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> | |
70626904 | 40 | #include <urcu/rculist.h> |
b8aa1682 JD |
41 | #include <unistd.h> |
42 | #include <fcntl.h> | |
f8be1183 | 43 | #include <strings.h> |
b8aa1682 JD |
44 | |
45 | #include <lttng/lttng.h> | |
46 | #include <common/common.h> | |
47 | #include <common/compat/poll.h> | |
48 | #include <common/compat/socket.h> | |
f263b7fd | 49 | #include <common/compat/endian.h> |
e8fa9fb0 | 50 | #include <common/compat/getenv.h> |
b8aa1682 | 51 | #include <common/defaults.h> |
3fd27398 | 52 | #include <common/daemonize.h> |
b8aa1682 JD |
53 | #include <common/futex.h> |
54 | #include <common/sessiond-comm/sessiond-comm.h> | |
55 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
56 | #include <common/sessiond-comm/relayd.h> |
57 | #include <common/uri.h> | |
a02de639 | 58 | #include <common/utils.h> |
d3ecc550 | 59 | #include <common/align.h> |
f40ef1d5 | 60 | #include <common/config/session-config.h> |
5312a3ed JG |
61 | #include <common/dynamic-buffer.h> |
62 | #include <common/buffer-view.h> | |
70626904 | 63 | #include <common/string-utils/format.h> |
b8aa1682 | 64 | |
0f907de1 | 65 | #include "cmd.h" |
d3e2ba59 | 66 | #include "ctf-trace.h" |
1c20f0e2 | 67 | #include "index.h" |
0f907de1 | 68 | #include "utils.h" |
b8aa1682 | 69 | #include "lttng-relayd.h" |
d3e2ba59 | 70 | #include "live.h" |
55706a7d | 71 | #include "health-relayd.h" |
9b5e0863 | 72 | #include "testpoint.h" |
2f8f53af | 73 | #include "viewer-stream.h" |
2a174661 DG |
74 | #include "session.h" |
75 | #include "stream.h" | |
58eb9381 | 76 | #include "connection.h" |
a44ca2ca | 77 | #include "tracefile-array.h" |
f056029c | 78 | #include "tcp_keep_alive.h" |
23c8ff50 | 79 | #include "sessiond-trace-chunks.h" |
b8aa1682 | 80 | |
4fc83d94 PP |
81 | static const char *help_msg = |
82 | #ifdef LTTNG_EMBED_HELP | |
83 | #include <lttng-relayd.8.h> | |
84 | #else | |
85 | NULL | |
86 | #endif | |
87 | ; | |
88 | ||
5569b118 JG |
89 | enum relay_connection_status { |
90 | RELAY_CONNECTION_STATUS_OK, | |
a9577b76 | 91 | /* An error occurred while processing an event on the connection. */ |
5569b118 JG |
92 | RELAY_CONNECTION_STATUS_ERROR, |
93 | /* Connection closed/shutdown cleanly. */ | |
94 | RELAY_CONNECTION_STATUS_CLOSED, | |
95 | }; | |
96 | ||
b8aa1682 | 97 | /* command line options */ |
0f907de1 | 98 | char *opt_output_path; |
b5218ffb | 99 | static int opt_daemon, opt_background; |
3fd27398 MD |
100 | |
101 | /* | |
102 | * We need to wait for listener and live listener threads, as well as | |
103 | * health check thread, before being ready to signal readiness. | |
104 | */ | |
105 | #define NR_LTTNG_RELAY_READY 3 | |
106 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
0848dba7 MD |
107 | |
108 | /* Size of receive buffer. */ | |
109 | #define RECV_DATA_BUFFER_SIZE 65536 | |
110 | ||
3fd27398 MD |
111 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ |
112 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
113 | ||
095a4ae5 MD |
114 | static struct lttng_uri *control_uri; |
115 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 116 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
117 | |
118 | const char *progname; | |
b8aa1682 | 119 | |
65931c8b | 120 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
121 | static int tracing_group_name_override; |
122 | ||
123 | const char * const config_section_name = "relayd"; | |
65931c8b | 124 | |
b8aa1682 JD |
125 | /* |
126 | * Quit pipe for all threads. This permits a single cancellation point | |
127 | * for all threads when receiving an event on the pipe. | |
128 | */ | |
0b242f62 | 129 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
130 | |
131 | /* | |
132 | * This pipe is used to inform the worker thread that a command is queued and | |
133 | * ready to be processed. | |
134 | */ | |
58eb9381 | 135 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 136 | |
26c9d55e | 137 | /* Shared between threads */ |
b8aa1682 JD |
138 | static int dispatch_thread_exit; |
139 | ||
140 | static pthread_t listener_thread; | |
141 | static pthread_t dispatcher_thread; | |
142 | static pthread_t worker_thread; | |
65931c8b | 143 | static pthread_t health_thread; |
b8aa1682 | 144 | |
7591bab1 MD |
145 | /* |
146 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
147 | * atomicity on 32-bit architectures. | |
148 | */ | |
149 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 150 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
151 | |
152 | /* | |
153 | * Relay command queue. | |
154 | * | |
155 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
156 | * queue. | |
157 | */ | |
58eb9381 | 158 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 | 159 | |
d3e2ba59 JD |
160 | /* Global relay stream hash table. */ |
161 | struct lttng_ht *relay_streams_ht; | |
162 | ||
92c6ca54 DG |
163 | /* Global relay viewer stream hash table. */ |
164 | struct lttng_ht *viewer_streams_ht; | |
165 | ||
7591bab1 MD |
166 | /* Global relay sessions hash table. */ |
167 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 168 | |
55706a7d | 169 | /* Relayd health monitoring */ |
eea7556c | 170 | struct health_app *health_relayd; |
55706a7d | 171 | |
23c8ff50 JG |
172 | struct sessiond_trace_chunk_registry *sessiond_trace_chunk_registry; |
173 | ||
cd60b05a JG |
174 | static struct option long_options[] = { |
175 | { "control-port", 1, 0, 'C', }, | |
176 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 177 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 178 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 179 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
180 | { "group", 1, 0, 'g', }, |
181 | { "help", 0, 0, 'h', }, | |
182 | { "output", 1, 0, 'o', }, | |
183 | { "verbose", 0, 0, 'v', }, | |
184 | { "config", 1, 0, 'f' }, | |
3a904098 | 185 | { "version", 0, 0, 'V' }, |
cd60b05a JG |
186 | { NULL, 0, 0, 0, }, |
187 | }; | |
188 | ||
3a904098 | 189 | static const char *config_ignore_options[] = { "help", "config", "version" }; |
cd60b05a | 190 | |
cd60b05a JG |
191 | /* |
192 | * Take an option from the getopt output and set it in the right variable to be | |
193 | * used later. | |
194 | * | |
195 | * Return 0 on success else a negative value. | |
196 | */ | |
7591bab1 | 197 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 198 | { |
cd60b05a JG |
199 | int ret; |
200 | ||
201 | switch (opt) { | |
202 | case 0: | |
203 | fprintf(stderr, "option %s", optname); | |
204 | if (arg) { | |
205 | fprintf(stderr, " with arg %s\n", arg); | |
206 | } | |
207 | break; | |
208 | case 'C': | |
e8fa9fb0 MD |
209 | if (lttng_is_setuid_setgid()) { |
210 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
211 | "-C, --control-port"); | |
212 | } else { | |
213 | ret = uri_parse(arg, &control_uri); | |
214 | if (ret < 0) { | |
215 | ERR("Invalid control URI specified"); | |
216 | goto end; | |
217 | } | |
218 | if (control_uri->port == 0) { | |
219 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
220 | } | |
cd60b05a JG |
221 | } |
222 | break; | |
223 | case 'D': | |
e8fa9fb0 MD |
224 | if (lttng_is_setuid_setgid()) { |
225 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
226 | "-D, -data-port"); | |
227 | } else { | |
228 | ret = uri_parse(arg, &data_uri); | |
229 | if (ret < 0) { | |
230 | ERR("Invalid data URI specified"); | |
231 | goto end; | |
232 | } | |
233 | if (data_uri->port == 0) { | |
234 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
235 | } | |
cd60b05a JG |
236 | } |
237 | break; | |
8d5c808e | 238 | case 'L': |
e8fa9fb0 MD |
239 | if (lttng_is_setuid_setgid()) { |
240 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
241 | "-L, -live-port"); | |
242 | } else { | |
243 | ret = uri_parse(arg, &live_uri); | |
244 | if (ret < 0) { | |
245 | ERR("Invalid live URI specified"); | |
246 | goto end; | |
247 | } | |
248 | if (live_uri->port == 0) { | |
249 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
250 | } | |
8d5c808e AM |
251 | } |
252 | break; | |
cd60b05a JG |
253 | case 'd': |
254 | opt_daemon = 1; | |
255 | break; | |
b5218ffb MD |
256 | case 'b': |
257 | opt_background = 1; | |
258 | break; | |
cd60b05a | 259 | case 'g': |
e8fa9fb0 MD |
260 | if (lttng_is_setuid_setgid()) { |
261 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
262 | "-g, --group"); | |
263 | } else { | |
264 | tracing_group_name = strdup(arg); | |
265 | if (tracing_group_name == NULL) { | |
266 | ret = -errno; | |
267 | PERROR("strdup"); | |
268 | goto end; | |
269 | } | |
270 | tracing_group_name_override = 1; | |
330a40bb | 271 | } |
cd60b05a JG |
272 | break; |
273 | case 'h': | |
4fc83d94 | 274 | ret = utils_show_help(8, "lttng-relayd", help_msg); |
655b5cc1 | 275 | if (ret) { |
4fc83d94 | 276 | ERR("Cannot show --help for `lttng-relayd`"); |
655b5cc1 PP |
277 | perror("exec"); |
278 | } | |
cd60b05a | 279 | exit(EXIT_FAILURE); |
3a904098 AW |
280 | case 'V': |
281 | fprintf(stdout, "%s\n", VERSION); | |
282 | exit(EXIT_SUCCESS); | |
cd60b05a | 283 | case 'o': |
e8fa9fb0 MD |
284 | if (lttng_is_setuid_setgid()) { |
285 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
286 | "-o, --output"); | |
287 | } else { | |
288 | ret = asprintf(&opt_output_path, "%s", arg); | |
289 | if (ret < 0) { | |
290 | ret = -errno; | |
291 | PERROR("asprintf opt_output_path"); | |
292 | goto end; | |
293 | } | |
cd60b05a JG |
294 | } |
295 | break; | |
296 | case 'v': | |
297 | /* Verbose level can increase using multiple -v */ | |
298 | if (arg) { | |
299 | lttng_opt_verbose = config_parse_value(arg); | |
300 | } else { | |
849e5b7b DG |
301 | /* Only 3 level of verbosity (-vvv). */ |
302 | if (lttng_opt_verbose < 3) { | |
303 | lttng_opt_verbose += 1; | |
304 | } | |
cd60b05a JG |
305 | } |
306 | break; | |
307 | default: | |
308 | /* Unknown option or other error. | |
309 | * Error is printed by getopt, just return */ | |
310 | ret = -1; | |
311 | goto end; | |
312 | } | |
313 | ||
314 | /* All good. */ | |
315 | ret = 0; | |
316 | ||
317 | end: | |
318 | return ret; | |
319 | } | |
320 | ||
321 | /* | |
322 | * config_entry_handler_cb used to handle options read from a config file. | |
f40ef1d5 | 323 | * See config_entry_handler_cb comment in common/config/session-config.h for the |
cd60b05a JG |
324 | * return value conventions. |
325 | */ | |
7591bab1 | 326 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
327 | { |
328 | int ret = 0, i; | |
329 | ||
330 | if (!entry || !entry->name || !entry->value) { | |
331 | ret = -EINVAL; | |
332 | goto end; | |
333 | } | |
334 | ||
335 | /* Check if the option is to be ignored */ | |
336 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
337 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
338 | goto end; | |
339 | } | |
340 | } | |
341 | ||
342 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
343 | /* Ignore if entry name is not fully matched. */ | |
344 | if (strcmp(entry->name, long_options[i].name)) { | |
345 | continue; | |
346 | } | |
347 | ||
348 | /* | |
7591bab1 MD |
349 | * If the option takes no argument on the command line, |
350 | * we have to check if the value is "true". We support | |
351 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
352 | */ |
353 | if (!long_options[i].has_arg) { | |
354 | ret = config_parse_value(entry->value); | |
355 | if (ret <= 0) { | |
356 | if (ret) { | |
357 | WARN("Invalid configuration value \"%s\" for option %s", | |
358 | entry->value, entry->name); | |
359 | } | |
360 | /* False, skip boolean config option. */ | |
361 | goto end; | |
362 | } | |
363 | } | |
364 | ||
365 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
366 | goto end; | |
367 | } | |
368 | ||
369 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
370 | entry->name); | |
371 | ||
372 | end: | |
373 | return ret; | |
374 | } | |
375 | ||
7591bab1 | 376 | static int set_options(int argc, char **argv) |
cd60b05a | 377 | { |
178a0557 | 378 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
379 | int orig_optopt = optopt, orig_optind = optind; |
380 | char *default_address, *optstring; | |
381 | const char *config_path = NULL; | |
382 | ||
383 | optstring = utils_generate_optstring(long_options, | |
384 | sizeof(long_options) / sizeof(struct option)); | |
385 | if (!optstring) { | |
178a0557 | 386 | retval = -ENOMEM; |
cd60b05a JG |
387 | goto exit; |
388 | } | |
389 | ||
390 | /* Check for the --config option */ | |
391 | ||
392 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
393 | &option_index)) != -1) { | |
394 | if (c == '?') { | |
178a0557 | 395 | retval = -EINVAL; |
cd60b05a JG |
396 | goto exit; |
397 | } else if (c != 'f') { | |
398 | continue; | |
399 | } | |
400 | ||
e8fa9fb0 MD |
401 | if (lttng_is_setuid_setgid()) { |
402 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
403 | "-f, --config"); | |
404 | } else { | |
405 | config_path = utils_expand_path(optarg); | |
406 | if (!config_path) { | |
407 | ERR("Failed to resolve path: %s", optarg); | |
408 | } | |
cd60b05a JG |
409 | } |
410 | } | |
411 | ||
412 | ret = config_get_section_entries(config_path, config_section_name, | |
413 | config_entry_handler, NULL); | |
414 | if (ret) { | |
415 | if (ret > 0) { | |
416 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 417 | } |
178a0557 | 418 | retval = -1; |
cd60b05a JG |
419 | goto exit; |
420 | } | |
b8aa1682 | 421 | |
cd60b05a JG |
422 | /* Reset getopt's global state */ |
423 | optopt = orig_optopt; | |
424 | optind = orig_optind; | |
b8aa1682 | 425 | while (1) { |
cd60b05a | 426 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
427 | if (c == -1) { |
428 | break; | |
429 | } | |
430 | ||
cd60b05a JG |
431 | ret = set_option(c, optarg, long_options[option_index].name); |
432 | if (ret < 0) { | |
178a0557 | 433 | retval = -1; |
b8aa1682 JD |
434 | goto exit; |
435 | } | |
436 | } | |
437 | ||
438 | /* assign default values */ | |
439 | if (control_uri == NULL) { | |
fa91dc52 MD |
440 | ret = asprintf(&default_address, |
441 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
442 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
443 | if (ret < 0) { |
444 | PERROR("asprintf default data address"); | |
178a0557 | 445 | retval = -1; |
b8aa1682 JD |
446 | goto exit; |
447 | } | |
448 | ||
449 | ret = uri_parse(default_address, &control_uri); | |
450 | free(default_address); | |
451 | if (ret < 0) { | |
452 | ERR("Invalid control URI specified"); | |
178a0557 | 453 | retval = -1; |
b8aa1682 JD |
454 | goto exit; |
455 | } | |
456 | } | |
457 | if (data_uri == NULL) { | |
fa91dc52 MD |
458 | ret = asprintf(&default_address, |
459 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
460 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
461 | if (ret < 0) { |
462 | PERROR("asprintf default data address"); | |
178a0557 | 463 | retval = -1; |
b8aa1682 JD |
464 | goto exit; |
465 | } | |
466 | ||
467 | ret = uri_parse(default_address, &data_uri); | |
468 | free(default_address); | |
469 | if (ret < 0) { | |
470 | ERR("Invalid data URI specified"); | |
178a0557 | 471 | retval = -1; |
b8aa1682 JD |
472 | goto exit; |
473 | } | |
474 | } | |
d3e2ba59 | 475 | if (live_uri == NULL) { |
fa91dc52 MD |
476 | ret = asprintf(&default_address, |
477 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
478 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
479 | if (ret < 0) { |
480 | PERROR("asprintf default viewer control address"); | |
178a0557 | 481 | retval = -1; |
d3e2ba59 JD |
482 | goto exit; |
483 | } | |
484 | ||
485 | ret = uri_parse(default_address, &live_uri); | |
486 | free(default_address); | |
487 | if (ret < 0) { | |
488 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 489 | retval = -1; |
d3e2ba59 JD |
490 | goto exit; |
491 | } | |
492 | } | |
b8aa1682 JD |
493 | |
494 | exit: | |
cd60b05a | 495 | free(optstring); |
178a0557 | 496 | return retval; |
b8aa1682 JD |
497 | } |
498 | ||
7591bab1 MD |
499 | static void print_global_objects(void) |
500 | { | |
501 | rcu_register_thread(); | |
502 | ||
503 | print_viewer_streams(); | |
504 | print_relay_streams(); | |
505 | print_sessions(); | |
506 | ||
507 | rcu_unregister_thread(); | |
508 | } | |
509 | ||
b8aa1682 JD |
510 | /* |
511 | * Cleanup the daemon | |
512 | */ | |
7591bab1 | 513 | static void relayd_cleanup(void) |
b8aa1682 | 514 | { |
7591bab1 MD |
515 | print_global_objects(); |
516 | ||
b8aa1682 JD |
517 | DBG("Cleaning up"); |
518 | ||
178a0557 MD |
519 | if (viewer_streams_ht) |
520 | lttng_ht_destroy(viewer_streams_ht); | |
521 | if (relay_streams_ht) | |
522 | lttng_ht_destroy(relay_streams_ht); | |
7591bab1 MD |
523 | if (sessions_ht) |
524 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 525 | |
095a4ae5 MD |
526 | /* free the dynamically allocated opt_output_path */ |
527 | free(opt_output_path); | |
528 | ||
a02de639 CB |
529 | /* Close thread quit pipes */ |
530 | utils_close_pipe(thread_quit_pipe); | |
531 | ||
710c1f73 DG |
532 | uri_free(control_uri); |
533 | uri_free(data_uri); | |
8d5c808e | 534 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
535 | |
536 | if (tracing_group_name_override) { | |
537 | free((void *) tracing_group_name); | |
538 | } | |
b8aa1682 JD |
539 | } |
540 | ||
541 | /* | |
542 | * Write to writable pipe used to notify a thread. | |
543 | */ | |
7591bab1 | 544 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 545 | { |
6cd525e8 | 546 | ssize_t ret; |
b8aa1682 | 547 | |
6cd525e8 MD |
548 | ret = lttng_write(wpipe, "!", 1); |
549 | if (ret < 1) { | |
b8aa1682 | 550 | PERROR("write poll pipe"); |
b4aacfdc | 551 | goto end; |
b8aa1682 | 552 | } |
b4aacfdc MD |
553 | ret = 0; |
554 | end: | |
b8aa1682 JD |
555 | return ret; |
556 | } | |
557 | ||
7591bab1 | 558 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 559 | { |
6cd525e8 | 560 | ssize_t ret; |
65931c8b | 561 | |
6cd525e8 MD |
562 | ret = lttng_write(pipe[1], "4", 1); |
563 | if (ret < 1) { | |
65931c8b | 564 | PERROR("write relay health quit"); |
b4aacfdc | 565 | goto end; |
65931c8b | 566 | } |
b4aacfdc MD |
567 | ret = 0; |
568 | end: | |
569 | return ret; | |
65931c8b MD |
570 | } |
571 | ||
b8aa1682 | 572 | /* |
b4aacfdc | 573 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 574 | */ |
b4aacfdc | 575 | int lttng_relay_stop_threads(void) |
b8aa1682 | 576 | { |
