Commit | Line | Data |
---|---|---|
b8aa1682 JD |
1 | /* |
2 | * Copyright (C) 2012 - Julien Desfossez <jdesfossez@efficios.com> | |
3 | * David Goulet <dgoulet@efficios.com> | |
cd60b05a | 4 | * 2013 - Jérémie Galarneau <jeremie.galarneau@efficios.com> |
7591bab1 | 5 | * 2015 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
b8aa1682 JD |
6 | * |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License, version 2 only, | |
9 | * as published by the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
14 | * more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; if not, write to the Free Software Foundation, Inc., | |
18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
19 | */ | |
20 | ||
6c1c0768 | 21 | #define _LGPL_SOURCE |
b8aa1682 JD |
22 | #include <getopt.h> |
23 | #include <grp.h> | |
24 | #include <limits.h> | |
25 | #include <pthread.h> | |
26 | #include <signal.h> | |
27 | #include <stdio.h> | |
28 | #include <stdlib.h> | |
29 | #include <string.h> | |
30 | #include <sys/mman.h> | |
31 | #include <sys/mount.h> | |
32 | #include <sys/resource.h> | |
33 | #include <sys/socket.h> | |
34 | #include <sys/stat.h> | |
35 | #include <sys/types.h> | |
36 | #include <sys/wait.h> | |
173af62f | 37 | #include <inttypes.h> |
b8aa1682 JD |
38 | #include <urcu/futex.h> |
39 | #include <urcu/uatomic.h> | |
40 | #include <unistd.h> | |
41 | #include <fcntl.h> | |
b8aa1682 JD |
42 | |
43 | #include <lttng/lttng.h> | |
44 | #include <common/common.h> | |
45 | #include <common/compat/poll.h> | |
46 | #include <common/compat/socket.h> | |
f263b7fd | 47 | #include <common/compat/endian.h> |
e8fa9fb0 | 48 | #include <common/compat/getenv.h> |
b8aa1682 | 49 | #include <common/defaults.h> |
3fd27398 | 50 | #include <common/daemonize.h> |
b8aa1682 JD |
51 | #include <common/futex.h> |
52 | #include <common/sessiond-comm/sessiond-comm.h> | |
53 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
54 | #include <common/sessiond-comm/relayd.h> |
55 | #include <common/uri.h> | |
a02de639 | 56 | #include <common/utils.h> |
f40ef1d5 | 57 | #include <common/config/session-config.h> |
44a2fbf1 JG |
58 | #include <common/dynamic-buffer.h> |
59 | #include <common/buffer-view.h> | |
7591bab1 | 60 | #include <urcu/rculist.h> |
b8aa1682 | 61 | |
8ff7624a | 62 | #include "version.h" |
0f907de1 | 63 | #include "cmd.h" |
d3e2ba59 | 64 | #include "ctf-trace.h" |
1c20f0e2 | 65 | #include "index.h" |
0f907de1 | 66 | #include "utils.h" |
b8aa1682 | 67 | #include "lttng-relayd.h" |
d3e2ba59 | 68 | #include "live.h" |
55706a7d | 69 | #include "health-relayd.h" |
9b5e0863 | 70 | #include "testpoint.h" |
2f8f53af | 71 | #include "viewer-stream.h" |
2a174661 DG |
72 | #include "session.h" |
73 | #include "stream.h" | |
58eb9381 | 74 | #include "connection.h" |
a44ca2ca | 75 | #include "tracefile-array.h" |
2fc6b1ab | 76 | #include "tcp_keep_alive.h" |
b8aa1682 | 77 | |
44a2fbf1 JG |
78 | enum relay_connection_status { |
79 | RELAY_CONNECTION_STATUS_OK, | |
80 | /* An error occured while processing an event on the connection. */ | |
81 | RELAY_CONNECTION_STATUS_ERROR, | |
82 | /* Connection closed/shutdown cleanly. */ | |
83 | RELAY_CONNECTION_STATUS_CLOSED, | |
84 | }; | |
85 | ||
b8aa1682 | 86 | /* command line options */ |
76375580 | 87 | char *opt_output_path, *opt_working_directory; |
8ff7624a | 88 | static int opt_daemon, opt_background, opt_print_version; |
b4d507ca JR |
89 | int opt_group_output_by_session; |
90 | int opt_group_output_by_host; | |
3fd27398 MD |
91 | |
92 | /* | |
93 | * We need to wait for listener and live listener threads, as well as | |
94 | * health check thread, before being ready to signal readiness. | |
95 | */ | |
96 | #define NR_LTTNG_RELAY_READY 3 | |
97 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
0848dba7 MD |
98 | |
99 | /* Size of receive buffer. */ | |
100 | #define RECV_DATA_BUFFER_SIZE 65536 | |
101 | ||
3fd27398 MD |
102 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ |
103 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
104 | ||
095a4ae5 MD |
105 | static struct lttng_uri *control_uri; |
106 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 107 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
108 | |
109 | const char *progname; | |
b8aa1682 | 110 | |
65931c8b | 111 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
112 | static int tracing_group_name_override; |
113 | ||
114 | const char * const config_section_name = "relayd"; | |
65931c8b | 115 | |
b8aa1682 JD |
116 | /* |
117 | * Quit pipe for all threads. This permits a single cancellation point | |
118 | * for all threads when receiving an event on the pipe. | |
119 | */ | |
0b242f62 | 120 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
121 | |
122 | /* | |
123 | * This pipe is used to inform the worker thread that a command is queued and | |
124 | * ready to be processed. | |
125 | */ | |
58eb9381 | 126 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 127 | |
26c9d55e | 128 | /* Shared between threads */ |
b8aa1682 JD |
129 | static int dispatch_thread_exit; |
130 | ||
131 | static pthread_t listener_thread; | |
132 | static pthread_t dispatcher_thread; | |
133 | static pthread_t worker_thread; | |
65931c8b | 134 | static pthread_t health_thread; |
b8aa1682 | 135 | |
7591bab1 MD |
136 | /* |
137 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
138 | * atomicity on 32-bit architectures. | |
139 | */ | |
140 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 141 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
142 | |
143 | /* | |
144 | * Relay command queue. | |
145 | * | |
146 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
147 | * queue. | |
148 | */ | |
58eb9381 | 149 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 | 150 | |
d3e2ba59 JD |
151 | /* Global relay stream hash table. */ |
152 | struct lttng_ht *relay_streams_ht; | |
153 | ||
92c6ca54 DG |
154 | /* Global relay viewer stream hash table. */ |
155 | struct lttng_ht *viewer_streams_ht; | |
156 | ||
7591bab1 MD |
157 | /* Global relay sessions hash table. */ |
158 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 159 | |
55706a7d | 160 | /* Relayd health monitoring */ |
eea7556c | 161 | struct health_app *health_relayd; |
55706a7d | 162 | |
cd60b05a JG |
163 | static struct option long_options[] = { |
164 | { "control-port", 1, 0, 'C', }, | |
165 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 166 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 167 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 168 | { "background", 0, 0, 'b', }, |
cd60b05a JG |
169 | { "group", 1, 0, 'g', }, |
170 | { "help", 0, 0, 'h', }, | |
171 | { "output", 1, 0, 'o', }, | |
172 | { "verbose", 0, 0, 'v', }, | |
173 | { "config", 1, 0, 'f' }, | |
297af7ff | 174 | { "version", 0, 0, 'V' }, |
76375580 | 175 | { "working-directory", 1, 0, 'w', }, |
b4d507ca JR |
176 | { "group-output-by-session", 0, 0, 's', }, |
177 | { "group-output-by-host", 0, 0, 'p', }, | |
cd60b05a JG |
178 | { NULL, 0, 0, 0, }, |
179 | }; | |
180 | ||
297af7ff | 181 | static const char *config_ignore_options[] = { "help", "config", "version" }; |
cd60b05a | 182 | |
8ff7624a JR |
183 | static void print_version(void) { |
184 | fprintf(stdout, "%s\n", VERSION); | |
185 | } | |
186 | ||
187 | static void relayd_config_log(void) | |
188 | { | |
189 | DBG("LTTng-relayd " VERSION " - " VERSION_NAME "%s%s", | |
190 | GIT_VERSION[0] == '\0' ? "" : " - " GIT_VERSION, | |
191 | EXTRA_VERSION_NAME[0] == '\0' ? "" : " - " EXTRA_VERSION_NAME); | |
192 | if (EXTRA_VERSION_DESCRIPTION[0] != '\0') { | |
193 | DBG("LTTng-relayd extra version description:\n\t" EXTRA_VERSION_DESCRIPTION "\n"); | |
194 | } | |
2987f328 JR |
195 | if (EXTRA_VERSION_PATCHES[0] != '\0') { |
196 | DBG("LTTng-relayd extra patches:\n\t" EXTRA_VERSION_PATCHES "\n"); | |
197 | } | |
8ff7624a JR |
198 | } |
199 | ||
cd60b05a JG |
200 | /* |
201 | * Take an option from the getopt output and set it in the right variable to be | |
202 | * used later. | |
203 | * | |
204 | * Return 0 on success else a negative value. | |
205 | */ | |
7591bab1 | 206 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 207 | { |
cd60b05a JG |
208 | int ret; |
209 | ||
210 | switch (opt) { | |
211 | case 0: | |
212 | fprintf(stderr, "option %s", optname); | |
213 | if (arg) { | |
214 | fprintf(stderr, " with arg %s\n", arg); | |
215 | } | |
216 | break; | |
217 | case 'C': | |
e8fa9fb0 MD |
218 | if (lttng_is_setuid_setgid()) { |
219 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
220 | "-C, --control-port"); | |
221 | } else { | |
222 | ret = uri_parse(arg, &control_uri); | |
223 | if (ret < 0) { | |
224 | ERR("Invalid control URI specified"); | |
225 | goto end; | |
226 | } | |
227 | if (control_uri->port == 0) { | |
228 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
229 | } | |
cd60b05a JG |
230 | } |
231 | break; | |
232 | case 'D': | |
e8fa9fb0 MD |
233 | if (lttng_is_setuid_setgid()) { |
234 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
235 | "-D, -data-port"); | |
236 | } else { | |
237 | ret = uri_parse(arg, &data_uri); | |
238 | if (ret < 0) { | |
239 | ERR("Invalid data URI specified"); | |
240 | goto end; | |
241 | } | |
242 | if (data_uri->port == 0) { | |
243 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
244 | } | |
cd60b05a JG |
245 | } |
246 | break; | |
8d5c808e | 247 | case 'L': |
e8fa9fb0 MD |
248 | if (lttng_is_setuid_setgid()) { |
249 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
250 | "-L, -live-port"); | |
251 | } else { | |
252 | ret = uri_parse(arg, &live_uri); | |
253 | if (ret < 0) { | |
254 | ERR("Invalid live URI specified"); | |
255 | goto end; | |
256 | } | |
257 | if (live_uri->port == 0) { | |
258 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
259 | } | |
8d5c808e AM |
260 | } |
261 | break; | |
cd60b05a JG |
262 | case 'd': |
263 | opt_daemon = 1; | |
264 | break; | |
b5218ffb MD |
265 | case 'b': |
266 | opt_background = 1; | |
267 | break; | |
cd60b05a | 268 | case 'g': |
e8fa9fb0 MD |
269 | if (lttng_is_setuid_setgid()) { |
270 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
271 | "-g, --group"); | |
272 | } else { | |
273 | tracing_group_name = strdup(arg); | |
274 | if (tracing_group_name == NULL) { | |
275 | ret = -errno; | |
276 | PERROR("strdup"); | |
277 | goto end; | |
278 | } | |
279 | tracing_group_name_override = 1; | |
330a40bb | 280 | } |
cd60b05a JG |
281 | break; |
282 | case 'h': | |
655b5cc1 PP |
283 | ret = utils_show_man_page(8, "lttng-relayd"); |
284 | if (ret) { | |
285 | ERR("Cannot view man page lttng-relayd(8)"); | |
286 | perror("exec"); | |
287 | } | |
cd60b05a | 288 | exit(EXIT_FAILURE); |
297af7ff | 289 | case 'V': |
8ff7624a JR |
290 | opt_print_version = 1; |
291 | break; | |
cd60b05a | 292 | case 'o': |
e8fa9fb0 MD |
293 | if (lttng_is_setuid_setgid()) { |
294 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
295 | "-o, --output"); | |
296 | } else { | |
297 | ret = asprintf(&opt_output_path, "%s", arg); | |
298 | if (ret < 0) { | |
299 | ret = -errno; | |
300 | PERROR("asprintf opt_output_path"); | |
301 | goto end; | |
302 | } | |
cd60b05a JG |
303 | } |
304 | break; | |
76375580 JR |
305 | case 'w': |
306 | if (lttng_is_setuid_setgid()) { | |
307 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
308 | "-w, --working-directory"); | |
309 | } else { | |
310 | ret = asprintf(&opt_working_directory, "%s", arg); | |
311 | if (ret < 0) { | |
312 | ret = -errno; | |
313 | PERROR("asprintf working_directory"); | |
314 | goto end; | |
315 | } | |
316 | } | |
317 | break; | |
318 | ||
cd60b05a JG |
319 | case 'v': |
320 | /* Verbose level can increase using multiple -v */ | |
321 | if (arg) { | |
322 | lttng_opt_verbose = config_parse_value(arg); | |
323 | } else { | |
849e5b7b DG |
324 | /* Only 3 level of verbosity (-vvv). */ |
325 | if (lttng_opt_verbose < 3) { | |
326 | lttng_opt_verbose += 1; | |
327 | } | |
cd60b05a JG |
328 | } |
329 | break; | |
b4d507ca JR |
330 | case 's': |
331 | if (opt_group_output_by_host) { | |
332 | ERR("Cannot set --group-output-by-session, --group-output-by-host already defined"); | |
333 | exit(EXIT_FAILURE); | |
334 | } | |
335 | opt_group_output_by_session = 1; | |
336 | break; | |
337 | case 'p': | |
338 | if (opt_group_output_by_session) { | |
339 | ERR("Cannot set --group-output-by-host, --group-output-by-session already defined"); | |
340 | exit(EXIT_FAILURE); | |
341 | } | |
342 | opt_group_output_by_host = 1; | |
343 | break; | |
cd60b05a JG |
344 | default: |
345 | /* Unknown option or other error. | |
346 | * Error is printed by getopt, just return */ | |
347 | ret = -1; | |
348 | goto end; | |
349 | } | |
350 | ||
351 | /* All good. */ | |
352 | ret = 0; | |
353 | ||
354 | end: | |
355 | return ret; | |
356 | } | |
357 | ||
358 | /* | |
359 | * config_entry_handler_cb used to handle options read from a config file. | |
f40ef1d5 | 360 | * See config_entry_handler_cb comment in common/config/session-config.h for the |
cd60b05a JG |
361 | * return value conventions. |
362 | */ | |
7591bab1 | 363 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
364 | { |
365 | int ret = 0, i; | |
366 | ||
367 | if (!entry || !entry->name || !entry->value) { | |
368 | ret = -EINVAL; | |
369 | goto end; | |
370 | } | |
371 | ||
372 | /* Check if the option is to be ignored */ | |
373 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
374 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
375 | goto end; | |
376 | } | |
377 | } | |
378 | ||
379 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
380 | /* Ignore if entry name is not fully matched. */ | |
381 | if (strcmp(entry->name, long_options[i].name)) { | |
382 | continue; | |
383 | } | |
384 | ||
385 | /* | |
7591bab1 MD |
386 | * If the option takes no argument on the command line, |
387 | * we have to check if the value is "true". We support | |
388 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
389 | */ |
390 | if (!long_options[i].has_arg) { | |
391 | ret = config_parse_value(entry->value); | |
392 | if (ret <= 0) { | |
393 | if (ret) { | |
394 | WARN("Invalid configuration value \"%s\" for option %s", | |
395 | entry->value, entry->name); | |
396 | } | |
397 | /* False, skip boolean config option. */ | |
398 | goto end; | |
399 | } | |
400 | } | |
401 | ||
402 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
403 | goto end; | |
404 | } | |
405 | ||
406 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
407 | entry->name); | |
408 | ||
409 | end: | |
410 | return ret; | |
411 | } | |
412 | ||
c7cc870e JR |
413 | static void parse_env_options(void) |
414 | { | |
415 | char *value = NULL; | |
416 | ||
417 | value = lttng_secure_getenv(DEFAULT_LTTNG_RELAYD_WORKING_DIRECTORY_ENV); | |
418 | if (value) { | |
419 | opt_working_directory = value; | |
420 | } | |
421 | } | |
422 | ||
7591bab1 | 423 | static int set_options(int argc, char **argv) |
cd60b05a | 424 | { |
178a0557 | 425 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
426 | int orig_optopt = optopt, orig_optind = optind; |
427 | char *default_address, *optstring; | |
428 | const char *config_path = NULL; | |
429 | ||
430 | optstring = utils_generate_optstring(long_options, | |
431 | sizeof(long_options) / sizeof(struct option)); | |
432 | if (!optstring) { | |
178a0557 | 433 | retval = -ENOMEM; |
cd60b05a JG |
434 | goto exit; |
435 | } | |
436 | ||
437 | /* Check for the --config option */ | |
438 | ||
439 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
440 | &option_index)) != -1) { | |
441 | if (c == '?') { | |
178a0557 | 442 | retval = -EINVAL; |
cd60b05a JG |
443 | goto exit; |
444 | } else if (c != 'f') { | |
445 | continue; | |
446 | } | |
447 | ||
e8fa9fb0 MD |
448 | if (lttng_is_setuid_setgid()) { |
449 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
450 | "-f, --config"); | |
451 | } else { | |
452 | config_path = utils_expand_path(optarg); | |
453 | if (!config_path) { | |
454 | ERR("Failed to resolve path: %s", optarg); | |
455 | } | |
cd60b05a JG |
456 | } |
