a2fbf1 | 2733 | |
7591bab1 | 2734 | stream_put(stream); |
b8aa1682 | 2735 | end: |
44a2fbf1 JG |
2736 | return status; |
2737 | } | |
2738 | ||
2739 | /* | |
2740 | * relay_process_data: Process the data received on the data socket | |
2741 | */ | |
2742 | static enum relay_connection_status relay_process_data( | |
2743 | struct relay_connection *conn) | |
2744 | { | |
2745 | enum relay_connection_status status; | |
2746 | ||
2747 | switch (conn->protocol.data.state_id) { | |
2748 | case DATA_CONNECTION_STATE_RECEIVE_HEADER: | |
2749 | status = relay_process_data_receive_header(conn); | |
2750 | break; | |
2751 | case DATA_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
2752 | status = relay_process_data_receive_payload(conn); | |
2753 | break; | |
2754 | default: | |
2755 | ERR("Unexpected data connection communication state."); | |
2756 | abort(); | |
2757 | } | |
2758 | ||
2759 | return status; | |
b8aa1682 JD |
2760 | } |
2761 | ||
7591bab1 | 2762 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2763 | { |
2764 | int ret; | |
2765 | ||
58eb9381 | 2766 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2767 | |
2768 | ret = close(pollfd); | |
2769 | if (ret < 0) { | |
2770 | ERR("Closing pollfd %d", pollfd); | |
2771 | } | |
2772 | } | |
2773 | ||
7591bab1 MD |
2774 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2775 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2776 | { |
7591bab1 | 2777 | const char *type_str; |
2a174661 | 2778 | |
7591bab1 MD |
2779 | switch (conn->type) { |
2780 | case RELAY_DATA: | |
2781 | type_str = "Data"; | |
2782 | break; | |
2783 | case RELAY_CONTROL: | |
2784 | type_str = "Control"; | |
2785 | break; | |
2786 | case RELAY_VIEWER_COMMAND: | |
2787 | type_str = "Viewer Command"; | |
2788 | break; | |
2789 | case RELAY_VIEWER_NOTIFICATION: | |
2790 | type_str = "Viewer Notification"; | |
2791 | break; | |
2792 | default: | |
2793 | type_str = "Unknown"; | |
9d1bbf21 | 2794 | } |
7591bab1 MD |
2795 | cleanup_connection_pollfd(events, pollfd); |
2796 | connection_put(conn); | |
2797 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2798 | } |
2799 | ||
2800 | /* | |
2801 | * This thread does the actual work | |
2802 | */ | |
7591bab1 | 2803 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2804 | { |
beaad64c DG |
2805 | int ret, err = -1, last_seen_data_fd = -1; |
2806 | uint32_t nb_fd; | |
b8aa1682 JD |
2807 | struct lttng_poll_event events; |
2808 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2809 | struct lttng_ht_iter iter; |
90e7d72f | 2810 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2811 | |
2812 | DBG("[thread] Relay worker started"); | |
2813 | ||
9d1bbf21 MD |
2814 | rcu_register_thread(); |
2815 | ||
55706a7d MD |
2816 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2817 | ||
9b5e0863 MD |
2818 | if (testpoint(relayd_thread_worker)) { |
2819 | goto error_testpoint; | |
2820 | } | |
2821 | ||
f385ae0a MD |
2822 | health_code_update(); |
2823 | ||
b8aa1682 JD |
2824 | /* table of connections indexed on socket */ |
2825 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2826 | if (!relay_connections_ht) { |
2827 | goto relay_connections_ht_error; | |
2828 | } | |
b8aa1682 | 2829 | |
b8aa1682 JD |
2830 | ret = create_thread_poll_set(&events, 2); |
2831 | if (ret < 0) { | |
2832 | goto error_poll_create; | |
2833 | } | |
2834 | ||
58eb9381 | 2835 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2836 | if (ret < 0) { |
2837 | goto error; | |
2838 | } | |
2839 | ||
beaad64c | 2840 | restart: |
b8aa1682 | 2841 | while (1) { |
beaad64c DG |
2842 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2843 | ||
f385ae0a MD |
2844 | health_code_update(); |
2845 | ||