b4aacfdc | 577 | int retval = 0; |
b8aa1682 JD |
578 | |
579 | /* Stopping all threads */ | |
580 | DBG("Terminating all threads"); | |
b4aacfdc | 581 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 582 | ERR("write error on thread quit pipe"); |
b4aacfdc | 583 | retval = -1; |
b8aa1682 JD |
584 | } |
585 | ||
b4aacfdc MD |
586 | if (notify_health_quit_pipe(health_quit_pipe)) { |
587 | ERR("write error on health quit pipe"); | |
588 | } | |
65931c8b | 589 | |
b8aa1682 | 590 | /* Dispatch thread */ |
26c9d55e | 591 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 592 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 593 | |
b4aacfdc | 594 | if (relayd_live_stop()) { |
178a0557 | 595 | ERR("Error stopping live threads"); |
b4aacfdc | 596 | retval = -1; |
178a0557 | 597 | } |
b4aacfdc | 598 | return retval; |
b8aa1682 JD |
599 | } |
600 | ||
601 | /* | |
602 | * Signal handler for the daemon | |
603 | * | |
604 | * Simply stop all worker threads, leaving main() return gracefully after | |
605 | * joining all threads and calling cleanup(). | |
606 | */ | |
7591bab1 | 607 | static void sighandler(int sig) |
b8aa1682 JD |
608 | { |
609 | switch (sig) { | |
b8aa1682 JD |
610 | case SIGINT: |
611 | DBG("SIGINT caught"); | |
b4aacfdc MD |
612 | if (lttng_relay_stop_threads()) { |
613 | ERR("Error stopping threads"); | |
614 | } | |
b8aa1682 JD |
615 | break; |
616 | case SIGTERM: | |
617 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
618 | if (lttng_relay_stop_threads()) { |
619 | ERR("Error stopping threads"); | |
620 | } | |
b8aa1682 | 621 | break; |
3fd27398 MD |
622 | case SIGUSR1: |
623 | CMM_STORE_SHARED(recv_child_signal, 1); | |
624 | break; | |
b8aa1682 JD |
625 | default: |
626 | break; | |
627 | } | |
628 | } | |
629 | ||
630 | /* | |
631 | * Setup signal handler for : | |
632 | * SIGINT, SIGTERM, SIGPIPE | |
633 | */ | |
7591bab1 | 634 | static int set_signal_handler(void) |
b8aa1682 JD |
635 | { |
636 | int ret = 0; | |
637 | struct sigaction sa; | |
638 | sigset_t sigset; | |
639 | ||
640 | if ((ret = sigemptyset(&sigset)) < 0) { | |
641 | PERROR("sigemptyset"); | |
642 | return ret; | |
643 | } | |
644 | ||
b8aa1682 JD |
645 | sa.sa_mask = sigset; |
646 | sa.sa_flags = 0; | |
0072e5e2 MD |
647 | |
648 | sa.sa_handler = sighandler; | |
b8aa1682 JD |
649 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { |
650 | PERROR("sigaction"); | |
651 | return ret; | |
652 | } | |
653 | ||
654 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
655 | PERROR("sigaction"); | |
656 | return ret; | |
657 | } | |
658 | ||
0072e5e2 | 659 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
b8aa1682 JD |
660 | PERROR("sigaction"); |
661 | return ret; | |
662 | } | |
663 | ||
0072e5e2 MD |
664 | sa.sa_handler = SIG_IGN; |
665 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
3fd27398 MD |
666 | PERROR("sigaction"); |
667 | return ret; | |
668 | } | |
669 | ||
670 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
671 | |
672 | return ret; | |
673 | } | |
674 | ||
3fd27398 MD |
675 | void lttng_relay_notify_ready(void) |
676 | { | |
677 | /* Notify the parent of the fork() process that we are ready. */ | |
678 | if (opt_daemon || opt_background) { | |
679 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
680 | kill(child_ppid, SIGUSR1); | |
681 | } | |
682 | } | |
683 | } | |
684 | ||
b8aa1682 JD |
685 | /* |
686 | * Init thread quit pipe. | |
687 | * | |
688 | * Return -1 on error or 0 if all pipes are created. | |
689 | */ | |
7591bab1 | 690 | static int init_thread_quit_pipe(void) |
b8aa1682 | 691 | { |
a02de639 | 692 | int ret; |
b8aa1682 | 693 | |
a02de639 | 694 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 695 | |
b8aa1682 JD |
696 | return ret; |
697 | } | |
698 | ||
699 | /* | |
700 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
701 | */ | |
7591bab1 | 702 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
703 | { |
704 | int ret; | |
705 | ||
706 | if (events == NULL || size == 0) { | |
707 | ret = -1; | |
708 | goto error; | |
709 | } | |
710 | ||
711 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
712 | if (ret < 0) { | |
713 | goto error; | |
714 | } | |
715 | ||
716 | /* Add quit pipe */ | |
c7759e6a | 717 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
718 | if (ret < 0) { |
719 | goto error; | |
720 | } | |
721 | ||
722 | return 0; | |
723 | ||
724 | error: | |
725 | return ret; | |
726 | } | |
727 | ||
728 | /* | |
729 | * Check if the thread quit pipe was triggered. | |
730 | * | |
731 | * Return 1 if it was triggered else 0; | |
732 | */ | |
7591bab1 | 733 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
734 | { |
735 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
736 | return 1; | |
737 | } | |
738 | ||
739 | return 0; | |
740 | } | |
741 | ||
742 | /* | |
743 | * Create and init socket from uri. | |
744 | */ | |
7591bab1 | 745 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
746 | { |
747 | int ret; | |
748 | struct lttcomm_sock *sock = NULL; | |
749 | ||
750 | sock = lttcomm_alloc_sock_from_uri(uri); | |
751 | if (sock == NULL) { | |
752 | ERR("Allocating socket"); | |
753 | goto error; | |
754 | } | |
755 | ||
756 | ret = lttcomm_create_sock(sock); | |
757 | if (ret < 0) { | |
758 | goto error; | |
759 | } | |
760 | DBG("Listening on sock %d", sock->fd); | |
761 | ||
762 | ret = sock->ops->bind(sock); | |
763 | if (ret < 0) { | |
2288467f | 764 | PERROR("Failed to bind socket"); |
b8aa1682 JD |
765 | goto error; |
766 | } | |
767 | ||
768 | ret = sock->ops->listen(sock, -1); | |
769 | if (ret < 0) { | |
770 | goto error; | |
771 | ||
772 | } | |
773 | ||
774 | return sock; | |
775 | ||
776 | error: | |
777 | if (sock) { | |
778 | lttcomm_destroy_sock(sock); | |
779 | } | |
780 | return NULL; | |
781 | } | |
782 | ||
783 | /* | |
784 | * This thread manages the listening for new connections on the network | |
785 | */ | |
7591bab1 | 786 | static void *relay_thread_listener(void *data) |
b8aa1682 | 787 | { |
095a4ae5 | 788 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
789 | uint32_t revents, nb_fd; |
790 | struct lttng_poll_event events; | |
791 | struct lttcomm_sock *control_sock, *data_sock; | |
792 | ||
b8aa1682 JD |
793 | DBG("[thread] Relay listener started"); |
794 | ||
55706a7d MD |
795 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
796 | ||
f385ae0a MD |
797 | health_code_update(); |
798 | ||
7591bab1 | 799 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 800 | if (!control_sock) { |
095a4ae5 | 801 | goto error_sock_control; |
b8aa1682 JD |
802 | } |
803 | ||
7591bab1 | 804 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 805 | if (!data_sock) { |
095a4ae5 | 806 | goto error_sock_relay; |
b8aa1682 JD |
807 | } |
808 | ||
809 | /* | |
7591bab1 MD |
810 | * Pass 3 as size here for the thread quit pipe, control and |
811 | * data socket. | |
b8aa1682 JD |
812 | */ |
813 | ret = create_thread_poll_set(&events, 3); | |
814 | if (ret < 0) { | |
815 | goto error_create_poll; | |
816 | } | |
817 | ||
818 | /* Add the control socket */ | |
819 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
820 | if (ret < 0) { | |
821 | goto error_poll_add; | |
822 | } | |
823 | ||
824 | /* Add the data socket */ | |
825 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
826 | if (ret < 0) { | |
827 | goto error_poll_add; | |
828 | } | |
829 | ||
3fd27398 MD |
830 | lttng_relay_notify_ready(); |
831 | ||
9b5e0863 MD |
832 | if (testpoint(relayd_thread_listener)) { |
833 | goto error_testpoint; | |
834 | } | |
835 | ||
b8aa1682 | 836 | while (1) { |
f385ae0a MD |
837 | health_code_update(); |
838 | ||
b8aa1682 JD |
839 | DBG("Listener accepting connections"); |
840 | ||
b8aa1682 | 841 | restart: |
f385ae0a | 842 | health_poll_entry(); |
b8aa1682 | 843 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 844 | health_poll_exit(); |
b8aa1682 JD |
845 | if (ret < 0) { |
846 | /* | |
847 | * Restart interrupted system call. | |
848 | */ | |
849 | if (errno == EINTR) { | |
850 | goto restart; | |
851 | } | |
852 | goto error; | |
853 | } | |
854 | ||
0d9c5d77 DG |
855 | nb_fd = ret; |
856 | ||
b8aa1682 JD |
857 | DBG("Relay new connection received"); |
858 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
859 | health_code_update(); |
860 | ||
b8aa1682 JD |
861 | /* Fetch once the poll data */ |
862 | revents = LTTNG_POLL_GETEV(&events, i); | |
863 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
864 | ||
865 | /* Thread quit pipe has been closed. Killing thread. */ | |
866 | ret = check_thread_quit_pipe(pollfd, revents); | |
867 | if (ret) { | |
095a4ae5 MD |
868 | err = 0; |
869 | goto exit; | |
b8aa1682 JD |
870 | } |
871 | ||
03e43155 | 872 | if (revents & LPOLLIN) { |
4b7f17b2 | 873 | /* |
7591bab1 MD |
874 | * A new connection is requested, therefore a |
875 | * sessiond/consumerd connection is allocated in | |
876 | * this thread, enqueued to a global queue and | |
877 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 878 | */ |
58eb9381 DG |
879 | int val = 1; |
880 | struct relay_connection *new_conn; | |
4b7f17b2 | 881 | struct lttcomm_sock *newsock; |
7591bab1 | 882 | enum connection_type type; |
b8aa1682 JD |
883 | |
884 | if (pollfd == data_sock->fd) { | |
7591bab1 | 885 | type = RELAY_DATA; |
b8aa1682 | 886 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
887 | DBG("Relay data connection accepted, socket %d", |
888 | newsock->fd); | |
4b7f17b2 MD |
889 | } else { |
890 | assert(pollfd == control_sock->fd); | |
7591bab1 | 891 | type = RELAY_CONTROL; |
b8aa1682 | 892 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
893 | DBG("Relay control connection accepted, socket %d", |
894 | newsock->fd); | |
b8aa1682 | 895 | } |
58eb9381 DG |
896 | if (!newsock) { |
897 | PERROR("accepting sock"); | |
58eb9381 DG |
898 | goto error; |
899 | } | |
900 | ||
901 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
902 | sizeof(val)); | |
b8aa1682 JD |
903 | if (ret < 0) { |
904 | PERROR("setsockopt inet"); | |
4b7f17b2 | 905 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
906 | goto error; |
907 | } | |
f056029c JR |
908 | |
909 | ret = socket_apply_keep_alive_config(newsock->fd); | |
910 | if (ret < 0) { | |
911 | ERR("Failed to apply TCP keep-alive configuration on socket (%i)", | |
912 | newsock->fd); | |
913 | lttcomm_destroy_sock(newsock); | |
914 | goto error; | |
915 | } | |
916 | ||
7591bab1 MD |
917 | new_conn = connection_create(newsock, type); |
918 | if (!new_conn) { | |
919 | lttcomm_destroy_sock(newsock); | |
920 | goto error; | |
921 | } | |
58eb9381 DG |
922 | |
923 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
924 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
925 | &new_conn->qnode); | |
b8aa1682 JD |
926 | |
927 | /* | |
7591bab1 MD |
928 | * Wake the dispatch queue futex. |
929 | * Implicit memory barrier with the | |
930 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 931 | */ |
58eb9381 | 932 | futex_nto1_wake(&relay_conn_queue.futex); |
03e43155 MD |
933 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
934 | ERR("socket poll error"); | |
935 | goto error; | |
936 | } else { | |
937 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
938 | goto error; | |
b8aa1682 JD |
939 | } |
940 | } | |
941 | } | |
942 | ||
095a4ae5 | 943 | exit: |
b8aa1682 JD |
944 | error: |
945 | error_poll_add: | |
9b5e0863 | 946 | error_testpoint: |
b8aa1682 JD |
947 | lttng_poll_clean(&events); |
948 | error_create_poll: | |
095a4ae5 MD |
949 | if (data_sock->fd >= 0) { |
950 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
951 | if (ret) { |
952 | PERROR("close"); | |
953 | } | |
b8aa1682 | 954 | } |
095a4ae5 MD |
955 | lttcomm_destroy_sock(data_sock); |
956 | error_sock_relay: | |
957 | if (control_sock->fd >= 0) { | |
958 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
959 | if (ret) { |
960 | PERROR("close"); | |
961 | } | |
b8aa1682 | 962 | } |
095a4ae5 MD |
963 | lttcomm_destroy_sock(control_sock); |
964 | error_sock_control: | |
965 | if (err) { | |
f385ae0a MD |
966 | health_error(); |
967 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 968 | } |
55706a7d | 969 | health_unregister(health_relayd); |
b8aa1682 | 970 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 971 | lttng_relay_stop_threads(); |
b8aa1682 JD |
972 | return NULL; |
973 | } | |
974 | ||
975 | /* | |
976 | * This thread manages the dispatching of the requests to worker threads | |
977 | */ | |
7591bab1 | 978 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 979 | { |
6cd525e8 MD |
980 | int err = -1; |
981 | ssize_t ret; | |
8bdee6e2 | 982 | struct cds_wfcq_node *node; |
58eb9381 | 983 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
984 | |
985 | DBG("[thread] Relay dispatcher started"); | |
986 | ||
55706a7d MD |
987 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
988 | ||
9b5e0863 MD |
989 | if (testpoint(relayd_thread_dispatcher)) { |
990 | goto error_testpoint; | |
991 | } | |
992 | ||
f385ae0a MD |
993 | health_code_update(); |
994 | ||
0ed3b1a8 | 995 | for (;;) { |
f385ae0a MD |
996 | health_code_update(); |
997 | ||
b8aa1682 | 998 | /* Atomically prepare the queue futex */ |
58eb9381 | 999 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 | 1000 | |
0ed3b1a8 MD |
1001 | if (CMM_LOAD_SHARED(dispatch_thread_exit)) { |
1002 | break; | |
1003 | } | |
1004 | ||
b8aa1682 | 1005 | do { |
f385ae0a MD |
1006 | health_code_update(); |
1007 | ||
b8aa1682 | 1008 | /* Dequeue commands */ |
8bdee6e2 SM |
1009 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
1010 | &relay_conn_queue.tail); | |
b8aa1682 JD |
1011 | if (node == NULL) { |
1012 | DBG("Woken up but nothing in the relay command queue"); | |
1013 | /* Continue thread execution */ | |
1014 | break; | |
1015 | } | |
58eb9381 | 1016 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 1017 | |
58eb9381 | 1018 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1019 | |
1020 | /* | |
7591bab1 MD |
1021 | * Inform worker thread of the new request. This |
1022 | * call is blocking so we can be assured that | |
1023 | * the data will be read at some point in time | |
1024 | * or wait to the end of the world :) | |
b8aa1682 | 1025 | */ |
58eb9381 DG |
1026 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1027 | if (ret < 0) { | |
1028 | PERROR("write connection pipe"); | |
7591bab1 | 1029 | connection_put(new_conn); |
b8aa1682 JD |
1030 | goto error; |
1031 | } | |
1032 | } while (node != NULL); | |
1033 | ||
1034 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1035 | health_poll_entry(); |
58eb9381 | 1036 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1037 | health_poll_exit(); |
b8aa1682 JD |
1038 | } |
1039 | ||
f385ae0a MD |
1040 | /* Normal exit, no error */ |
1041 | err = 0; | |
1042 | ||
b8aa1682 | 1043 | error: |
9b5e0863 | 1044 | error_testpoint: |
f385ae0a MD |
1045 | if (err) { |
1046 | health_error(); | |
1047 | ERR("Health error occurred in %s", __func__); | |
1048 | } | |
55706a7d | 1049 | health_unregister(health_relayd); |
b8aa1682 | 1050 | DBG("Dispatch thread dying"); |
b4aacfdc | 1051 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1052 | return NULL; |
1053 | } | |
1054 | ||
298a25ca JG |
1055 | static bool session_streams_have_index(const struct relay_session *session) |
1056 | { | |
1057 | return session->minor >= 4 && !session->snapshot; | |
1058 | } | |
1059 | ||
c5b6f4f0 DG |
1060 | /* |
1061 | * Handle the RELAYD_CREATE_SESSION command. | |
1062 | * | |
1063 | * On success, send back the session id or else return a negative value. | |
1064 | */ | |
5312a3ed JG |
1065 | static int relay_create_session(const struct lttcomm_relayd_hdr *recv_hdr, |
1066 | struct relay_connection *conn, | |
1067 | const struct lttng_buffer_view *payload) | |
c5b6f4f0 | 1068 | { |
5312a3ed JG |
1069 | int ret = 0; |
1070 | ssize_t send_ret; | |
4c6885d2 | 1071 | struct relay_session *session = NULL; |
ecd1a12f | 1072 | struct lttcomm_relayd_create_session_reply_2_11 reply = {}; |
1e791a74 JG |
1073 | char session_name[LTTNG_NAME_MAX] = {}; |
1074 | char hostname[LTTNG_HOST_NAME_MAX] = {}; | |
7591bab1 MD |
1075 | uint32_t live_timer = 0; |
1076 | bool snapshot = false; | |
46ef2188 | 1077 | bool session_name_contains_creation_timestamp = false; |
23c8ff50 | 1078 | /* Left nil for peers < 2.11. */ |
6fa5fe7c | 1079 | char base_path[LTTNG_PATH_MAX] = {}; |
23c8ff50 | 1080 | lttng_uuid sessiond_uuid = {}; |
1e791a74 JG |
1081 | LTTNG_OPTIONAL(uint64_t) id_sessiond = {}; |
1082 | LTTNG_OPTIONAL(uint64_t) current_chunk_id = {}; | |
db1da059 | 1083 | LTTNG_OPTIONAL(time_t) creation_time = {}; |
ecd1a12f MD |
1084 | struct lttng_dynamic_buffer reply_payload; |
1085 | ||
1086 | lttng_dynamic_buffer_init(&reply_payload); | |
c5b6f4f0 | 1087 | |
f86f6389 JR |
1088 | if (conn->minor < 4) { |
1089 | /* From 2.1 to 2.3 */ | |
1090 | ret = 0; | |
1091 | } else if (conn->minor >= 4 && conn->minor < 11) { | |
1092 | /* From 2.4 to 2.10 */ | |
5312a3ed | 1093 | ret = cmd_create_session_2_4(payload, session_name, |
7591bab1 | 1094 | hostname, &live_timer, &snapshot); |
f86f6389 | 1095 | } else { |
84fa4db5 | 1096 | bool has_current_chunk; |
db1da059 JG |
1097 | uint64_t current_chunk_id_value; |
1098 | time_t creation_time_value; | |
1099 | uint64_t id_sessiond_value; | |
84fa4db5 | 1100 | |
f86f6389 | 1101 | /* From 2.11 to ... */ |
db1da059 | 1102 | ret = cmd_create_session_2_11(payload, session_name, hostname, |
6fa5fe7c | 1103 | base_path, &live_timer, &snapshot, &id_sessiond_value, |
db1da059 | 1104 | sessiond_uuid, &has_current_chunk, |
46ef2188 MD |
1105 | ¤t_chunk_id_value, &creation_time_value, |
1106 | &session_name_contains_creation_timestamp); | |
23c8ff50 JG |
1107 | if (lttng_uuid_is_nil(sessiond_uuid)) { |
1108 | /* The nil UUID is reserved for pre-2.11 clients. */ | |
1109 | ERR("Illegal nil UUID announced by peer in create session command"); | |
1110 | ret = -1; | |
1111 | goto send_reply; | |
1112 | } | |
db1da059 JG |
1113 | LTTNG_OPTIONAL_SET(&id_sessiond, id_sessiond_value); |
1114 | LTTNG_OPTIONAL_SET(&creation_time, creation_time_value); | |
1115 | if (has_current_chunk) { | |
1116 | LTTNG_OPTIONAL_SET(¤t_chunk_id, | |
1117 | current_chunk_id_value); | |
1118 | } | |
7591bab1 | 1119 | } |
f86f6389 | 1120 | |
7591bab1 MD |
1121 | if (ret < 0) { |