457 | } | |
458 | ||
459 | ret = config_get_section_entries(config_path, config_section_name, | |
460 | config_entry_handler, NULL); | |
461 | if (ret) { | |
462 | if (ret > 0) { | |
463 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 464 | } |
178a0557 | 465 | retval = -1; |
cd60b05a JG |
466 | goto exit; |
467 | } | |
b8aa1682 | 468 | |
cd60b05a JG |
469 | /* Reset getopt's global state */ |
470 | optopt = orig_optopt; | |
471 | optind = orig_optind; | |
b8aa1682 | 472 | while (1) { |
cd60b05a | 473 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
474 | if (c == -1) { |
475 | break; | |
476 | } | |
477 | ||
cd60b05a JG |
478 | ret = set_option(c, optarg, long_options[option_index].name); |
479 | if (ret < 0) { | |
178a0557 | 480 | retval = -1; |
b8aa1682 JD |
481 | goto exit; |
482 | } | |
483 | } | |
484 | ||
485 | /* assign default values */ | |
486 | if (control_uri == NULL) { | |
fa91dc52 MD |
487 | ret = asprintf(&default_address, |
488 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
489 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
490 | if (ret < 0) { |
491 | PERROR("asprintf default data address"); | |
178a0557 | 492 | retval = -1; |
b8aa1682 JD |
493 | goto exit; |
494 | } | |
495 | ||
496 | ret = uri_parse(default_address, &control_uri); | |
497 | free(default_address); | |
498 | if (ret < 0) { | |
499 | ERR("Invalid control URI specified"); | |
178a0557 | 500 | retval = -1; |
b8aa1682 JD |
501 | goto exit; |
502 | } | |
503 | } | |
504 | if (data_uri == NULL) { | |
fa91dc52 MD |
505 | ret = asprintf(&default_address, |
506 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
507 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
508 | if (ret < 0) { |
509 | PERROR("asprintf default data address"); | |
178a0557 | 510 | retval = -1; |
b8aa1682 JD |
511 | goto exit; |
512 | } | |
513 | ||
514 | ret = uri_parse(default_address, &data_uri); | |
515 | free(default_address); | |
516 | if (ret < 0) { | |
517 | ERR("Invalid data URI specified"); | |
178a0557 | 518 | retval = -1; |
b8aa1682 JD |
519 | goto exit; |
520 | } | |
521 | } | |
d3e2ba59 | 522 | if (live_uri == NULL) { |
fa91dc52 MD |
523 | ret = asprintf(&default_address, |
524 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
525 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
526 | if (ret < 0) { |
527 | PERROR("asprintf default viewer control address"); | |
178a0557 | 528 | retval = -1; |
d3e2ba59 JD |
529 | goto exit; |
530 | } | |
531 | ||
532 | ret = uri_parse(default_address, &live_uri); | |
533 | free(default_address); | |
534 | if (ret < 0) { | |
535 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 536 | retval = -1; |
d3e2ba59 JD |
537 | goto exit; |
538 | } | |
539 | } | |
b8aa1682 | 540 | |
b4d507ca JR |
541 | if (!opt_group_output_by_session && !opt_group_output_by_host) { |
542 | /* Group by host by default */ | |
543 | opt_group_output_by_host = 1; | |
544 | } | |
545 | ||
b8aa1682 | 546 | exit: |
cd60b05a | 547 | free(optstring); |
178a0557 | 548 | return retval; |
b8aa1682 JD |
549 | } |
550 | ||
7591bab1 MD |
551 | static void print_global_objects(void) |
552 | { | |
553 | rcu_register_thread(); | |
554 | ||
555 | print_viewer_streams(); | |
556 | print_relay_streams(); | |
557 | print_sessions(); | |
558 | ||
559 | rcu_unregister_thread(); | |
560 | } | |
561 | ||
b8aa1682 JD |
562 | /* |
563 | * Cleanup the daemon | |
564 | */ | |
7591bab1 | 565 | static void relayd_cleanup(void) |
b8aa1682 | 566 | { |
7591bab1 MD |
567 | print_global_objects(); |
568 | ||
b8aa1682 JD |
569 | DBG("Cleaning up"); |
570 | ||
178a0557 MD |
571 | if (viewer_streams_ht) |
572 | lttng_ht_destroy(viewer_streams_ht); | |
573 | if (relay_streams_ht) | |
574 | lttng_ht_destroy(relay_streams_ht); | |
7591bab1 MD |
575 | if (sessions_ht) |
576 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 577 | |
095a4ae5 MD |
578 | /* free the dynamically allocated opt_output_path */ |
579 | free(opt_output_path); | |
580 | ||
a02de639 CB |
581 | /* Close thread quit pipes */ |
582 | utils_close_pipe(thread_quit_pipe); | |
583 | ||
710c1f73 DG |
584 | uri_free(control_uri); |
585 | uri_free(data_uri); | |
8d5c808e | 586 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
587 | |
588 | if (tracing_group_name_override) { | |
589 | free((void *) tracing_group_name); | |
590 | } | |
b8aa1682 JD |
591 | } |
592 | ||
593 | /* | |
594 | * Write to writable pipe used to notify a thread. | |
595 | */ | |
7591bab1 | 596 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 597 | { |
6cd525e8 | 598 | ssize_t ret; |
b8aa1682 | 599 | |
6cd525e8 MD |
600 | ret = lttng_write(wpipe, "!", 1); |
601 | if (ret < 1) { | |
b8aa1682 | 602 | PERROR("write poll pipe"); |
b4aacfdc | 603 | goto end; |
b8aa1682 | 604 | } |
b4aacfdc MD |
605 | ret = 0; |
606 | end: | |
b8aa1682 JD |
607 | return ret; |
608 | } | |
609 | ||
7591bab1 | 610 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 611 | { |
6cd525e8 | 612 | ssize_t ret; |
65931c8b | 613 | |
6cd525e8 MD |
614 | ret = lttng_write(pipe[1], "4", 1); |
615 | if (ret < 1) { | |
65931c8b | 616 | PERROR("write relay health quit"); |
b4aacfdc | 617 | goto end; |
65931c8b | 618 | } |
b4aacfdc MD |
619 | ret = 0; |
620 | end: | |
621 | return ret; | |
65931c8b MD |
622 | } |
623 | ||
b8aa1682 | 624 | /* |
b4aacfdc | 625 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 626 | */ |
b4aacfdc | 627 | int lttng_relay_stop_threads(void) |
b8aa1682 | 628 | { |
b4aacfdc | 629 | int retval = 0; |
b8aa1682 JD |
630 | |
631 | /* Stopping all threads */ | |
632 | DBG("Terminating all threads"); | |
b4aacfdc | 633 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 634 | ERR("write error on thread quit pipe"); |
b4aacfdc | 635 | retval = -1; |
b8aa1682 JD |
636 | } |
637 | ||
b4aacfdc MD |
638 | if (notify_health_quit_pipe(health_quit_pipe)) { |
639 | ERR("write error on health quit pipe"); | |
640 | } | |
65931c8b | 641 | |
b8aa1682 | 642 | /* Dispatch thread */ |
26c9d55e | 643 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 644 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 645 | |
b4aacfdc | 646 | if (relayd_live_stop()) { |
178a0557 | 647 | ERR("Error stopping live threads"); |
b4aacfdc | 648 | retval = -1; |
178a0557 | 649 | } |
b4aacfdc | 650 | return retval; |
b8aa1682 JD |
651 | } |
652 | ||
653 | /* | |
654 | * Signal handler for the daemon | |
655 | * | |
656 | * Simply stop all worker threads, leaving main() return gracefully after | |
657 | * joining all threads and calling cleanup(). | |
658 | */ | |
7591bab1 | 659 | static void sighandler(int sig) |
b8aa1682 JD |
660 | { |
661 | switch (sig) { | |
b8aa1682 JD |
662 | case SIGINT: |
663 | DBG("SIGINT caught"); | |
b4aacfdc MD |
664 | if (lttng_relay_stop_threads()) { |
665 | ERR("Error stopping threads"); | |
666 | } | |
b8aa1682 JD |
667 | break; |
668 | case SIGTERM: | |
669 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
670 | if (lttng_relay_stop_threads()) { |
671 | ERR("Error stopping threads"); | |
672 | } | |
b8aa1682 | 673 | break; |
3fd27398 MD |
674 | case SIGUSR1: |
675 | CMM_STORE_SHARED(recv_child_signal, 1); | |
676 | break; | |
b8aa1682 JD |
677 | default: |
678 | break; | |
679 | } | |
680 | } | |
681 | ||
682 | /* | |
683 | * Setup signal handler for : | |
684 | * SIGINT, SIGTERM, SIGPIPE | |
685 | */ | |
7591bab1 | 686 | static int set_signal_handler(void) |
b8aa1682 JD |
687 | { |
688 | int ret = 0; | |
689 | struct sigaction sa; | |
690 | sigset_t sigset; | |
691 | ||
692 | if ((ret = sigemptyset(&sigset)) < 0) { | |
693 | PERROR("sigemptyset"); | |
694 | return ret; | |
695 | } | |
696 | ||
b8aa1682 JD |
697 | sa.sa_mask = sigset; |
698 | sa.sa_flags = 0; | |
0072e5e2 MD |
699 | |
700 | sa.sa_handler = sighandler; | |
b8aa1682 JD |
701 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { |
702 | PERROR("sigaction"); | |
703 | return ret; | |
704 | } | |
705 | ||
706 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
707 | PERROR("sigaction"); | |
708 | return ret; | |
709 | } | |
710 | ||
0072e5e2 | 711 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
b8aa1682 JD |
712 | PERROR("sigaction"); |
713 | return ret; | |
714 | } | |
715 | ||
0072e5e2 MD |
716 | sa.sa_handler = SIG_IGN; |
717 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
3fd27398 MD |
718 | PERROR("sigaction"); |
719 | return ret; | |
720 | } | |
721 | ||
722 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
723 | |
724 | return ret; | |
725 | } | |
726 | ||
3fd27398 MD |
727 | void lttng_relay_notify_ready(void) |
728 | { | |
729 | /* Notify the parent of the fork() process that we are ready. */ | |
730 | if (opt_daemon || opt_background) { | |
731 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
732 | kill(child_ppid, SIGUSR1); | |
733 | } | |
734 | } | |
735 | } | |
736 | ||
b8aa1682 JD |
737 | /* |
738 | * Init thread quit pipe. | |
739 | * | |
740 | * Return -1 on error or 0 if all pipes are created. | |
741 | */ | |
7591bab1 | 742 | static int init_thread_quit_pipe(void) |
b8aa1682 | 743 | { |
a02de639 | 744 | int ret; |
b8aa1682 | 745 | |
a02de639 | 746 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 747 | |
b8aa1682 JD |
748 | return ret; |
749 | } | |
750 | ||
751 | /* | |
752 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
753 | */ | |
7591bab1 | 754 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
755 | { |
756 | int ret; | |
757 | ||
758 | if (events == NULL || size == 0) { | |
759 | ret = -1; | |
760 | goto error; | |
761 | } | |
762 | ||
763 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
764 | if (ret < 0) { | |
765 | goto error; | |
766 | } | |
767 | ||
768 | /* Add quit pipe */ | |
c7759e6a | 769 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
770 | if (ret < 0) { |
771 | goto error; | |
772 | } | |
773 | ||
774 | return 0; | |
775 | ||
776 | error: | |
777 | return ret; | |
778 | } | |
779 | ||
780 | /* | |
781 | * Check if the thread quit pipe was triggered. | |
782 | * | |
783 | * Return 1 if it was triggered else 0; | |
784 | */ | |
7591bab1 | 785 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
786 | { |
787 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
788 | return 1; | |
789 | } | |
790 | ||
791 | return 0; | |
792 | } | |
793 | ||
794 | /* | |
795 | * Create and init socket from uri. | |
796 | */ | |
7591bab1 | 797 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
798 | { |
799 | int ret; | |
800 | struct lttcomm_sock *sock = NULL; | |
801 | ||
802 | sock = lttcomm_alloc_sock_from_uri(uri); | |
803 | if (sock == NULL) { | |
804 | ERR("Allocating socket"); | |
805 | goto error; | |
806 | } | |
807 | ||
808 | ret = lttcomm_create_sock(sock); | |
809 | if (ret < 0) { | |
810 | goto error; | |
811 | } | |
812 | DBG("Listening on sock %d", sock->fd); | |
813 | ||
814 | ret = sock->ops->bind(sock); | |
815 | if (ret < 0) { | |
816 | goto error; | |
817 | } | |
818 | ||
819 | ret = sock->ops->listen(sock, -1); | |
820 | if (ret < 0) { | |
821 | goto error; | |
822 | ||
823 | } | |
824 | ||
825 | return sock; | |
826 | ||
827 | error: | |
828 | if (sock) { | |
829 | lttcomm_destroy_sock(sock); | |
830 | } | |
831 | return NULL; | |
832 | } | |
833 | ||
834 | /* | |
835 | * This thread manages the listening for new connections on the network | |
836 | */ | |
7591bab1 | 837 | static void *relay_thread_listener(void *data) |
b8aa1682 | 838 | { |
095a4ae5 | 839 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
840 | uint32_t revents, nb_fd; |
841 | struct lttng_poll_event events; | |
842 | struct lttcomm_sock *control_sock, *data_sock; | |
843 | ||
b8aa1682 JD |
844 | DBG("[thread] Relay listener started"); |
845 | ||
55706a7d MD |
846 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
847 | ||
f385ae0a MD |
848 | health_code_update(); |
849 | ||
7591bab1 | 850 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 851 | if (!control_sock) { |
095a4ae5 | 852 | goto error_sock_control; |
b8aa1682 JD |
853 | } |
854 | ||
7591bab1 | 855 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 856 | if (!data_sock) { |
095a4ae5 | 857 | goto error_sock_relay; |
b8aa1682 JD |
858 | } |
859 | ||
860 | /* | |
7591bab1 MD |
861 | * Pass 3 as size here for the thread quit pipe, control and |
862 | * data socket. | |
b8aa1682 JD |
863 | */ |
864 | ret = create_thread_poll_set(&events, 3); | |
865 | if (ret < 0) { | |
866 | goto error_create_poll; | |
867 | } | |
868 | ||
869 | /* Add the control socket */ | |
870 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
871 | if (ret < 0) { | |
872 | goto error_poll_add; | |
873 | } | |
874 | ||
875 | /* Add the data socket */ | |
876 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
877 | if (ret < 0) { | |
878 | goto error_poll_add; | |
879 | } | |
880 | ||
3fd27398 MD |
881 | lttng_relay_notify_ready(); |
882 | ||
9b5e0863 MD |
883 | if (testpoint(relayd_thread_listener)) { |
884 | goto error_testpoint; | |
885 | } | |
886 | ||
b8aa1682 | 887 | while (1) { |
f385ae0a MD |
888 | health_code_update(); |
889 | ||
b8aa1682 JD |
890 | DBG("Listener accepting connections"); |
891 | ||
b8aa1682 | 892 | restart: |
f385ae0a | 893 | health_poll_entry(); |
b8aa1682 | 894 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 895 | health_poll_exit(); |
b8aa1682 JD |
896 | if (ret < 0) { |
897 | /* | |
898 | * Restart interrupted system call. | |
899 | */ | |
900 | if (errno == EINTR) { | |
901 | goto restart; | |
902 | } | |
903 | goto error; | |
904 | } | |
905 | ||
0d9c5d77 DG |
906 | nb_fd = ret; |
907 | ||
b8aa1682 JD |
908 | DBG("Relay new connection received"); |
909 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
910 | health_code_update(); |
911 | ||
b8aa1682 JD |
912 | /* Fetch once the poll data */ |
913 | revents = LTTNG_POLL_GETEV(&events, i); | |
914 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
915 | ||
fd20dac9 | 916 | if (!revents) { |
7591bab1 MD |
917 | /* |
918 | * No activity for this FD (poll | |
919 | * implementation). | |
920 | */ | |
fd20dac9 MD |
921 | continue; |
922 | } | |
923 | ||
b8aa1682 JD |
924 | /* Thread quit pipe has been closed. Killing thread. */ |
925 | ret = check_thread_quit_pipe(pollfd, revents); | |
926 | if (ret) { | |
095a4ae5 MD |
927 | err = 0; |
928 | goto exit; | |
b8aa1682 JD |
929 | } |
930 | ||
03e43155 | 931 | if (revents & LPOLLIN) { |
4b7f17b2 | 932 | /* |
7591bab1 MD |
933 | * A new connection is requested, therefore a |
934 | * sessiond/consumerd connection is allocated in | |
935 | * this thread, enqueued to a global queue and | |
936 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 937 | */ |
58eb9381 DG |
938 | int val = 1; |
939 | struct relay_connection *new_conn; | |
4b7f17b2 | 940 | struct lttcomm_sock *newsock; |
7591bab1 | 941 | enum connection_type type; |
b8aa1682 JD |
942 | |
943 | if (pollfd == data_sock->fd) { | |
7591bab1 | 944 | type = RELAY_DATA; |
b8aa1682 | 945 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
946 | DBG("Relay data connection accepted, socket %d", |
947 | newsock->fd); | |
4b7f17b2 MD |
948 | } else { |
949 | assert(pollfd == control_sock->fd); | |
7591bab1 | 950 | type = RELAY_CONTROL; |
b8aa1682 | 951 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
952 | DBG("Relay control connection accepted, socket %d", |
953 | newsock->fd); | |
b8aa1682 | 954 | } |
58eb9381 DG |
955 | if (!newsock) { |
956 | PERROR("accepting sock"); | |
58eb9381 DG |
957 | goto error; |
958 | } | |
959 | ||
960 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
961 | sizeof(val)); | |
b8aa1682 JD |
962 | if (ret < 0) { |
963 | PERROR("setsockopt inet"); | |
4b7f17b2 | 964 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
965 | goto error; |
966 | } | |
2fc6b1ab JR |
967 | |
968 | ret = socket_apply_keep_alive_config(newsock->fd); | |
969 | if (ret < 0) { | |
970 | ERR("Failed to apply TCP keep-alive configuration on socket (%i)", | |
971 | newsock->fd); | |
972 | lttcomm_destroy_sock(newsock); | |
973 | goto error; | |
974 | } | |
975 | ||
7591bab1 MD |