b8aa1682 | 2846 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2847 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2848 | health_poll_entry(); |
b8aa1682 | 2849 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2850 | health_poll_exit(); |
b8aa1682 JD |
2851 | if (ret < 0) { |
2852 | /* | |
2853 | * Restart interrupted system call. | |
2854 | */ | |
2855 | if (errno == EINTR) { | |
2856 | goto restart; | |
2857 | } | |
2858 | goto error; | |
2859 | } | |
2860 | ||
0d9c5d77 DG |
2861 | nb_fd = ret; |
2862 | ||
beaad64c | 2863 | /* |
7591bab1 MD |
2864 | * Process control. The control connection is |
2865 | * prioritized so we don't starve it with high | |
2866 | * throughput tracing data on the data connection. | |
beaad64c | 2867 | */ |
b8aa1682 JD |
2868 | for (i = 0; i < nb_fd; i++) { |
2869 | /* Fetch once the poll data */ | |
beaad64c DG |
2870 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2871 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2872 | |
f385ae0a MD |
2873 | health_code_update(); |
2874 | ||
fd20dac9 | 2875 | if (!revents) { |
7591bab1 MD |
2876 | /* |
2877 | * No activity for this FD (poll | |
2878 | * implementation). | |
2879 | */ | |
fd20dac9 MD |
2880 | continue; |
2881 | } | |
2882 | ||
b8aa1682 JD |
2883 | /* Thread quit pipe has been closed. Killing thread. */ |
2884 | ret = check_thread_quit_pipe(pollfd, revents); | |
2885 | if (ret) { | |
095a4ae5 MD |
2886 | err = 0; |
2887 | goto exit; | |
b8aa1682 JD |
2888 | } |
2889 | ||
58eb9381 DG |
2890 | /* Inspect the relay conn pipe for new connection */ |
2891 | if (pollfd == relay_conn_pipe[0]) { | |
03e43155 | 2892 | if (revents & LPOLLIN) { |
90e7d72f JG |
2893 | struct relay_connection *conn; |
2894 | ||
58eb9381 | 2895 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2896 | if (ret < 0) { |
2897 | goto error; | |
2898 | } | |
58eb9381 DG |
2899 | lttng_poll_add(&events, conn->sock->fd, |
2900 | LPOLLIN | LPOLLRDHUP); | |
7591bab1 | 2901 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2902 | DBG("Connection socket %d added", conn->sock->fd); |
03e43155 MD |
2903 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2904 | ERR("Relay connection pipe error"); | |
2905 | goto error; | |
2906 | } else { | |
2907 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
2908 | goto error; | |
b8aa1682 | 2909 | } |
58eb9381 | 2910 | } else { |
90e7d72f JG |
2911 | struct relay_connection *ctrl_conn; |
2912 | ||
7591bab1 | 2913 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 2914 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 2915 | assert(ctrl_conn); |
58eb9381 | 2916 | |
03e43155 MD |
2917 | if (ctrl_conn->type == RELAY_DATA) { |
2918 | if (revents & LPOLLIN) { | |
beaad64c DG |
2919 | /* |
2920 | * Flag the last seen data fd not deleted. It will be | |
2921 | * used as the last seen fd if any fd gets deleted in | |
2922 | * this first loop. | |
2923 | */ | |
2924 | last_notdel_data_fd = pollfd; | |
2925 | } | |
03e43155 MD |
2926 | goto put_ctrl_connection; |
2927 | } | |
2928 | assert(ctrl_conn->type == RELAY_CONTROL); | |
2929 | ||
2930 | if (revents & LPOLLIN) { | |
44a2fbf1 JG |
2931 | enum relay_connection_status status; |
2932 | ||
2933 | status = relay_process_control(ctrl_conn); | |
2934 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
2935 | /* | |
2936 | * On socket error flag the session as aborted to force | |
2937 | * the cleanup of its stream otherwise it can leak | |
2938 | * during the lifetime of the relayd. | |
2939 | * | |
2940 | * This prevents situations in which streams can be | |
2941 | * left opened because an index was received, the | |
2942 | * control connection is closed, and the data | |
2943 | * connection is closed (uncleanly) before the packet's | |