1122 | goto send_reply; | |
d3e2ba59 JD |
1123 | } |
1124 | ||
6fa5fe7c | 1125 | session = session_create(session_name, hostname, base_path, live_timer, |
1e791a74 JG |
1126 | snapshot, sessiond_uuid, |
1127 | id_sessiond.is_set ? &id_sessiond.value : NULL, | |
1128 | current_chunk_id.is_set ? ¤t_chunk_id.value : NULL, | |
db1da059 | 1129 | creation_time.is_set ? &creation_time.value : NULL, |
46ef2188 MD |
1130 | conn->major, conn->minor, |
1131 | session_name_contains_creation_timestamp); | |
7591bab1 MD |
1132 | if (!session) { |
1133 | ret = -1; | |
1134 | goto send_reply; | |
1135 | } | |
1136 | assert(!conn->session); | |
1137 | conn->session = session; | |
c5b6f4f0 DG |
1138 | DBG("Created session %" PRIu64, session->id); |
1139 | ||
ecd1a12f | 1140 | reply.generic.session_id = htobe64(session->id); |
7591bab1 MD |
1141 | |
1142 | send_reply: | |
c5b6f4f0 | 1143 | if (ret < 0) { |
ecd1a12f | 1144 | reply.generic.ret_code = htobe32(LTTNG_ERR_FATAL); |
c5b6f4f0 | 1145 | } else { |
ecd1a12f | 1146 | reply.generic.ret_code = htobe32(LTTNG_OK); |
c5b6f4f0 DG |
1147 | } |
1148 | ||
ecd1a12f MD |
1149 | if (conn->minor < 11) { |
1150 | /* From 2.1 to 2.10 */ | |
1151 | ret = lttng_dynamic_buffer_append(&reply_payload, | |
1152 | &reply.generic, sizeof(reply.generic)); | |
1153 | if (ret) { | |
1154 | ERR("Failed to append \"create session\" command reply header to payload buffer"); | |
1155 | ret = -1; | |
1156 | goto end; | |
1157 | } | |
1158 | } else { | |
1159 | const uint32_t output_path_length = | |
1160 | strlen(session->output_path) + 1; | |
1161 | ||
1162 | reply.output_path_length = htobe32(output_path_length); | |
1163 | ret = lttng_dynamic_buffer_append(&reply_payload, &reply, | |
1164 | sizeof(reply)); | |
1165 | if (ret) { | |
1166 | ERR("Failed to append \"create session\" command reply header to payload buffer"); | |
1167 | goto end; | |
1168 | } | |
1169 | ||
1170 | ret = lttng_dynamic_buffer_append(&reply_payload, | |
1171 | session->output_path, output_path_length); | |
1172 | if (ret) { | |
1173 | ERR("Failed to append \"create session\" command reply path to payload buffer"); | |
1174 | goto end; | |
1175 | } | |
1176 | } | |
1177 | ||
1178 | send_ret = conn->sock->ops->sendmsg(conn->sock, reply_payload.data, | |
1179 | reply_payload.size, 0); | |
1180 | if (send_ret < (ssize_t) reply_payload.size) { | |
1181 | ERR("Failed to send \"create session\" command reply of %zu bytes (ret = %zd)", | |
1182 | reply_payload.size, send_ret); | |
5312a3ed | 1183 | ret = -1; |
c5b6f4f0 | 1184 | } |
ecd1a12f | 1185 | end: |
4c6885d2 JG |
1186 | if (ret < 0 && session) { |
1187 | session_put(session); | |
1188 | } | |
ecd1a12f | 1189 | lttng_dynamic_buffer_reset(&reply_payload); |
c5b6f4f0 DG |
1190 | return ret; |
1191 | } | |
1192 | ||
a4baae1b JD |
1193 | /* |
1194 | * When we have received all the streams and the metadata for a channel, | |
1195 | * we make them visible to the viewer threads. | |
1196 | */ | |
7591bab1 | 1197 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1198 | { |
7591bab1 MD |
1199 | struct relay_stream *stream; |
1200 | struct relay_session *session = conn->session; | |
a4baae1b | 1201 | |
7591bab1 MD |
1202 | /* |
1203 | * We publish all streams belonging to a session atomically wrt | |
1204 | * session lock. | |
1205 | */ | |
1206 | pthread_mutex_lock(&session->lock); | |
1207 | rcu_read_lock(); | |
1208 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1209 | recv_node) { | |
1210 | stream_publish(stream); | |
a4baae1b | 1211 | } |
7591bab1 | 1212 | rcu_read_unlock(); |
a4baae1b | 1213 | |
7591bab1 MD |
1214 | /* |
1215 | * Inform the viewer that there are new streams in the session. | |
1216 | */ | |
1217 | if (session->viewer_attached) { | |
1218 | uatomic_set(&session->new_streams, 1); | |
1219 | } | |
1220 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1221 | } |
1222 | ||
348a81dc JG |
1223 | static int conform_channel_path(char *channel_path) |
1224 | { | |
1225 | int ret = 0; | |
1226 | ||
1227 | if (strstr("../", channel_path)) { | |
1228 | ERR("Refusing channel path as it walks up the path hierarchy: \"%s\"", | |
1229 | channel_path); | |
1230 | ret = -1; | |
1231 | goto end; | |
1232 | } | |
1233 | ||
1234 | if (*channel_path == '/') { | |
1235 | const size_t len = strlen(channel_path); | |
1236 | ||
1237 | /* | |
1238 | * Channel paths from peers prior to 2.11 are expressed as an | |
1239 | * absolute path that is, in reality, relative to the relay | |
1240 | * daemon's output directory. Remove the leading slash so it | |
1241 | * is correctly interpreted as a relative path later on. | |
1242 | * | |
1243 | * len (and not len - 1) is used to copy the trailing NULL. | |
1244 | */ | |
1245 | bcopy(channel_path + 1, channel_path, len); | |
1246 | } | |
1247 | end: | |
1248 | return ret; | |
1249 | } | |
1250 | ||
b8aa1682 JD |
1251 | /* |
1252 | * relay_add_stream: allocate a new stream for a session | |
1253 | */ | |
5312a3ed JG |
1254 | static int relay_add_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1255 | struct relay_connection *conn, | |
1256 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1257 | { |
7591bab1 MD |
1258 | int ret; |
1259 | ssize_t send_ret; | |
58eb9381 | 1260 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1261 | struct relay_stream *stream = NULL; |
1262 | struct lttcomm_relayd_status_stream reply; | |
4030a636 | 1263 | struct ctf_trace *trace = NULL; |
7591bab1 MD |
1264 | uint64_t stream_handle = -1ULL; |
1265 | char *path_name = NULL, *channel_name = NULL; | |
1266 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
348a81dc | 1267 | LTTNG_OPTIONAL(uint64_t) stream_chunk_id = {}; |
b8aa1682 | 1268 | |
5312a3ed | 1269 | if (!session || !conn->version_check_done) { |
b8aa1682 JD |
1270 | ERR("Trying to add a stream before version check"); |
1271 | ret = -1; | |
1272 | goto end_no_session; | |
1273 | } | |
1274 | ||
2f21a469 JR |
1275 | if (session->minor == 1) { |
1276 | /* For 2.1 */ | |
5312a3ed | 1277 | ret = cmd_recv_stream_2_1(payload, &path_name, |
7591bab1 | 1278 | &channel_name); |
2f21a469 JR |
1279 | } else if (session->minor > 1 && session->minor < 11) { |
1280 | /* From 2.2 to 2.10 */ | |
5312a3ed | 1281 | ret = cmd_recv_stream_2_2(payload, &path_name, |
7591bab1 | 1282 | &channel_name, &tracefile_size, &tracefile_count); |
2f21a469 JR |
1283 | } else { |
1284 | /* From 2.11 to ... */ | |
1285 | ret = cmd_recv_stream_2_11(payload, &path_name, | |
0b50e4b3 JG |
1286 | &channel_name, &tracefile_size, &tracefile_count, |
1287 | &stream_chunk_id.value); | |
1288 | stream_chunk_id.is_set = true; | |
0f907de1 | 1289 | } |
2f21a469 | 1290 | |
0f907de1 | 1291 | if (ret < 0) { |
7591bab1 | 1292 | goto send_reply; |
b8aa1682 JD |
1293 | } |
1294 | ||
348a81dc JG |
1295 | if (conform_channel_path(path_name)) { |
1296 | goto send_reply; | |
1297 | } | |
1298 | ||
7591bab1 | 1299 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1300 | if (!trace) { |
7591bab1 | 1301 | goto send_reply; |
2a174661 | 1302 | } |
7591bab1 | 1303 | /* This stream here has one reference on the trace. */ |
2a174661 | 1304 | |
7591bab1 MD |
1305 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1306 | stream_handle = ++last_relay_stream_id; | |
1307 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1308 | |
7591bab1 MD |
1309 | /* We pass ownership of path_name and channel_name. */ |
1310 | stream = stream_create(trace, stream_handle, path_name, | |
348a81dc | 1311 | channel_name, tracefile_size, tracefile_count); |
7591bab1 MD |
1312 | path_name = NULL; |
1313 | channel_name = NULL; | |
a4baae1b | 1314 | |
2a174661 | 1315 | /* |
7591bab1 MD |
1316 | * Streams are the owners of their trace. Reference to trace is |
1317 | * kept within stream_create(). | |
2a174661 | 1318 | */ |
7591bab1 | 1319 | ctf_trace_put(trace); |
d3e2ba59 | 1320 | |
7591bab1 | 1321 | send_reply: |
53efb85a | 1322 | memset(&reply, 0, sizeof(reply)); |
7591bab1 MD |
1323 | reply.handle = htobe64(stream_handle); |
1324 | if (!stream) { | |
f73fabfd | 1325 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1326 | } else { |
f73fabfd | 1327 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1328 | } |
5af40280 | 1329 | |
58eb9381 | 1330 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 | 1331 | sizeof(struct lttcomm_relayd_status_stream), 0); |
5312a3ed JG |
1332 | if (send_ret < (ssize_t) sizeof(reply)) { |
1333 | ERR("Failed to send \"add stream\" command reply (ret = %zd)", | |
1334 | send_ret); | |
1335 | ret = -1; | |
b8aa1682 JD |
1336 | } |
1337 | ||
1338 | end_no_session: | |
7591bab1 MD |
1339 | free(path_name); |
1340 | free(channel_name); | |
0f907de1 | 1341 | return ret; |
b8aa1682 JD |
1342 | } |
1343 | ||
173af62f DG |
1344 | /* |
1345 | * relay_close_stream: close a specific stream | |
1346 | */ | |
5312a3ed JG |
1347 | static int relay_close_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1348 | struct relay_connection *conn, | |
1349 | const struct lttng_buffer_view *payload) | |
173af62f | 1350 | { |
5312a3ed JG |
1351 | int ret; |
1352 | ssize_t send_ret; | |
58eb9381 | 1353 | struct relay_session *session = conn->session; |
173af62f DG |
1354 | struct lttcomm_relayd_close_stream stream_info; |
1355 | struct lttcomm_relayd_generic_reply reply; | |
1356 | struct relay_stream *stream; | |
173af62f DG |
1357 | |
1358 | DBG("Close stream received"); | |
1359 | ||
5312a3ed | 1360 | if (!session || !conn->version_check_done) { |
173af62f DG |
1361 | ERR("Trying to close a stream before version check"); |
1362 | ret = -1; | |
1363 | goto end_no_session; | |
1364 | } | |
1365 | ||
5312a3ed JG |
1366 | if (payload->size < sizeof(stream_info)) { |
1367 | ERR("Unexpected payload size in \"relay_close_stream\": expected >= %zu bytes, got %zu bytes", | |
1368 | sizeof(stream_info), payload->size); | |
173af62f DG |
1369 | ret = -1; |
1370 | goto end_no_session; | |
1371 | } | |
5312a3ed JG |
1372 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1373 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1374 | stream_info.last_net_seq_num = be64toh(stream_info.last_net_seq_num); | |
173af62f | 1375 | |
5312a3ed | 1376 | stream = stream_get_by_id(stream_info.stream_id); |
173af62f DG |
1377 | if (!stream) { |
1378 | ret = -1; | |
7591bab1 | 1379 | goto end; |
173af62f | 1380 | } |
77f7bd85 MD |
1381 | |
1382 | /* | |
1383 | * Set last_net_seq_num before the close flag. Required by data | |
1384 | * pending check. | |
1385 | */ | |
7591bab1 | 1386 | pthread_mutex_lock(&stream->lock); |
5312a3ed | 1387 | stream->last_net_seq_num = stream_info.last_net_seq_num; |
77f7bd85 MD |
1388 | pthread_mutex_unlock(&stream->lock); |
1389 | ||
bda7c7b9 JG |
1390 | /* |
1391 | * This is one of the conditions which may trigger a stream close | |
1392 | * with the others being: | |
1393 | * 1) A close command is received for a stream | |
1394 | * 2) The control connection owning the stream is closed | |
1395 | * 3) We have received all of the stream's data _after_ a close | |
1396 | * request. | |
1397 | */ | |
1398 | try_stream_close(stream); | |
7591bab1 MD |
1399 | if (stream->is_metadata) { |
1400 | struct relay_viewer_stream *vstream; | |
173af62f | 1401 | |
7591bab1 MD |
1402 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1403 | if (vstream) { | |
1404 | if (vstream->metadata_sent == stream->metadata_received) { | |
1405 | /* | |
1406 | * Since all the metadata has been sent to the | |
1407 | * viewer and that we have a request to close | |
1408 | * its stream, we can safely teardown the | |
1409 | * corresponding metadata viewer stream. | |
1410 | */ | |
1411 | viewer_stream_put(vstream); | |
1412 | } | |
1413 | /* Put local reference. */ | |
1414 | viewer_stream_put(vstream); | |
1415 | } | |
1416 | } | |
7591bab1 | 1417 | stream_put(stream); |
5312a3ed | 1418 | ret = 0; |
173af62f | 1419 | |
7591bab1 | 1420 | end: |
53efb85a | 1421 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1422 | if (ret < 0) { |
f73fabfd | 1423 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1424 | } else { |
f73fabfd | 1425 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1426 | } |
58eb9381 | 1427 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f | 1428 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
5312a3ed JG |
1429 | if (send_ret < (ssize_t) sizeof(reply)) { |
1430 | ERR("Failed to send \"close stream\" command reply (ret = %zd)", | |
1431 | send_ret); | |
1432 | ret = -1; | |
173af62f DG |
1433 | } |
1434 | ||
1435 | end_no_session: | |
1436 | return ret; | |
1437 | } | |
1438 | ||
93ec662e JD |
1439 | /* |
1440 | * relay_reset_metadata: reset a metadata stream | |
1441 | */ | |
1442 | static | |
5312a3ed JG |
1443 | int relay_reset_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1444 | struct relay_connection *conn, | |
1445 | const struct lttng_buffer_view *payload) | |
93ec662e | 1446 | { |
5312a3ed JG |
1447 | int ret; |
1448 | ssize_t send_ret; | |
93ec662e JD |
1449 | struct relay_session *session = conn->session; |
1450 | struct lttcomm_relayd_reset_metadata stream_info; | |
1451 | struct lttcomm_relayd_generic_reply reply; | |
1452 | struct relay_stream *stream; | |
1453 | ||
1454 | DBG("Reset metadata received"); | |
1455 | ||
5312a3ed | 1456 | if (!session || !conn->version_check_done) { |
93ec662e JD |
1457 | ERR("Trying to reset a metadata stream before version check"); |
1458 | ret = -1; | |
1459 | goto end_no_session; | |
1460 | } | |
1461 | ||
5312a3ed JG |
1462 | if (payload->size < sizeof(stream_info)) { |
1463 | ERR("Unexpected payload size in \"relay_reset_metadata\": expected >= %zu bytes, got %zu bytes", | |
1464 | sizeof(stream_info), payload->size); | |
93ec662e JD |
1465 | ret = -1; |
1466 | goto end_no_session; | |
1467 | } | |
5312a3ed JG |
1468 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1469 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1470 | stream_info.version = be64toh(stream_info.version); | |
1471 | ||
1472 | DBG("Update metadata to version %" PRIu64, stream_info.version); | |
93ec662e JD |
1473 | |
1474 | /* Unsupported for live sessions for now. */ | |
1475 | if (session->live_timer != 0) { | |
1476 | ret = -1; | |
1477 | goto end; | |
1478 | } | |
1479 | ||
5312a3ed | 1480 | stream = stream_get_by_id(stream_info.stream_id); |
93ec662e JD |
1481 | if (!stream) { |
1482 | ret = -1; | |
1483 | goto end; | |
1484 | } | |
1485 | pthread_mutex_lock(&stream->lock); | |
1486 | if (!stream->is_metadata) { | |
1487 | ret = -1; | |
1488 | goto end_unlock; | |
1489 | } | |
1490 | ||
c35f9726 | 1491 | ret = stream_reset_file(stream); |
93ec662e | 1492 | if (ret < 0) { |
c35f9726 JG |
1493 | ERR("Failed to reset metadata stream %" PRIu64 |
1494 | ": stream_path = %s, channel = %s", | |
1495 | stream->stream_handle, stream->path_name, | |
1496 | stream->channel_name); | |
93ec662e JD |
1497 | goto end_unlock; |
1498 | } | |
93ec662e JD |
1499 | end_unlock: |
1500 | pthread_mutex_unlock(&stream->lock); | |
1501 | stream_put(stream); | |
1502 | ||
1503 | end: | |
1504 | memset(&reply, 0, sizeof(reply)); | |
1505 | if (ret < 0) { | |
1506 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1507 | } else { | |
1508 | reply.ret_code = htobe32(LTTNG_OK); | |
1509 | } | |
1510 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1511 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed JG |
1512 | if (send_ret < (ssize_t) sizeof(reply)) { |
1513 | ERR("Failed to send \"reset metadata\" command reply (ret = %zd)", | |
1514 | send_ret); | |
1515 | ret = -1; | |
93ec662e JD |
1516 | } |
1517 | ||
1518 | end_no_session: | |
1519 | return ret; | |
1520 | } | |
1521 | ||
b8aa1682 JD |
1522 | /* |
1523 | * relay_unknown_command: send -1 if received unknown command | |
1524 | */ | |
7591bab1 | 1525 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1526 | { |
1527 | struct lttcomm_relayd_generic_reply reply; | |
5312a3ed | 1528 | ssize_t send_ret; |
b8aa1682 | 1529 | |
53efb85a | 1530 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1531 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
5312a3ed JG |
1532 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1533 | if (send_ret < sizeof(reply)) { | |
1534 | ERR("Failed to send \"unknown command\" command reply (ret = %zd)", send_ret); | |
b8aa1682 JD |
1535 | } |
1536 | } | |
1537 | ||
1538 | /* | |
1539 | * relay_start: send an acknowledgment to the client to tell if we are | |
1540 | * ready to receive data. We are ready if a session is established. | |
1541 | */ | |
5312a3ed JG |
1542 | static int relay_start(const struct lttcomm_relayd_hdr *recv_hdr, |
1543 | struct relay_connection *conn, | |
1544 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1545 | { |
5312a3ed JG |
1546 | int ret = 0; |
1547 | ssize_t send_ret; | |
b8aa1682 | 1548 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1549 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1550 | |
1551 | if (!session) { | |
1552 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1553 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1554 | } |
1555 | ||
53efb85a | 1556 | memset(&reply, 0, sizeof(reply)); |
5312a3ed JG |
1557 | reply.ret_code = htobe32(LTTNG_OK); |
1558 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1559 | sizeof(reply), 0); | |
1560 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1561 | ERR("Failed to send \"relay_start\" command reply (ret = %zd)", | |
1562 | send_ret); | |
1563 | ret = -1; | |
b8aa1682 JD |
1564 | } |
1565 | ||
1566 | return ret; | |
1567 | } | |
1568 | ||
b8aa1682 | 1569 | /* |
7591bab1 | 1570 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1571 | */ |
5312a3ed JG |
1572 | static int relay_recv_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1573 | struct relay_connection *conn, | |
1574 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1575 | { |
32d1569c | 1576 | int ret = 0; |
58eb9381 | 1577 | struct relay_session *session = conn->session; |
5312a3ed | 1578 | struct lttcomm_relayd_metadata_payload metadata_payload_header; |
b8aa1682 | 1579 | struct relay_stream *metadata_stream; |
5312a3ed | 1580 | uint64_t metadata_payload_size; |
c35f9726 | 1581 | struct lttng_buffer_view packet_view; |
b8aa1682 JD |
1582 | |
1583 | if (!session) { | |
1584 | ERR("Metadata sent before version check"); | |