976 | new_conn = connection_create(newsock, type); |
977 | if (!new_conn) { | |
978 | lttcomm_destroy_sock(newsock); | |
979 | goto error; | |
980 | } | |
58eb9381 DG |
981 | |
982 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
983 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
984 | &new_conn->qnode); | |
b8aa1682 JD |
985 | |
986 | /* | |
7591bab1 MD |
987 | * Wake the dispatch queue futex. |
988 | * Implicit memory barrier with the | |
989 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 990 | */ |
58eb9381 | 991 | futex_nto1_wake(&relay_conn_queue.futex); |
03e43155 MD |
992 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
993 | ERR("socket poll error"); | |
994 | goto error; | |
995 | } else { | |
996 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
997 | goto error; | |
b8aa1682 JD |
998 | } |
999 | } | |
1000 | } | |
1001 | ||
095a4ae5 | 1002 | exit: |
b8aa1682 JD |
1003 | error: |
1004 | error_poll_add: | |
9b5e0863 | 1005 | error_testpoint: |
b8aa1682 JD |
1006 | lttng_poll_clean(&events); |
1007 | error_create_poll: | |
095a4ae5 MD |
1008 | if (data_sock->fd >= 0) { |
1009 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
1010 | if (ret) { |
1011 | PERROR("close"); | |
1012 | } | |
b8aa1682 | 1013 | } |
095a4ae5 MD |
1014 | lttcomm_destroy_sock(data_sock); |
1015 | error_sock_relay: | |
1016 | if (control_sock->fd >= 0) { | |
1017 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
1018 | if (ret) { |
1019 | PERROR("close"); | |
1020 | } | |
b8aa1682 | 1021 | } |
095a4ae5 MD |
1022 | lttcomm_destroy_sock(control_sock); |
1023 | error_sock_control: | |
1024 | if (err) { | |
f385ae0a MD |
1025 | health_error(); |
1026 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 1027 | } |
55706a7d | 1028 | health_unregister(health_relayd); |
b8aa1682 | 1029 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 1030 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1031 | return NULL; |
1032 | } | |
1033 | ||
1034 | /* | |
1035 | * This thread manages the dispatching of the requests to worker threads | |
1036 | */ | |
7591bab1 | 1037 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 1038 | { |
6cd525e8 MD |
1039 | int err = -1; |
1040 | ssize_t ret; | |
8bdee6e2 | 1041 | struct cds_wfcq_node *node; |
58eb9381 | 1042 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
1043 | |
1044 | DBG("[thread] Relay dispatcher started"); | |
1045 | ||
55706a7d MD |
1046 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
1047 | ||
9b5e0863 MD |
1048 | if (testpoint(relayd_thread_dispatcher)) { |
1049 | goto error_testpoint; | |
1050 | } | |
1051 | ||
f385ae0a MD |
1052 | health_code_update(); |
1053 | ||
68924dc1 | 1054 | for (;;) { |
f385ae0a MD |
1055 | health_code_update(); |
1056 | ||
b8aa1682 | 1057 | /* Atomically prepare the queue futex */ |
58eb9381 | 1058 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 | 1059 | |
68924dc1 MD |
1060 | if (CMM_LOAD_SHARED(dispatch_thread_exit)) { |
1061 | break; | |
1062 | } | |
1063 | ||
b8aa1682 | 1064 | do { |
f385ae0a MD |
1065 | health_code_update(); |
1066 | ||
b8aa1682 | 1067 | /* Dequeue commands */ |
8bdee6e2 SM |
1068 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
1069 | &relay_conn_queue.tail); | |
b8aa1682 JD |
1070 | if (node == NULL) { |
1071 | DBG("Woken up but nothing in the relay command queue"); | |
1072 | /* Continue thread execution */ | |
1073 | break; | |
1074 | } | |
58eb9381 | 1075 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 1076 | |
58eb9381 | 1077 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1078 | |
1079 | /* | |
7591bab1 MD |
1080 | * Inform worker thread of the new request. This |
1081 | * call is blocking so we can be assured that | |
1082 | * the data will be read at some point in time | |
1083 | * or wait to the end of the world :) | |
b8aa1682 | 1084 | */ |
58eb9381 DG |
1085 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1086 | if (ret < 0) { | |
1087 | PERROR("write connection pipe"); | |
7591bab1 | 1088 | connection_put(new_conn); |
b8aa1682 JD |
1089 | goto error; |
1090 | } | |
1091 | } while (node != NULL); | |
1092 | ||
1093 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1094 | health_poll_entry(); |
58eb9381 | 1095 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1096 | health_poll_exit(); |
b8aa1682 JD |
1097 | } |
1098 | ||
f385ae0a MD |
1099 | /* Normal exit, no error */ |
1100 | err = 0; | |
1101 | ||
b8aa1682 | 1102 | error: |
9b5e0863 | 1103 | error_testpoint: |
f385ae0a MD |
1104 | if (err) { |
1105 | health_error(); | |
1106 | ERR("Health error occurred in %s", __func__); | |
1107 | } | |
55706a7d | 1108 | health_unregister(health_relayd); |
b8aa1682 | 1109 | DBG("Dispatch thread dying"); |
b4aacfdc | 1110 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1111 | return NULL; |
1112 | } | |
1113 | ||
b8aa1682 | 1114 | /* |
7591bab1 | 1115 | * Set index data from the control port to a given index object. |
b8aa1682 | 1116 | */ |
7591bab1 | 1117 | static int set_index_control_data(struct relay_index *index, |
234cd636 JD |
1118 | struct lttcomm_relayd_index *data, |
1119 | struct relay_connection *conn) | |
1c20f0e2 | 1120 | { |
7591bab1 | 1121 | struct ctf_packet_index index_data; |
1c20f0e2 JD |
1122 | |
1123 | /* | |
44a2fbf1 | 1124 | * The index on disk is encoded in big endian. |
1c20f0e2 | 1125 | */ |
44a2fbf1 JG |
1126 | index_data.packet_size = htobe64(data->packet_size); |
1127 | index_data.content_size = htobe64(data->content_size); | |
1128 | index_data.timestamp_begin = htobe64(data->timestamp_begin); | |
1129 | index_data.timestamp_end = htobe64(data->timestamp_end); | |
1130 | index_data.events_discarded = htobe64(data->events_discarded); | |
1131 | index_data.stream_id = htobe64(data->stream_id); | |
234cd636 JD |
1132 | |
1133 | if (conn->minor >= 8) { | |
44a2fbf1 JG |
1134 | index->index_data.stream_instance_id = htobe64(data->stream_instance_id); |
1135 | index->index_data.packet_seq_num = htobe64(data->packet_seq_num); | |
234cd636 JD |
1136 | } |
1137 | ||
7591bab1 | 1138 | return relay_index_set_data(index, &index_data); |
1c20f0e2 JD |
1139 | } |
1140 | ||
c5b6f4f0 DG |
1141 | /* |
1142 | * Handle the RELAYD_CREATE_SESSION command. | |
1143 | * | |
1144 | * On success, send back the session id or else return a negative value. | |
1145 | */ | |
44a2fbf1 JG |
1146 | static int relay_create_session(const struct lttcomm_relayd_hdr *recv_hdr, |
1147 | struct relay_connection *conn, | |
1148 | const struct lttng_buffer_view *payload) | |
c5b6f4f0 | 1149 | { |
44a2fbf1 JG |
1150 | int ret = 0; |
1151 | ssize_t send_ret; | |
c5b6f4f0 DG |
1152 | struct relay_session *session; |
1153 | struct lttcomm_relayd_status_session reply; | |
36d2e35d | 1154 | char session_name[LTTNG_NAME_MAX]; |
9cf3eb20 | 1155 | char hostname[LTTNG_HOST_NAME_MAX]; |
7591bab1 MD |
1156 | uint32_t live_timer = 0; |
1157 | bool snapshot = false; | |
c5b6f4f0 | 1158 | |
36d2e35d | 1159 | memset(session_name, 0, LTTNG_NAME_MAX); |
9cf3eb20 | 1160 | memset(hostname, 0, LTTNG_HOST_NAME_MAX); |
c5b6f4f0 DG |
1161 | |
1162 | memset(&reply, 0, sizeof(reply)); | |
1163 | ||
58eb9381 | 1164 | switch (conn->minor) { |
2a174661 DG |
1165 | case 1: |
1166 | case 2: | |
1167 | case 3: | |
1168 | break; | |
1169 | case 4: /* LTTng sessiond 2.4 */ | |
1170 | default: | |
44a2fbf1 | 1171 | ret = cmd_create_session_2_4(payload, session_name, |
7591bab1 MD |
1172 | hostname, &live_timer, &snapshot); |
1173 | } | |
1174 | if (ret < 0) { | |
1175 | goto send_reply; | |
d3e2ba59 JD |
1176 | } |
1177 | ||
7591bab1 MD |
1178 | session = session_create(session_name, hostname, live_timer, |
1179 | snapshot, conn->major, conn->minor); | |
1180 | if (!session) { | |
1181 | ret = -1; | |
1182 | goto send_reply; | |
1183 | } | |
1184 | assert(!conn->session); | |
1185 | conn->session = session; | |
c5b6f4f0 DG |
1186 | DBG("Created session %" PRIu64, session->id); |
1187 | ||
7591bab1 MD |
1188 | reply.session_id = htobe64(session->id); |
1189 | ||
1190 | send_reply: | |
c5b6f4f0 DG |
1191 | if (ret < 0) { |
1192 | reply.ret_code = htobe32(LTTNG_ERR_FATAL); | |
1193 | } else { | |
1194 | reply.ret_code = htobe32(LTTNG_OK); | |
1195 | } | |
1196 | ||
58eb9381 | 1197 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
44a2fbf1 JG |
1198 | if (send_ret < (ssize_t) sizeof(reply)) { |
1199 | ERR("Failed to send \"create session\" command reply (ret = %zd)", | |
1200 | send_ret); | |
1201 | ret = -1; | |
c5b6f4f0 DG |
1202 | } |
1203 | ||
1204 | return ret; | |
1205 | } | |
1206 | ||
a4baae1b JD |
1207 | /* |
1208 | * When we have received all the streams and the metadata for a channel, | |
1209 | * we make them visible to the viewer threads. | |
1210 | */ | |
7591bab1 | 1211 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1212 | { |
7591bab1 MD |
1213 | struct relay_stream *stream; |
1214 | struct relay_session *session = conn->session; | |
a4baae1b | 1215 | |
7591bab1 MD |
1216 | /* |
1217 | * We publish all streams belonging to a session atomically wrt | |
1218 | * session lock. | |
1219 | */ | |
1220 | pthread_mutex_lock(&session->lock); | |
1221 | rcu_read_lock(); | |
1222 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1223 | recv_node) { | |
1224 | stream_publish(stream); | |
a4baae1b | 1225 | } |
7591bab1 | 1226 | rcu_read_unlock(); |
a4baae1b | 1227 | |
7591bab1 MD |
1228 | /* |
1229 | * Inform the viewer that there are new streams in the session. | |
1230 | */ | |
1231 | if (session->viewer_attached) { | |
1232 | uatomic_set(&session->new_streams, 1); | |
1233 | } | |
1234 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1235 | } |
1236 | ||
b8aa1682 JD |
1237 | /* |
1238 | * relay_add_stream: allocate a new stream for a session | |
1239 | */ | |
44a2fbf1 JG |
1240 | static int relay_add_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1241 | struct relay_connection *conn, | |
1242 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1243 | { |
7591bab1 MD |
1244 | int ret; |
1245 | ssize_t send_ret; | |
58eb9381 | 1246 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1247 | struct relay_stream *stream = NULL; |
1248 | struct lttcomm_relayd_status_stream reply; | |
4030a636 | 1249 | struct ctf_trace *trace = NULL; |
7591bab1 MD |
1250 | uint64_t stream_handle = -1ULL; |
1251 | char *path_name = NULL, *channel_name = NULL; | |
1252 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
b8aa1682 | 1253 | |
44a2fbf1 | 1254 | if (!session || !conn->version_check_done) { |
b8aa1682 JD |
1255 | ERR("Trying to add a stream before version check"); |
1256 | ret = -1; | |
1257 | goto end_no_session; | |
1258 | } | |
1259 | ||
7591bab1 MD |
1260 | switch (session->minor) { |
1261 | case 1: /* LTTng sessiond 2.1. Allocates path_name and channel_name. */ | |
44a2fbf1 | 1262 | ret = cmd_recv_stream_2_1(payload, &path_name, |
b4d507ca | 1263 | &channel_name, session); |
0f907de1 | 1264 | break; |
7591bab1 | 1265 | case 2: /* LTTng sessiond 2.2. Allocates path_name and channel_name. */ |
0f907de1 | 1266 | default: |
44a2fbf1 | 1267 | ret = cmd_recv_stream_2_2(payload, &path_name, |
b4d507ca JR |
1268 | &channel_name, &tracefile_size, &tracefile_count, |
1269 | session); | |
0f907de1 JD |
1270 | break; |
1271 | } | |
1272 | if (ret < 0) { | |
7591bab1 | 1273 | goto send_reply; |
b8aa1682 JD |
1274 | } |
1275 | ||
7591bab1 | 1276 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1277 | if (!trace) { |
7591bab1 | 1278 | goto send_reply; |
2a174661 | 1279 | } |
7591bab1 | 1280 | /* This stream here has one reference on the trace. */ |
2a174661 | 1281 | |
7591bab1 MD |
1282 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1283 | stream_handle = ++last_relay_stream_id; | |
1284 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1285 | |
7591bab1 MD |
1286 | /* We pass ownership of path_name and channel_name. */ |
1287 | stream = stream_create(trace, stream_handle, path_name, | |
1288 | channel_name, tracefile_size, tracefile_count); | |
1289 | path_name = NULL; | |
1290 | channel_name = NULL; | |
a4baae1b | 1291 | |
2a174661 | 1292 | /* |
7591bab1 MD |
1293 | * Streams are the owners of their trace. Reference to trace is |
1294 | * kept within stream_create(). | |
2a174661 | 1295 | */ |
7591bab1 | 1296 | ctf_trace_put(trace); |
d3e2ba59 | 1297 | |
7591bab1 | 1298 | send_reply: |
53efb85a | 1299 | memset(&reply, 0, sizeof(reply)); |
7591bab1 MD |
1300 | reply.handle = htobe64(stream_handle); |
1301 | if (!stream) { | |
f73fabfd | 1302 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1303 | } else { |
f73fabfd | 1304 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1305 | } |
5af40280 | 1306 | |
58eb9381 | 1307 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 | 1308 | sizeof(struct lttcomm_relayd_status_stream), 0); |
44a2fbf1 JG |
1309 | if (send_ret < (ssize_t) sizeof(reply)) { |
1310 | ERR("Failed to send \"add stream\" command reply (ret = %zd)", | |
1311 | send_ret); | |
1312 | ret = -1; | |
b8aa1682 JD |
1313 | } |
1314 | ||
1315 | end_no_session: | |
7591bab1 MD |
1316 | free(path_name); |
1317 | free(channel_name); | |
0f907de1 | 1318 | return ret; |
b8aa1682 JD |
1319 | } |
1320 | ||
173af62f DG |
1321 | /* |
1322 | * relay_close_stream: close a specific stream | |
1323 | */ | |
44a2fbf1 JG |
1324 | static int relay_close_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1325 | struct relay_connection *conn, | |
1326 | const struct lttng_buffer_view *payload) | |
173af62f | 1327 | { |
44a2fbf1 JG |
1328 | int ret; |
1329 | ssize_t send_ret; | |
58eb9381 | 1330 | struct relay_session *session = conn->session; |
173af62f DG |
1331 | struct lttcomm_relayd_close_stream stream_info; |
1332 | struct lttcomm_relayd_generic_reply reply; | |
1333 | struct relay_stream *stream; | |
173af62f DG |
1334 | |
1335 | DBG("Close stream received"); | |
1336 | ||
44a2fbf1 | 1337 | if (!session || !conn->version_check_done) { |
173af62f DG |
1338 | ERR("Trying to close a stream before version check"); |
1339 | ret = -1; | |
1340 | goto end_no_session; | |
1341 | } | |
1342 | ||
44a2fbf1 JG |
1343 | if (payload->size < sizeof(stream_info)) { |
1344 | ERR("Unexpected payload size in \"relay_close_stream\": expected >= %zu bytes, got %zu bytes", | |
1345 | sizeof(stream_info), payload->size); | |
173af62f DG |
1346 | ret = -1; |
1347 | goto end_no_session; | |
1348 | } | |
44a2fbf1 JG |
1349 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1350 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1351 | stream_info.last_net_seq_num = be64toh(stream_info.last_net_seq_num); | |
173af62f | 1352 | |
44a2fbf1 | 1353 | stream = stream_get_by_id(stream_info.stream_id); |
173af62f DG |
1354 | if (!stream) { |
1355 | ret = -1; | |
7591bab1 | 1356 | goto end; |
173af62f | 1357 | } |
77f7bd85 MD |
1358 | |
1359 | /* | |
1360 | * Set last_net_seq_num before the close flag. Required by data | |
1361 | * pending check. | |
1362 | */ | |
7591bab1 | 1363 | pthread_mutex_lock(&stream->lock); |
44a2fbf1 | 1364 | stream->last_net_seq_num = stream_info.last_net_seq_num; |
77f7bd85 MD |
1365 | pthread_mutex_unlock(&stream->lock); |
1366 | ||
bda7c7b9 JG |
1367 | /* |
1368 | * This is one of the conditions which may trigger a stream close | |
1369 | * with the others being: | |
1370 | * 1) A close command is received for a stream | |
1371 | * 2) The control connection owning the stream is closed | |
1372 | * 3) We have received all of the stream's data _after_ a close | |
1373 | * request. | |
1374 | */ | |
1375 | try_stream_close(stream); | |
7591bab1 MD |
1376 | if (stream->is_metadata) { |
1377 | struct relay_viewer_stream *vstream; | |
173af62f | 1378 | |
7591bab1 MD |
1379 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1380 | if (vstream) { | |
1381 | if (vstream->metadata_sent == stream->metadata_received) { | |
1382 | /* | |
1383 | * Since all the metadata has been sent to the | |
1384 | * viewer and that we have a request to close | |
1385 | * its stream, we can safely teardown the | |
1386 | * corresponding metadata viewer stream. | |
1387 | */ | |