2944 | * data provided. | |
2945 | * | |
2946 | * Since the control connection encountered an error, | |
2947 | * it is okay to be conservative and close the | |
2948 | * session right now as we can't rely on the protocol | |
2949 | * being respected anymore. | |
2950 | */ | |
2951 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
2952 | session_abort(ctrl_conn->session); | |
03e43155 | 2953 | } |
44a2fbf1 JG |
2954 | |
2955 | /* Clear the connection on error or close. */ | |
2956 | relay_thread_close_connection(&events, | |
2957 | pollfd, | |
2958 | ctrl_conn); | |
03e43155 | 2959 | } |
44a2fbf1 | 2960 | seen_control = 1; |
03e43155 MD |
2961 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2962 | relay_thread_close_connection(&events, | |
2963 | pollfd, ctrl_conn); | |
2964 | if (last_seen_data_fd == pollfd) { | |
2965 | last_seen_data_fd = last_notdel_data_fd; | |
2966 | } | |
58eb9381 | 2967 | } else { |
03e43155 MD |
2968 | ERR("Unexpected poll events %u for control sock %d", |
2969 | revents, pollfd); | |
2970 | connection_put(ctrl_conn); | |
2971 | goto error; | |
beaad64c | 2972 | } |
03e43155 | 2973 | put_ctrl_connection: |
7591bab1 | 2974 | connection_put(ctrl_conn); |
beaad64c DG |
2975 | } |
2976 | } | |
2977 | ||
2978 | /* | |
2979 | * The last loop handled a control request, go back to poll to make | |
2980 | * sure we prioritise the control socket. | |
2981 | */ | |
2982 | if (seen_control) { | |
2983 | continue; | |
2984 | } | |
2985 | ||
2986 | if (last_seen_data_fd >= 0) { | |
2987 | for (i = 0; i < nb_fd; i++) { | |
2988 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
2989 | |
2990 | health_code_update(); | |
2991 | ||
beaad64c DG |
2992 | if (last_seen_data_fd == pollfd) { |
2993 | idx = i; | |
2994 | break; | |
2995 | } | |
2996 | } | |
2997 | } | |
2998 | ||
2999 | /* Process data connection. */ | |
3000 | for (i = idx + 1; i < nb_fd; i++) { | |
3001 | /* Fetch the poll data. */ | |
3002 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
3003 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 3004 | struct relay_connection *data_conn; |
beaad64c | 3005 | |
f385ae0a MD |
3006 | health_code_update(); |
3007 | ||
fd20dac9 MD |
3008 | if (!revents) { |
3009 | /* No activity for this FD (poll implementation). */ | |
3010 | continue; | |
3011 | } | |
3012 | ||
beaad64c | 3013 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 3014 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
3015 | continue; |
3016 | } | |
3017 | ||
7591bab1 | 3018 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 3019 | if (!data_conn) { |
fd20dac9 | 3020 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
3021 | continue; |
3022 | } | |
03e43155 MD |
3023 | if (data_conn->type == RELAY_CONTROL) { |
3024 | goto put_data_connection; | |
3025 | } | |
3026 | assert(data_conn->type == RELAY_DATA); | |
fd20dac9 MD |
3027 | |
3028 | if (revents & LPOLLIN) { | |
44a2fbf1 JG |
3029 | enum relay_connection_status status; |
3030 | ||
3031 | status = relay_process_data(data_conn); | |
3032 | /* Connection closed or error. */ | |
3033 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
3034 | /* | |
3035 | * On socket error flag the session as aborted to force | |
3036 | * the cleanup of its stream otherwise it can leak | |
3037 | * during the lifetime of the relayd. | |
3038 | * | |
3039 | * This prevents situations in which streams can be | |
3040 | * left opened because an index was received, the | |
3041 | * control connection is closed, and the data | |
3042 | * connection is closed (uncleanly) before the packet's | |
3043 | * data provided. | |
3044 | * | |
3045 | * Since the data connection encountered an error, | |
3046 | * it is okay to be conservative and close the | |