1585 | ret = -1; | |
1586 | goto end; | |
1587 | } | |
1588 | ||
5312a3ed | 1589 | if (recv_hdr->data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { |
f6416125 MD |
1590 | ERR("Incorrect data size"); |
1591 | ret = -1; | |
1592 | goto end; | |
1593 | } | |
5312a3ed JG |
1594 | metadata_payload_size = recv_hdr->data_size - |
1595 | sizeof(struct lttcomm_relayd_metadata_payload); | |
f6416125 | 1596 | |
5312a3ed JG |
1597 | memcpy(&metadata_payload_header, payload->data, |
1598 | sizeof(metadata_payload_header)); | |
1599 | metadata_payload_header.stream_id = be64toh( | |
1600 | metadata_payload_header.stream_id); | |
1601 | metadata_payload_header.padding_size = be32toh( | |
1602 | metadata_payload_header.padding_size); | |
9d1bbf21 | 1603 | |
5312a3ed | 1604 | metadata_stream = stream_get_by_id(metadata_payload_header.stream_id); |
b8aa1682 JD |
1605 | if (!metadata_stream) { |
1606 | ret = -1; | |
7591bab1 | 1607 | goto end; |
b8aa1682 JD |
1608 | } |
1609 | ||
c35f9726 JG |
1610 | packet_view = lttng_buffer_view_from_view(payload, |
1611 | sizeof(metadata_payload_header), metadata_payload_size); | |
1612 | if (!packet_view.data) { | |
1613 | ERR("Invalid metadata packet length announced by header"); | |
b8aa1682 | 1614 | ret = -1; |
7591bab1 | 1615 | goto end_put; |
b8aa1682 | 1616 | } |
1d4dfdef | 1617 | |
c35f9726 JG |
1618 | pthread_mutex_lock(&metadata_stream->lock); |
1619 | ret = stream_write(metadata_stream, &packet_view, | |
5312a3ed | 1620 | metadata_payload_header.padding_size); |
c35f9726 JG |
1621 | pthread_mutex_unlock(&metadata_stream->lock); |
1622 | if (ret){ | |
5312a3ed | 1623 | ret = -1; |
7591bab1 | 1624 | goto end_put; |
1d4dfdef | 1625 | } |
7591bab1 | 1626 | end_put: |
7591bab1 | 1627 | stream_put(metadata_stream); |
b8aa1682 JD |
1628 | end: |
1629 | return ret; | |
1630 | } | |
1631 | ||
1632 | /* | |
1633 | * relay_send_version: send relayd version number | |
1634 | */ | |
5312a3ed JG |
1635 | static int relay_send_version(const struct lttcomm_relayd_hdr *recv_hdr, |
1636 | struct relay_connection *conn, | |
1637 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1638 | { |
7f51dcba | 1639 | int ret; |
5312a3ed | 1640 | ssize_t send_ret; |
092b6259 | 1641 | struct lttcomm_relayd_version reply, msg; |
87cb6359 | 1642 | bool compatible = true; |
b8aa1682 | 1643 | |
5312a3ed | 1644 | conn->version_check_done = true; |
b8aa1682 | 1645 | |
092b6259 | 1646 | /* Get version from the other side. */ |
5312a3ed JG |
1647 | if (payload->size < sizeof(msg)) { |
1648 | ERR("Unexpected payload size in \"relay_send_version\": expected >= %zu bytes, got %zu bytes", | |
1649 | sizeof(msg), payload->size); | |
092b6259 | 1650 | ret = -1; |
092b6259 DG |
1651 | goto end; |
1652 | } | |
1653 | ||
5312a3ed JG |
1654 | memcpy(&msg, payload->data, sizeof(msg)); |
1655 | msg.major = be32toh(msg.major); | |
1656 | msg.minor = be32toh(msg.minor); | |
1657 | ||
53efb85a | 1658 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1659 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1660 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1661 | |
1662 | /* Major versions must be the same */ | |
5312a3ed | 1663 | if (reply.major != msg.major) { |
6151a90f | 1664 | DBG("Incompatible major versions (%u vs %u), deleting session", |
5312a3ed | 1665 | reply.major, msg.major); |
87cb6359 | 1666 | compatible = false; |
d4519fa3 JD |
1667 | } |
1668 | ||
58eb9381 | 1669 | conn->major = reply.major; |
0f907de1 | 1670 | /* We adapt to the lowest compatible version */ |
5312a3ed | 1671 | if (reply.minor <= msg.minor) { |
58eb9381 | 1672 | conn->minor = reply.minor; |
0f907de1 | 1673 | } else { |
5312a3ed | 1674 | conn->minor = msg.minor; |
0f907de1 JD |
1675 | } |
1676 | ||
6151a90f JD |
1677 | reply.major = htobe32(reply.major); |
1678 | reply.minor = htobe32(reply.minor); | |
5312a3ed JG |
1679 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
1680 | sizeof(reply), 0); | |
1681 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1682 | ERR("Failed to send \"send version\" command reply (ret = %zd)", | |
1683 | send_ret); | |
1684 | ret = -1; | |
1685 | goto end; | |
1686 | } else { | |
1687 | ret = 0; | |
6151a90f JD |
1688 | } |
1689 | ||
87cb6359 JD |
1690 | if (!compatible) { |
1691 | ret = -1; | |
1692 | goto end; | |
1693 | } | |
1694 | ||
58eb9381 DG |
1695 | DBG("Version check done using protocol %u.%u", conn->major, |
1696 | conn->minor); | |
b8aa1682 JD |
1697 | |
1698 | end: | |
1699 | return ret; | |
1700 | } | |
1701 | ||
c8f59ee5 | 1702 | /* |
6d805429 | 1703 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1704 | */ |
5312a3ed JG |
1705 | static int relay_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1706 | struct relay_connection *conn, | |
1707 | const struct lttng_buffer_view *payload) | |
c8f59ee5 | 1708 | { |
58eb9381 | 1709 | struct relay_session *session = conn->session; |
6d805429 | 1710 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1711 | struct lttcomm_relayd_generic_reply reply; |
1712 | struct relay_stream *stream; | |
5312a3ed | 1713 | ssize_t send_ret; |
c8f59ee5 | 1714 | int ret; |
298a25ca | 1715 | uint64_t stream_seq; |
c8f59ee5 | 1716 | |
6d805429 | 1717 | DBG("Data pending command received"); |
c8f59ee5 | 1718 | |
5312a3ed | 1719 | if (!session || !conn->version_check_done) { |
c8f59ee5 DG |
1720 | ERR("Trying to check for data before version check"); |
1721 | ret = -1; | |
1722 | goto end_no_session; | |
1723 | } | |
1724 | ||
5312a3ed JG |
1725 | if (payload->size < sizeof(msg)) { |
1726 | ERR("Unexpected payload size in \"relay_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1727 | sizeof(msg), payload->size); | |
c8f59ee5 DG |
1728 | ret = -1; |
1729 | goto end_no_session; | |
1730 | } | |
5312a3ed JG |
1731 | memcpy(&msg, payload->data, sizeof(msg)); |
1732 | msg.stream_id = be64toh(msg.stream_id); | |
1733 | msg.last_net_seq_num = be64toh(msg.last_net_seq_num); | |
c8f59ee5 | 1734 | |
5312a3ed | 1735 | stream = stream_get_by_id(msg.stream_id); |
de91f48a | 1736 | if (stream == NULL) { |
c8f59ee5 | 1737 | ret = -1; |
7591bab1 | 1738 | goto end; |
c8f59ee5 DG |
1739 | } |
1740 | ||
7591bab1 MD |
1741 | pthread_mutex_lock(&stream->lock); |
1742 | ||
298a25ca JG |
1743 | if (session_streams_have_index(session)) { |
1744 | /* | |
1745 | * Ensure that both the index and stream data have been | |
1746 | * flushed up to the requested point. | |
1747 | */ | |
a8f9f353 | 1748 | stream_seq = min(stream->prev_data_seq, stream->prev_index_seq); |
298a25ca | 1749 | } else { |
a8f9f353 | 1750 | stream_seq = stream->prev_data_seq; |
298a25ca | 1751 | } |
a8f9f353 | 1752 | DBG("Data pending for stream id %" PRIu64 ": prev_data_seq %" PRIu64 |
298a25ca JG |
1753 | ", prev_index_seq %" PRIu64 |
1754 | ", and last_seq %" PRIu64, msg.stream_id, | |
a8f9f353 | 1755 | stream->prev_data_seq, stream->prev_index_seq, |
298a25ca | 1756 | msg.last_net_seq_num); |
c8f59ee5 | 1757 | |
33832e64 | 1758 | /* Avoid wrapping issue */ |
298a25ca | 1759 | if (((int64_t) (stream_seq - msg.last_net_seq_num)) >= 0) { |
6d805429 | 1760 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1761 | ret = 0; |
6d805429 DG |
1762 | } else { |
1763 | /* Data still being streamed thus pending */ | |
1764 | ret = 1; | |
c8f59ee5 DG |
1765 | } |
1766 | ||
7591bab1 MD |
1767 | stream->data_pending_check_done = true; |
1768 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1769 | |
7591bab1 MD |
1770 | stream_put(stream); |
1771 | end: | |
c8f59ee5 | 1772 | |
53efb85a | 1773 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1774 | reply.ret_code = htobe32(ret); |
5312a3ed JG |
1775 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1776 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1777 | ERR("Failed to send \"data pending\" command reply (ret = %zd)", | |
1778 | send_ret); | |
1779 | ret = -1; | |
c8f59ee5 DG |
1780 | } |
1781 | ||
1782 | end_no_session: | |
1783 | return ret; | |
1784 | } | |
1785 | ||
1786 | /* | |
1787 | * Wait for the control socket to reach a quiescent state. | |
1788 | * | |
7591bab1 MD |
1789 | * Note that for now, when receiving this command from the session |
1790 | * daemon, this means that every subsequent commands or data received on | |
1791 | * the control socket has been handled. So, this is why we simply return | |
1792 | * OK here. | |
c8f59ee5 | 1793 | */ |
5312a3ed JG |
1794 | static int relay_quiescent_control(const struct lttcomm_relayd_hdr *recv_hdr, |
1795 | struct relay_connection *conn, | |
1796 | const struct lttng_buffer_view *payload) | |
c8f59ee5 DG |
1797 | { |
1798 | int ret; | |
5312a3ed | 1799 | ssize_t send_ret; |
ad7051c0 | 1800 | struct relay_stream *stream; |
ad7051c0 | 1801 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1802 | struct lttcomm_relayd_generic_reply reply; |
1803 | ||
1804 | DBG("Checking quiescent state on control socket"); | |
1805 | ||
5312a3ed | 1806 | if (!conn->session || !conn->version_check_done) { |
ad7051c0 DG |
1807 | ERR("Trying to check for data before version check"); |
1808 | ret = -1; | |
1809 | goto end_no_session; | |
1810 | } | |
1811 | ||
5312a3ed JG |
1812 | if (payload->size < sizeof(msg)) { |
1813 | ERR("Unexpected payload size in \"relay_quiescent_control\": expected >= %zu bytes, got %zu bytes", | |
1814 | sizeof(msg), payload->size); | |
ad7051c0 DG |
1815 | ret = -1; |
1816 | goto end_no_session; | |
1817 | } | |
5312a3ed JG |
1818 | memcpy(&msg, payload->data, sizeof(msg)); |
1819 | msg.stream_id = be64toh(msg.stream_id); | |
ad7051c0 | 1820 | |
5312a3ed | 1821 | stream = stream_get_by_id(msg.stream_id); |
7591bab1 MD |
1822 | if (!stream) { |
1823 | goto reply; | |
1824 | } | |
1825 | pthread_mutex_lock(&stream->lock); | |
1826 | stream->data_pending_check_done = true; | |
1827 | pthread_mutex_unlock(&stream->lock); | |
5312a3ed JG |
1828 | |
1829 | DBG("Relay quiescent control pending flag set to %" PRIu64, msg.stream_id); | |
7591bab1 MD |
1830 | stream_put(stream); |
1831 | reply: | |
53efb85a | 1832 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1833 | reply.ret_code = htobe32(LTTNG_OK); |
5312a3ed JG |
1834 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1835 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1836 | ERR("Failed to send \"quiescent control\" command reply (ret = %zd)", | |
1837 | send_ret); | |
1838 | ret = -1; | |
1839 | } else { | |
1840 | ret = 0; | |
c8f59ee5 DG |
1841 | } |
1842 | ||
ad7051c0 | 1843 | end_no_session: |
c8f59ee5 DG |
1844 | return ret; |
1845 | } | |
1846 | ||
f7079f67 | 1847 | /* |
7591bab1 MD |
1848 | * Initialize a data pending command. This means that a consumer is about |
1849 | * to ask for data pending for each stream it holds. Simply iterate over | |
1850 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1851 | * |
1852 | * This command returns to the client a LTTNG_OK code. | |
1853 | */ | |
5312a3ed JG |
1854 | static int relay_begin_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1855 | struct relay_connection *conn, | |
1856 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
1857 | { |
1858 | int ret; | |
5312a3ed | 1859 | ssize_t send_ret; |
f7079f67 DG |
1860 | struct lttng_ht_iter iter; |
1861 | struct lttcomm_relayd_begin_data_pending msg; | |
1862 | struct lttcomm_relayd_generic_reply reply; | |
1863 | struct relay_stream *stream; | |
f7079f67 DG |
1864 | |
1865 | assert(recv_hdr); | |
58eb9381 | 1866 | assert(conn); |
f7079f67 DG |
1867 | |
1868 | DBG("Init streams for data pending"); | |
1869 | ||
5312a3ed | 1870 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
1871 | ERR("Trying to check for data before version check"); |
1872 | ret = -1; | |
1873 | goto end_no_session; | |
1874 | } | |
1875 | ||
5312a3ed JG |
1876 | if (payload->size < sizeof(msg)) { |
1877 | ERR("Unexpected payload size in \"relay_begin_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1878 | sizeof(msg), payload->size); | |
f7079f67 DG |
1879 | ret = -1; |
1880 | goto end_no_session; | |
1881 | } | |
5312a3ed JG |
1882 | memcpy(&msg, payload->data, sizeof(msg)); |
1883 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 DG |
1884 | |
1885 | /* | |
7591bab1 MD |
1886 | * Iterate over all streams to set the begin data pending flag. |
1887 | * For now, the streams are indexed by stream handle so we have | |
1888 | * to iterate over all streams to find the one associated with | |
1889 | * the right session_id. | |
f7079f67 DG |
1890 | */ |
1891 | rcu_read_lock(); | |
d3e2ba59 | 1892 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1893 | node.node) { |
7591bab1 MD |
1894 | if (!stream_get(stream)) { |
1895 | continue; | |
1896 | } | |
5312a3ed | 1897 | if (stream->trace->session->id == msg.session_id) { |
7591bab1 MD |
1898 | pthread_mutex_lock(&stream->lock); |
1899 | stream->data_pending_check_done = false; | |
1900 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1901 | DBG("Set begin data pending flag to stream %" PRIu64, |
1902 | stream->stream_handle); | |
1903 | } | |
7591bab1 | 1904 | stream_put(stream); |
f7079f67 DG |
1905 | } |
1906 | rcu_read_unlock(); | |
1907 | ||
53efb85a | 1908 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1909 | /* All good, send back reply. */ |
1910 | reply.ret_code = htobe32(LTTNG_OK); | |
1911 | ||
5312a3ed JG |
1912 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1913 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1914 | ERR("Failed to send \"begin data pending\" command reply (ret = %zd)", | |
1915 | send_ret); | |
1916 | ret = -1; | |
1917 | } else { | |
1918 | ret = 0; | |
f7079f67 DG |
1919 | } |
1920 | ||
1921 | end_no_session: | |
1922 | return ret; | |
1923 | } | |
1924 | ||
1925 | /* | |
7591bab1 MD |
1926 | * End data pending command. This will check, for a given session id, if |
1927 | * each stream associated with it has its data_pending_check_done flag | |
1928 | * set. If not, this means that the client lost track of the stream but | |
1929 | * the data is still being streamed on our side. In this case, we inform | |
1930 | * the client that data is in flight. | |
f7079f67 DG |
1931 | * |
1932 | * Return to the client if there is data in flight or not with a ret_code. | |
1933 | */ | |
5312a3ed JG |
1934 | static int relay_end_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1935 | struct relay_connection *conn, | |
1936 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
1937 | { |
1938 | int ret; | |
5312a3ed | 1939 | ssize_t send_ret; |
f7079f67 DG |
1940 | struct lttng_ht_iter iter; |
1941 | struct lttcomm_relayd_end_data_pending msg; | |
1942 | struct lttcomm_relayd_generic_reply reply; | |
1943 | struct relay_stream *stream; | |
f7079f67 DG |
1944 | uint32_t is_data_inflight = 0; |
1945 | ||
f7079f67 DG |
1946 | DBG("End data pending command"); |
1947 | ||
5312a3ed | 1948 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
1949 | ERR("Trying to check for data before version check"); |
1950 | ret = -1; | |
1951 | goto end_no_session; | |
1952 | } | |
1953 | ||
5312a3ed JG |
1954 | if (payload->size < sizeof(msg)) { |
1955 | ERR("Unexpected payload size in \"relay_end_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1956 | sizeof(msg), payload->size); | |
f7079f67 DG |
1957 | ret = -1; |
1958 | goto end_no_session; | |
1959 | } | |
5312a3ed JG |
1960 | memcpy(&msg, payload->data, sizeof(msg)); |
1961 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 | 1962 | |
7591bab1 MD |
1963 | /* |
1964 | * Iterate over all streams to see if the begin data pending | |
1965 | * flag is set. | |
1966 | */ | |
f7079f67 | 1967 | rcu_read_lock(); |
d3e2ba59 | 1968 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1969 | node.node) { |
7591bab1 MD |
1970 | if (!stream_get(stream)) { |
1971 | continue; | |
1972 | } | |
5312a3ed | 1973 | if (stream->trace->session->id != msg.session_id) { |
7591bab1 MD |
1974 | stream_put(stream); |
1975 | continue; | |
1976 | } | |
1977 | pthread_mutex_lock(&stream->lock); | |
1978 | if (!stream->data_pending_check_done) { | |
298a25ca JG |
1979 | uint64_t stream_seq; |
1980 | ||
1981 | if (session_streams_have_index(conn->session)) { | |
1982 | /* | |
1983 | * Ensure that both the index and stream data have been | |
1984 | * flushed up to the requested point. | |
1985 | */ | |
a8f9f353 | 1986 | stream_seq = min(stream->prev_data_seq, stream->prev_index_seq); |
298a25ca | 1987 | } else { |
a8f9f353 | 1988 | stream_seq = stream->prev_data_seq; |
298a25ca JG |
1989 | } |
1990 | if (!stream->closed || !(((int64_t) (stream_seq - stream->last_net_seq_num)) >= 0)) { | |
7591bab1 MD |
1991 | is_data_inflight = 1; |
1992 | DBG("Data is still in flight for stream %" PRIu64, | |
1993 | stream->stream_handle); | |
1994 | pthread_mutex_unlock(&stream->lock); | |
1995 | stream_put(stream); | |
1996 | break; | |
1997 | } | |
f7079f67 | 1998 | } |
7591bab1 MD |
1999 | pthread_mutex_unlock(&stream->lock); |
2000 | stream_put(stream); | |
f7079f67 DG |
2001 | } |
2002 | rcu_read_unlock(); | |
2003 | ||
53efb85a | 2004 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
2005 | /* All good, send back reply. */ |
2006 | reply.ret_code = htobe32(is_data_inflight); | |
2007 | ||
5312a3ed JG |
2008 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2009 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2010 | ERR("Failed to send \"end data pending\" command reply (ret = %zd)", | |
2011 | send_ret); | |
2012 | ret = -1; | |
2013 | } else { | |
2014 | ret = 0; | |
f7079f67 DG |
2015 | } |
2016 | ||
2017 | end_no_session: | |
2018 | return ret; | |
2019 | } | |
2020 | ||
1c20f0e2 JD |
2021 | /* |
2022 | * Receive an index for a specific stream. | |
2023 | * | |
2024 | * Return 0 on success else a negative value. | |
2025 | */ | |
5312a3ed JG |
2026 | static int relay_recv_index(const struct lttcomm_relayd_hdr *recv_hdr, |
2027 | struct relay_connection *conn, | |
2028 | const struct lttng_buffer_view *payload) | |
1c20f0e2 | 2029 | { |
5312a3ed JG |
2030 | int ret; |
2031 | ssize_t send_ret; | |
58eb9381 | 2032 | struct relay_session *session = conn->session; |
1c20f0e2 | 2033 | struct lttcomm_relayd_index index_info; |
1c20f0e2 JD |
2034 | struct lttcomm_relayd_generic_reply reply; |
2035 | struct relay_stream *stream; | |
f8f3885c | 2036 | size_t msg_len; |