1388 | viewer_stream_put(vstream); | |
1389 | } | |
1390 | /* Put local reference. */ | |
1391 | viewer_stream_put(vstream); | |
1392 | } | |
1393 | } | |
7591bab1 | 1394 | stream_put(stream); |
44a2fbf1 | 1395 | ret = 0; |
173af62f | 1396 | |
7591bab1 | 1397 | end: |
53efb85a | 1398 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1399 | if (ret < 0) { |
f73fabfd | 1400 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1401 | } else { |
f73fabfd | 1402 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1403 | } |
58eb9381 | 1404 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f | 1405 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
44a2fbf1 JG |
1406 | if (send_ret < (ssize_t) sizeof(reply)) { |
1407 | ERR("Failed to send \"close stream\" command reply (ret = %zd)", | |
1408 | send_ret); | |
1409 | ret = -1; | |
173af62f DG |
1410 | } |
1411 | ||
1412 | end_no_session: | |
1413 | return ret; | |
1414 | } | |
1415 | ||
93ec662e JD |
1416 | /* |
1417 | * relay_reset_metadata: reset a metadata stream | |
1418 | */ | |
1419 | static | |
44a2fbf1 JG |
1420 | int relay_reset_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1421 | struct relay_connection *conn, | |
1422 | const struct lttng_buffer_view *payload) | |
93ec662e | 1423 | { |
44a2fbf1 JG |
1424 | int ret; |
1425 | ssize_t send_ret; | |
93ec662e JD |
1426 | struct relay_session *session = conn->session; |
1427 | struct lttcomm_relayd_reset_metadata stream_info; | |
1428 | struct lttcomm_relayd_generic_reply reply; | |
1429 | struct relay_stream *stream; | |
1430 | ||
1431 | DBG("Reset metadata received"); | |
1432 | ||
44a2fbf1 | 1433 | if (!session || !conn->version_check_done) { |
93ec662e JD |
1434 | ERR("Trying to reset a metadata stream before version check"); |
1435 | ret = -1; | |
1436 | goto end_no_session; | |
1437 | } | |
1438 | ||
44a2fbf1 JG |
1439 | if (payload->size < sizeof(stream_info)) { |
1440 | ERR("Unexpected payload size in \"relay_reset_metadata\": expected >= %zu bytes, got %zu bytes", | |
1441 | sizeof(stream_info), payload->size); | |
93ec662e JD |
1442 | ret = -1; |
1443 | goto end_no_session; | |
1444 | } | |
44a2fbf1 JG |
1445 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1446 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1447 | stream_info.version = be64toh(stream_info.version); | |
1448 | ||
1449 | DBG("Update metadata to version %" PRIu64, stream_info.version); | |
93ec662e JD |
1450 | |
1451 | /* Unsupported for live sessions for now. */ | |
1452 | if (session->live_timer != 0) { | |
1453 | ret = -1; | |
1454 | goto end; | |
1455 | } | |
1456 | ||
44a2fbf1 | 1457 | stream = stream_get_by_id(stream_info.stream_id); |
93ec662e JD |
1458 | if (!stream) { |
1459 | ret = -1; | |
1460 | goto end; | |
1461 | } | |
1462 | pthread_mutex_lock(&stream->lock); | |
1463 | if (!stream->is_metadata) { | |
1464 | ret = -1; | |
1465 | goto end_unlock; | |
1466 | } | |
1467 | ||
1468 | ret = utils_rotate_stream_file(stream->path_name, stream->channel_name, | |
1469 | 0, 0, -1, -1, stream->stream_fd->fd, NULL, | |
1470 | &stream->stream_fd->fd); | |
1471 | if (ret < 0) { | |
1472 | ERR("Failed to rotate metadata file %s of channel %s", | |
1473 | stream->path_name, stream->channel_name); | |
1474 | goto end_unlock; | |
1475 | } | |
1476 | ||
1477 | end_unlock: | |
1478 | pthread_mutex_unlock(&stream->lock); | |
1479 | stream_put(stream); | |
1480 | ||
1481 | end: | |
1482 | memset(&reply, 0, sizeof(reply)); | |
1483 | if (ret < 0) { | |
1484 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1485 | } else { | |
1486 | reply.ret_code = htobe32(LTTNG_OK); | |
1487 | } | |
1488 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1489 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
44a2fbf1 JG |
1490 | if (send_ret < (ssize_t) sizeof(reply)) { |
1491 | ERR("Failed to send \"reset metadata\" command reply (ret = %zd)", | |
1492 | send_ret); | |
1493 | ret = -1; | |
93ec662e JD |
1494 | } |
1495 | ||
1496 | end_no_session: | |
1497 | return ret; | |
1498 | } | |
1499 | ||
b8aa1682 JD |
1500 | /* |
1501 | * relay_unknown_command: send -1 if received unknown command | |
1502 | */ | |
7591bab1 | 1503 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1504 | { |
1505 | struct lttcomm_relayd_generic_reply reply; | |
44a2fbf1 | 1506 | ssize_t send_ret; |
b8aa1682 | 1507 | |
53efb85a | 1508 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1509 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
44a2fbf1 JG |
1510 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1511 | if (send_ret < sizeof(reply)) { | |
1512 | ERR("Failed to send \"unknown command\" command reply (ret = %zd)", send_ret); | |
b8aa1682 JD |
1513 | } |
1514 | } | |
1515 | ||
1516 | /* | |
1517 | * relay_start: send an acknowledgment to the client to tell if we are | |
1518 | * ready to receive data. We are ready if a session is established. | |
1519 | */ | |
44a2fbf1 JG |
1520 | static int relay_start(const struct lttcomm_relayd_hdr *recv_hdr, |
1521 | struct relay_connection *conn, | |
1522 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1523 | { |
44a2fbf1 JG |
1524 | int ret = 0; |
1525 | ssize_t send_ret; | |
b8aa1682 | 1526 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1527 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1528 | |
1529 | if (!session) { | |
1530 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1531 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1532 | } |
1533 | ||
53efb85a | 1534 | memset(&reply, 0, sizeof(reply)); |
44a2fbf1 JG |
1535 | reply.ret_code = htobe32(LTTNG_OK); |
1536 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1537 | sizeof(reply), 0); | |
1538 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1539 | ERR("Failed to send \"relay_start\" command reply (ret = %zd)", | |
1540 | send_ret); | |
1541 | ret = -1; | |
b8aa1682 JD |
1542 | } |
1543 | ||
1544 | return ret; | |
1545 | } | |
1546 | ||
1d4dfdef DG |
1547 | /* |
1548 | * Append padding to the file pointed by the file descriptor fd. | |
1549 | */ | |
1550 | static int write_padding_to_file(int fd, uint32_t size) | |
1551 | { | |
6cd525e8 | 1552 | ssize_t ret = 0; |
1d4dfdef DG |
1553 | char *zeros; |
1554 | ||
1555 | if (size == 0) { | |
1556 | goto end; | |
1557 | } | |
1558 | ||
1559 | zeros = zmalloc(size); | |
1560 | if (zeros == NULL) { | |
1561 | PERROR("zmalloc zeros for padding"); | |
1562 | ret = -1; | |
1563 | goto end; | |
1564 | } | |
1565 | ||
6cd525e8 MD |
1566 | ret = lttng_write(fd, zeros, size); |
1567 | if (ret < size) { | |
1d4dfdef DG |
1568 | PERROR("write padding to file"); |
1569 | } | |
1570 | ||
e986c7a1 DG |
1571 | free(zeros); |
1572 | ||
1d4dfdef DG |
1573 | end: |
1574 | return ret; | |
1575 | } | |
1576 | ||
b8aa1682 | 1577 | /* |
7591bab1 | 1578 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1579 | */ |
44a2fbf1 JG |
1580 | static int relay_recv_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1581 | struct relay_connection *conn, | |
1582 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1583 | { |
32d1569c | 1584 | int ret = 0; |
6cd525e8 | 1585 | ssize_t size_ret; |
58eb9381 | 1586 | struct relay_session *session = conn->session; |
44a2fbf1 | 1587 | struct lttcomm_relayd_metadata_payload metadata_payload_header; |
b8aa1682 | 1588 | struct relay_stream *metadata_stream; |
44a2fbf1 | 1589 | uint64_t metadata_payload_size; |
b8aa1682 JD |
1590 | |
1591 | if (!session) { | |
1592 | ERR("Metadata sent before version check"); | |
1593 | ret = -1; | |
1594 | goto end; | |
1595 | } | |
1596 | ||
44a2fbf1 | 1597 | if (recv_hdr->data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { |
f6416125 MD |
1598 | ERR("Incorrect data size"); |
1599 | ret = -1; | |
1600 | goto end; | |
1601 | } | |
44a2fbf1 JG |
1602 | metadata_payload_size = recv_hdr->data_size - |
1603 | sizeof(struct lttcomm_relayd_metadata_payload); | |
f6416125 | 1604 | |
44a2fbf1 JG |
1605 | memcpy(&metadata_payload_header, payload->data, |
1606 | sizeof(metadata_payload_header)); | |
1607 | metadata_payload_header.stream_id = be64toh( | |
1608 | metadata_payload_header.stream_id); | |
1609 | metadata_payload_header.padding_size = be32toh( | |
1610 | metadata_payload_header.padding_size); | |
c617c0c6 | 1611 | |
44a2fbf1 | 1612 | metadata_stream = stream_get_by_id(metadata_payload_header.stream_id); |
b8aa1682 JD |
1613 | if (!metadata_stream) { |
1614 | ret = -1; | |
7591bab1 | 1615 | goto end; |
b8aa1682 JD |
1616 | } |
1617 | ||
7591bab1 MD |
1618 | pthread_mutex_lock(&metadata_stream->lock); |
1619 | ||
44a2fbf1 JG |
1620 | size_ret = lttng_write(metadata_stream->stream_fd->fd, |
1621 | payload->data + sizeof(metadata_payload_header), | |
1622 | metadata_payload_size); | |
1623 | if (size_ret < metadata_payload_size) { | |
b8aa1682 JD |
1624 | ERR("Relay error writing metadata on file"); |
1625 | ret = -1; | |
7591bab1 | 1626 | goto end_put; |
b8aa1682 | 1627 | } |
1d4dfdef | 1628 | |
32d1569c | 1629 | size_ret = write_padding_to_file(metadata_stream->stream_fd->fd, |
44a2fbf1 JG |
1630 | metadata_payload_header.padding_size); |
1631 | if (size_ret < (int64_t) metadata_payload_header.padding_size) { | |
1632 | ret = -1; | |
7591bab1 | 1633 | goto end_put; |
1d4dfdef | 1634 | } |
2a174661 | 1635 | |
7591bab1 | 1636 | metadata_stream->metadata_received += |
44a2fbf1 | 1637 | metadata_payload_size + metadata_payload_header.padding_size; |
7591bab1 MD |
1638 | DBG2("Relay metadata written. Updated metadata_received %" PRIu64, |
1639 | metadata_stream->metadata_received); | |
1d4dfdef | 1640 | |
7591bab1 MD |
1641 | end_put: |
1642 | pthread_mutex_unlock(&metadata_stream->lock); | |
1643 | stream_put(metadata_stream); | |
b8aa1682 JD |
1644 | end: |
1645 | return ret; | |
1646 | } | |
1647 | ||
1648 | /* | |
1649 | * relay_send_version: send relayd version number | |
1650 | */ | |
44a2fbf1 JG |
1651 | static int relay_send_version(const struct lttcomm_relayd_hdr *recv_hdr, |
1652 | struct relay_connection *conn, | |
1653 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1654 | { |
7f51dcba | 1655 | int ret; |
44a2fbf1 | 1656 | ssize_t send_ret; |
092b6259 | 1657 | struct lttcomm_relayd_version reply, msg; |
40096142 | 1658 | bool compatible = true; |
b8aa1682 | 1659 | |
44a2fbf1 | 1660 | conn->version_check_done = true; |
b8aa1682 | 1661 | |
092b6259 | 1662 | /* Get version from the other side. */ |
44a2fbf1 JG |
1663 | if (payload->size < sizeof(msg)) { |
1664 | ERR("Unexpected payload size in \"relay_send_version\": expected >= %zu bytes, got %zu bytes", | |
1665 | sizeof(msg), payload->size); | |
092b6259 | 1666 | ret = -1; |
092b6259 DG |
1667 | goto end; |
1668 | } | |
1669 | ||
44a2fbf1 JG |
1670 | memcpy(&msg, payload->data, sizeof(msg)); |
1671 | msg.major = be32toh(msg.major); | |
1672 | msg.minor = be32toh(msg.minor); | |
1673 | ||
53efb85a | 1674 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1675 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1676 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1677 | |
1678 | /* Major versions must be the same */ | |
44a2fbf1 | 1679 | if (reply.major != msg.major) { |
6151a90f | 1680 | DBG("Incompatible major versions (%u vs %u), deleting session", |
44a2fbf1 | 1681 | reply.major, msg.major); |
40096142 | 1682 | compatible = false; |
d4519fa3 JD |
1683 | } |
1684 | ||
58eb9381 | 1685 | conn->major = reply.major; |
0f907de1 | 1686 | /* We adapt to the lowest compatible version */ |
44a2fbf1 | 1687 | if (reply.minor <= msg.minor) { |
58eb9381 | 1688 | conn->minor = reply.minor; |
0f907de1 | 1689 | } else { |
44a2fbf1 | 1690 | conn->minor = msg.minor; |
0f907de1 JD |
1691 | } |
1692 | ||
6151a90f JD |
1693 | reply.major = htobe32(reply.major); |
1694 | reply.minor = htobe32(reply.minor); | |
44a2fbf1 JG |
1695 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
1696 | sizeof(reply), 0); | |
1697 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1698 | ERR("Failed to send \"send version\" command reply (ret = %zd)", | |
1699 | send_ret); | |
1700 | ret = -1; | |
1701 | goto end; | |
1702 | } else { | |
1703 | ret = 0; | |
6151a90f JD |
1704 | } |
1705 | ||
40096142 JD |
1706 | if (!compatible) { |
1707 | ret = -1; | |
1708 | goto end; | |
1709 | } | |
1710 | ||
58eb9381 DG |
1711 | DBG("Version check done using protocol %u.%u", conn->major, |
1712 | conn->minor); | |
b8aa1682 JD |
1713 | |
1714 | end: | |
1715 | return ret; | |
1716 | } | |
1717 | ||
c8f59ee5 | 1718 | /* |
6d805429 | 1719 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1720 | */ |
44a2fbf1 JG |
1721 | static int relay_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1722 | struct relay_connection *conn, | |
1723 | const struct lttng_buffer_view *payload) | |
c8f59ee5 | 1724 | { |
58eb9381 | 1725 | struct relay_session *session = conn->session; |
6d805429 | 1726 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1727 | struct lttcomm_relayd_generic_reply reply; |
1728 | struct relay_stream *stream; | |
44a2fbf1 | 1729 | ssize_t send_ret; |
c8f59ee5 | 1730 | int ret; |
c8f59ee5 | 1731 | |
6d805429 | 1732 | DBG("Data pending command received"); |
c8f59ee5 | 1733 | |
44a2fbf1 | 1734 | if (!session || !conn->version_check_done) { |
c8f59ee5 DG |
1735 | ERR("Trying to check for data before version check"); |
1736 | ret = -1; | |
1737 | goto end_no_session; | |
1738 | } | |
1739 | ||
44a2fbf1 JG |
1740 | if (payload->size < sizeof(msg)) { |
1741 | ERR("Unexpected payload size in \"relay_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1742 | sizeof(msg), payload->size); | |
c8f59ee5 DG |
1743 | ret = -1; |
1744 | goto end_no_session; | |
1745 | } | |
44a2fbf1 JG |
1746 | memcpy(&msg, payload->data, sizeof(msg)); |
1747 | msg.stream_id = be64toh(msg.stream_id); | |
1748 | msg.last_net_seq_num = be64toh(msg.last_net_seq_num); | |
c8f59ee5 | 1749 | |
44a2fbf1 | 1750 | stream = stream_get_by_id(msg.stream_id); |
de91f48a | 1751 | if (stream == NULL) { |
c8f59ee5 | 1752 | ret = -1; |
7591bab1 | 1753 | goto end; |
c8f59ee5 DG |
1754 | } |
1755 | ||
7591bab1 MD |
1756 | pthread_mutex_lock(&stream->lock); |
1757 | ||
6d805429 | 1758 | DBG("Data pending for stream id %" PRIu64 " prev_seq %" PRIu64 |
44a2fbf1 JG |
1759 | " and last_seq %" PRIu64, msg.stream_id, |
1760 | stream->prev_seq, msg.last_net_seq_num); | |
c8f59ee5 | 1761 | |
33832e64 | 1762 | /* Avoid wrapping issue */ |
44a2fbf1 | 1763 | if (((int64_t) (stream->prev_seq - msg.last_net_seq_num)) >= 0) { |
6d805429 | 1764 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1765 | ret = 0; |
6d805429 DG |
1766 | } else { |
1767 | /* Data still being streamed thus pending */ | |
1768 | ret = 1; | |
c8f59ee5 DG |
1769 | } |
1770 | ||
7591bab1 MD |
1771 | stream->data_pending_check_done = true; |
1772 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1773 | |
7591bab1 MD |
1774 | stream_put(stream); |
1775 | end: | |
c8f59ee5 | 1776 | |
53efb85a | 1777 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1778 | reply.ret_code = htobe32(ret); |
44a2fbf1 JG |
1779 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1780 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1781 | ERR("Failed to send \"data pending\" command reply (ret = %zd)", | |
1782 | send_ret); | |
1783 | ret = -1; | |
c8f59ee5 DG |
1784 | } |
1785 | ||
1786 | end_no_session: | |
1787 | return ret; | |
1788 | } | |
1789 | ||
1790 | /* | |
1791 | * Wait for the control socket to reach a quiescent state. | |
1792 | * | |
7591bab1 MD |
1793 | * Note that for now, when receiving this command from the session |
1794 | * daemon, this means that every subsequent commands or data received on | |
1795 | * the control socket has been handled. So, this is why we simply return | |
1796 | * OK here. | |
c8f59ee5 | 1797 | */ |
44a2fbf1 JG |
1798 | static int relay_quiescent_control(const struct lttcomm_relayd_hdr *recv_hdr, |
1799 | struct relay_connection *conn, | |
1800 | const struct lttng_buffer_view *payload) | |
c8f59ee5 DG |
1801 | { |
1802 | int ret; | |
44a2fbf1 | 1803 | ssize_t send_ret; |