3047 | * session right now as we can't rely on the protocol | |
3048 | * being respected anymore. | |
3049 | */ | |
3050 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3051 | session_abort(data_conn->session); | |
3052 | } | |
7591bab1 | 3053 | relay_thread_close_connection(&events, pollfd, |
03e43155 | 3054 | data_conn); |
fd20dac9 MD |
3055 | /* |
3056 | * Every goto restart call sets the last seen fd where | |
3057 | * here we don't really care since we gracefully | |
3058 | * continue the loop after the connection is deleted. | |
3059 | */ | |
3060 | } else { | |
3061 | /* Keep last seen port. */ | |
3062 | last_seen_data_fd = pollfd; | |
7591bab1 | 3063 | connection_put(data_conn); |
fd20dac9 | 3064 | goto restart; |
b8aa1682 | 3065 | } |
03e43155 MD |
3066 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3067 | relay_thread_close_connection(&events, pollfd, | |
3068 | data_conn); | |
3069 | } else { | |
3070 | ERR("Unknown poll events %u for data sock %d", | |
3071 | revents, pollfd); | |
b8aa1682 | 3072 | } |
03e43155 | 3073 | put_data_connection: |
7591bab1 | 3074 | connection_put(data_conn); |
b8aa1682 | 3075 | } |
beaad64c | 3076 | last_seen_data_fd = -1; |
b8aa1682 JD |
3077 | } |
3078 | ||
f385ae0a MD |
3079 | /* Normal exit, no error */ |
3080 | ret = 0; | |
3081 | ||
095a4ae5 | 3082 | exit: |
b8aa1682 | 3083 | error: |
58eb9381 | 3084 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 3085 | rcu_read_lock(); |
90e7d72f JG |
3086 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
3087 | destroy_conn, | |
58eb9381 | 3088 | sock_n.node) { |
f385ae0a | 3089 | health_code_update(); |
06a68401 | 3090 | |
44a2fbf1 | 3091 | session_abort(destroy_conn->session); |
06a68401 | 3092 | |
7591bab1 MD |
3093 | /* |
3094 | * No need to grab another ref, because we own | |
3095 | * destroy_conn. | |
3096 | */ | |
3097 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
3098 | destroy_conn); | |
b8aa1682 | 3099 | } |
94d49140 | 3100 | rcu_read_unlock(); |
7591bab1 MD |
3101 | |
3102 | lttng_poll_clean(&events); | |
7d2f7452 | 3103 | error_poll_create: |
b8aa1682 | 3104 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 3105 | relay_connections_ht_error: |
58eb9381 DG |
3106 | /* Close relay conn pipes */ |
3107 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
3108 | if (err) { |
3109 | DBG("Thread exited with error"); | |
3110 | } | |
b8aa1682 | 3111 | DBG("Worker thread cleanup complete"); |
9b5e0863 | 3112 | error_testpoint: |
f385ae0a MD |
3113 | if (err) { |
3114 | health_error(); | |
3115 | ERR("Health error occurred in %s", __func__); | |
3116 | } | |
3117 | health_unregister(health_relayd); | |
9d1bbf21 | 3118 | rcu_unregister_thread(); |
b4aacfdc | 3119 | lttng_relay_stop_threads(); |
b8aa1682 JD |
3120 | return NULL; |
3121 | } | |
3122 | ||
3123 | /* | |
3124 | * Create the relay command pipe to wake thread_manage_apps. | |
3125 | * Closed in cleanup(). | |
3126 | */ | |
58eb9381 | 3127 | static int create_relay_conn_pipe(void) |
b8aa1682 | 3128 | { |
a02de639 | 3129 | int ret; |
b8aa1682 | 3130 | |
58eb9381 | 3131 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 3132 | |
b8aa1682 JD |
3133 | return ret; |
3134 | } | |
3135 | ||
3136 | /* | |
3137 | * main | |
3138 | */ | |
3139 | int main(int argc, char **argv) | |
3140 | { | |
178a0557 | 3141 | int ret = 0, retval = 0; |
b8aa1682 JD |
3142 | void *status; |
3143 | ||
b8aa1682 JD |
3144 | /* Parse arguments */ |
3145 | progname = argv[0]; | |
178a0557 MD |
3146 | if (set_options(argc, argv)) { |
3147 | retval = -1; | |
3148 | goto exit_options; | |
b8aa1682 JD |
3149 | } |
3150 | ||
178a0557 MD |
3151 | if (set_signal_handler()) { |
3152 | retval = -1; | |