1c20f0e2 | 2037 | |
58eb9381 | 2038 | assert(conn); |
1c20f0e2 JD |
2039 | |
2040 | DBG("Relay receiving index"); | |
2041 | ||
5312a3ed | 2042 | if (!session || !conn->version_check_done) { |
1c20f0e2 JD |
2043 | ERR("Trying to close a stream before version check"); |
2044 | ret = -1; | |
2045 | goto end_no_session; | |
2046 | } | |
2047 | ||
f8f3885c MD |
2048 | msg_len = lttcomm_relayd_index_len( |
2049 | lttng_to_index_major(conn->major, conn->minor), | |
2050 | lttng_to_index_minor(conn->major, conn->minor)); | |
5312a3ed JG |
2051 | if (payload->size < msg_len) { |
2052 | ERR("Unexpected payload size in \"relay_recv_index\": expected >= %zu bytes, got %zu bytes", | |
2053 | msg_len, payload->size); | |
1c20f0e2 JD |
2054 | ret = -1; |
2055 | goto end_no_session; | |
2056 | } | |
5312a3ed JG |
2057 | memcpy(&index_info, payload->data, msg_len); |
2058 | index_info.relay_stream_id = be64toh(index_info.relay_stream_id); | |
2059 | index_info.net_seq_num = be64toh(index_info.net_seq_num); | |
2060 | index_info.packet_size = be64toh(index_info.packet_size); | |
2061 | index_info.content_size = be64toh(index_info.content_size); | |
2062 | index_info.timestamp_begin = be64toh(index_info.timestamp_begin); | |
2063 | index_info.timestamp_end = be64toh(index_info.timestamp_end); | |
2064 | index_info.events_discarded = be64toh(index_info.events_discarded); | |
2065 | index_info.stream_id = be64toh(index_info.stream_id); | |
81df238b JR |
2066 | |
2067 | if (conn->minor >= 8) { | |
2068 | index_info.stream_instance_id = | |
2069 | be64toh(index_info.stream_instance_id); | |
2070 | index_info.packet_seq_num = be64toh(index_info.packet_seq_num); | |
2071 | } | |
5312a3ed JG |
2072 | |
2073 | stream = stream_get_by_id(index_info.relay_stream_id); | |
1c20f0e2 | 2074 | if (!stream) { |
7591bab1 | 2075 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 2076 | ret = -1; |
7591bab1 | 2077 | goto end; |
1c20f0e2 | 2078 | } |
d3e2ba59 | 2079 | |
c35f9726 JG |
2080 | pthread_mutex_lock(&stream->lock); |
2081 | ret = stream_add_index(stream, &index_info); | |
2082 | pthread_mutex_unlock(&stream->lock); | |
2083 | if (ret) { | |
7591bab1 MD |
2084 | goto end_stream_put; |
2085 | } | |
1c20f0e2 | 2086 | |
7591bab1 | 2087 | end_stream_put: |
7591bab1 | 2088 | stream_put(stream); |
7591bab1 | 2089 | end: |
53efb85a | 2090 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
2091 | if (ret < 0) { |
2092 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
2093 | } else { | |
2094 | reply.ret_code = htobe32(LTTNG_OK); | |
2095 | } | |
58eb9381 | 2096 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed JG |
2097 | if (send_ret < (ssize_t) sizeof(reply)) { |
2098 | ERR("Failed to send \"recv index\" command reply (ret = %zd)", send_ret); | |
2099 | ret = -1; | |
1c20f0e2 JD |
2100 | } |
2101 | ||
2102 | end_no_session: | |
2103 | return ret; | |
2104 | } | |
2105 | ||
a4baae1b JD |
2106 | /* |
2107 | * Receive the streams_sent message. | |
2108 | * | |
2109 | * Return 0 on success else a negative value. | |
2110 | */ | |
5312a3ed JG |
2111 | static int relay_streams_sent(const struct lttcomm_relayd_hdr *recv_hdr, |
2112 | struct relay_connection *conn, | |
2113 | const struct lttng_buffer_view *payload) | |
a4baae1b | 2114 | { |
5312a3ed JG |
2115 | int ret; |
2116 | ssize_t send_ret; | |
a4baae1b JD |
2117 | struct lttcomm_relayd_generic_reply reply; |
2118 | ||
58eb9381 | 2119 | assert(conn); |
a4baae1b JD |
2120 | |
2121 | DBG("Relay receiving streams_sent"); | |
2122 | ||
5312a3ed | 2123 | if (!conn->session || !conn->version_check_done) { |
a4baae1b JD |
2124 | ERR("Trying to close a stream before version check"); |
2125 | ret = -1; | |
2126 | goto end_no_session; | |
2127 | } | |
2128 | ||
2129 | /* | |
7591bab1 MD |
2130 | * Publish every pending stream in the connection recv list which are |
2131 | * now ready to be used by the viewer. | |
4a9daf17 | 2132 | */ |
7591bab1 | 2133 | publish_connection_local_streams(conn); |
4a9daf17 | 2134 | |
53efb85a | 2135 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2136 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2137 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed JG |
2138 | if (send_ret < (ssize_t) sizeof(reply)) { |
2139 | ERR("Failed to send \"streams sent\" command reply (ret = %zd)", | |
2140 | send_ret); | |
2141 | ret = -1; | |
a4baae1b JD |
2142 | } else { |
2143 | /* Success. */ | |
2144 | ret = 0; | |
2145 | } | |
2146 | ||
2147 | end_no_session: | |
2148 | return ret; | |
2149 | } | |
2150 | ||
d3ecc550 | 2151 | /* |
c35f9726 JG |
2152 | * relay_rotate_session_stream: rotate a stream to a new tracefile for the |
2153 | * session rotation feature (not the tracefile rotation feature). | |
d3ecc550 | 2154 | */ |
c35f9726 JG |
2155 | static int relay_rotate_session_streams( |
2156 | const struct lttcomm_relayd_hdr *recv_hdr, | |
5312a3ed JG |
2157 | struct relay_connection *conn, |
2158 | const struct lttng_buffer_view *payload) | |
d3ecc550 | 2159 | { |
30b9d5ab | 2160 | int ret = 0; |
c35f9726 | 2161 | uint32_t i; |
5312a3ed | 2162 | ssize_t send_ret; |
c35f9726 | 2163 | enum lttng_error_code reply_code = LTTNG_ERR_UNK; |
d3ecc550 | 2164 | struct relay_session *session = conn->session; |
c35f9726 JG |
2165 | struct lttcomm_relayd_rotate_streams rotate_streams; |
2166 | struct lttcomm_relayd_generic_reply reply = {}; | |
2167 | struct relay_stream *stream = NULL; | |
2168 | const size_t header_len = sizeof(struct lttcomm_relayd_rotate_streams); | |
2169 | struct lttng_trace_chunk *next_trace_chunk = NULL; | |
2170 | struct lttng_buffer_view stream_positions; | |
70626904 JG |
2171 | char chunk_id_buf[MAX_INT_DEC_LEN(uint64_t)]; |
2172 | const char *chunk_id_str = "none"; | |
d3ecc550 | 2173 | |
d3ecc550 JD |
2174 | if (!session || !conn->version_check_done) { |
2175 | ERR("Trying to rotate a stream before version check"); | |
2176 | ret = -1; | |
2177 | goto end_no_reply; | |
2178 | } | |
2179 | ||
2180 | if (session->major == 2 && session->minor < 11) { | |
2181 | ERR("Unsupported feature before 2.11"); | |
2182 | ret = -1; | |
2183 | goto end_no_reply; | |
2184 | } | |
2185 | ||
5312a3ed JG |
2186 | if (payload->size < header_len) { |
2187 | ERR("Unexpected payload size in \"relay_rotate_session_stream\": expected >= %zu bytes, got %zu bytes", | |
2188 | header_len, payload->size); | |
d3ecc550 JD |
2189 | ret = -1; |
2190 | goto end_no_reply; | |
2191 | } | |
2192 | ||
c35f9726 | 2193 | memcpy(&rotate_streams, payload->data, header_len); |
5312a3ed | 2194 | |
c35f9726 JG |
2195 | /* Convert header to host endianness. */ |
2196 | rotate_streams = (typeof(rotate_streams)) { | |
2197 | .stream_count = be32toh(rotate_streams.stream_count), | |
2198 | .new_chunk_id = (typeof(rotate_streams.new_chunk_id)) { | |
2199 | .is_set = !!rotate_streams.new_chunk_id.is_set, | |
2200 | .value = be64toh(rotate_streams.new_chunk_id.value), | |
2201 | } | |
2202 | }; | |
d3ecc550 | 2203 | |
c35f9726 JG |
2204 | if (rotate_streams.new_chunk_id.is_set) { |
2205 | /* | |
2206 | * Retrieve the trace chunk the stream must transition to. As | |
2207 | * per the protocol, this chunk should have been created | |
2208 | * before this command is received. | |
2209 | */ | |
2210 | next_trace_chunk = sessiond_trace_chunk_registry_get_chunk( | |
2211 | sessiond_trace_chunk_registry, | |
2212 | session->sessiond_uuid, session->id, | |
2213 | rotate_streams.new_chunk_id.value); | |
2214 | if (!next_trace_chunk) { | |
2215 | char uuid_str[UUID_STR_LEN]; | |
2216 | ||
2217 | lttng_uuid_to_str(session->sessiond_uuid, uuid_str); | |
2218 | ERR("Unknown next trace chunk in ROTATE_STREAMS command: sessiond_uuid = {%s}, session_id = %" PRIu64 | |
2219 | ", trace_chunk_id = %" PRIu64, | |
2220 | uuid_str, session->id, | |
2221 | rotate_streams.new_chunk_id.value); | |
2222 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2223 | ret = -1; | |
2224 | goto end; | |
2225 | } | |
70626904 JG |
2226 | |
2227 | ret = snprintf(chunk_id_buf, sizeof(chunk_id_buf), "%" PRIu64, | |
2228 | rotate_streams.new_chunk_id.value); | |
2229 | if (ret < 0 || ret >= sizeof(chunk_id_buf)) { | |
2230 | chunk_id_str = "formatting error"; | |
2231 | } else { | |
2232 | chunk_id_str = chunk_id_buf; | |
2233 | } | |
ecd1a12f | 2234 | session->has_rotated = true; |
d3ecc550 JD |
2235 | } |
2236 | ||
70626904 JG |
2237 | DBG("Rotate %" PRIu32 " streams of session \"%s\" to chunk \"%s\"", |
2238 | rotate_streams.stream_count, session->session_name, | |
2239 | chunk_id_str); | |
2240 | ||
c35f9726 JG |
2241 | stream_positions = lttng_buffer_view_from_view(payload, |
2242 | sizeof(rotate_streams), -1); | |
2243 | if (!stream_positions.data || | |
2244 | stream_positions.size < | |
2245 | (rotate_streams.stream_count * | |
2246 | sizeof(struct lttcomm_relayd_stream_rotation_position))) { | |
2247 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
d3ecc550 | 2248 | ret = -1; |
5312a3ed | 2249 | goto end; |
d3ecc550 JD |
2250 | } |
2251 | ||
c35f9726 JG |
2252 | for (i = 0; i < rotate_streams.stream_count; i++) { |
2253 | struct lttcomm_relayd_stream_rotation_position *position_comm = | |
2254 | &((typeof(position_comm)) stream_positions.data)[i]; | |
2255 | const struct lttcomm_relayd_stream_rotation_position pos = { | |
2256 | .stream_id = be64toh(position_comm->stream_id), | |
2257 | .rotate_at_seq_num = be64toh( | |
2258 | position_comm->rotate_at_seq_num), | |
2259 | }; | |
5312a3ed | 2260 | |
c35f9726 JG |
2261 | stream = stream_get_by_id(pos.stream_id); |
2262 | if (!stream) { | |
2263 | reply_code = LTTNG_ERR_INVALID; | |
2264 | ret = -1; | |
2265 | goto end; | |
c6db3843 JG |
2266 | } |
2267 | ||
c35f9726 JG |
2268 | pthread_mutex_lock(&stream->lock); |
2269 | ret = stream_set_pending_rotation(stream, next_trace_chunk, | |
2270 | pos.rotate_at_seq_num); | |
2271 | pthread_mutex_unlock(&stream->lock); | |
2272 | if (ret) { | |
2273 | reply_code = LTTNG_ERR_FILE_CREATION_ERROR; | |
2274 | goto end; | |
c6db3843 | 2275 | } |
c35f9726 JG |
2276 | |
2277 | stream_put(stream); | |
2278 | stream = NULL; | |
d3ecc550 JD |
2279 | } |
2280 | ||
c35f9726 | 2281 | reply_code = LTTNG_OK; |
d3ecc550 | 2282 | end: |
c35f9726 JG |
2283 | if (stream) { |
2284 | stream_put(stream); | |
d3ecc550 | 2285 | } |
c35f9726 JG |
2286 | |
2287 | reply.ret_code = htobe32((uint32_t) reply_code); | |
d3ecc550 JD |
2288 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
2289 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed JG |
2290 | if (send_ret < (ssize_t) sizeof(reply)) { |
2291 | ERR("Failed to send \"rotate session stream\" command reply (ret = %zd)", | |
2292 | send_ret); | |
2293 | ret = -1; | |
d3ecc550 JD |
2294 | } |
2295 | ||
2fdaeb7e | 2296 | ret = 0; |
d3ecc550 | 2297 | end_no_reply: |
c35f9726 | 2298 | lttng_trace_chunk_put(next_trace_chunk); |
d3ecc550 JD |
2299 | return ret; |
2300 | } | |
2301 | ||
e5add6d0 | 2302 | |
e5add6d0 JG |
2303 | |
2304 | /* | |
2305 | * relay_create_trace_chunk: create a new trace chunk | |
2306 | */ | |
2307 | static int relay_create_trace_chunk(const struct lttcomm_relayd_hdr *recv_hdr, | |
2308 | struct relay_connection *conn, | |
2309 | const struct lttng_buffer_view *payload) | |
2310 | { | |
2311 | int ret = 0; | |
2312 | ssize_t send_ret; | |
2313 | struct relay_session *session = conn->session; | |
2314 | struct lttcomm_relayd_create_trace_chunk *msg; | |
2315 | struct lttcomm_relayd_generic_reply reply = {}; | |
2316 | struct lttng_buffer_view header_view; | |
2317 | struct lttng_buffer_view chunk_name_view; | |
2318 | struct lttng_trace_chunk *chunk = NULL, *published_chunk = NULL; | |
2319 | enum lttng_error_code reply_code = LTTNG_OK; | |
2320 | enum lttng_trace_chunk_status chunk_status; | |
2321 | struct lttng_directory_handle session_output; | |
2322 | ||
2323 | if (!session || !conn->version_check_done) { | |
2324 | ERR("Trying to create a trace chunk before version check"); | |
2325 | ret = -1; | |
2326 | goto end_no_reply; | |
2327 | } | |
2328 | ||
2329 | if (session->major == 2 && session->minor < 11) { | |
2330 | ERR("Chunk creation command is unsupported before 2.11"); | |
2331 | ret = -1; | |
2332 | goto end_no_reply; | |
2333 | } | |
2334 | ||
2335 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
2336 | if (!header_view.data) { | |
2337 | ERR("Failed to receive payload of chunk creation command"); | |
2338 | ret = -1; | |
2339 | goto end_no_reply; | |
2340 | } | |
2341 | ||
2342 | /* Convert to host endianness. */ | |
2343 | msg = (typeof(msg)) header_view.data; | |
2344 | msg->chunk_id = be64toh(msg->chunk_id); | |
2345 | msg->creation_timestamp = be64toh(msg->creation_timestamp); | |
2346 | msg->override_name_length = be32toh(msg->override_name_length); | |
2347 | ||
2348 | chunk = lttng_trace_chunk_create( | |
2349 | msg->chunk_id, msg->creation_timestamp); | |
2350 | if (!chunk) { | |
2351 | ERR("Failed to create trace chunk in trace chunk creation command"); | |
2352 | ret = -1; | |
2353 | reply_code = LTTNG_ERR_NOMEM; | |
2354 | goto end; | |
2355 | } | |
2356 | ||
2357 | if (msg->override_name_length) { | |
2358 | const char *name; | |
2359 | ||
2360 | chunk_name_view = lttng_buffer_view_from_view(payload, | |
2361 | sizeof(*msg), | |
2362 | msg->override_name_length); | |
2363 | name = chunk_name_view.data; | |
2364 | if (!name || name[msg->override_name_length - 1]) { | |
2365 | ERR("Failed to receive payload of chunk creation command"); | |
2366 | ret = -1; | |
2367 | reply_code = LTTNG_ERR_INVALID; | |
2368 | goto end; | |
2369 | } | |
2370 | ||
2371 | chunk_status = lttng_trace_chunk_override_name( | |
2372 | chunk, chunk_name_view.data); | |
2373 | switch (chunk_status) { | |
2374 | case LTTNG_TRACE_CHUNK_STATUS_OK: | |
2375 | break; | |
2376 | case LTTNG_TRACE_CHUNK_STATUS_INVALID_ARGUMENT: | |
2377 | ERR("Failed to set the name of new trace chunk in trace chunk creation command (invalid name)"); | |
2378 | reply_code = LTTNG_ERR_INVALID; | |
2379 | ret = -1; | |
2380 | goto end; | |
2381 | default: | |
2382 | ERR("Failed to set the name of new trace chunk in trace chunk creation command (unknown error)"); | |
2383 | reply_code = LTTNG_ERR_UNK; | |
2384 | ret = -1; | |
2385 | goto end; | |
2386 | } | |
2387 | } | |
2388 | ||
e5add6d0 JG |
2389 | chunk_status = lttng_trace_chunk_set_credentials_current_user(chunk); |
2390 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2391 | reply_code = LTTNG_ERR_UNK; | |
2392 | ret = -1; | |
2393 | goto end; | |
2394 | } | |
2395 | ||
0ccc0411 JG |
2396 | ret = session_init_output_directory_handle( |
2397 | conn->session, &session_output); | |
2398 | if (ret) { | |
2399 | reply_code = LTTNG_ERR_CREATE_DIR_FAIL; | |
2400 | goto end; | |
2401 | } | |
e5add6d0 | 2402 | chunk_status = lttng_trace_chunk_set_as_owner(chunk, &session_output); |
0ccc0411 | 2403 | lttng_directory_handle_fini(&session_output); |
e5add6d0 JG |
2404 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
2405 | reply_code = LTTNG_ERR_UNK; | |
2406 | ret = -1; | |
2407 | goto end; | |
2408 | } | |
2409 | ||
2410 | published_chunk = sessiond_trace_chunk_registry_publish_chunk( | |
2411 | sessiond_trace_chunk_registry, | |
2412 | conn->session->sessiond_uuid, | |
2413 | conn->session->id, | |
2414 | chunk); | |
2415 | if (!published_chunk) { | |
2416 | char uuid_str[UUID_STR_LEN]; | |
2417 | ||
2418 | lttng_uuid_to_str(conn->session->sessiond_uuid, uuid_str); | |
2419 | ERR("Failed to publish chunk: sessiond_uuid = %s, session_id = %" PRIu64 ", chunk_id = %" PRIu64, | |
2420 | uuid_str, | |
2421 | conn->session->id, | |
2422 | msg->chunk_id); | |
2423 | ret = -1; | |
2424 | reply_code = LTTNG_ERR_NOMEM; | |
2425 | goto end; | |
2426 | } | |
2427 | ||
2428 | pthread_mutex_lock(&conn->session->lock); | |
62bad3bf JG |
2429 | if (conn->session->pending_closure_trace_chunk) { |
2430 | /* | |
2431 | * Invalid; this means a second create_trace_chunk command was | |
2432 | * received before a close_trace_chunk. | |
2433 | */ | |
2434 | ERR("Invalid trace chunk close command received; a trace chunk is already waiting for a trace chunk close command"); | |
2435 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2436 | ret = -1; | |
2437 | goto end_unlock_session; | |
2438 | } | |
2439 | conn->session->pending_closure_trace_chunk = | |
2440 | conn->session->current_trace_chunk; | |
e5add6d0 | 2441 | conn->session->current_trace_chunk = published_chunk; |
e5add6d0 | 2442 | published_chunk = NULL; |
62bad3bf | 2443 | end_unlock_session: |
c35f9726 | 2444 | pthread_mutex_unlock(&conn->session->lock); |
e5add6d0 JG |
2445 | end: |
2446 | reply.ret_code = htobe32((uint32_t) reply_code); | |
2447 | send_ret = conn->sock->ops->sendmsg(conn->sock, | |
2448 | &reply, | |
2449 | sizeof(struct lttcomm_relayd_generic_reply), | |
2450 | 0); | |
2451 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2452 | ERR("Failed to send \"create trace chunk\" command reply (ret = %zd)", | |
2453 | send_ret); | |
2454 | ret = -1; | |
2455 | } | |
2456 | end_no_reply: | |
2457 | lttng_trace_chunk_put(chunk); | |
2458 | lttng_trace_chunk_put(published_chunk); | |
e5add6d0 JG |
2459 | return ret; |
2460 | } | |
2461 | ||
bbc4768c JG |
2462 | /* |
2463 | * relay_close_trace_chunk: close a trace chunk | |
2464 | */ | |
2465 | static int relay_close_trace_chunk(const struct lttcomm_relayd_hdr *recv_hdr, | |
2466 | struct relay_connection *conn, | |
2467 | const struct lttng_buffer_view *payload) | |
2468 | { | |
2469 | int ret = 0; | |
2470 | ssize_t send_ret; | |
2471 | struct relay_session *session = conn->session; | |
2472 | struct lttcomm_relayd_close_trace_chunk *msg; | |
ecd1a12f | 2473 | struct lttcomm_relayd_close_trace_chunk_reply reply = {}; |
bbc4768c JG |
2474 | struct lttng_buffer_view header_view; |
2475 | struct lttng_trace_chunk *chunk = NULL; | |
2476 | enum lttng_error_code reply_code = LTTNG_OK; | |
2477 | enum lttng_trace_chunk_status chunk_status; | |
2478 | uint64_t chunk_id; | |
c35f9726 | 2479 | LTTNG_OPTIONAL(enum lttng_trace_chunk_command_type) close_command = {}; |
bbc4768c | 2480 | time_t close_timestamp; |
ecd1a12f MD |
2481 | char closed_trace_chunk_path[LTTNG_PATH_MAX]; |
2482 | size_t path_length = 0; | |
2483 | const char *chunk_name = NULL; | |
2484 | struct lttng_dynamic_buffer reply_payload; | |
2485 | ||