ad7051c0 | 1804 | struct relay_stream *stream; |
ad7051c0 | 1805 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1806 | struct lttcomm_relayd_generic_reply reply; |
1807 | ||
1808 | DBG("Checking quiescent state on control socket"); | |
1809 | ||
44a2fbf1 | 1810 | if (!conn->session || !conn->version_check_done) { |
ad7051c0 DG |
1811 | ERR("Trying to check for data before version check"); |
1812 | ret = -1; | |
1813 | goto end_no_session; | |
1814 | } | |
1815 | ||
44a2fbf1 JG |
1816 | if (payload->size < sizeof(msg)) { |
1817 | ERR("Unexpected payload size in \"relay_quiescent_control\": expected >= %zu bytes, got %zu bytes", | |
1818 | sizeof(msg), payload->size); | |
ad7051c0 DG |
1819 | ret = -1; |
1820 | goto end_no_session; | |
1821 | } | |
44a2fbf1 JG |
1822 | memcpy(&msg, payload->data, sizeof(msg)); |
1823 | msg.stream_id = be64toh(msg.stream_id); | |
ad7051c0 | 1824 | |
44a2fbf1 | 1825 | stream = stream_get_by_id(msg.stream_id); |
7591bab1 MD |
1826 | if (!stream) { |
1827 | goto reply; | |
1828 | } | |
1829 | pthread_mutex_lock(&stream->lock); | |
1830 | stream->data_pending_check_done = true; | |
1831 | pthread_mutex_unlock(&stream->lock); | |
44a2fbf1 JG |
1832 | |
1833 | DBG("Relay quiescent control pending flag set to %" PRIu64, msg.stream_id); | |
7591bab1 MD |
1834 | stream_put(stream); |
1835 | reply: | |
53efb85a | 1836 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1837 | reply.ret_code = htobe32(LTTNG_OK); |
44a2fbf1 JG |
1838 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1839 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1840 | ERR("Failed to send \"quiescent control\" command reply (ret = %zd)", | |
1841 | send_ret); | |
1842 | ret = -1; | |
1843 | } else { | |
1844 | ret = 0; | |
c8f59ee5 DG |
1845 | } |
1846 | ||
ad7051c0 | 1847 | end_no_session: |
c8f59ee5 DG |
1848 | return ret; |
1849 | } | |
1850 | ||
f7079f67 | 1851 | /* |
7591bab1 MD |
1852 | * Initialize a data pending command. This means that a consumer is about |
1853 | * to ask for data pending for each stream it holds. Simply iterate over | |
1854 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
1855 | * |
1856 | * This command returns to the client a LTTNG_OK code. | |
1857 | */ | |
44a2fbf1 JG |
1858 | static int relay_begin_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1859 | struct relay_connection *conn, | |
1860 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
1861 | { |
1862 | int ret; | |
44a2fbf1 | 1863 | ssize_t send_ret; |
f7079f67 DG |
1864 | struct lttng_ht_iter iter; |
1865 | struct lttcomm_relayd_begin_data_pending msg; | |
1866 | struct lttcomm_relayd_generic_reply reply; | |
1867 | struct relay_stream *stream; | |
f7079f67 DG |
1868 | |
1869 | assert(recv_hdr); | |
58eb9381 | 1870 | assert(conn); |
f7079f67 DG |
1871 | |
1872 | DBG("Init streams for data pending"); | |
1873 | ||
44a2fbf1 | 1874 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
1875 | ERR("Trying to check for data before version check"); |
1876 | ret = -1; | |
1877 | goto end_no_session; | |
1878 | } | |
1879 | ||
44a2fbf1 JG |
1880 | if (payload->size < sizeof(msg)) { |
1881 | ERR("Unexpected payload size in \"relay_begin_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1882 | sizeof(msg), payload->size); | |
f7079f67 DG |
1883 | ret = -1; |
1884 | goto end_no_session; | |
1885 | } | |
44a2fbf1 JG |
1886 | memcpy(&msg, payload->data, sizeof(msg)); |
1887 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 DG |
1888 | |
1889 | /* | |
7591bab1 MD |
1890 | * Iterate over all streams to set the begin data pending flag. |
1891 | * For now, the streams are indexed by stream handle so we have | |
1892 | * to iterate over all streams to find the one associated with | |
1893 | * the right session_id. | |
f7079f67 DG |
1894 | */ |
1895 | rcu_read_lock(); | |
d3e2ba59 | 1896 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1897 | node.node) { |
7591bab1 MD |
1898 | if (!stream_get(stream)) { |
1899 | continue; | |
1900 | } | |
44a2fbf1 | 1901 | if (stream->trace->session->id == msg.session_id) { |
7591bab1 MD |
1902 | pthread_mutex_lock(&stream->lock); |
1903 | stream->data_pending_check_done = false; | |
1904 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
1905 | DBG("Set begin data pending flag to stream %" PRIu64, |
1906 | stream->stream_handle); | |
1907 | } | |
7591bab1 | 1908 | stream_put(stream); |
f7079f67 DG |
1909 | } |
1910 | rcu_read_unlock(); | |
1911 | ||
53efb85a | 1912 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1913 | /* All good, send back reply. */ |
1914 | reply.ret_code = htobe32(LTTNG_OK); | |
1915 | ||
44a2fbf1 JG |
1916 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1917 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1918 | ERR("Failed to send \"begin data pending\" command reply (ret = %zd)", | |
1919 | send_ret); | |
1920 | ret = -1; | |
1921 | } else { | |
1922 | ret = 0; | |
f7079f67 DG |
1923 | } |
1924 | ||
1925 | end_no_session: | |
1926 | return ret; | |
1927 | } | |
1928 | ||
1929 | /* | |
7591bab1 MD |
1930 | * End data pending command. This will check, for a given session id, if |
1931 | * each stream associated with it has its data_pending_check_done flag | |
1932 | * set. If not, this means that the client lost track of the stream but | |
1933 | * the data is still being streamed on our side. In this case, we inform | |
1934 | * the client that data is in flight. | |
f7079f67 DG |
1935 | * |
1936 | * Return to the client if there is data in flight or not with a ret_code. | |
1937 | */ | |
44a2fbf1 JG |
1938 | static int relay_end_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1939 | struct relay_connection *conn, | |
1940 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
1941 | { |
1942 | int ret; | |
44a2fbf1 | 1943 | ssize_t send_ret; |
f7079f67 DG |
1944 | struct lttng_ht_iter iter; |
1945 | struct lttcomm_relayd_end_data_pending msg; | |
1946 | struct lttcomm_relayd_generic_reply reply; | |
1947 | struct relay_stream *stream; | |
f7079f67 DG |
1948 | uint32_t is_data_inflight = 0; |
1949 | ||
f7079f67 DG |
1950 | DBG("End data pending command"); |
1951 | ||
44a2fbf1 | 1952 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
1953 | ERR("Trying to check for data before version check"); |
1954 | ret = -1; | |
1955 | goto end_no_session; | |
1956 | } | |
1957 | ||
44a2fbf1 JG |
1958 | if (payload->size < sizeof(msg)) { |
1959 | ERR("Unexpected payload size in \"relay_end_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1960 | sizeof(msg), payload->size); | |
f7079f67 DG |
1961 | ret = -1; |
1962 | goto end_no_session; | |
1963 | } | |
44a2fbf1 JG |
1964 | memcpy(&msg, payload->data, sizeof(msg)); |
1965 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 | 1966 | |
7591bab1 MD |
1967 | /* |
1968 | * Iterate over all streams to see if the begin data pending | |
1969 | * flag is set. | |
1970 | */ | |
f7079f67 | 1971 | rcu_read_lock(); |
d3e2ba59 | 1972 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 1973 | node.node) { |
7591bab1 MD |
1974 | if (!stream_get(stream)) { |
1975 | continue; | |
1976 | } | |
44a2fbf1 | 1977 | if (stream->trace->session->id != msg.session_id) { |
7591bab1 MD |
1978 | stream_put(stream); |
1979 | continue; | |
1980 | } | |
1981 | pthread_mutex_lock(&stream->lock); | |
1982 | if (!stream->data_pending_check_done) { | |
1983 | if (!stream->closed || !(((int64_t) (stream->prev_seq - stream->last_net_seq_num)) >= 0)) { | |
1984 | is_data_inflight = 1; | |
1985 | DBG("Data is still in flight for stream %" PRIu64, | |
1986 | stream->stream_handle); | |
1987 | pthread_mutex_unlock(&stream->lock); | |
1988 | stream_put(stream); | |
1989 | break; | |
1990 | } | |
f7079f67 | 1991 | } |
7591bab1 MD |
1992 | pthread_mutex_unlock(&stream->lock); |
1993 | stream_put(stream); | |
f7079f67 DG |
1994 | } |
1995 | rcu_read_unlock(); | |
1996 | ||
53efb85a | 1997 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
1998 | /* All good, send back reply. */ |
1999 | reply.ret_code = htobe32(is_data_inflight); | |
2000 | ||
44a2fbf1 JG |
2001 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2002 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2003 | ERR("Failed to send \"end data pending\" command reply (ret = %zd)", | |
2004 | send_ret); | |
2005 | ret = -1; | |
2006 | } else { | |
2007 | ret = 0; | |
f7079f67 DG |
2008 | } |
2009 | ||
2010 | end_no_session: | |
2011 | return ret; | |
2012 | } | |
2013 | ||
1c20f0e2 JD |
2014 | /* |
2015 | * Receive an index for a specific stream. | |
2016 | * | |
2017 | * Return 0 on success else a negative value. | |
2018 | */ | |
44a2fbf1 JG |
2019 | static int relay_recv_index(const struct lttcomm_relayd_hdr *recv_hdr, |
2020 | struct relay_connection *conn, | |
2021 | const struct lttng_buffer_view *payload) | |
1c20f0e2 | 2022 | { |
44a2fbf1 JG |
2023 | int ret; |
2024 | ssize_t send_ret; | |
58eb9381 | 2025 | struct relay_session *session = conn->session; |
1c20f0e2 | 2026 | struct lttcomm_relayd_index index_info; |
7591bab1 | 2027 | struct relay_index *index; |
1c20f0e2 JD |
2028 | struct lttcomm_relayd_generic_reply reply; |
2029 | struct relay_stream *stream; | |
e0547b83 | 2030 | size_t msg_len; |
1c20f0e2 | 2031 | |
58eb9381 | 2032 | assert(conn); |
1c20f0e2 JD |
2033 | |
2034 | DBG("Relay receiving index"); | |
2035 | ||
44a2fbf1 | 2036 | if (!session || !conn->version_check_done) { |
1c20f0e2 JD |
2037 | ERR("Trying to close a stream before version check"); |
2038 | ret = -1; | |
2039 | goto end_no_session; | |
2040 | } | |
2041 | ||
e0547b83 MD |
2042 | msg_len = lttcomm_relayd_index_len( |
2043 | lttng_to_index_major(conn->major, conn->minor), | |
2044 | lttng_to_index_minor(conn->major, conn->minor)); | |
44a2fbf1 JG |
2045 | if (payload->size < msg_len) { |
2046 | ERR("Unexpected payload size in \"relay_recv_index\": expected >= %zu bytes, got %zu bytes", | |
2047 | msg_len, payload->size); | |
1c20f0e2 JD |
2048 | ret = -1; |
2049 | goto end_no_session; | |
2050 | } | |
44a2fbf1 JG |
2051 | memcpy(&index_info, payload->data, msg_len); |
2052 | index_info.relay_stream_id = be64toh(index_info.relay_stream_id); | |
2053 | index_info.net_seq_num = be64toh(index_info.net_seq_num); | |
2054 | index_info.packet_size = be64toh(index_info.packet_size); | |
2055 | index_info.content_size = be64toh(index_info.content_size); | |
2056 | index_info.timestamp_begin = be64toh(index_info.timestamp_begin); | |
2057 | index_info.timestamp_end = be64toh(index_info.timestamp_end); | |
2058 | index_info.events_discarded = be64toh(index_info.events_discarded); | |
2059 | index_info.stream_id = be64toh(index_info.stream_id); | |
1c20f0e2 | 2060 | |
44a2fbf1 JG |
2061 | if (conn->minor >= 8) { |
2062 | index_info.stream_instance_id = | |
2063 | be64toh(index_info.stream_instance_id); | |
2064 | index_info.packet_seq_num = be64toh(index_info.packet_seq_num); | |
2065 | } | |
1c20f0e2 | 2066 | |
44a2fbf1 | 2067 | stream = stream_get_by_id(index_info.relay_stream_id); |
1c20f0e2 | 2068 | if (!stream) { |
7591bab1 | 2069 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 2070 | ret = -1; |
7591bab1 | 2071 | goto end; |
1c20f0e2 | 2072 | } |
7591bab1 | 2073 | pthread_mutex_lock(&stream->lock); |
1c20f0e2 | 2074 | |
d3e2ba59 JD |
2075 | /* Live beacon handling */ |
2076 | if (index_info.packet_size == 0) { | |
7591bab1 MD |
2077 | DBG("Received live beacon for stream %" PRIu64, |
2078 | stream->stream_handle); | |
d3e2ba59 JD |
2079 | |
2080 | /* | |
7591bab1 MD |
2081 | * Only flag a stream inactive when it has already |
2082 | * received data and no indexes are in flight. | |
d3e2ba59 | 2083 | */ |
a44ca2ca | 2084 | if (stream->index_received_seqcount > 0 |
7591bab1 | 2085 | && stream->indexes_in_flight == 0) { |
44a2fbf1 | 2086 | stream->beacon_ts_end = index_info.timestamp_end; |
d3e2ba59 JD |
2087 | } |
2088 | ret = 0; | |
7591bab1 | 2089 | goto end_stream_put; |
d3e2ba59 JD |
2090 | } else { |
2091 | stream->beacon_ts_end = -1ULL; | |
2092 | } | |
2093 | ||
528f2ffa | 2094 | if (stream->ctf_stream_id == -1ULL) { |
44a2fbf1 | 2095 | stream->ctf_stream_id = index_info.stream_id; |
528f2ffa | 2096 | } |
44a2fbf1 | 2097 | index = relay_index_get_by_id_or_create(stream, index_info.net_seq_num); |
7591bab1 MD |
2098 | if (!index) { |
2099 | ret = -1; | |
2100 | ERR("relay_index_get_by_id_or_create index NULL"); | |
2101 | goto end_stream_put; | |
1c20f0e2 | 2102 | } |
234cd636 | 2103 | if (set_index_control_data(index, &index_info, conn)) { |
7591bab1 MD |
2104 | ERR("set_index_control_data error"); |
2105 | relay_index_put(index); | |
2106 | ret = -1; | |
2107 | goto end_stream_put; | |
2108 | } | |
2109 | ret = relay_index_try_flush(index); | |
2110 | if (ret == 0) { | |
a44ca2ca MD |
2111 | tracefile_array_commit_seq(stream->tfa); |
2112 | stream->index_received_seqcount++; | |
7591bab1 MD |
2113 | } else if (ret > 0) { |
2114 | /* no flush. */ | |
2115 | ret = 0; | |
2116 | } else { | |
2117 | ERR("relay_index_try_flush error %d", ret); | |
2118 | relay_index_put(index); | |
2119 | ret = -1; | |
1c20f0e2 JD |
2120 | } |
2121 | ||
7591bab1 MD |
2122 | end_stream_put: |
2123 | pthread_mutex_unlock(&stream->lock); | |
2124 | stream_put(stream); | |
2125 | ||
2126 | end: | |
1c20f0e2 | 2127 | |
53efb85a | 2128 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
2129 | if (ret < 0) { |
2130 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
2131 | } else { | |
2132 | reply.ret_code = htobe32(LTTNG_OK); | |
2133 | } | |
58eb9381 | 2134 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
44a2fbf1 JG |
2135 | if (send_ret < (ssize_t) sizeof(reply)) { |
2136 | ERR("Failed to send \"recv index\" command reply (ret = %zd)", send_ret); | |
2137 | ret = -1; | |
1c20f0e2 JD |
2138 | } |
2139 | ||
2140 | end_no_session: | |
2141 | return ret; | |
2142 | } | |
2143 | ||
a4baae1b JD |
2144 | /* |
2145 | * Receive the streams_sent message. | |
2146 | * | |
2147 | * Return 0 on success else a negative value. | |
2148 | */ | |
44a2fbf1 JG |
2149 | static int relay_streams_sent(const struct lttcomm_relayd_hdr *recv_hdr, |
2150 | struct relay_connection *conn, | |
2151 | const struct lttng_buffer_view *payload) | |
a4baae1b | 2152 | { |
44a2fbf1 JG |
2153 | int ret; |
2154 | ssize_t send_ret; | |
a4baae1b JD |
2155 | struct lttcomm_relayd_generic_reply reply; |
2156 | ||
58eb9381 | 2157 | assert(conn); |
a4baae1b JD |
2158 | |
2159 | DBG("Relay receiving streams_sent"); | |
2160 | ||
44a2fbf1 | 2161 | if (!conn->session || !conn->version_check_done) { |
a4baae1b JD |
2162 | ERR("Trying to close a stream before version check"); |
2163 | ret = -1; | |
2164 | goto end_no_session; | |
2165 | } | |
2166 | ||
2167 | /* | |
7591bab1 MD |
2168 | * Publish every pending stream in the connection recv list which are |
2169 | * now ready to be used by the viewer. | |
4a9daf17 | 2170 | */ |
7591bab1 | 2171 | publish_connection_local_streams(conn); |
4a9daf17 | 2172 | |
53efb85a | 2173 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2174 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2175 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
44a2fbf1 JG |
2176 | if (send_ret < (ssize_t) sizeof(reply)) { |
2177 | ERR("Failed to send \"streams sent\" command reply (ret = %zd)", | |
2178 | send_ret); | |
2179 | ret = -1; | |
a4baae1b JD |
2180 | } else { |
2181 | /* Success. */ | |
2182 | ret = 0; | |
2183 | } | |
2184 | ||
2185 | end_no_session: | |
2186 | return ret; | |
2187 | } | |
2188 | ||
44a2fbf1 JG |
2189 | #define DBG_CMD(cmd_name, conn) \ |
2190 | DBG3("Processing \"%s\" command for socket %i", cmd_name, conn->sock->fd); | |
2191 | ||
2192 | static int relay_process_control_command(struct relay_connection *conn, | |
2193 | const struct lttcomm_relayd_hdr *header, | |
2194 | const struct lttng_buffer_view *payload) | |
b8aa1682 JD |
2195 | { |
2196 | int ret = 0; | |
2197 | ||
44a2fbf1 | 2198 | switch (header->cmd) { |
b8aa1682 | 2199 | case RELAYD_CREATE_SESSION: |
44a2fbf1 JG |
2200 | DBG_CMD("RELAYD_CREATE_SESSION", conn); |