3153 | goto exit_options; | |
b8aa1682 JD |
3154 | } |
3155 | ||
4d513a50 DG |
3156 | /* Try to create directory if -o, --output is specified. */ |
3157 | if (opt_output_path) { | |
994fa64f DG |
3158 | if (*opt_output_path != '/') { |
3159 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
3160 | retval = -1; |
3161 | goto exit_options; | |
994fa64f DG |
3162 | } |
3163 | ||
d77dded2 JG |
3164 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
3165 | -1, -1); | |
4d513a50 DG |
3166 | if (ret < 0) { |
3167 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
3168 | retval = -1; |
3169 | goto exit_options; | |
4d513a50 DG |
3170 | } |
3171 | } | |
3172 | ||
b8aa1682 | 3173 | /* Daemonize */ |
b5218ffb | 3174 | if (opt_daemon || opt_background) { |
3fd27398 MD |
3175 | int i; |
3176 | ||
3177 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
3178 | !opt_background); | |
b8aa1682 | 3179 | if (ret < 0) { |
178a0557 MD |
3180 | retval = -1; |
3181 | goto exit_options; | |
b8aa1682 | 3182 | } |
3fd27398 MD |
3183 | |
3184 | /* | |
3185 | * We are in the child. Make sure all other file | |
3186 | * descriptors are closed, in case we are called with | |
3187 | * more opened file descriptors than the standard ones. | |
3188 | */ | |
3189 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
3190 | (void) close(i); | |
3191 | } | |
3192 | } | |
3193 | ||
178a0557 MD |
3194 | /* Initialize thread health monitoring */ |
3195 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
3196 | if (!health_relayd) { | |
3197 | PERROR("health_app_create error"); | |
3198 | retval = -1; | |
3199 | goto exit_health_app_create; | |
3200 | } | |
3201 | ||
3fd27398 | 3202 | /* Create thread quit pipe */ |
178a0557 MD |
3203 | if (init_thread_quit_pipe()) { |
3204 | retval = -1; | |
3205 | goto exit_init_data; | |
b8aa1682 JD |
3206 | } |
3207 | ||
b8aa1682 | 3208 | /* Setup the thread apps communication pipe. */ |
178a0557 MD |
3209 | if (create_relay_conn_pipe()) { |
3210 | retval = -1; | |
3211 | goto exit_init_data; | |
b8aa1682 JD |
3212 | } |
3213 | ||
3214 | /* Init relay command queue. */ | |
8bdee6e2 | 3215 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 | 3216 | |
554831e7 MD |
3217 | /* Initialize communication library */ |
3218 | lttcomm_init(); | |
87e45c13 | 3219 | lttcomm_inet_init(); |
554831e7 | 3220 | |
d3e2ba59 | 3221 | /* tables of sessions indexed by session ID */ |
7591bab1 MD |
3222 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3223 | if (!sessions_ht) { | |
178a0557 MD |
3224 | retval = -1; |
3225 | goto exit_init_data; | |
d3e2ba59 JD |
3226 | } |
3227 | ||
3228 | /* tables of streams indexed by stream ID */ | |
2a174661 | 3229 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 3230 | if (!relay_streams_ht) { |
178a0557 MD |
3231 | retval = -1; |
3232 | goto exit_init_data; | |
d3e2ba59 JD |
3233 | } |
3234 | ||
3235 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
3236 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3237 | if (!viewer_streams_ht) { | |
178a0557 MD |
3238 | retval = -1; |
3239 | goto exit_init_data; | |
55706a7d MD |
3240 | } |
3241 | ||
65931c8b | 3242 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
3243 | if (ret) { |
3244 | retval = -1; | |
3245 | goto exit_health_quit_pipe; | |
65931c8b MD |
3246 | } |
3247 | ||
3248 | /* Create thread to manage the client socket */ | |
1a1a34b4 | 3249 | ret = pthread_create(&health_thread, default_pthread_attr(), |
65931c8b | 3250 | thread_manage_health, (void *) NULL); |
178a0557 MD |
3251 | if (ret) { |
3252 | errno = ret; | |
65931c8b | 3253 | PERROR("pthread_create health"); |
178a0557 MD |
3254 | retval = -1; |
3255 | goto exit_health_thread; | |
65931c8b MD |
3256 | } |
3257 | ||