2486 | lttng_dynamic_buffer_init(&reply_payload); | |
bbc4768c JG |
2487 | |
2488 | if (!session || !conn->version_check_done) { | |
2489 | ERR("Trying to close a trace chunk before version check"); | |
2490 | ret = -1; | |
2491 | goto end_no_reply; | |
2492 | } | |
2493 | ||
2494 | if (session->major == 2 && session->minor < 11) { | |
2495 | ERR("Chunk close command is unsupported before 2.11"); | |
2496 | ret = -1; | |
2497 | goto end_no_reply; | |
2498 | } | |
2499 | ||
2500 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
2501 | if (!header_view.data) { | |
2502 | ERR("Failed to receive payload of chunk close command"); | |
2503 | ret = -1; | |
2504 | goto end_no_reply; | |
2505 | } | |
2506 | ||
2507 | /* Convert to host endianness. */ | |
2508 | msg = (typeof(msg)) header_view.data; | |
2509 | chunk_id = be64toh(msg->chunk_id); | |
2510 | close_timestamp = (time_t) be64toh(msg->close_timestamp); | |
2511 | close_command = (typeof(close_command)){ | |
2512 | .value = be32toh(msg->close_command.value), | |
2513 | .is_set = msg->close_command.is_set, | |
2514 | }; | |
2515 | ||
2516 | chunk = sessiond_trace_chunk_registry_get_chunk( | |
2517 | sessiond_trace_chunk_registry, | |
2518 | conn->session->sessiond_uuid, | |
2519 | conn->session->id, | |
2520 | chunk_id); | |
2521 | if (!chunk) { | |
2522 | char uuid_str[UUID_STR_LEN]; | |
2523 | ||
2524 | lttng_uuid_to_str(conn->session->sessiond_uuid, uuid_str); | |
2525 | ERR("Failed to find chunk to close: sessiond_uuid = %s, session_id = %" PRIu64 ", chunk_id = %" PRIu64, | |
2526 | uuid_str, | |
2527 | conn->session->id, | |
2528 | msg->chunk_id); | |
2529 | ret = -1; | |
2530 | reply_code = LTTNG_ERR_NOMEM; | |
2531 | goto end; | |
2532 | } | |
2533 | ||
62bad3bf JG |
2534 | pthread_mutex_lock(&session->lock); |
2535 | if (session->pending_closure_trace_chunk && | |
2536 | session->pending_closure_trace_chunk != chunk) { | |
2537 | ERR("Trace chunk close command for session \"%s\" does not target the trace chunk pending closure", | |
2538 | session->session_name); | |
2539 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2540 | ret = -1; | |
2541 | goto end_unlock_session; | |
2542 | } | |
2543 | ||
bbc4768c JG |
2544 | chunk_status = lttng_trace_chunk_set_close_timestamp( |
2545 | chunk, close_timestamp); | |
2546 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2547 | ERR("Failed to set trace chunk close timestamp"); | |
2548 | ret = -1; | |
2549 | reply_code = LTTNG_ERR_UNK; | |
62bad3bf | 2550 | goto end_unlock_session; |
bbc4768c JG |
2551 | } |
2552 | ||
2553 | if (close_command.is_set) { | |
2554 | chunk_status = lttng_trace_chunk_set_close_command( | |
2555 | chunk, close_command.value); | |
2556 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2557 | ret = -1; | |
2558 | reply_code = LTTNG_ERR_INVALID; | |
62bad3bf | 2559 | goto end_unlock_session; |
bbc4768c JG |
2560 | } |
2561 | } | |
ecd1a12f MD |
2562 | chunk_status = lttng_trace_chunk_get_name(chunk, &chunk_name, NULL); |
2563 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2564 | ERR("Failed to get chunk name"); | |
2565 | ret = -1; | |
2566 | reply_code = LTTNG_ERR_UNK; | |
2567 | goto end_unlock_session; | |
2568 | } | |
2569 | if (!session->has_rotated && !session->snapshot) { | |
2570 | ret = lttng_strncpy(closed_trace_chunk_path, | |
2571 | session->output_path, | |
2572 | sizeof(closed_trace_chunk_path)); | |
2573 | if (ret) { | |
2574 | ERR("Failed to send trace chunk path: path length of %zu bytes exceeds the maximal allowed length of %zu bytes", | |
2575 | strlen(session->output_path), | |
2576 | sizeof(closed_trace_chunk_path)); | |
2577 | reply_code = LTTNG_ERR_NOMEM; | |
2578 | ret = -1; | |
2579 | goto end_unlock_session; | |
2580 | } | |
2581 | } else { | |
2582 | if (session->snapshot) { | |
2583 | ret = snprintf(closed_trace_chunk_path, | |
2584 | sizeof(closed_trace_chunk_path), | |
2585 | "%s/%s", session->output_path, | |
2586 | chunk_name); | |
2587 | } else { | |
2588 | ret = snprintf(closed_trace_chunk_path, | |
2589 | sizeof(closed_trace_chunk_path), | |
2590 | "%s/" DEFAULT_ARCHIVED_TRACE_CHUNKS_DIRECTORY | |
2591 | "/%s", | |
2592 | session->output_path, chunk_name); | |
2593 | } | |
2594 | if (ret < 0 || ret == sizeof(closed_trace_chunk_path)) { | |
2595 | ERR("Failed to format closed trace chunk resulting path"); | |
2596 | reply_code = ret < 0 ? LTTNG_ERR_UNK : LTTNG_ERR_NOMEM; | |
2597 | ret = -1; | |
2598 | goto end_unlock_session; | |
2599 | } | |
2600 | } | |
2601 | DBG("Reply chunk path on close: %s", closed_trace_chunk_path); | |
2602 | path_length = strlen(closed_trace_chunk_path) + 1; | |
2603 | if (path_length > UINT32_MAX) { | |
2604 | ERR("Closed trace chunk path exceeds the maximal length allowed by the protocol"); | |
2605 | ret = -1; | |
2606 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2607 | goto end_unlock_session; | |
2608 | } | |
bbc4768c | 2609 | |
c35f9726 JG |
2610 | if (session->current_trace_chunk == chunk) { |
2611 | /* | |
2612 | * After a trace chunk close command, no new streams | |
2613 | * referencing the chunk may be created. Hence, on the | |
2614 | * event that no new trace chunk have been created for | |
2615 | * the session, the reference to the current trace chunk | |
2616 | * is released in order to allow it to be reclaimed when | |
2617 | * the last stream releases its reference to it. | |
2618 | */ | |
2619 | lttng_trace_chunk_put(session->current_trace_chunk); | |
2620 | session->current_trace_chunk = NULL; | |
2621 | } | |
62bad3bf JG |
2622 | lttng_trace_chunk_put(session->pending_closure_trace_chunk); |
2623 | session->pending_closure_trace_chunk = NULL; | |
2624 | end_unlock_session: | |
c35f9726 JG |
2625 | pthread_mutex_unlock(&session->lock); |
2626 | ||
bbc4768c | 2627 | end: |
ecd1a12f MD |
2628 | reply.generic.ret_code = htobe32((uint32_t) reply_code); |
2629 | reply.path_length = htobe32((uint32_t) path_length); | |
2630 | ret = lttng_dynamic_buffer_append( | |
2631 | &reply_payload, &reply, sizeof(reply)); | |
2632 | if (ret) { | |
2633 | ERR("Failed to append \"close trace chunk\" command reply header to payload buffer"); | |
2634 | goto end_no_reply; | |
2635 | } | |
2636 | ||
2637 | if (reply_code == LTTNG_OK) { | |
2638 | ret = lttng_dynamic_buffer_append(&reply_payload, | |
2639 | closed_trace_chunk_path, path_length); | |
2640 | if (ret) { | |
2641 | ERR("Failed to append \"close trace chunk\" command reply path to payload buffer"); | |
2642 | goto end_no_reply; | |
2643 | } | |
2644 | } | |
2645 | ||
bbc4768c | 2646 | send_ret = conn->sock->ops->sendmsg(conn->sock, |
ecd1a12f MD |
2647 | reply_payload.data, |
2648 | reply_payload.size, | |
bbc4768c | 2649 | 0); |
ecd1a12f MD |
2650 | if (send_ret < reply_payload.size) { |
2651 | ERR("Failed to send \"close trace chunk\" command reply of %zu bytes (ret = %zd)", | |
2652 | reply_payload.size, send_ret); | |
bbc4768c | 2653 | ret = -1; |
ecd1a12f | 2654 | goto end_no_reply; |
bbc4768c JG |
2655 | } |
2656 | end_no_reply: | |
2657 | lttng_trace_chunk_put(chunk); | |
ecd1a12f | 2658 | lttng_dynamic_buffer_reset(&reply_payload); |
bbc4768c JG |
2659 | return ret; |
2660 | } | |
2661 | ||
c35f9726 JG |
2662 | /* |
2663 | * relay_trace_chunk_exists: check if a trace chunk exists | |
2664 | */ | |
2665 | static int relay_trace_chunk_exists(const struct lttcomm_relayd_hdr *recv_hdr, | |
2666 | struct relay_connection *conn, | |
2667 | const struct lttng_buffer_view *payload) | |
2668 | { | |
2669 | int ret = 0; | |
2670 | ssize_t send_ret; | |
2671 | struct relay_session *session = conn->session; | |
2672 | struct lttcomm_relayd_trace_chunk_exists *msg; | |
2673 | struct lttcomm_relayd_trace_chunk_exists_reply reply = {}; | |
2674 | struct lttng_buffer_view header_view; | |
2675 | struct lttng_trace_chunk *chunk = NULL; | |
2676 | uint64_t chunk_id; | |
2677 | ||
2678 | if (!session || !conn->version_check_done) { | |
2679 | ERR("Trying to close a trace chunk before version check"); | |
2680 | ret = -1; | |
2681 | goto end_no_reply; | |
2682 | } | |
2683 | ||
2684 | if (session->major == 2 && session->minor < 11) { | |
2685 | ERR("Chunk close command is unsupported before 2.11"); | |
2686 | ret = -1; | |
2687 | goto end_no_reply; | |
2688 | } | |
2689 | ||
2690 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
2691 | if (!header_view.data) { | |
2692 | ERR("Failed to receive payload of chunk close command"); | |
2693 | ret = -1; | |
2694 | goto end_no_reply; | |
2695 | } | |
2696 | ||
2697 | /* Convert to host endianness. */ | |
2698 | msg = (typeof(msg)) header_view.data; | |
2699 | chunk_id = be64toh(msg->chunk_id); | |
2700 | ||
2701 | chunk = sessiond_trace_chunk_registry_get_chunk( | |
2702 | sessiond_trace_chunk_registry, | |
2703 | conn->session->sessiond_uuid, | |
2704 | conn->session->id, | |
2705 | chunk_id); | |
2706 | ||
2707 | reply = (typeof(reply)) { | |
2708 | .generic.ret_code = htobe32((uint32_t) LTTNG_OK), | |
2709 | .trace_chunk_exists = !!chunk, | |
2710 | }; | |
2711 | send_ret = conn->sock->ops->sendmsg(conn->sock, | |
2712 | &reply, sizeof(reply), 0); | |
2713 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2714 | ERR("Failed to send \"create trace chunk\" command reply (ret = %zd)", | |
2715 | send_ret); | |
2716 | ret = -1; | |
2717 | } | |
2718 | end_no_reply: | |
2719 | lttng_trace_chunk_put(chunk); | |
2720 | return ret; | |
2721 | } | |
2722 | ||
5312a3ed JG |
2723 | #define DBG_CMD(cmd_name, conn) \ |
2724 | DBG3("Processing \"%s\" command for socket %i", cmd_name, conn->sock->fd); | |
2725 | ||
2726 | static int relay_process_control_command(struct relay_connection *conn, | |
2727 | const struct lttcomm_relayd_hdr *header, | |
2728 | const struct lttng_buffer_view *payload) | |
b8aa1682 JD |
2729 | { |
2730 | int ret = 0; | |
2731 | ||
5312a3ed | 2732 | switch (header->cmd) { |
b8aa1682 | 2733 | case RELAYD_CREATE_SESSION: |
5312a3ed JG |
2734 | DBG_CMD("RELAYD_CREATE_SESSION", conn); |
2735 | ret = relay_create_session(header, conn, payload); | |
b8aa1682 | 2736 | break; |
b8aa1682 | 2737 | case RELAYD_ADD_STREAM: |
5312a3ed JG |
2738 | DBG_CMD("RELAYD_ADD_STREAM", conn); |
2739 | ret = relay_add_stream(header, conn, payload); | |
b8aa1682 JD |
2740 | break; |
2741 | case RELAYD_START_DATA: | |
5312a3ed JG |
2742 | DBG_CMD("RELAYD_START_DATA", conn); |
2743 | ret = relay_start(header, conn, payload); | |
b8aa1682 JD |
2744 | break; |
2745 | case RELAYD_SEND_METADATA: | |
5312a3ed JG |
2746 | DBG_CMD("RELAYD_SEND_METADATA", conn); |
2747 | ret = relay_recv_metadata(header, conn, payload); | |
b8aa1682 JD |
2748 | break; |
2749 | case RELAYD_VERSION: | |
5312a3ed JG |
2750 | DBG_CMD("RELAYD_VERSION", conn); |
2751 | ret = relay_send_version(header, conn, payload); | |
b8aa1682 | 2752 | break; |
173af62f | 2753 | case RELAYD_CLOSE_STREAM: |
5312a3ed JG |
2754 | DBG_CMD("RELAYD_CLOSE_STREAM", conn); |
2755 | ret = relay_close_stream(header, conn, payload); | |
173af62f | 2756 | break; |
6d805429 | 2757 | case RELAYD_DATA_PENDING: |
5312a3ed JG |
2758 | DBG_CMD("RELAYD_DATA_PENDING", conn); |
2759 | ret = relay_data_pending(header, conn, payload); | |
c8f59ee5 DG |
2760 | break; |
2761 | case RELAYD_QUIESCENT_CONTROL: | |
5312a3ed JG |
2762 | DBG_CMD("RELAYD_QUIESCENT_CONTROL", conn); |
2763 | ret = relay_quiescent_control(header, conn, payload); | |
c8f59ee5 | 2764 | break; |
f7079f67 | 2765 | case RELAYD_BEGIN_DATA_PENDING: |
5312a3ed JG |
2766 | DBG_CMD("RELAYD_BEGIN_DATA_PENDING", conn); |
2767 | ret = relay_begin_data_pending(header, conn, payload); | |
f7079f67 DG |
2768 | break; |
2769 | case RELAYD_END_DATA_PENDING: | |
5312a3ed JG |
2770 | DBG_CMD("RELAYD_END_DATA_PENDING", conn); |
2771 | ret = relay_end_data_pending(header, conn, payload); | |
f7079f67 | 2772 | break; |
1c20f0e2 | 2773 | case RELAYD_SEND_INDEX: |
5312a3ed JG |
2774 | DBG_CMD("RELAYD_SEND_INDEX", conn); |
2775 | ret = relay_recv_index(header, conn, payload); | |
1c20f0e2 | 2776 | break; |
a4baae1b | 2777 | case RELAYD_STREAMS_SENT: |
5312a3ed JG |
2778 | DBG_CMD("RELAYD_STREAMS_SENT", conn); |
2779 | ret = relay_streams_sent(header, conn, payload); | |
a4baae1b | 2780 | break; |
93ec662e | 2781 | case RELAYD_RESET_METADATA: |
5312a3ed JG |
2782 | DBG_CMD("RELAYD_RESET_METADATA", conn); |
2783 | ret = relay_reset_metadata(header, conn, payload); | |
93ec662e | 2784 | break; |
c35f9726 JG |
2785 | case RELAYD_ROTATE_STREAMS: |
2786 | DBG_CMD("RELAYD_ROTATE_STREAMS", conn); | |
2787 | ret = relay_rotate_session_streams(header, conn, payload); | |
d3ecc550 | 2788 | break; |
e5add6d0 JG |
2789 | case RELAYD_CREATE_TRACE_CHUNK: |
2790 | DBG_CMD("RELAYD_CREATE_TRACE_CHUNK", conn); | |
2791 | ret = relay_create_trace_chunk(header, conn, payload); | |
2792 | break; | |
bbc4768c JG |
2793 | case RELAYD_CLOSE_TRACE_CHUNK: |
2794 | DBG_CMD("RELAYD_CLOSE_TRACE_CHUNK", conn); | |
2795 | ret = relay_close_trace_chunk(header, conn, payload); | |
2796 | break; | |
c35f9726 JG |
2797 | case RELAYD_TRACE_CHUNK_EXISTS: |
2798 | DBG_CMD("RELAYD_TRACE_CHUNK_EXISTS", conn); | |
2799 | ret = relay_trace_chunk_exists(header, conn, payload); | |
2800 | break; | |
b8aa1682 JD |
2801 | case RELAYD_UPDATE_SYNC_INFO: |
2802 | default: | |
5312a3ed | 2803 | ERR("Received unknown command (%u)", header->cmd); |
58eb9381 | 2804 | relay_unknown_command(conn); |
b8aa1682 JD |
2805 | ret = -1; |
2806 | goto end; | |
2807 | } | |
2808 | ||
2809 | end: | |
2810 | return ret; | |
2811 | } | |
2812 | ||
5569b118 JG |
2813 | static enum relay_connection_status relay_process_control_receive_payload( |
2814 | struct relay_connection *conn) | |
5312a3ed JG |
2815 | { |
2816 | int ret = 0; | |
5569b118 | 2817 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
2818 | struct lttng_dynamic_buffer *reception_buffer = |
2819 | &conn->protocol.ctrl.reception_buffer; | |
2820 | struct ctrl_connection_state_receive_payload *state = | |
2821 | &conn->protocol.ctrl.state.receive_payload; | |
2822 | struct lttng_buffer_view payload_view; | |
2823 | ||
2824 | if (state->left_to_receive == 0) { | |
2825 | /* Short-circuit for payload-less commands. */ | |
2826 | goto reception_complete; | |
2827 | } | |
2828 | ||
2829 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2830 | reception_buffer->data + state->received, | |
2831 | state->left_to_receive, MSG_DONTWAIT); | |
2832 | if (ret < 0) { | |
5569b118 JG |
2833 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
2834 | PERROR("Unable to receive command payload on sock %d", | |
2835 | conn->sock->fd); | |
2836 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2837 | } | |
5312a3ed JG |
2838 | goto end; |
2839 | } else if (ret == 0) { | |
2840 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 2841 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
2842 | goto end; |
2843 | } | |
2844 | ||
2845 | assert(ret > 0); | |
2846 | assert(ret <= state->left_to_receive); | |
2847 | ||
2848 | state->left_to_receive -= ret; | |
2849 | state->received += ret; | |
2850 | ||
2851 | if (state->left_to_receive > 0) { | |
2852 | /* | |
2853 | * Can't transition to the protocol's next state, wait to | |
2854 | * receive the rest of the header. | |
2855 | */ | |
2856 | DBG3("Partial reception of control connection protocol payload (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2857 | state->received, state->left_to_receive, | |
2858 | conn->sock->fd); | |
5312a3ed JG |
2859 | goto end; |
2860 | } | |
2861 | ||
2862 | reception_complete: | |
2863 | DBG("Done receiving control command payload: fd = %i, payload size = %" PRIu64 " bytes", | |
2864 | conn->sock->fd, state->received); | |
2865 | /* | |
2866 | * The payload required to process the command has been received. | |
2867 | * A view to the reception buffer is forwarded to the various | |
2868 | * commands and the state of the control is reset on success. | |
2869 | * | |
2870 | * Commands are responsible for sending their reply to the peer. | |
2871 | */ | |
2872 | payload_view = lttng_buffer_view_from_dynamic_buffer(reception_buffer, | |
2873 | 0, -1); | |
2874 | ret = relay_process_control_command(conn, | |
2875 | &state->header, &payload_view); | |
2876 | if (ret < 0) { | |
5569b118 | 2877 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
2878 | goto end; |
2879 | } | |
2880 | ||
2881 | ret = connection_reset_protocol_state(conn); | |
5569b118 JG |
2882 | if (ret) { |
2883 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2884 | } | |
5312a3ed | 2885 | end: |
5569b118 | 2886 | return status; |
5312a3ed JG |
2887 | } |
2888 | ||
5569b118 JG |
2889 | static enum relay_connection_status relay_process_control_receive_header( |
2890 | struct relay_connection *conn) | |
5312a3ed JG |
2891 | { |
2892 | int ret = 0; | |
5569b118 | 2893 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
2894 | struct lttcomm_relayd_hdr header; |
2895 | struct lttng_dynamic_buffer *reception_buffer = | |
2896 | &conn->protocol.ctrl.reception_buffer; | |
2897 | struct ctrl_connection_state_receive_header *state = | |
2898 | &conn->protocol.ctrl.state.receive_header; | |
2899 | ||
2900 | assert(state->left_to_receive != 0); | |
2901 | ||
2902 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2903 | reception_buffer->data + state->received, | |
2904 | state->left_to_receive, MSG_DONTWAIT); | |
2905 | if (ret < 0) { | |
5569b118 JG |
2906 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
2907 | PERROR("Unable to receive control command header on sock %d", | |
2908 | conn->sock->fd); | |
2909 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2910 | } | |
5312a3ed JG |
2911 | goto end; |
2912 | } else if (ret == 0) { | |
2913 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 2914 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
2915 | goto end; |
2916 | } | |
2917 | ||
2918 | assert(ret > 0); | |
2919 | assert(ret <= state->left_to_receive); | |
2920 | ||
2921 | state->left_to_receive -= ret; | |
2922 | state->received += ret; | |
2923 | ||
2924 | if (state->left_to_receive > 0) { | |
2925 | /* | |
2926 | * Can't transition to the protocol's next state, wait to | |
2927 | * receive the rest of the header. | |
2928 | */ | |
2929 | DBG3("Partial reception of control connection protocol header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2930 | state->received, state->left_to_receive, | |