2201 | ret = relay_create_session(header, conn, payload); | |
b8aa1682 | 2202 | break; |
b8aa1682 | 2203 | case RELAYD_ADD_STREAM: |
44a2fbf1 JG |
2204 | DBG_CMD("RELAYD_ADD_STREAM", conn); |
2205 | ret = relay_add_stream(header, conn, payload); | |
b8aa1682 JD |
2206 | break; |
2207 | case RELAYD_START_DATA: | |
44a2fbf1 JG |
2208 | DBG_CMD("RELAYD_START_DATA", conn); |
2209 | ret = relay_start(header, conn, payload); | |
b8aa1682 JD |
2210 | break; |
2211 | case RELAYD_SEND_METADATA: | |
44a2fbf1 JG |
2212 | DBG_CMD("RELAYD_SEND_METADATA", conn); |
2213 | ret = relay_recv_metadata(header, conn, payload); | |
b8aa1682 JD |
2214 | break; |
2215 | case RELAYD_VERSION: | |
44a2fbf1 JG |
2216 | DBG_CMD("RELAYD_VERSION", conn); |
2217 | ret = relay_send_version(header, conn, payload); | |
b8aa1682 | 2218 | break; |
173af62f | 2219 | case RELAYD_CLOSE_STREAM: |
44a2fbf1 JG |
2220 | DBG_CMD("RELAYD_CLOSE_STREAM", conn); |
2221 | ret = relay_close_stream(header, conn, payload); | |
173af62f | 2222 | break; |
6d805429 | 2223 | case RELAYD_DATA_PENDING: |
44a2fbf1 JG |
2224 | DBG_CMD("RELAYD_DATA_PENDING", conn); |
2225 | ret = relay_data_pending(header, conn, payload); | |
c8f59ee5 DG |
2226 | break; |
2227 | case RELAYD_QUIESCENT_CONTROL: | |
44a2fbf1 JG |
2228 | DBG_CMD("RELAYD_QUIESCENT_CONTROL", conn); |
2229 | ret = relay_quiescent_control(header, conn, payload); | |
c8f59ee5 | 2230 | break; |
f7079f67 | 2231 | case RELAYD_BEGIN_DATA_PENDING: |
44a2fbf1 JG |
2232 | DBG_CMD("RELAYD_BEGIN_DATA_PENDING", conn); |
2233 | ret = relay_begin_data_pending(header, conn, payload); | |
f7079f67 DG |
2234 | break; |
2235 | case RELAYD_END_DATA_PENDING: | |
44a2fbf1 JG |
2236 | DBG_CMD("RELAYD_END_DATA_PENDING", conn); |
2237 | ret = relay_end_data_pending(header, conn, payload); | |
f7079f67 | 2238 | break; |
1c20f0e2 | 2239 | case RELAYD_SEND_INDEX: |
44a2fbf1 JG |
2240 | DBG_CMD("RELAYD_SEND_INDEX", conn); |
2241 | ret = relay_recv_index(header, conn, payload); | |
1c20f0e2 | 2242 | break; |
a4baae1b | 2243 | case RELAYD_STREAMS_SENT: |
44a2fbf1 JG |
2244 | DBG_CMD("RELAYD_STREAMS_SENT", conn); |
2245 | ret = relay_streams_sent(header, conn, payload); | |
a4baae1b | 2246 | break; |
93ec662e | 2247 | case RELAYD_RESET_METADATA: |
44a2fbf1 JG |
2248 | DBG_CMD("RELAYD_RESET_METADATA", conn); |
2249 | ret = relay_reset_metadata(header, conn, payload); | |
93ec662e | 2250 | break; |
b8aa1682 JD |
2251 | case RELAYD_UPDATE_SYNC_INFO: |
2252 | default: | |
44a2fbf1 | 2253 | ERR("Received unknown command (%u)", header->cmd); |
58eb9381 | 2254 | relay_unknown_command(conn); |
b8aa1682 JD |
2255 | ret = -1; |
2256 | goto end; | |
2257 | } | |
2258 | ||
2259 | end: | |
2260 | return ret; | |
2261 | } | |
2262 | ||
44a2fbf1 JG |
2263 | static enum relay_connection_status relay_process_control_receive_payload( |
2264 | struct relay_connection *conn) | |
2265 | { | |
2266 | int ret = 0; | |
2267 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2268 | struct lttng_dynamic_buffer *reception_buffer = | |
2269 | &conn->protocol.ctrl.reception_buffer; | |
2270 | struct ctrl_connection_state_receive_payload *state = | |
2271 | &conn->protocol.ctrl.state.receive_payload; | |
2272 | struct lttng_buffer_view payload_view; | |
2273 | ||
2274 | if (state->left_to_receive == 0) { | |
2275 | /* Short-circuit for payload-less commands. */ | |
2276 | goto reception_complete; | |
2277 | } | |
2278 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2279 | reception_buffer->data + state->received, | |
2280 | state->left_to_receive, MSG_DONTWAIT); | |
2281 | if (ret < 0) { | |
2282 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2283 | PERROR("Unable to receive command payload on sock %d", | |
2284 | conn->sock->fd); | |
2285 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2286 | } | |
2287 | goto end; | |
2288 | } else if (ret == 0) { | |
2289 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
2290 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
2291 | goto end; | |
2292 | } | |
2293 | ||
2294 | assert(ret > 0); | |
2295 | assert(ret <= state->left_to_receive); | |
2296 | ||
2297 | state->left_to_receive -= ret; | |
2298 | state->received += ret; | |
2299 | ||
2300 | if (state->left_to_receive > 0) { | |
2301 | /* | |
2302 | * Can't transition to the protocol's next state, wait to | |
2303 | * receive the rest of the header. | |
2304 | */ | |
2305 | DBG3("Partial reception of control connection protocol payload (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2306 | state->received, state->left_to_receive, | |
2307 | conn->sock->fd); | |
2308 | goto end; | |
2309 | } | |
2310 | ||
2311 | reception_complete: | |
2312 | DBG("Done receiving control command payload: fd = %i, payload size = %" PRIu64 " bytes", | |
2313 | conn->sock->fd, state->received); | |
2314 | /* | |
2315 | * The payload required to process the command has been received. | |
2316 | * A view to the reception buffer is forwarded to the various | |
2317 | * commands and the state of the control is reset on success. | |
2318 | * | |
2319 | * Commands are responsible for sending their reply to the peer. | |
2320 | */ | |
2321 | payload_view = lttng_buffer_view_from_dynamic_buffer(reception_buffer, | |
2322 | 0, -1); | |
2323 | ret = relay_process_control_command(conn, | |
2324 | &state->header, &payload_view); | |
2325 | if (ret < 0) { | |
2326 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2327 | goto end; | |
2328 | } | |
2329 | ||
2330 | ret = connection_reset_protocol_state(conn); | |
2331 | if (ret) { | |
2332 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2333 | } | |
2334 | end: | |
2335 | return status; | |
2336 | } | |
2337 | ||
2338 | static enum relay_connection_status relay_process_control_receive_header( | |
2339 | struct relay_connection *conn) | |
2340 | { | |
2341 | int ret = 0; | |
2342 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2343 | struct lttcomm_relayd_hdr header; | |
2344 | struct lttng_dynamic_buffer *reception_buffer = | |
2345 | &conn->protocol.ctrl.reception_buffer; | |
2346 | struct ctrl_connection_state_receive_header *state = | |
2347 | &conn->protocol.ctrl.state.receive_header; | |
2348 | ||
2349 | assert(state->left_to_receive != 0); | |
2350 | ||
2351 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2352 | reception_buffer->data + state->received, | |
2353 | state->left_to_receive, MSG_DONTWAIT); | |
2354 | if (ret < 0) { | |
2355 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2356 | PERROR("Unable to receive control command header on sock %d", | |
2357 | conn->sock->fd); | |
2358 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2359 | } | |
2360 | goto end; | |
2361 | } else if (ret == 0) { | |
2362 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
2363 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
2364 | goto end; | |
2365 | } | |
2366 | ||
2367 | assert(ret > 0); | |
2368 | assert(ret <= state->left_to_receive); | |
2369 | ||
2370 | state->left_to_receive -= ret; | |
2371 | state->received += ret; | |
2372 | ||
2373 | if (state->left_to_receive > 0) { | |
2374 | /* | |
2375 | * Can't transition to the protocol's next state, wait to | |
2376 | * receive the rest of the header. | |
2377 | */ | |
2378 | DBG3("Partial reception of control connection protocol header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2379 | state->received, state->left_to_receive, | |
2380 | conn->sock->fd); | |
2381 | goto end; | |
2382 | } | |
2383 | ||
2384 | /* Transition to next state: receiving the command's payload. */ | |
2385 | conn->protocol.ctrl.state_id = | |
2386 | CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
2387 | memcpy(&header, reception_buffer->data, sizeof(header)); | |
2388 | header.circuit_id = be64toh(header.circuit_id); | |
2389 | header.data_size = be64toh(header.data_size); | |
2390 | header.cmd = be32toh(header.cmd); | |
2391 | header.cmd_version = be32toh(header.cmd_version); | |
2392 | memcpy(&conn->protocol.ctrl.state.receive_payload.header, | |
2393 | &header, sizeof(header)); | |
2394 | ||
2395 | DBG("Done receiving control command header: fd = %i, cmd = %" PRIu32 ", cmd_version = %" PRIu32 ", payload size = %" PRIu64 " bytes", | |
2396 | conn->sock->fd, header.cmd, header.cmd_version, | |
2397 | header.data_size); | |
2398 | ||
2399 | if (header.data_size > DEFAULT_NETWORK_RELAYD_CTRL_MAX_PAYLOAD_SIZE) { | |
2400 | ERR("Command header indicates a payload (%" PRIu64 " bytes) that exceeds the maximal payload size allowed on a control connection.", | |
2401 | header.data_size); | |
2402 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2403 | goto end; | |
2404 | } | |
2405 | ||
2406 | conn->protocol.ctrl.state.receive_payload.left_to_receive = | |
2407 | header.data_size; | |
2408 | conn->protocol.ctrl.state.receive_payload.received = 0; | |
2409 | ret = lttng_dynamic_buffer_set_size(reception_buffer, | |
2410 | header.data_size); | |
2411 | if (ret) { | |
2412 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2413 | goto end; | |
2414 | } | |
2415 | ||
2416 | if (header.data_size == 0) { | |
2417 | /* | |
2418 | * Manually invoke the next state as the poll loop | |
2419 | * will not wake-up to allow us to proceed further. | |
2420 | */ | |
2421 | status = relay_process_control_receive_payload(conn); | |
2422 | } | |
2423 | end: | |
2424 | return status; | |
2425 | } | |
2426 | ||
2427 | /* | |
2428 | * Process the commands received on the control socket | |
2429 | */ | |
2430 | static enum relay_connection_status relay_process_control( | |
2431 | struct relay_connection *conn) | |
2432 | { | |
2433 | enum relay_connection_status status; | |
2434 | ||
2435 | switch (conn->protocol.ctrl.state_id) { | |
2436 | case CTRL_CONNECTION_STATE_RECEIVE_HEADER: | |
2437 | status = relay_process_control_receive_header(conn); | |
2438 | break; | |
2439 | case CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
2440 | status = relay_process_control_receive_payload(conn); | |
2441 | break; | |
2442 | default: | |
2443 | ERR("Unknown control connection protocol state encountered."); | |
2444 | abort(); | |
2445 | } | |
2446 | ||
2447 | return status; | |
2448 | } | |
2449 | ||
7d2f7452 DG |
2450 | /* |
2451 | * Handle index for a data stream. | |
2452 | * | |
7591bab1 | 2453 | * Called with the stream lock held. |
7d2f7452 DG |
2454 | * |
2455 | * Return 0 on success else a negative value. | |
2456 | */ | |
2457 | static int handle_index_data(struct relay_stream *stream, uint64_t net_seq_num, | |
44a2fbf1 | 2458 | bool rotate_index) |
7d2f7452 | 2459 | { |
7591bab1 MD |
2460 | int ret = 0; |
2461 | uint64_t data_offset; | |
2462 | struct relay_index *index; | |
7d2f7452 | 2463 | |
7d2f7452 DG |
2464 | /* Get data offset because we are about to update the index. */ |
2465 | data_offset = htobe64(stream->tracefile_size_current); | |
2466 | ||
65d66b19 MD |
2467 | DBG("handle_index_data: stream %" PRIu64 " net_seq_num %" PRIu64 " data offset %" PRIu64, |
2468 | stream->stream_handle, net_seq_num, stream->tracefile_size_current); | |
7591bab1 | 2469 | |
7d2f7452 | 2470 | /* |
7591bab1 MD |
2471 | * Lookup for an existing index for that stream id/sequence |
2472 | * number. If it exists, the control thread has already received the | |
2473 | * data for it, thus we need to write it to disk. | |
7d2f7452 | 2474 | */ |
7591bab1 | 2475 | index = relay_index_get_by_id_or_create(stream, net_seq_num); |
7d2f7452 | 2476 | if (!index) { |
7591bab1 MD |
2477 | ret = -1; |
2478 | goto end; | |
7d2f7452 DG |
2479 | } |
2480 | ||
e0547b83 MD |
2481 | if (rotate_index || !stream->index_file) { |
2482 | uint32_t major, minor; | |
7591bab1 | 2483 | |
e0547b83 MD |
2484 | /* Put ref on previous index_file. */ |
2485 | if (stream->index_file) { | |
2486 | lttng_index_file_put(stream->index_file); | |
2487 | stream->index_file = NULL; | |
7d2f7452 | 2488 | } |
e0547b83 MD |
2489 | major = stream->trace->session->major; |
2490 | minor = stream->trace->session->minor; | |
2491 | stream->index_file = lttng_index_file_create(stream->path_name, | |
2492 | stream->channel_name, | |
7591bab1 | 2493 | -1, -1, stream->tracefile_size, |
e0547b83 MD |
2494 | tracefile_array_get_file_index_head(stream->tfa), |
2495 | lttng_to_index_major(major, minor), | |
2496 | lttng_to_index_minor(major, minor)); | |
2497 | if (!stream->index_file) { | |
7591bab1 MD |
2498 | ret = -1; |
2499 | /* Put self-ref for this index due to error. */ | |
2500 | relay_index_put(index); | |
e0547b83 | 2501 | index = NULL; |
7591bab1 | 2502 | goto end; |
7d2f7452 | 2503 | } |
7d2f7452 DG |
2504 | } |
2505 | ||
e0547b83 | 2506 | if (relay_index_set_file(index, stream->index_file, data_offset)) { |
7591bab1 MD |
2507 | ret = -1; |
2508 | /* Put self-ref for this index due to error. */ | |
2509 | relay_index_put(index); | |
e0547b83 | 2510 | index = NULL; |
7591bab1 | 2511 | goto end; |
7d2f7452 DG |
2512 | } |
2513 | ||
7591bab1 MD |
2514 | ret = relay_index_try_flush(index); |
2515 | if (ret == 0) { | |
a44ca2ca MD |
2516 | tracefile_array_commit_seq(stream->tfa); |
2517 | stream->index_received_seqcount++; | |
7591bab1 MD |
2518 | } else if (ret > 0) { |
2519 | /* No flush. */ | |
2520 | ret = 0; | |
2521 | } else { | |
2522 | /* Put self-ref for this index due to error. */ | |
2523 | relay_index_put(index); | |
e0547b83 | 2524 | index = NULL; |
7591bab1 MD |
2525 | ret = -1; |
2526 | } | |
2527 | end: | |
7d2f7452 DG |
2528 | return ret; |
2529 | } | |
2530 | ||
44a2fbf1 JG |
2531 | static enum relay_connection_status relay_process_data_receive_header( |
2532 | struct relay_connection *conn) | |
b8aa1682 | 2533 | { |
44a2fbf1 JG |
2534 | int ret; |
2535 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2536 | struct data_connection_state_receive_header *state = | |
2537 | &conn->protocol.data.state.receive_header; | |
2538 | struct lttcomm_relayd_data_hdr header; | |
b8aa1682 | 2539 | struct relay_stream *stream; |
b8aa1682 | 2540 | |
44a2fbf1 JG |
2541 | assert(state->left_to_receive != 0); |
2542 | ||
2543 | ret = conn->sock->ops->recvmsg(conn->sock, | |
2544 | state->header_reception_buffer + state->received, | |
2545 | state->left_to_receive, MSG_DONTWAIT); | |
2546 | if (ret < 0) { | |
2547 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2548 | PERROR("Unable to receive data header on sock %d", conn->sock->fd); | |
2549 | status = RELAY_CONNECTION_STATUS_ERROR; | |
a6cd2b97 | 2550 | } |
44a2fbf1 JG |
2551 | goto end; |
2552 | } else if (ret == 0) { | |
2553 | /* Orderly shutdown. Not necessary to print an error. */ | |
2554 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
2555 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
b8aa1682 JD |
2556 | goto end; |
2557 | } | |
2558 | ||
44a2fbf1 JG |
2559 | assert(ret > 0); |
2560 | assert(ret <= state->left_to_receive); | |
2561 | ||
2562 | state->left_to_receive -= ret; | |
2563 | state->received += ret; | |
2564 | ||
2565 | if (state->left_to_receive > 0) { | |
2566 | /* | |
2567 | * Can't transition to the protocol's next state, wait to | |
2568 | * receive the rest of the header. | |
2569 | */ | |
2570 | DBG3("Partial reception of data connection header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
2571 | state->received, state->left_to_receive, | |
2572 | conn->sock->fd); | |
2573 | ret = 0; | |
7591bab1 | 2574 | goto end; |
b8aa1682 | 2575 | } |
b8aa1682 | 2576 | |
44a2fbf1 JG |
2577 | /* Transition to next state: receiving the payload. */ |
2578 | conn->protocol.data.state_id = DATA_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
173af62f | 2579 | |
44a2fbf1 JG |
2580 | memcpy(&header, state->header_reception_buffer, sizeof(header)); |
2581 | header.circuit_id = be64toh(header.circuit_id); | |
2582 | header.stream_id = be64toh(header.stream_id); | |
2583 | header.data_size = be32toh(header.data_size); | |
2584 | header.net_seq_num = be64toh(header.net_seq_num); | |
2585 | header.padding_size = be32toh(header.padding_size); | |
2586 | memcpy(&conn->protocol.data.state.receive_payload.header, &header, sizeof(header)); | |