b8aa1682 | 3258 | /* Setup the dispatcher thread */ |
1a1a34b4 | 3259 | ret = pthread_create(&dispatcher_thread, default_pthread_attr(), |
b8aa1682 | 3260 | relay_thread_dispatcher, (void *) NULL); |
178a0557 MD |
3261 | if (ret) { |
3262 | errno = ret; | |
b8aa1682 | 3263 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
3264 | retval = -1; |
3265 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
3266 | } |
3267 | ||
3268 | /* Setup the worker thread */ | |
1a1a34b4 | 3269 | ret = pthread_create(&worker_thread, default_pthread_attr(), |
7591bab1 | 3270 | relay_thread_worker, NULL); |
178a0557 MD |
3271 | if (ret) { |
3272 | errno = ret; | |
b8aa1682 | 3273 | PERROR("pthread_create worker"); |
178a0557 MD |
3274 | retval = -1; |
3275 | goto exit_worker_thread; | |
b8aa1682 JD |
3276 | } |
3277 | ||
3278 | /* Setup the listener thread */ | |
1a1a34b4 | 3279 | ret = pthread_create(&listener_thread, default_pthread_attr(), |
b8aa1682 | 3280 | relay_thread_listener, (void *) NULL); |
178a0557 MD |
3281 | if (ret) { |
3282 | errno = ret; | |
b8aa1682 | 3283 | PERROR("pthread_create listener"); |
178a0557 MD |
3284 | retval = -1; |
3285 | goto exit_listener_thread; | |
b8aa1682 JD |
3286 | } |
3287 | ||
7591bab1 | 3288 | ret = relayd_live_create(live_uri); |
178a0557 | 3289 | if (ret) { |
d3e2ba59 | 3290 | ERR("Starting live viewer threads"); |
178a0557 | 3291 | retval = -1; |
50138f51 | 3292 | goto exit_live; |
d3e2ba59 JD |
3293 | } |
3294 | ||
178a0557 MD |
3295 | /* |
3296 | * This is where we start awaiting program completion (e.g. through | |
3297 | * signal that asks threads to teardown). | |
3298 | */ | |
3299 | ||
3300 | ret = relayd_live_join(); | |
3301 | if (ret) { | |
3302 | retval = -1; | |
3303 | } | |
50138f51 | 3304 | exit_live: |
178a0557 | 3305 | |
b8aa1682 | 3306 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
3307 | if (ret) { |
3308 | errno = ret; | |
3309 | PERROR("pthread_join listener_thread"); | |
3310 | retval = -1; | |
b8aa1682 JD |
3311 | } |
3312 | ||
178a0557 | 3313 | exit_listener_thread: |
b8aa1682 | 3314 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
3315 | if (ret) { |
3316 | errno = ret; | |
3317 | PERROR("pthread_join worker_thread"); | |
3318 | retval = -1; | |
b8aa1682 JD |
3319 | } |
3320 | ||
178a0557 | 3321 | exit_worker_thread: |
b8aa1682 | 3322 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
3323 | if (ret) { |
3324 | errno = ret; | |
3325 | PERROR("pthread_join dispatcher_thread"); | |
3326 | retval = -1; | |
b8aa1682 | 3327 | } |
178a0557 | 3328 | exit_dispatcher_thread: |
42415026 | 3329 | |
65931c8b | 3330 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
3331 | if (ret) { |
3332 | errno = ret; | |
3333 | PERROR("pthread_join health_thread"); | |
3334 | retval = -1; | |
65931c8b | 3335 | } |
178a0557 | 3336 | exit_health_thread: |
65931c8b | 3337 | |
65931c8b | 3338 | utils_close_pipe(health_quit_pipe); |
178a0557 | 3339 | exit_health_quit_pipe: |
65931c8b | 3340 | |
178a0557 | 3341 | exit_init_data: |
55706a7d | 3342 | health_app_destroy(health_relayd); |
55706a7d | 3343 | exit_health_app_create: |
178a0557 | 3344 | exit_options: |
2668465a MD |
3345 | /* |
3346 | * Wait for all pending call_rcu work to complete before tearing | |
3347 | * down data structures. call_rcu worker may be trying to | |
3348 | * perform lookups in those structures. | |
3349 | */ | |
3350 | rcu_barrier(); | |
7591bab1 MD |
3351 | relayd_cleanup(); |
3352 | ||
3353 | /* Ensure all prior call_rcu are done. */ | |
3354 | rcu_barrier(); | |
d3e2ba59 | 3355 | |
178a0557 | 3356 | if (!retval) { |
b8aa1682 | 3357 | exit(EXIT_SUCCESS); |
178a0557 MD |
3358 | } else { |
3359 | exit(EXIT_FAILURE); | |
b8aa1682 | 3360 | } |
b8aa1682 | 3361 | } |