2931 | conn->sock->fd); | |
5312a3ed JG |
2932 | goto end; |
2933 | } | |
2934 | ||
2935 | /* Transition to next state: receiving the command's payload. */ | |
2936 | conn->protocol.ctrl.state_id = | |
2937 | CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
2938 | memcpy(&header, reception_buffer->data, sizeof(header)); | |
2939 | header.circuit_id = be64toh(header.circuit_id); | |
2940 | header.data_size = be64toh(header.data_size); | |
2941 | header.cmd = be32toh(header.cmd); | |
2942 | header.cmd_version = be32toh(header.cmd_version); | |
2943 | memcpy(&conn->protocol.ctrl.state.receive_payload.header, | |
2944 | &header, sizeof(header)); | |
2945 | ||
2946 | DBG("Done receiving control command header: fd = %i, cmd = %" PRIu32 ", cmd_version = %" PRIu32 ", payload size = %" PRIu64 " bytes", | |
2947 | conn->sock->fd, header.cmd, header.cmd_version, | |
2948 | header.data_size); | |
2949 | ||
715e6fb1 | 2950 | if (header.data_size > DEFAULT_NETWORK_RELAYD_CTRL_MAX_PAYLOAD_SIZE) { |
5312a3ed JG |
2951 | ERR("Command header indicates a payload (%" PRIu64 " bytes) that exceeds the maximal payload size allowed on a control connection.", |
2952 | header.data_size); | |
5569b118 | 2953 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
2954 | goto end; |
2955 | } | |
2956 | ||
2957 | conn->protocol.ctrl.state.receive_payload.left_to_receive = | |
2958 | header.data_size; | |
2959 | conn->protocol.ctrl.state.receive_payload.received = 0; | |
2960 | ret = lttng_dynamic_buffer_set_size(reception_buffer, | |
2961 | header.data_size); | |
2962 | if (ret) { | |
5569b118 | 2963 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
2964 | goto end; |
2965 | } | |
2966 | ||
2967 | if (header.data_size == 0) { | |
2968 | /* | |
2969 | * Manually invoke the next state as the poll loop | |
2970 | * will not wake-up to allow us to proceed further. | |
2971 | */ | |
5569b118 | 2972 | status = relay_process_control_receive_payload(conn); |
5312a3ed JG |
2973 | } |
2974 | end: | |
5569b118 | 2975 | return status; |
5312a3ed JG |
2976 | } |
2977 | ||
2978 | /* | |
2979 | * Process the commands received on the control socket | |
2980 | */ | |
5569b118 JG |
2981 | static enum relay_connection_status relay_process_control( |
2982 | struct relay_connection *conn) | |
5312a3ed | 2983 | { |
5569b118 | 2984 | enum relay_connection_status status; |
5312a3ed JG |
2985 | |
2986 | switch (conn->protocol.ctrl.state_id) { | |
2987 | case CTRL_CONNECTION_STATE_RECEIVE_HEADER: | |
5569b118 | 2988 | status = relay_process_control_receive_header(conn); |
5312a3ed JG |
2989 | break; |
2990 | case CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
5569b118 | 2991 | status = relay_process_control_receive_payload(conn); |
5312a3ed JG |
2992 | break; |
2993 | default: | |
2994 | ERR("Unknown control connection protocol state encountered."); | |
2995 | abort(); | |
2996 | } | |
2997 | ||
5569b118 | 2998 | return status; |
5312a3ed JG |
2999 | } |
3000 | ||
5569b118 JG |
3001 | static enum relay_connection_status relay_process_data_receive_header( |
3002 | struct relay_connection *conn) | |
b8aa1682 | 3003 | { |
5312a3ed | 3004 | int ret; |
5569b118 | 3005 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3006 | struct data_connection_state_receive_header *state = |
3007 | &conn->protocol.data.state.receive_header; | |
3008 | struct lttcomm_relayd_data_hdr header; | |
b8aa1682 | 3009 | struct relay_stream *stream; |
5312a3ed JG |
3010 | |
3011 | assert(state->left_to_receive != 0); | |
3012 | ||
3013 | ret = conn->sock->ops->recvmsg(conn->sock, | |
3014 | state->header_reception_buffer + state->received, | |
3015 | state->left_to_receive, MSG_DONTWAIT); | |
3016 | if (ret < 0) { | |
5569b118 JG |
3017 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
3018 | PERROR("Unable to receive data header on sock %d", conn->sock->fd); | |
3019 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3020 | } | |
5312a3ed JG |
3021 | goto end; |
3022 | } else if (ret == 0) { | |
3023 | /* Orderly shutdown. Not necessary to print an error. */ | |
3024 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3025 | status = RELAY_CONNECTION_STATUS_CLOSED; |
b8aa1682 JD |
3026 | goto end; |
3027 | } | |
3028 | ||
5312a3ed JG |
3029 | assert(ret > 0); |
3030 | assert(ret <= state->left_to_receive); | |
3031 | ||
3032 | state->left_to_receive -= ret; | |
3033 | state->received += ret; | |
3034 | ||
3035 | if (state->left_to_receive > 0) { | |
3036 | /* | |
3037 | * Can't transition to the protocol's next state, wait to | |
3038 | * receive the rest of the header. | |
3039 | */ | |
3040 | DBG3("Partial reception of data connection header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3041 | state->received, state->left_to_receive, | |
3042 | conn->sock->fd); | |
7591bab1 | 3043 | goto end; |
b8aa1682 | 3044 | } |
b8aa1682 | 3045 | |
5312a3ed JG |
3046 | /* Transition to next state: receiving the payload. */ |
3047 | conn->protocol.data.state_id = DATA_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
173af62f | 3048 | |
5312a3ed JG |
3049 | memcpy(&header, state->header_reception_buffer, sizeof(header)); |
3050 | header.circuit_id = be64toh(header.circuit_id); | |
3051 | header.stream_id = be64toh(header.stream_id); | |
3052 | header.data_size = be32toh(header.data_size); | |
3053 | header.net_seq_num = be64toh(header.net_seq_num); | |
3054 | header.padding_size = be32toh(header.padding_size); | |
3055 | memcpy(&conn->protocol.data.state.receive_payload.header, &header, sizeof(header)); | |
3056 | ||
3057 | conn->protocol.data.state.receive_payload.left_to_receive = | |
3058 | header.data_size; | |
3059 | conn->protocol.data.state.receive_payload.received = 0; | |
3060 | conn->protocol.data.state.receive_payload.rotate_index = false; | |
3061 | ||
3062 | DBG("Received data connection header on fd %i: circuit_id = %" PRIu64 ", stream_id = %" PRIu64 ", data_size = %" PRIu32 ", net_seq_num = %" PRIu64 ", padding_size = %" PRIu32, | |
3063 | conn->sock->fd, header.circuit_id, | |
3064 | header.stream_id, header.data_size, | |
3065 | header.net_seq_num, header.padding_size); | |
3066 | ||
3067 | stream = stream_get_by_id(header.stream_id); | |
3068 | if (!stream) { | |
3069 | DBG("relay_process_data_receive_payload: Cannot find stream %" PRIu64, | |
3070 | header.stream_id); | |
5569b118 JG |
3071 | /* Protocol error. */ |
3072 | status = RELAY_CONNECTION_STATUS_ERROR; | |
5312a3ed JG |
3073 | goto end; |
3074 | } | |
b8aa1682 | 3075 | |
7591bab1 | 3076 | pthread_mutex_lock(&stream->lock); |
c35f9726 JG |
3077 | /* Prepare stream for the reception of a new packet. */ |
3078 | ret = stream_init_packet(stream, header.data_size, | |
3079 | &conn->protocol.data.state.receive_payload.rotate_index); | |
3080 | pthread_mutex_unlock(&stream->lock); | |
3081 | if (ret) { | |
3082 | ERR("Failed to rotate stream output file"); | |
3083 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3084 | goto end_stream_unlock; | |
1c20f0e2 JD |
3085 | } |
3086 | ||
5312a3ed | 3087 | end_stream_unlock: |
5312a3ed JG |
3088 | stream_put(stream); |
3089 | end: | |
5569b118 | 3090 | return status; |
5312a3ed JG |
3091 | } |
3092 | ||
5569b118 JG |
3093 | static enum relay_connection_status relay_process_data_receive_payload( |
3094 | struct relay_connection *conn) | |
5312a3ed JG |
3095 | { |
3096 | int ret; | |
5569b118 | 3097 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3098 | struct relay_stream *stream; |
3099 | struct data_connection_state_receive_payload *state = | |
3100 | &conn->protocol.data.state.receive_payload; | |
3101 | const size_t chunk_size = RECV_DATA_BUFFER_SIZE; | |
3102 | char data_buffer[chunk_size]; | |
3103 | bool partial_recv = false; | |
3104 | bool new_stream = false, close_requested = false, index_flushed = false; | |
3105 | uint64_t left_to_receive = state->left_to_receive; | |
3106 | struct relay_session *session; | |
3107 | ||
fd0f1e3e JR |
3108 | DBG3("Receiving data for stream id %" PRIu64 " seqnum %" PRIu64 ", %" PRIu64" bytes received, %" PRIu64 " bytes left to receive", |
3109 | state->header.stream_id, state->header.net_seq_num, | |
3110 | state->received, left_to_receive); | |
3111 | ||
5312a3ed JG |
3112 | stream = stream_get_by_id(state->header.stream_id); |
3113 | if (!stream) { | |
5569b118 | 3114 | /* Protocol error. */ |
fd0f1e3e | 3115 | ERR("relay_process_data_receive_payload: cannot find stream %" PRIu64, |
5312a3ed | 3116 | state->header.stream_id); |
5569b118 | 3117 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed | 3118 | goto end; |
1c20f0e2 JD |
3119 | } |
3120 | ||
5312a3ed JG |
3121 | pthread_mutex_lock(&stream->lock); |
3122 | session = stream->trace->session; | |
fd0f1e3e JR |
3123 | if (!conn->session) { |
3124 | ret = connection_set_session(conn, session); | |
3125 | if (ret) { | |
3126 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3127 | goto end_stream_unlock; | |
3128 | } | |
3129 | } | |
5312a3ed JG |
3130 | |
3131 | /* | |
3132 | * The size of the "chunk" received on any iteration is bounded by: | |
3133 | * - the data left to receive, | |
3134 | * - the data immediately available on the socket, | |
3135 | * - the on-stack data buffer | |
3136 | */ | |
3137 | while (left_to_receive > 0 && !partial_recv) { | |
5312a3ed | 3138 | size_t recv_size = min(left_to_receive, chunk_size); |
c35f9726 | 3139 | struct lttng_buffer_view packet_chunk; |
5312a3ed JG |
3140 | |
3141 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, | |
3142 | recv_size, MSG_DONTWAIT); | |
3143 | if (ret < 0) { | |
5569b118 JG |
3144 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
3145 | PERROR("Socket %d error", conn->sock->fd); | |
3146 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3147 | } | |
0848dba7 | 3148 | goto end_stream_unlock; |
5312a3ed JG |
3149 | } else if (ret == 0) { |
3150 | /* No more data ready to be consumed on socket. */ | |
3151 | DBG3("No more data ready for consumption on data socket of stream id %" PRIu64, | |
3152 | state->header.stream_id); | |
5569b118 | 3153 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3154 | break; |
3155 | } else if (ret < (int) recv_size) { | |
3156 | /* | |
3157 | * All the data available on the socket has been | |
3158 | * consumed. | |
3159 | */ | |
3160 | partial_recv = true; | |
c35f9726 | 3161 | recv_size = ret; |
0848dba7 MD |
3162 | } |
3163 | ||
c35f9726 JG |
3164 | packet_chunk = lttng_buffer_view_init(data_buffer, |
3165 | 0, recv_size); | |
3166 | assert(packet_chunk.data); | |
5312a3ed | 3167 | |
c35f9726 JG |
3168 | ret = stream_write(stream, &packet_chunk, 0); |
3169 | if (ret) { | |
0848dba7 | 3170 | ERR("Relay error writing data to file"); |
5569b118 | 3171 | status = RELAY_CONNECTION_STATUS_ERROR; |
0848dba7 MD |
3172 | goto end_stream_unlock; |
3173 | } | |
3174 | ||
5312a3ed JG |
3175 | left_to_receive -= recv_size; |
3176 | state->received += recv_size; | |
3177 | state->left_to_receive = left_to_receive; | |
5312a3ed JG |
3178 | } |
3179 | ||
3180 | if (state->left_to_receive > 0) { | |
3181 | /* | |
3182 | * Did not receive all the data expected, wait for more data to | |
3183 | * become available on the socket. | |
3184 | */ | |
3185 | DBG3("Partial receive on data connection of stream id %" PRIu64 ", %" PRIu64 " bytes received, %" PRIu64 " bytes left to receive", | |
3186 | state->header.stream_id, state->received, | |
3187 | state->left_to_receive); | |
5312a3ed | 3188 | goto end_stream_unlock; |
0848dba7 | 3189 | } |
5ab7344e | 3190 | |
c35f9726 JG |
3191 | ret = stream_write(stream, NULL, state->header.padding_size); |
3192 | if (ret) { | |
5569b118 | 3193 | status = RELAY_CONNECTION_STATUS_ERROR; |
7591bab1 | 3194 | goto end_stream_unlock; |
1d4dfdef | 3195 | } |
5312a3ed | 3196 | |
298a25ca | 3197 | if (session_streams_have_index(session)) { |
c35f9726 JG |
3198 | ret = stream_update_index(stream, state->header.net_seq_num, |
3199 | state->rotate_index, &index_flushed, | |
3200 | state->header.data_size + state->header.padding_size); | |
5312a3ed | 3201 | if (ret < 0) { |
c35f9726 | 3202 | ERR("Failed to update index: stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
5312a3ed JG |
3203 | stream->stream_handle, |
3204 | state->header.net_seq_num, ret); | |
5569b118 | 3205 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3206 | goto end_stream_unlock; |
3207 | } | |
3208 | } | |
3209 | ||
a8f9f353 | 3210 | if (stream->prev_data_seq == -1ULL) { |
c0bae11d MD |
3211 | new_stream = true; |
3212 | } | |
3213 | ||
c35f9726 JG |
3214 | ret = stream_complete_packet(stream, state->header.data_size + |
3215 | state->header.padding_size, state->header.net_seq_num, | |
3216 | index_flushed); | |
3217 | if (ret) { | |
3218 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3219 | goto end_stream_unlock; | |
3220 | } | |
5312a3ed JG |
3221 | |
3222 | /* | |
3223 | * Resetting the protocol state (to RECEIVE_HEADER) will trash the | |
3224 | * contents of *state which are aliased (union) to the same location as | |
3225 | * the new state. Don't use it beyond this point. | |
3226 | */ | |
3227 | connection_reset_protocol_state(conn); | |
3228 | state = NULL; | |
173af62f | 3229 | |
7591bab1 | 3230 | end_stream_unlock: |
bda7c7b9 | 3231 | close_requested = stream->close_requested; |
7591bab1 | 3232 | pthread_mutex_unlock(&stream->lock); |
5312a3ed | 3233 | if (close_requested && left_to_receive == 0) { |
bda7c7b9 JG |
3234 | try_stream_close(stream); |
3235 | } | |
3236 | ||
c0bae11d MD |
3237 | if (new_stream) { |
3238 | pthread_mutex_lock(&session->lock); | |
3239 | uatomic_set(&session->new_streams, 1); | |
3240 | pthread_mutex_unlock(&session->lock); | |
3241 | } | |
5312a3ed | 3242 | |
7591bab1 | 3243 | stream_put(stream); |
b8aa1682 | 3244 | end: |
5569b118 | 3245 | return status; |
b8aa1682 JD |
3246 | } |
3247 | ||
5312a3ed JG |
3248 | /* |
3249 | * relay_process_data: Process the data received on the data socket | |
3250 | */ | |
5569b118 JG |
3251 | static enum relay_connection_status relay_process_data( |
3252 | struct relay_connection *conn) | |
5312a3ed | 3253 | { |
5569b118 | 3254 | enum relay_connection_status status; |
5312a3ed JG |
3255 | |
3256 | switch (conn->protocol.data.state_id) { | |
3257 | case DATA_CONNECTION_STATE_RECEIVE_HEADER: | |
5569b118 | 3258 | status = relay_process_data_receive_header(conn); |
5312a3ed JG |
3259 | break; |
3260 | case DATA_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
5569b118 | 3261 | status = relay_process_data_receive_payload(conn); |
5312a3ed JG |
3262 | break; |
3263 | default: | |
3264 | ERR("Unexpected data connection communication state."); | |
3265 | abort(); | |
3266 | } | |
3267 | ||
5569b118 | 3268 | return status; |
5312a3ed JG |
3269 | } |
3270 | ||
7591bab1 | 3271 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
3272 | { |
3273 | int ret; | |
3274 | ||
58eb9381 | 3275 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
3276 | |
3277 | ret = close(pollfd); | |
3278 | if (ret < 0) { | |
3279 | ERR("Closing pollfd %d", pollfd); | |
3280 | } | |
3281 | } | |
3282 | ||
7591bab1 MD |
3283 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
3284 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 3285 | { |
7591bab1 | 3286 | const char *type_str; |
2a174661 | 3287 | |
7591bab1 MD |
3288 | switch (conn->type) { |
3289 | case RELAY_DATA: | |
3290 | type_str = "Data"; | |
3291 | break; | |
3292 | case RELAY_CONTROL: | |
3293 | type_str = "Control"; | |
3294 | break; | |
3295 | case RELAY_VIEWER_COMMAND: | |
3296 | type_str = "Viewer Command"; | |
3297 | break; | |
3298 | case RELAY_VIEWER_NOTIFICATION: | |
3299 | type_str = "Viewer Notification"; | |
3300 | break; | |
3301 | default: | |
3302 | type_str = "Unknown"; | |
9d1bbf21 | 3303 | } |
7591bab1 MD |
3304 | cleanup_connection_pollfd(events, pollfd); |
3305 | connection_put(conn); | |
3306 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
3307 | } |
3308 | ||
3309 | /* | |
3310 | * This thread does the actual work | |
3311 | */ | |
7591bab1 | 3312 | static void *relay_thread_worker(void *data) |
b8aa1682 | 3313 | { |
beaad64c DG |
3314 | int ret, err = -1, last_seen_data_fd = -1; |
3315 | uint32_t nb_fd; | |
b8aa1682 JD |
3316 | struct lttng_poll_event events; |
3317 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 3318 | struct lttng_ht_iter iter; |
90e7d72f | 3319 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
3320 | |
3321 | DBG("[thread] Relay worker started"); | |
3322 | ||
9d1bbf21 MD |
3323 | rcu_register_thread(); |
3324 | ||
55706a7d MD |
3325 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
3326 | ||
9b5e0863 MD |
3327 | if (testpoint(relayd_thread_worker)) { |
3328 | goto error_testpoint; | |
3329 | } | |
3330 | ||
f385ae0a MD |
3331 | health_code_update(); |
3332 | ||
b8aa1682 JD |
3333 | /* table of connections indexed on socket */ |
3334 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
3335 | if (!relay_connections_ht) { |
3336 | goto relay_connections_ht_error; | |
3337 | } | |
b8aa1682 | 3338 | |
b8aa1682 JD |
3339 | ret = create_thread_poll_set(&events, 2); |
3340 | if (ret < 0) { | |
3341 | goto error_poll_create; | |
3342 | } | |
3343 | ||
58eb9381 | 3344 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
3345 | if (ret < 0) { |
3346 | goto error; | |
3347 | } | |
3348 | ||
beaad64c | 3349 | restart: |
b8aa1682 | 3350 | while (1) { |
beaad64c DG |
3351 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
3352 | ||
f385ae0a MD |
3353 | health_code_update(); |
3354 | ||
b8aa1682 | 3355 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 3356 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 3357 | health_poll_entry(); |
b8aa1682 | 3358 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 3359 | health_poll_exit(); |
b8aa1682 JD |
3360 | if (ret < 0) { |
3361 | /* | |
3362 | * Restart interrupted system call. | |
3363 | */ | |
3364 | if (errno == EINTR) { | |
3365 | goto restart; | |
3366 | } | |
3367 | goto error; | |
3368 | } | |
3369 | ||
0d9c5d77 DG |
3370 | nb_fd = ret; |
3371 | ||
beaad64c | 3372 | /* |
7591bab1 MD |
3373 | * Process control. The control connection is |
3374 | * prioritized so we don't starve it with high | |
3375 | * throughput tracing data on the data connection. | |