2587 | ||
2588 | conn->protocol.data.state.receive_payload.left_to_receive = | |
2589 | header.data_size; | |
2590 | conn->protocol.data.state.receive_payload.received = 0; | |
2591 | conn->protocol.data.state.receive_payload.rotate_index = false; | |
2592 | ||
2593 | DBG("Received data connection header on fd %i: circuit_id = %" PRIu64 ", stream_id = %" PRIu64 ", data_size = %" PRIu32 ", net_seq_num = %" PRIu64 ", padding_size = %" PRIu32, | |
2594 | conn->sock->fd, header.circuit_id, | |
2595 | header.stream_id, header.data_size, | |
2596 | header.net_seq_num, header.padding_size); | |
2597 | ||
2598 | stream = stream_get_by_id(header.stream_id); | |
2599 | if (!stream) { | |
2600 | DBG("relay_process_data_receive_payload: Cannot find stream %" PRIu64, | |
2601 | header.stream_id); | |
2602 | /* Protocol error. */ | |
2603 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2604 | goto end; | |
2605 | } | |
b8aa1682 | 2606 | |
7591bab1 MD |
2607 | pthread_mutex_lock(&stream->lock); |
2608 | ||
1c20f0e2 | 2609 | /* Check if a rotation is needed. */ |
0f907de1 | 2610 | if (stream->tracefile_size > 0 && |
44a2fbf1 | 2611 | (stream->tracefile_size_current + header.data_size) > |
0f907de1 | 2612 | stream->tracefile_size) { |
a44ca2ca MD |
2613 | uint64_t old_id, new_id; |
2614 | ||
2615 | old_id = tracefile_array_get_file_index_head(stream->tfa); | |
2616 | tracefile_array_file_rotate(stream->tfa); | |
2617 | ||
2618 | /* new_id is updated by utils_rotate_stream_file. */ | |
2619 | new_id = old_id; | |
6b6b9a5a | 2620 | |
7591bab1 MD |
2621 | ret = utils_rotate_stream_file(stream->path_name, |
2622 | stream->channel_name, stream->tracefile_size, | |
2623 | stream->tracefile_count, -1, | |
2624 | -1, stream->stream_fd->fd, | |
a44ca2ca | 2625 | &new_id, &stream->stream_fd->fd); |
0f907de1 | 2626 | if (ret < 0) { |
44a2fbf1 JG |
2627 | ERR("Failed to rotate stream output file"); |
2628 | status = RELAY_CONNECTION_STATUS_ERROR; | |
7591bab1 MD |
2629 | goto end_stream_unlock; |
2630 | } | |
44a2fbf1 | 2631 | |
7591bab1 MD |
2632 | /* |
2633 | * Reset current size because we just performed a stream | |
2634 | * rotation. | |
2635 | */ | |
a6976990 | 2636 | stream->tracefile_size_current = 0; |
44a2fbf1 | 2637 | conn->protocol.data.state.receive_payload.rotate_index = true; |
1c20f0e2 JD |
2638 | } |
2639 | ||
44a2fbf1 JG |
2640 | ret = 0; |
2641 | end_stream_unlock: | |
2642 | pthread_mutex_unlock(&stream->lock); | |
2643 | stream_put(stream); | |
2644 | end: | |
2645 | return status; | |
2646 | } | |
2647 | ||
2648 | static enum relay_connection_status relay_process_data_receive_payload( | |
2649 | struct relay_connection *conn) | |
2650 | { | |
2651 | int ret; | |
2652 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; | |
2653 | struct relay_stream *stream; | |
2654 | struct data_connection_state_receive_payload *state = | |
2655 | &conn->protocol.data.state.receive_payload; | |
2656 | const size_t chunk_size = RECV_DATA_BUFFER_SIZE; | |
2657 | char data_buffer[chunk_size]; | |
2658 | bool partial_recv = false; | |
2659 | bool new_stream = false, close_requested = false; | |
2660 | uint64_t left_to_receive = state->left_to_receive; | |
2661 | struct relay_session *session; | |
2662 | ||
2663 | DBG3("Receiving data for stream id %" PRIu64 " seqnum %" PRIu64 ", %" PRIu64" bytes received, %" PRIu64 " bytes left to receive", | |
2664 | state->header.stream_id, state->header.net_seq_num, | |
2665 | state->received, left_to_receive); | |
2666 | ||
2667 | stream = stream_get_by_id(state->header.stream_id); | |
2668 | if (!stream) { | |
2669 | /* Protocol error. */ | |
2670 | ERR("relay_process_data_receive_payload: cannot find stream %" PRIu64, | |
2671 | state->header.stream_id); | |
2672 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2673 | goto end; | |
2674 | } | |
2675 | ||
2676 | pthread_mutex_lock(&stream->lock); | |
2677 | session = stream->trace->session; | |
2678 | if (!conn->session) { | |
2679 | ret = connection_set_session(conn, session); | |
2680 | if (ret) { | |
2681 | status = RELAY_CONNECTION_STATUS_ERROR; | |
7591bab1 | 2682 | goto end_stream_unlock; |
1c20f0e2 | 2683 | } |
1c20f0e2 JD |
2684 | } |
2685 | ||
44a2fbf1 JG |
2686 | /* |
2687 | * The size of the "chunk" received on any iteration is bounded by: | |
2688 | * - the data left to receive, | |
2689 | * - the data immediately available on the socket, | |
2690 | * - the on-stack data buffer | |
2691 | */ | |
2692 | while (left_to_receive > 0 && !partial_recv) { | |
2693 | ssize_t write_ret; | |
2694 | size_t recv_size = min(left_to_receive, chunk_size); | |
1d4dfdef | 2695 | |
44a2fbf1 JG |
2696 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, |
2697 | recv_size, MSG_DONTWAIT); | |
2698 | if (ret < 0) { | |
2699 | if (errno != EAGAIN && errno != EWOULDBLOCK) { | |
2700 | PERROR("Socket %d error", conn->sock->fd); | |
2701 | status = RELAY_CONNECTION_STATUS_ERROR; | |
0848dba7 | 2702 | } |
0848dba7 | 2703 | goto end_stream_unlock; |
44a2fbf1 JG |
2704 | } else if (ret == 0) { |
2705 | /* No more data ready to be consumed on socket. */ | |
2706 | DBG3("No more data ready for consumption on data socket of stream id %" PRIu64, | |
2707 | state->header.stream_id); | |
2708 | status = RELAY_CONNECTION_STATUS_CLOSED; | |
2709 | break; | |
2710 | } else if (ret < (int) recv_size) { | |
2711 | /* | |
2712 | * All the data available on the socket has been | |
2713 | * consumed. | |
2714 | */ | |
2715 | partial_recv = true; | |
0848dba7 MD |
2716 | } |
2717 | ||
44a2fbf1 JG |
2718 | recv_size = ret; |
2719 | ||
0848dba7 | 2720 | /* Write data to stream output fd. */ |
44a2fbf1 | 2721 | write_ret = lttng_write(stream->stream_fd->fd, data_buffer, |
0848dba7 | 2722 | recv_size); |
44a2fbf1 | 2723 | if (write_ret < (ssize_t) recv_size) { |
0848dba7 | 2724 | ERR("Relay error writing data to file"); |
44a2fbf1 | 2725 | status = RELAY_CONNECTION_STATUS_ERROR; |
0848dba7 MD |
2726 | goto end_stream_unlock; |
2727 | } | |
2728 | ||
44a2fbf1 JG |
2729 | left_to_receive -= recv_size; |
2730 | state->received += recv_size; | |
2731 | state->left_to_receive = left_to_receive; | |
2732 | ||
0848dba7 | 2733 | DBG2("Relay wrote %zd bytes to tracefile for stream id %" PRIu64, |
44a2fbf1 JG |
2734 | write_ret, stream->stream_handle); |
2735 | } | |
2736 | ||
2737 | if (state->left_to_receive > 0) { | |
2738 | /* | |
2739 | * Did not receive all the data expected, wait for more data to | |
2740 | * become available on the socket. | |
2741 | */ | |
2742 | DBG3("Partial receive on data connection of stream id %" PRIu64 ", %" PRIu64 " bytes received, %" PRIu64 " bytes left to receive", | |
2743 | state->header.stream_id, state->received, | |
2744 | state->left_to_receive); | |
2745 | goto end_stream_unlock; | |
0848dba7 | 2746 | } |
5ab7344e | 2747 | |
7591bab1 | 2748 | ret = write_padding_to_file(stream->stream_fd->fd, |
44a2fbf1 JG |
2749 | state->header.padding_size); |
2750 | if ((int64_t) ret < (int64_t) state->header.padding_size) { | |
65d66b19 | 2751 | ERR("write_padding_to_file: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
44a2fbf1 JG |
2752 | stream->stream_handle, |
2753 | state->header.net_seq_num, ret); | |
2754 | status = RELAY_CONNECTION_STATUS_ERROR; | |
7591bab1 | 2755 | goto end_stream_unlock; |
1d4dfdef | 2756 | } |
44a2fbf1 JG |
2757 | |
2758 | ||
2759 | if (session->minor >= 4 && !session->snapshot) { | |
2760 | ret = handle_index_data(stream, state->header.net_seq_num, | |
2761 | state->rotate_index); | |
2762 | if (ret < 0) { | |
2763 | ERR("handle_index_data: fail stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", | |
2764 | stream->stream_handle, | |
2765 | state->header.net_seq_num, ret); | |
2766 | status = RELAY_CONNECTION_STATUS_ERROR; | |
2767 | goto end_stream_unlock; | |
2768 | } | |
2769 | } | |
2770 | ||
2771 | stream->tracefile_size_current += state->header.data_size + | |
2772 | state->header.padding_size; | |
2773 | ||
c0bae11d MD |
2774 | if (stream->prev_seq == -1ULL) { |
2775 | new_stream = true; | |
2776 | } | |
2777 | ||
44a2fbf1 JG |
2778 | stream->prev_seq = state->header.net_seq_num; |
2779 | ||
2780 | /* | |
2781 | * Resetting the protocol state (to RECEIVE_HEADER) will trash the | |
2782 | * contents of *state which are aliased (union) to the same location as | |
2783 | * the new state. Don't use it beyond this point. | |
2784 | */ | |
2785 | connection_reset_protocol_state(conn); | |
2786 | state = NULL; | |
173af62f | 2787 | |
7591bab1 | 2788 | end_stream_unlock: |
bda7c7b9 | 2789 | close_requested = stream->close_requested; |
7591bab1 | 2790 | pthread_mutex_unlock(&stream->lock); |
44a2fbf1 | 2791 | if (close_requested && left_to_receive == 0) { |
bda7c7b9 JG |
2792 | try_stream_close(stream); |
2793 | } | |
2794 | ||
c0bae11d MD |
2795 | if (new_stream) { |
2796 | pthread_mutex_lock(&session->lock); | |
2797 | uatomic_set(&session->new_streams, 1); | |
2798 | pthread_mutex_unlock(&session->lock); | |
2799 | } | |
44a2fbf1 | 2800 | |
7591bab1 | 2801 | stream_put(stream); |
b8aa1682 | 2802 | end: |
44a2fbf1 JG |
2803 | return status; |
2804 | } | |
2805 | ||
2806 | /* | |
2807 | * relay_process_data: Process the data received on the data socket | |
2808 | */ | |
2809 | static enum relay_connection_status relay_process_data( | |
2810 | struct relay_connection *conn) | |
2811 | { | |
2812 | enum relay_connection_status status; | |
2813 | ||
2814 | switch (conn->protocol.data.state_id) { | |
2815 | case DATA_CONNECTION_STATE_RECEIVE_HEADER: | |
2816 | status = relay_process_data_receive_header(conn); | |
2817 | break; | |
2818 | case DATA_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
2819 | status = relay_process_data_receive_payload(conn); | |
2820 | break; | |
2821 | default: | |
2822 | ERR("Unexpected data connection communication state."); | |
2823 | abort(); | |
2824 | } | |
2825 | ||
2826 | return status; | |
b8aa1682 JD |
2827 | } |
2828 | ||
7591bab1 | 2829 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
2830 | { |
2831 | int ret; | |
2832 | ||
58eb9381 | 2833 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
2834 | |
2835 | ret = close(pollfd); | |
2836 | if (ret < 0) { | |
2837 | ERR("Closing pollfd %d", pollfd); | |
2838 | } | |
2839 | } | |
2840 | ||
7591bab1 MD |
2841 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
2842 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 2843 | { |
7591bab1 | 2844 | const char *type_str; |
2a174661 | 2845 | |
7591bab1 MD |
2846 | switch (conn->type) { |
2847 | case RELAY_DATA: | |
2848 | type_str = "Data"; | |
2849 | break; | |
2850 | case RELAY_CONTROL: | |
2851 | type_str = "Control"; | |
2852 | break; | |
2853 | case RELAY_VIEWER_COMMAND: | |
2854 | type_str = "Viewer Command"; | |
2855 | break; | |
2856 | case RELAY_VIEWER_NOTIFICATION: | |
2857 | type_str = "Viewer Notification"; | |
2858 | break; | |
2859 | default: | |
2860 | type_str = "Unknown"; | |
9d1bbf21 | 2861 | } |
7591bab1 MD |
2862 | cleanup_connection_pollfd(events, pollfd); |
2863 | connection_put(conn); | |
2864 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
2865 | } |
2866 | ||
2867 | /* | |
2868 | * This thread does the actual work | |
2869 | */ | |
7591bab1 | 2870 | static void *relay_thread_worker(void *data) |
b8aa1682 | 2871 | { |
beaad64c DG |
2872 | int ret, err = -1, last_seen_data_fd = -1; |
2873 | uint32_t nb_fd; | |
b8aa1682 JD |
2874 | struct lttng_poll_event events; |
2875 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 2876 | struct lttng_ht_iter iter; |
90e7d72f | 2877 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
2878 | |
2879 | DBG("[thread] Relay worker started"); | |
2880 | ||
9d1bbf21 MD |
2881 | rcu_register_thread(); |
2882 | ||
55706a7d MD |
2883 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
2884 | ||
9b5e0863 MD |
2885 | if (testpoint(relayd_thread_worker)) { |
2886 | goto error_testpoint; | |
2887 | } | |
2888 | ||
f385ae0a MD |
2889 | health_code_update(); |
2890 | ||
b8aa1682 JD |
2891 | /* table of connections indexed on socket */ |
2892 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
2893 | if (!relay_connections_ht) { |
2894 | goto relay_connections_ht_error; | |
2895 | } | |
b8aa1682 | 2896 | |
b8aa1682 JD |
2897 | ret = create_thread_poll_set(&events, 2); |
2898 | if (ret < 0) { | |
2899 | goto error_poll_create; | |
2900 | } | |
2901 | ||
58eb9381 | 2902 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
2903 | if (ret < 0) { |
2904 | goto error; | |
2905 | } | |
2906 | ||
beaad64c | 2907 | restart: |
b8aa1682 | 2908 | while (1) { |
beaad64c DG |
2909 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
2910 | ||
f385ae0a MD |
2911 | health_code_update(); |
2912 | ||
b8aa1682 | 2913 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 2914 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 2915 | health_poll_entry(); |
b8aa1682 | 2916 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 2917 | health_poll_exit(); |
b8aa1682 JD |
2918 | if (ret < 0) { |
2919 | /* | |
2920 | * Restart interrupted system call. | |
2921 | */ | |
2922 | if (errno == EINTR) { | |
2923 | goto restart; | |
2924 | } | |
2925 | goto error; | |
2926 | } | |
2927 | ||
0d9c5d77 DG |
2928 | nb_fd = ret; |
2929 | ||
beaad64c | 2930 | /* |
7591bab1 MD |
2931 | * Process control. The control connection is |
2932 | * prioritized so we don't starve it with high | |
2933 | * throughput tracing data on the data connection. | |
beaad64c | 2934 | */ |
b8aa1682 JD |
2935 | for (i = 0; i < nb_fd; i++) { |
2936 | /* Fetch once the poll data */ | |
beaad64c DG |
2937 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
2938 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 2939 | |
f385ae0a MD |
2940 | health_code_update(); |
2941 | ||
fd20dac9 | 2942 | if (!revents) { |
7591bab1 MD |
2943 | /* |
2944 | * No activity for this FD (poll | |
2945 | * implementation). | |
2946 | */ | |
fd20dac9 MD |
2947 | continue; |
2948 | } | |
2949 | ||
b8aa1682 JD |
2950 | /* Thread quit pipe has been closed. Killing thread. */ |
2951 | ret = check_thread_quit_pipe(pollfd, revents); | |
2952 | if (ret) { | |
095a4ae5 MD |
2953 | err = 0; |
2954 | goto exit; | |
b8aa1682 JD |
2955 | } |
2956 | ||
58eb9381 DG |
2957 | /* Inspect the relay conn pipe for new connection */ |
2958 | if (pollfd == relay_conn_pipe[0]) { | |
03e43155 | 2959 | if (revents & LPOLLIN) { |
90e7d72f JG |
2960 | struct relay_connection *conn; |
2961 | ||
58eb9381 | 2962 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
2963 | if (ret < 0) { |
2964 | goto error; | |
2965 | } | |
58eb9381 DG |
2966 | lttng_poll_add(&events, conn->sock->fd, |
2967 | LPOLLIN | LPOLLRDHUP); | |
7591bab1 | 2968 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 2969 | DBG("Connection socket %d added", conn->sock->fd); |
03e43155 MD |
2970 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
2971 | ERR("Relay connection pipe error"); | |
2972 | goto error; | |
2973 | } else { | |
2974 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
2975 | goto error; | |
b8aa1682 | 2976 | } |
58eb9381 | 2977 | } else { |
90e7d72f JG |
2978 | struct relay_connection *ctrl_conn; |
2979 | ||
7591bab1 | 2980 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 2981 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 2982 | assert(ctrl_conn); |
58eb9381 | 2983 | |
03e43155 MD |
2984 | if (ctrl_conn->type == RELAY_DATA) { |
2985 | if (revents & LPOLLIN) { | |
beaad64c DG |
2986 | /* |
2987 | * Flag the last seen data fd not deleted. It will be | |
2988 | * used as the last seen fd if any fd gets deleted in | |
2989 | * this first loop. | |
2990 | */ | |
2991 | last_notdel_data_fd = pollfd; | |
2992 | } | |
03e43155 MD |
2993 | goto put_ctrl_connection; |
2994 | } | |
2995 | assert(ctrl_conn->type == RELAY_CONTROL); | |
2996 | ||
2997 | if (revents & LPOLLIN) { | |
44a2fbf1 JG |
2998 | enum relay_connection_status status; |
2999 | ||
3000 | status = relay_process_control(ctrl_conn); | |
3001 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
3002 | /* | |
3003 | * On socket error flag the session as aborted to force | |
3004 | * the cleanup of its stream otherwise it can leak | |
3005 | * during the lifetime of the relayd. | |
3006 | * | |
3007 | * This prevents situations in which streams can be | |
3008 | * left opened because an index was received, the | |
3009 | * control connection is closed, and the data | |
3010 | * connection is closed (uncleanly) before the packet's | |
3011 | * data provided. | |
3012 | * | |
3013 | * Since the control connection encountered an error, | |
3014 | * it is okay to be conservative and close the | |
3015 | * session right now as we can't rely on the protocol | |
3016 | * being respected anymore. | |
3017 | */ | |
3018 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3019 | session_abort(ctrl_conn->session); | |
03e43155 | 3020 | } |
44a2fbf1 JG |
3021 | |
3022 | /* Clear the connection on error or close. */ | |
3023 | relay_thread_close_connection(&events, | |
3024 | pollfd, | |
3025 | ctrl_conn); | |
03e43155 | 3026 | } |
44a2fbf1 | 3027 | seen_control = 1; |
03e43155 MD |
3028 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3029 | relay_thread_close_connection(&events, | |
3030 | pollfd, ctrl_conn); | |
3031 | if (last_seen_data_fd == pollfd) { | |
3032 | last_seen_data_fd = last_notdel_data_fd; | |
3033 | } | |
58eb9381 | 3034 | } else { |
03e43155 MD |
3035 | ERR("Unexpected poll events %u for control sock %d", |
3036 | revents, pollfd); | |
3037 | connection_put(ctrl_conn); | |
3038 | goto error; | |
beaad64c | 3039 | } |
03e43155 | 3040 | put_ctrl_connection: |
7591bab1 | 3041 | connection_put(ctrl_conn); |
beaad64c DG |
3042 | } |
3043 | } | |
3044 | ||
3045 | /* | |
3046 | * The last loop handled a control request, go back to poll to make | |
3047 | * sure we prioritise the control socket. | |
3048 | */ | |
3049 | if (seen_control) { | |
3050 | continue; | |
3051 | } | |
3052 | ||
3053 | if (last_seen_data_fd >= 0) { | |
3054 | for (i = 0; i < nb_fd; i++) { | |
3055 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
3056 | |
3057 | health_code_update(); | |
3058 | ||
beaad64c DG |
3059 | if (last_seen_data_fd == pollfd) { |
3060 | idx = i; | |
3061 | break; | |
3062 | } | |
3063 | } | |
3064 | } | |
3065 | ||
3066 | /* Process data connection. */ | |
3067 | for (i = idx + 1; i < nb_fd; i++) { | |
3068 | /* Fetch the poll data. */ | |
3069 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
3070 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 3071 | struct relay_connection *data_conn; |
beaad64c | 3072 | |
f385ae0a MD |
3073 | health_code_update(); |
3074 | ||
fd20dac9 MD |
3075 | if (!revents) { |
3076 | /* No activity for this FD (poll implementation). */ | |
3077 | continue; | |
3078 | } | |
3079 | ||
beaad64c | 3080 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 3081 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
3082 | continue; |
3083 | } | |
3084 | ||
7591bab1 | 3085 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 3086 | if (!data_conn) { |
fd20dac9 | 3087 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
3088 | continue; |
3089 | } | |
03e43155 MD |
3090 | if (data_conn->type == RELAY_CONTROL) { |
3091 | goto put_data_connection; | |
3092 | } | |
3093 | assert(data_conn->type == RELAY_DATA); | |
fd20dac9 MD |
3094 | |
3095 | if (revents & LPOLLIN) { | |
44a2fbf1 JG |
3096 | enum relay_connection_status status; |
3097 | ||
3098 | status = relay_process_data(data_conn); | |
3099 | /* Connection closed or error. */ | |
3100 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
3101 | /* | |
3102 | * On socket error flag the session as aborted to force | |
3103 | * the cleanup of its stream otherwise it can leak | |
3104 | * during the lifetime of the relayd. | |
3105 | * | |
3106 | * This prevents situations in which streams can be | |
3107 | * left opened because an index was received, the | |
3108 | * control connection is closed, and the data | |
3109 | * connection is closed (uncleanly) before the packet's | |
3110 | * data provided. | |
3111 | * | |
3112 | * Since the data connection encountered an error, | |
3113 | * it is okay to be conservative and close the | |
3114 | * session right now as we can't rely on the protocol | |
3115 | * being respected anymore. | |
3116 | */ | |
3117 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3118 | session_abort(data_conn->session); | |
3119 | } | |
7591bab1 | 3120 | relay_thread_close_connection(&events, pollfd, |
03e43155 | 3121 | data_conn); |
fd20dac9 MD |
3122 | /* |
3123 | * Every goto restart call sets the last seen fd where | |
3124 | * here we don't really care since we gracefully | |
3125 | * continue the loop after the connection is deleted. | |
3126 | */ | |
3127 | } else { | |
3128 | /* Keep last seen port. */ | |
3129 | last_seen_data_fd = pollfd; | |
7591bab1 | 3130 | connection_put(data_conn); |
fd20dac9 | 3131 | goto restart; |
b8aa1682 | 3132 | } |
03e43155 MD |
3133 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3134 | relay_thread_close_connection(&events, pollfd, | |
3135 | data_conn); | |
3136 | } else { | |
3137 | ERR("Unknown poll events %u for data sock %d", | |
3138 | revents, pollfd); | |
b8aa1682 | 3139 | } |
03e43155 | 3140 | put_data_connection: |
7591bab1 | 3141 | connection_put(data_conn); |
b8aa1682 | 3142 | } |
beaad64c | 3143 | last_seen_data_fd = -1; |
b8aa1682 JD |
3144 | } |
3145 | ||
f385ae0a MD |
3146 | /* Normal exit, no error */ |
3147 | ret = 0; | |
3148 | ||
095a4ae5 | 3149 | exit: |
b8aa1682 | 3150 | error: |
58eb9381 | 3151 | /* Cleanup reamaining connection object. */ |
9d1bbf21 | 3152 | rcu_read_lock(); |
90e7d72f JG |
3153 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
3154 | destroy_conn, | |
58eb9381 | 3155 | sock_n.node) { |
f385ae0a | 3156 | health_code_update(); |
06a68401 | 3157 | |
44a2fbf1 | 3158 | session_abort(destroy_conn->session); |
06a68401 | 3159 | |
7591bab1 MD |
3160 | /* |
3161 | * No need to grab another ref, because we own | |
3162 | * destroy_conn. | |
3163 | */ | |
3164 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
3165 | destroy_conn); | |
b8aa1682 | 3166 | } |
94d49140 | 3167 | rcu_read_unlock(); |
7591bab1 MD |
3168 | |
3169 | lttng_poll_clean(&events); | |
7d2f7452 | 3170 | error_poll_create: |
b8aa1682 | 3171 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 3172 | relay_connections_ht_error: |
58eb9381 DG |
3173 | /* Close relay conn pipes */ |
3174 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
3175 | if (err) { |
3176 | DBG("Thread exited with error"); | |
3177 | } | |
b8aa1682 | 3178 | DBG("Worker thread cleanup complete"); |
9b5e0863 | 3179 | error_testpoint: |
f385ae0a MD |
3180 | if (err) { |
3181 | health_error(); | |
3182 | ERR("Health error occurred in %s", __func__); | |
3183 | } | |
3184 | health_unregister(health_relayd); | |
9d1bbf21 | 3185 | rcu_unregister_thread(); |
b4aacfdc | 3186 | lttng_relay_stop_threads(); |
b8aa1682 JD |
3187 | return NULL; |
3188 | } | |
3189 | ||
3190 | /* | |
3191 | * Create the relay command pipe to wake thread_manage_apps. | |
3192 | * Closed in cleanup(). | |
3193 | */ | |
58eb9381 | 3194 | static int create_relay_conn_pipe(void) |
b8aa1682 | 3195 | { |
a02de639 | 3196 | int ret; |
b8aa1682 | 3197 | |
58eb9381 | 3198 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 3199 | |
b8aa1682 JD |
3200 | return ret; |
3201 | } | |
3202 | ||
3203 | /* | |
3204 | * main | |
3205 | */ | |
3206 | int main(int argc, char **argv) | |
3207 | { | |
178a0557 | 3208 | int ret = 0, retval = 0; |
b8aa1682 JD |
3209 | void *status; |
3210 | ||
c7cc870e JR |
3211 | /* Parse environment variables */ |
3212 | parse_env_options(); | |
3213 | ||
3214 | /* | |
3215 | * Parse arguments. | |
3216 | * Command line arguments overwrite environment. | |
3217 | */ | |
b8aa1682 | 3218 | progname = argv[0]; |
178a0557 MD |
3219 | if (set_options(argc, argv)) { |
3220 | retval = -1; | |
3221 | goto exit_options; | |
b8aa1682 JD |
3222 | } |
3223 | ||
178a0557 MD |
3224 | if (set_signal_handler()) { |
3225 | retval = -1; | |
3226 | goto exit_options; | |
b8aa1682 JD |
3227 | } |
3228 | ||
8ff7624a JR |
3229 | relayd_config_log(); |
3230 | ||
3231 | if (opt_print_version) { | |
3232 | print_version(); | |
3233 | retval = 0; | |
3234 | goto exit_options; | |
3235 | } | |
3236 | ||
4d513a50 DG |
3237 | /* Try to create directory if -o, --output is specified. */ |
3238 | if (opt_output_path) { | |
994fa64f DG |
3239 | if (*opt_output_path != '/') { |
3240 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
3241 | retval = -1; |
3242 | goto exit_options; | |
994fa64f DG |
3243 | } |
3244 | ||
d77dded2 JG |
3245 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
3246 | -1, -1); | |
4d513a50 DG |
3247 | if (ret < 0) { |
3248 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
3249 | retval = -1; |
3250 | goto exit_options; | |
4d513a50 DG |
3251 | } |
3252 | } | |
3253 | ||
b8aa1682 | 3254 | /* Daemonize */ |
b5218ffb | 3255 | if (opt_daemon || opt_background) { |
3fd27398 MD |
3256 | int i; |
3257 | ||
3258 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
3259 | !opt_background); | |
b8aa1682 | 3260 | if (ret < 0) { |
178a0557 MD |
3261 | retval = -1; |
3262 | goto exit_options; | |
b8aa1682 | 3263 | } |
3fd27398 MD |
3264 | |
3265 | /* | |
3266 | * We are in the child. Make sure all other file | |
3267 | * descriptors are closed, in case we are called with | |
3268 | * more opened file descriptors than the standard ones. | |
3269 | */ | |
3270 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
3271 | (void) close(i); | |
3272 | } | |
3273 | } | |
3274 | ||
76375580 JR |
3275 | |
3276 | if (opt_working_directory) { | |
3277 | ret = utils_change_working_dir(opt_working_directory); | |
3278 | if (ret) { | |
3279 | ERR("Changing working directory"); | |
3280 | goto exit_options; | |
3281 | } | |
3282 | } | |
3283 | ||
178a0557 MD |
3284 | /* Initialize thread health monitoring */ |
3285 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
3286 | if (!health_relayd) { | |
3287 | PERROR("health_app_create error"); | |
3288 | retval = -1; | |
3289 | goto exit_health_app_create; | |
3290 | } | |
3291 | ||
3fd27398 | 3292 | /* Create thread quit pipe */ |
178a0557 MD |
3293 | if (init_thread_quit_pipe()) { |
3294 | retval = -1; | |
3295 | goto exit_init_data; | |
b8aa1682 JD |
3296 | } |
3297 | ||
b8aa1682 | 3298 | /* Setup the thread apps communication pipe. */ |
178a0557 MD |
3299 | if (create_relay_conn_pipe()) { |
3300 | retval = -1; | |
3301 | goto exit_init_data; | |
b8aa1682 JD |
3302 | } |
3303 | ||
3304 | /* Init relay command queue. */ | |
8bdee6e2 | 3305 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 | 3306 | |
554831e7 MD |
3307 | /* Initialize communication library */ |
3308 | lttcomm_init(); | |
87e45c13 | 3309 | lttcomm_inet_init(); |
554831e7 | 3310 | |
d3e2ba59 | 3311 | /* tables of sessions indexed by session ID */ |
7591bab1 MD |
3312 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3313 | if (!sessions_ht) { | |
178a0557 MD |
3314 | retval = -1; |
3315 | goto exit_init_data; | |
d3e2ba59 JD |
3316 | } |
3317 | ||
3318 | /* tables of streams indexed by stream ID */ | |
2a174661 | 3319 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 3320 | if (!relay_streams_ht) { |
178a0557 MD |
3321 | retval = -1; |
3322 | goto exit_init_data; | |
d3e2ba59 JD |
3323 | } |
3324 | ||
3325 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
3326 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
3327 | if (!viewer_streams_ht) { | |
178a0557 MD |
3328 | retval = -1; |
3329 | goto exit_init_data; | |
55706a7d MD |
3330 | } |
3331 | ||
65931c8b | 3332 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
3333 | if (ret) { |
3334 | retval = -1; | |
3335 | goto exit_health_quit_pipe; | |
65931c8b MD |
3336 | } |
3337 | ||
3338 | /* Create thread to manage the client socket */ | |
1a1a34b4 | 3339 | ret = pthread_create(&health_thread, default_pthread_attr(), |
65931c8b | 3340 | thread_manage_health, (void *) NULL); |
178a0557 MD |
3341 | if (ret) { |
3342 | errno = ret; | |
65931c8b | 3343 | PERROR("pthread_create health"); |
178a0557 MD |
3344 | retval = -1; |
3345 | goto exit_health_thread; | |
65931c8b MD |
3346 | } |
3347 | ||
b8aa1682 | 3348 | /* Setup the dispatcher thread */ |
1a1a34b4 | 3349 | ret = pthread_create(&dispatcher_thread, default_pthread_attr(), |
b8aa1682 | 3350 | relay_thread_dispatcher, (void *) NULL); |
178a0557 MD |
3351 | if (ret) { |
3352 | errno = ret; | |
b8aa1682 | 3353 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
3354 | retval = -1; |
3355 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
3356 | } |
3357 | ||
3358 | /* Setup the worker thread */ | |
1a1a34b4 | 3359 | ret = pthread_create(&worker_thread, default_pthread_attr(), |
7591bab1 | 3360 | relay_thread_worker, NULL); |
178a0557 MD |
3361 | if (ret) { |
3362 | errno = ret; | |
b8aa1682 | 3363 | PERROR("pthread_create worker"); |
178a0557 MD |
3364 | retval = -1; |
3365 | goto exit_worker_thread; | |
b8aa1682 JD |
3366 | } |
3367 | ||
3368 | /* Setup the listener thread */ | |
1a1a34b4 | 3369 | ret = pthread_create(&listener_thread, default_pthread_attr(), |
b8aa1682 | 3370 | relay_thread_listener, (void *) NULL); |
178a0557 MD |
3371 | if (ret) { |
3372 | errno = ret; | |
b8aa1682 | 3373 | PERROR("pthread_create listener"); |
178a0557 MD |
3374 | retval = -1; |
3375 | goto exit_listener_thread; | |
b8aa1682 JD |
3376 | } |
3377 | ||
7591bab1 | 3378 | ret = relayd_live_create(live_uri); |
178a0557 | 3379 | if (ret) { |
d3e2ba59 | 3380 | ERR("Starting live viewer threads"); |
178a0557 | 3381 | retval = -1; |
50138f51 | 3382 | goto exit_live; |
d3e2ba59 JD |
3383 | } |
3384 | ||
178a0557 MD |
3385 | /* |
3386 | * This is where we start awaiting program completion (e.g. through | |
3387 | * signal that asks threads to teardown). | |
3388 | */ | |
3389 | ||
3390 | ret = relayd_live_join(); | |
3391 | if (ret) { | |
3392 | retval = -1; | |
3393 | } | |
50138f51 | 3394 | exit_live: |
178a0557 | 3395 | |
b8aa1682 | 3396 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
3397 | if (ret) { |
3398 | errno = ret; | |
3399 | PERROR("pthread_join listener_thread"); | |
3400 | retval = -1; | |
b8aa1682 JD |
3401 | } |
3402 | ||
178a0557 | 3403 | exit_listener_thread: |
b8aa1682 | 3404 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
3405 | if (ret) { |
3406 | errno = ret; | |
3407 | PERROR("pthread_join worker_thread"); | |
3408 | retval = -1; | |
b8aa1682 JD |
3409 | } |
3410 | ||
178a0557 | 3411 | exit_worker_thread: |
b8aa1682 | 3412 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
3413 | if (ret) { |
3414 | errno = ret; | |
3415 | PERROR("pthread_join dispatcher_thread"); | |
3416 | retval = -1; | |
b8aa1682 | 3417 | } |
178a0557 | 3418 | exit_dispatcher_thread: |
42415026 | 3419 | |
65931c8b | 3420 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
3421 | if (ret) { |
3422 | errno = ret; | |
3423 | PERROR("pthread_join health_thread"); | |
3424 | retval = -1; | |
65931c8b | 3425 | } |
178a0557 | 3426 | exit_health_thread: |
65931c8b | 3427 | |
65931c8b | 3428 | utils_close_pipe(health_quit_pipe); |
178a0557 | 3429 | exit_health_quit_pipe: |
65931c8b | 3430 | |
178a0557 | 3431 | exit_init_data: |
55706a7d | 3432 | health_app_destroy(health_relayd); |
55706a7d | 3433 | exit_health_app_create: |
178a0557 | 3434 | exit_options: |
2668465a MD |
3435 | /* |
3436 | * Wait for all pending call_rcu work to complete before tearing | |
3437 | * down data structures. call_rcu worker may be trying to | |
3438 | * perform lookups in those structures. | |
3439 | */ | |
3440 | rcu_barrier(); | |
7591bab1 MD |
3441 | relayd_cleanup(); |
3442 | ||
3443 | /* Ensure all prior call_rcu are done. */ | |
3444 | rcu_barrier(); | |
d3e2ba59 | 3445 | |
178a0557 | 3446 | if (!retval) { |
b8aa1682 | 3447 | exit(EXIT_SUCCESS); |
178a0557 MD |
3448 | } else { |
3449 | exit(EXIT_FAILURE); | |
b8aa1682 | 3450 | } |
b8aa1682 | 3451 | } |