beaad64c | 3376 | */ |
b8aa1682 JD |
3377 | for (i = 0; i < nb_fd; i++) { |
3378 | /* Fetch once the poll data */ | |
beaad64c DG |
3379 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
3380 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 3381 | |
f385ae0a MD |
3382 | health_code_update(); |
3383 | ||
b8aa1682 JD |
3384 | /* Thread quit pipe has been closed. Killing thread. */ |
3385 | ret = check_thread_quit_pipe(pollfd, revents); | |
3386 | if (ret) { | |
095a4ae5 MD |
3387 | err = 0; |
3388 | goto exit; | |
b8aa1682 JD |
3389 | } |
3390 | ||
58eb9381 DG |
3391 | /* Inspect the relay conn pipe for new connection */ |
3392 | if (pollfd == relay_conn_pipe[0]) { | |
03e43155 | 3393 | if (revents & LPOLLIN) { |
90e7d72f JG |
3394 | struct relay_connection *conn; |
3395 | ||
58eb9381 | 3396 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
3397 | if (ret < 0) { |
3398 | goto error; | |
3399 | } | |
58eb9381 DG |
3400 | lttng_poll_add(&events, conn->sock->fd, |
3401 | LPOLLIN | LPOLLRDHUP); | |
7591bab1 | 3402 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 3403 | DBG("Connection socket %d added", conn->sock->fd); |
03e43155 MD |
3404 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3405 | ERR("Relay connection pipe error"); | |
3406 | goto error; | |
3407 | } else { | |
3408 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
3409 | goto error; | |
b8aa1682 | 3410 | } |
58eb9381 | 3411 | } else { |
90e7d72f JG |
3412 | struct relay_connection *ctrl_conn; |
3413 | ||
7591bab1 | 3414 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 3415 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 3416 | assert(ctrl_conn); |
58eb9381 | 3417 | |
03e43155 MD |
3418 | if (ctrl_conn->type == RELAY_DATA) { |
3419 | if (revents & LPOLLIN) { | |
beaad64c DG |
3420 | /* |
3421 | * Flag the last seen data fd not deleted. It will be | |
3422 | * used as the last seen fd if any fd gets deleted in | |
3423 | * this first loop. | |
3424 | */ | |
3425 | last_notdel_data_fd = pollfd; | |
3426 | } | |
03e43155 MD |
3427 | goto put_ctrl_connection; |
3428 | } | |
3429 | assert(ctrl_conn->type == RELAY_CONTROL); | |
3430 | ||
3431 | if (revents & LPOLLIN) { | |
5569b118 JG |
3432 | enum relay_connection_status status; |
3433 | ||
3434 | status = relay_process_control(ctrl_conn); | |
3435 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
fd0f1e3e JR |
3436 | /* |
3437 | * On socket error flag the session as aborted to force | |
3438 | * the cleanup of its stream otherwise it can leak | |
3439 | * during the lifetime of the relayd. | |
3440 | * | |
3441 | * This prevents situations in which streams can be | |
3442 | * left opened because an index was received, the | |
3443 | * control connection is closed, and the data | |
3444 | * connection is closed (uncleanly) before the packet's | |
3445 | * data provided. | |
3446 | * | |
3447 | * Since the control connection encountered an error, | |
3448 | * it is okay to be conservative and close the | |
3449 | * session right now as we can't rely on the protocol | |
3450 | * being respected anymore. | |
3451 | */ | |
3452 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3453 | session_abort(ctrl_conn->session); | |
3454 | } | |
3455 | ||
5569b118 | 3456 | /* Clear the connection on error or close. */ |
5312a3ed JG |
3457 | relay_thread_close_connection(&events, |
3458 | pollfd, | |
03e43155 | 3459 | ctrl_conn); |
03e43155 | 3460 | } |
5312a3ed | 3461 | seen_control = 1; |
03e43155 MD |
3462 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3463 | relay_thread_close_connection(&events, | |
3464 | pollfd, ctrl_conn); | |
3465 | if (last_seen_data_fd == pollfd) { | |
3466 | last_seen_data_fd = last_notdel_data_fd; | |
3467 | } | |
58eb9381 | 3468 | } else { |
03e43155 MD |
3469 | ERR("Unexpected poll events %u for control sock %d", |
3470 | revents, pollfd); | |
3471 | connection_put(ctrl_conn); | |
3472 | goto error; | |
beaad64c | 3473 | } |
03e43155 | 3474 | put_ctrl_connection: |
7591bab1 | 3475 | connection_put(ctrl_conn); |
beaad64c DG |
3476 | } |
3477 | } | |
3478 | ||
3479 | /* | |
3480 | * The last loop handled a control request, go back to poll to make | |
3481 | * sure we prioritise the control socket. | |
3482 | */ | |
3483 | if (seen_control) { | |
3484 | continue; | |
3485 | } | |
3486 | ||
3487 | if (last_seen_data_fd >= 0) { | |
3488 | for (i = 0; i < nb_fd; i++) { | |
3489 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
3490 | |
3491 | health_code_update(); | |
3492 | ||
beaad64c DG |
3493 | if (last_seen_data_fd == pollfd) { |
3494 | idx = i; | |
3495 | break; | |
3496 | } | |
3497 | } | |
3498 | } | |
3499 | ||
3500 | /* Process data connection. */ | |
3501 | for (i = idx + 1; i < nb_fd; i++) { | |
3502 | /* Fetch the poll data. */ | |
3503 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
3504 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 3505 | struct relay_connection *data_conn; |
beaad64c | 3506 | |
f385ae0a MD |
3507 | health_code_update(); |
3508 | ||
fd20dac9 MD |
3509 | if (!revents) { |
3510 | /* No activity for this FD (poll implementation). */ | |
3511 | continue; | |
3512 | } | |
3513 | ||
beaad64c | 3514 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 3515 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
3516 | continue; |
3517 | } | |
3518 | ||
7591bab1 | 3519 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 3520 | if (!data_conn) { |
fd20dac9 | 3521 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
3522 | continue; |
3523 | } | |
03e43155 MD |
3524 | if (data_conn->type == RELAY_CONTROL) { |
3525 | goto put_data_connection; | |
3526 | } | |
3527 | assert(data_conn->type == RELAY_DATA); | |
fd20dac9 MD |
3528 | |
3529 | if (revents & LPOLLIN) { | |
5569b118 JG |
3530 | enum relay_connection_status status; |
3531 | ||
3532 | status = relay_process_data(data_conn); | |
3533 | /* Connection closed or error. */ | |
3534 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
fd0f1e3e JR |
3535 | /* |
3536 | * On socket error flag the session as aborted to force | |
3537 | * the cleanup of its stream otherwise it can leak | |
3538 | * during the lifetime of the relayd. | |
3539 | * | |
3540 | * This prevents situations in which streams can be | |
3541 | * left opened because an index was received, the | |
3542 | * control connection is closed, and the data | |
3543 | * connection is closed (uncleanly) before the packet's | |
3544 | * data provided. | |
3545 | * | |
3546 | * Since the data connection encountered an error, | |
3547 | * it is okay to be conservative and close the | |
3548 | * session right now as we can't rely on the protocol | |
3549 | * being respected anymore. | |
3550 | */ | |
3551 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3552 | session_abort(data_conn->session); | |
3553 | } | |
7591bab1 | 3554 | relay_thread_close_connection(&events, pollfd, |
03e43155 | 3555 | data_conn); |
fd20dac9 MD |
3556 | /* |
3557 | * Every goto restart call sets the last seen fd where | |
3558 | * here we don't really care since we gracefully | |
3559 | * continue the loop after the connection is deleted. | |
3560 | */ | |
3561 | } else { | |
3562 | /* Keep last seen port. */ | |
3563 | last_seen_data_fd = pollfd; | |
7591bab1 | 3564 | connection_put(data_conn); |
fd20dac9 | 3565 | goto restart; |
b8aa1682 | 3566 | } |
03e43155 MD |
3567 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3568 | relay_thread_close_connection(&events, pollfd, | |
3569 | data_conn); | |
3570 | } else { | |
3571 | ERR("Unknown poll events %u for data sock %d", | |
3572 | revents, pollfd); | |
b8aa1682 | 3573 | } |
03e43155 | 3574 | put_data_connection: |
7591bab1 | 3575 | connection_put(data_conn); |
b8aa1682 | 3576 | } |
beaad64c | 3577 | last_seen_data_fd = -1; |
b8aa1682 JD |
3578 | } |
3579 | ||
f385ae0a MD |
3580 | /* Normal exit, no error */ |
3581 | ret = 0; | |
3582 | ||
095a4ae5 | 3583 | exit: |
b8aa1682 | 3584 | error: |
71efa8ef | 3585 | /* Cleanup remaining connection object. */ |
9d1bbf21 | 3586 | rcu_read_lock(); |
90e7d72f JG |
3587 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
3588 | destroy_conn, | |
58eb9381 | 3589 | sock_n.node) { |
f385ae0a | 3590 | health_code_update(); |
98ba050e | 3591 | |
fd0f1e3e | 3592 | session_abort(destroy_conn->session); |
98ba050e | 3593 | |
7591bab1 MD |
3594 | /* |
3595 | * No need to grab another ref, because we own | |
3596 | * destroy_conn. | |
3597 | */ | |
3598 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
3599 | destroy_conn); | |
b8aa1682 | 3600 | } |
94d49140 | 3601 | rcu_read_unlock(); |
7591bab1 MD |
3602 | |
3603 | lttng_poll_clean(&events); | |
7d2f7452 | 3604 | error_poll_create: |
b8aa1682 | 3605 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 3606 | relay_connections_ht_error: |
58eb9381 DG |
3607 | /* Close relay conn pipes */ |
3608 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
3609 | if (err) { |
3610 | DBG("Thread exited with error"); | |
3611 | } | |
b8aa1682 | 3612 | DBG("Worker thread cleanup complete"); |
9b5e0863 | 3613 | error_testpoint: |
f385ae0a MD |
3614 | if (err) { |
3615 | health_error(); | |
3616 | ERR("Health error occurred in %s", __func__); | |
3617 | } | |
3618 | health_unregister(health_relayd); | |
9d1bbf21 | 3619 | rcu_unregister_thread(); |
b4aacfdc | 3620 | lttng_relay_stop_threads(); |
b8aa1682 JD |
3621 | return NULL; |
3622 | } | |
3623 | ||
3624 | /* | |
3625 | * Create the relay command pipe to wake thread_manage_apps. | |
3626 | * Closed in cleanup(). | |
3627 | */ | |
58eb9381 | 3628 | static int create_relay_conn_pipe(void) |
b8aa1682 | 3629 | { |
a02de639 | 3630 | int ret; |
b8aa1682 | 3631 | |
58eb9381 | 3632 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 3633 | |
b8aa1682 JD |
3634 | return ret; |
3635 | } | |
3636 | ||
3637 | /* | |
3638 | * main | |
3639 | */ | |
3640 | int main(int argc, char **argv) | |
3641 | { | |
178a0557 | 3642 | int ret = 0, retval = 0; |
b8aa1682 JD |
3643 | void *status; |
3644 | ||
b8aa1682 JD |
3645 | /* Parse arguments */ |
3646 | progname = argv[0]; | |
178a0557 MD |
3647 | if (set_options(argc, argv)) { |
3648 | retval = -1; | |
3649 | goto exit_options; | |
b8aa1682 JD |
3650 | } |
3651 | ||
178a0557 MD |
3652 | if (set_signal_handler()) { |
3653 | retval = -1; | |
3654 | goto exit_options; | |
b8aa1682 JD |
3655 | } |
3656 | ||
4d513a50 DG |
3657 | /* Try to create directory if -o, --output is specified. */ |
3658 | if (opt_output_path) { | |
994fa64f DG |
3659 | if (*opt_output_path != '/') { |
3660 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
3661 | retval = -1; |
3662 | goto exit_options; | |
994fa64f DG |
3663 | } |
3664 | ||
d77dded2 JG |
3665 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
3666 | -1, -1); | |
4d513a50 DG |
3667 | if (ret < 0) { |
3668 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
3669 | retval = -1; |
3670 | goto exit_options; | |
4d513a50 DG |
3671 | } |
3672 | } | |
3673 | ||
b8aa1682 | 3674 | /* Daemonize */ |
b5218ffb | 3675 | if (opt_daemon || opt_background) { |
3fd27398 MD |
3676 | int i; |
3677 | ||
3678 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
3679 | !opt_background); | |
b8aa1682 | 3680 | if (ret < 0) { |
178a0557 MD |
3681 | retval = -1; |
3682 | goto exit_options; | |
b8aa1682 | 3683 | } |
3fd27398 MD |
3684 | |
3685 | /* | |
3686 | * We are in the child. Make sure all other file | |
3687 | * descriptors are closed, in case we are called with | |
3688 | * more opened file descriptors than the standard ones. | |
3689 | */ | |
3690 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
3691 | (void) close(i); | |
3692 | } | |
3693 | } | |
3694 | ||
23c8ff50 JG |
3695 | sessiond_trace_chunk_registry = sessiond_trace_chunk_registry_create(); |
3696 | if (!sessiond_trace_chunk_registry) { | |
3697 | ERR("Failed to initialize session daemon trace chunk registry"); | |
3698 | retval = -1; | |
3699 | goto exit_sessiond_trace_chunk_registry; | |
3700 | } | |
3701 | ||
178a0557 MD |
3702 | /* Initialize thread health monitoring */ |
3703 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
3704 | if (!health_relayd) { | |
3705 | PERROR("health_app_create error"); | |
3706 | retval = -1; | |
3707 | goto exit_health_app_create; | |
3708 | } | |
3709 | ||
3fd27398 | 3710 | /* Create thread quit pipe */ |
178a0557 MD |
3711 | if (init_thread_quit_pipe()) { |
3712 | retval = -1; | |
3713 | goto exit_init_data; | |
b8aa1682 JD |
3714 | } |
3715 | ||
b8aa1682 | 3716 | /* Setup the thread apps communication pipe. */ |
178a0557 MD |
3717 | if (create_relay_conn_pipe()) { |
3718 | retval = -1; | |
3719 | goto exit_init_data; | |
b8aa1682 JD |
3720 | } |
3721 | ||
3722 | /* Init relay command queue. */ | |
8bdee6e2 | 3723 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 | 3724 | |
554831e7 MD |
3725 | /* Initialize communication library */ |
3726 | lttcomm_init(); | |
87e45c13 | 3727 | lttcomm_inet_init(); |
554831e7 | 3728 | |
d3e2ba59 | 3729 | /* tables of sessions indexed by session ID */ |
7591bab1 MD |
3730 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3731 | if (!sessions_ht) { | |
178a0557 MD |
3732 | retval = -1; |
3733 | goto exit_init_data; | |
d3e2ba59 JD |
3734 | } |
3735 | ||
3736 | /* tables of streams indexed by stream ID */ | |
2a174661 | 3737 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 3738 | if (!relay_streams_ht) { |
178a0557 MD |
3739 | retval = -1; |
3740 | goto exit_init_data; | |
d3e2ba59 JD |
3741 | } |
3742 | ||
3743 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
3744 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3745 | if (!viewer_streams_ht) { | |
178a0557 MD |
3746 | retval = -1; |
3747 | goto exit_init_data; | |
55706a7d MD |
3748 | } |
3749 | ||
65931c8b | 3750 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
3751 | if (ret) { |
3752 | retval = -1; | |
3753 | goto exit_health_quit_pipe; | |
65931c8b MD |
3754 | } |
3755 | ||
3756 | /* Create thread to manage the client socket */ | |
1a1a34b4 | 3757 | ret = pthread_create(&health_thread, default_pthread_attr(), |
65931c8b | 3758 | thread_manage_health, (void *) NULL); |
178a0557 MD |
3759 | if (ret) { |
3760 | errno = ret; | |
65931c8b | 3761 | PERROR("pthread_create health"); |
178a0557 MD |
3762 | retval = -1; |
3763 | goto exit_health_thread; | |
65931c8b MD |
3764 | } |
3765 | ||
b8aa1682 | 3766 | /* Setup the dispatcher thread */ |
1a1a34b4 | 3767 | ret = pthread_create(&dispatcher_thread, default_pthread_attr(), |
b8aa1682 | 3768 | relay_thread_dispatcher, (void *) NULL); |
178a0557 MD |
3769 | if (ret) { |
3770 | errno = ret; | |
b8aa1682 | 3771 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
3772 | retval = -1; |
3773 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
3774 | } |
3775 | ||
3776 | /* Setup the worker thread */ | |
1a1a34b4 | 3777 | ret = pthread_create(&worker_thread, default_pthread_attr(), |
7591bab1 | 3778 | relay_thread_worker, NULL); |
178a0557 MD |
3779 | if (ret) { |
3780 | errno = ret; | |
b8aa1682 | 3781 | PERROR("pthread_create worker"); |
178a0557 MD |
3782 | retval = -1; |
3783 | goto exit_worker_thread; | |
b8aa1682 JD |
3784 | } |
3785 | ||
3786 | /* Setup the listener thread */ | |
1a1a34b4 | 3787 | ret = pthread_create(&listener_thread, default_pthread_attr(), |
b8aa1682 | 3788 | relay_thread_listener, (void *) NULL); |
178a0557 MD |
3789 | if (ret) { |
3790 | errno = ret; | |
b8aa1682 | 3791 | PERROR("pthread_create listener"); |
178a0557 MD |
3792 | retval = -1; |
3793 | goto exit_listener_thread; | |
b8aa1682 JD |
3794 | } |
3795 | ||
7591bab1 | 3796 | ret = relayd_live_create(live_uri); |
178a0557 | 3797 | if (ret) { |
d3e2ba59 | 3798 | ERR("Starting live viewer threads"); |
178a0557 | 3799 | retval = -1; |
50138f51 | 3800 | goto exit_live; |
d3e2ba59 JD |
3801 | } |
3802 | ||
178a0557 MD |
3803 | /* |
3804 | * This is where we start awaiting program completion (e.g. through | |
3805 | * signal that asks threads to teardown). | |
3806 | */ | |
3807 | ||
3808 | ret = relayd_live_join(); | |
3809 | if (ret) { | |
3810 | retval = -1; | |
3811 | } | |
50138f51 | 3812 | exit_live: |
178a0557 | 3813 | |
b8aa1682 | 3814 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
3815 | if (ret) { |
3816 | errno = ret; | |
3817 | PERROR("pthread_join listener_thread"); | |
3818 | retval = -1; | |
b8aa1682 JD |
3819 | } |
3820 | ||
178a0557 | 3821 | exit_listener_thread: |
b8aa1682 | 3822 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
3823 | if (ret) { |
3824 | errno = ret; | |
3825 | PERROR("pthread_join worker_thread"); | |
3826 | retval = -1; | |
b8aa1682 JD |
3827 | } |
3828 | ||
178a0557 | 3829 | exit_worker_thread: |
b8aa1682 | 3830 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
3831 | if (ret) { |
3832 | errno = ret; | |
3833 | PERROR("pthread_join dispatcher_thread"); | |
3834 | retval = -1; | |
b8aa1682 | 3835 | } |
178a0557 | 3836 | exit_dispatcher_thread: |
42415026 | 3837 | |
65931c8b | 3838 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
3839 | if (ret) { |
3840 | errno = ret; | |
3841 | PERROR("pthread_join health_thread"); | |
3842 | retval = -1; | |
65931c8b | 3843 | } |
178a0557 | 3844 | exit_health_thread: |
65931c8b | 3845 | |
65931c8b | 3846 | utils_close_pipe(health_quit_pipe); |
178a0557 | 3847 | exit_health_quit_pipe: |
65931c8b | 3848 | |
178a0557 | 3849 | exit_init_data: |
55706a7d | 3850 | health_app_destroy(health_relayd); |
23c8ff50 | 3851 | sessiond_trace_chunk_registry_destroy(sessiond_trace_chunk_registry); |
55706a7d | 3852 | exit_health_app_create: |
23c8ff50 | 3853 | exit_sessiond_trace_chunk_registry: |
178a0557 | 3854 | exit_options: |
4d62fbf8 MD |
3855 | /* |
3856 | * Wait for all pending call_rcu work to complete before tearing | |
3857 | * down data structures. call_rcu worker may be trying to | |
3858 | * perform lookups in those structures. | |
3859 | */ | |
3860 | rcu_barrier(); | |
7591bab1 MD |
3861 | relayd_cleanup(); |
3862 | ||
3863 | /* Ensure all prior call_rcu are done. */ | |
3864 | rcu_barrier(); | |
d3e2ba59 | 3865 | |
178a0557 | 3866 | if (!retval) { |
b8aa1682 | 3867 | exit(EXIT_SUCCESS); |
178a0557 MD |
3868 | } else { |
3869 | exit(EXIT_FAILURE); | |
b8aa1682 | 3870 | } |
b8aa1682 | 3871 | } |